Preferred Citation: Quigley, John M., and Daniel L. Rubinfeld, editors American Domestic Priorities: An Economic Appraisal. Berkeley:  University of California Press,  c1985 1985. http://ark.cdlib.org/ark:/13030/ft4b69n90z/

American Domestic Priorities An Economic Appraisal Edited by John M. Quigley Daniel L. Rubinfeld UNIVERSITY OF CALIFORNIA PRESS Berkeley · Los Angeles · Oxford © 1985 The Regents of the University of California

Preferred Citation: Quigley, John M., and Daniel L. Rubinfeld, editors American Domestic Priorities: An Economic Appraisal. Berkeley:  University of California Press,  c1985 1985. http://ark.cdlib.org/ark:/13030/ft4b69n90z/

PREFACE

Domestic programs and their budgetary implications will be scrutinized closely during the 99th Congress and the second term of the Reagan administration. Can further cuts in these programs reduce massive federal deficits? Or must spending cuts be made either in social security entitlements or in defense appropriations? Can the Reagan administration succeed in shifting many public functions from the federal government to the states? Would such transfers affect public support for these programs? Is a shift away from federal responsibility and control desirable?

A forthright evaluation of domestic programs was conspicuously absent during the recent election campaign. Public debate about domestic priorities during the fall of 1984 was partisan and political, with ideological statements and misinformation the rule rather than the exception. In contrast, the urgency of the issue of the federal deficit, and of the unusually high interest rates the deficits cause, will now require elected officials to pay close attention to details of the domestic budget in defining public priorities for the next term. This will require facts about program operation, analysis of program outcomes, and knowledge of the budgetary consequences of policy alternatives.

This book provides the kind of analysis needed for this crucial debate about federal policy. It offers a serious and in-depth evaluation of domestic programs and priorities, with coverage of a broad range of issues, from education and welfare to urban transportation, from housing policy to environmental regulation. It supplies a framework for assessing the pro-

posals of the New Federalism and the consequences of continued trade deficits.

This book presents the views of a group of nationally prominent economists, including many who have served in policymaking positions in the administrations of both parties. It carefully summarizes the recent history of government policies and their outcomes. The authors review program priorities and offer proposals for the future. The analysis is addressed to a wide audience and is enriched by lively commentary and discussion by economists knowledgeable about each of the substantive programs.

The authors present no unified opinion about what our domestic priorities ought to be. There is, however, a general consensus as to appropriate directions in a number of areas. The analysts clearly sense that we ought to move to less intrusive federal command and control and also to greater state involvement in a number of programs such as transportation, education, and the environment. The New Federalism receives substantial support in many areas, not involving poverty and welfare, from a group of economists who supported a stronger federal role a decade ago.

Despite this intellectual shift, the authors present convincing evidence that the Reagan program is really a federal budget-cutting exercise in disguise. They have forceful and controversial ideas about desirable reforms. They believe that substantial reductions in expenditures would adversely affect the quality of domestic programs.

The economic perspective of the authors is an importance one, given the current policy debate and the budgetary emphasis. Economic analysis of the 1983 federal budget indicates that spending on national defense amounted to $201 billion, and spending on social security, veterans' benefits, and interest came to over $442 billion. With a federal budget of $820 billion and political promises not to cut defense or social security, the administration appears committed to cuts in the remaining $240 billion. It is hard to see how a deficit of roughly $200 billion can be removed by domestic cuts alone. The only options appear to be tax increases, large and growing federal deficits, or blind faith that growth in the economy will alleviate all problems.

Economists familiar with programs and outcomes clear up the often confused and confusing facts. For example, authors Sheldon Danziger and Daniel Feaster demonstrate irrefutably that poverty did increase under the first term of the Reagan administration. From 1978 to 1983 the poverty rate increased from 11.4 percent to 15.2 percent, while real dollars of federal aid to the poor decreased by more than one percent. Commentator Jennifer Wolch shows that many of those removed from

the poverty rolls were in serious need of assistance—these are the "service-dependent" poor, who suffered severely during the first Reagan term. Further budgetary cuts in welfare can only exacerbate their problems, problems which cannot be cured simply by the benefits of a growing economy.

Sherman Maisel's economic analysis of housing programs clarifies the effects of current subsidy policies and of the alternative programs proposed by the administration. Housing affordability is a spreading problem, argues Maisel, despite the fact that 75 percent of federal housing subsidies go to the non-poor. John Kain claims that the most pressing domestic social problem in America is discrimination in the housing market. His detailed analysis of the 1980 Census of Housing suggests that some gains have been made in reducing residential segregation.

Campaign press releases indicated that student test scores had improved as a result of government programs. Economist Richard Murnane provides a detailed analysis of outcomes and program effects. He finds that reading skills of students have improved over the past decade, but that math and science skills have declined substantially. How can this crisis in education be resolved, especially in light of the need for budgetary savings? Murnane argues that the program and its solution lie in the labor market for teachers and suggests some important, but inexpensive, reforms to make it operate more effectively.

The economic analyses of domestic programs in this book conclude that there are real opportunities to reduce the federal deficit—by applying the principles of the New Federalism to revenues as well as expenditures in the federal domestic budget, and by reducing and redirecting intergovernmental grants. In fact, recommendations presented here suggest that a large share of the current deficit could be eliminated by such reforms. The detailed discussion and commentary that follows the papers provides thought-provoking and valuable recommendations for action.

JOHN M. QUIGLEY
DANIEL L. RUBINFELD
BERKELEY, CALIFORNIA

ACKNOWLEDGMENTS

This book grew out of a conference on domestic public policy held at the University of California, Berkeley, in September 1984. The conference was supported financially by the Center for Real Estate and Urban Economics at Berkeley and encouraged intellectually by the Center's director, Kenneth T. Rosen. The Center continued to play an important role in the development of this book by providing financial and administrative support during the preparation process.

A number of individuals helped in manuscript preparation. We are grateful for the extraordinary efforts of Jo Magaraci of the Real Estate Center and Michelle Dethke of the Graduate School of Public Policy.

PART ONE—
THE UNITED STATES FEDERAL SYSTEM

An analysis of the United States federal system provides a useful framework for an evaluation of domestic budgetary programs. First, Robert Inman describes the Reagan administration's New Federalism, with its implied reallocation of functions between the states and the federal government. Edward Gramlich then addresses the fundamental normative question of federalism: Which levels of government ought to be assigned the right to levy taxes and to utilize other revenue-raising sources? In the Commentary, Henry Aaron discusses arguments for the important fiscal role of intergovernmental grants; George Break argues, against Gramlich, that the federal income-tax deduction for state and local taxes, although it may be less than efficient, does contribute to equity; and Julius Margolis provides a broad sociopolitical perspective on the rapid growth of government in the twentieth century.

Chapter One—
Fiscal Allocations in a Federalist Economy: Understanding the "New" Federalism

Robert P. Inman

From its constitutional beginnings to today, the United States public economy has been committed to the concept of federalism in the provision of public services. The use of multiple layers of government, with each higher level possessing rights of control over a lower level, is a founding principle of our fiscal structure.^[1] But while the constitutional commitment to a federalist system is clear, the precise structure and performance of that system are not.

Significant changes have taken place in our federalist fiscal system, both historically and in recent years. Scheiber (1966) has identified four stages of historical development of United States federalism: first, a period of "dualism" (1790–1860) in which states and the federal sector coexisted with essentially equivalent responsibility and powers; second, a time of "centralizing federalism" (1860–1933) as power began to gravitate to the federal level; third, a period of "cooperative federalism" growing out of the social programs to deal with the national crisis of the Great Depression (1933–1964); and finally, the recent period of "creative federalism" in which the federal government has taken an active policy interest in the specific problems of state and local governments. The first three periods can be characterized as times when the states (and their localities) made fiscal policy largely independently of direct federal interventions, while the recent period of creative federalism has involved the federal government directly in state and local fiscal affairs. Federal grants-in-aid and all their spending requirements, as well as the many new federal regulations

of state-local governments, have deeply affected the budgetary choices of the state-local sector.

In January 1982, President Reagan proposed a significant break with this trend toward federalization of our public economy. As originally presented in his budget message of that year, President Reagan's "new" federalism proposed: (1) that Medicaid become a fully federal program (which the administration hoped to curtail as part of its health-care-reform efforts); (2) that the states and localities assume responsibility for food stamp programs and Aid to Families with Dependent Children (AFDC); and (3) that more than sixty federal programs in education, community development, transportation, and social services be returned to the states, with the states receiving $28 billion a year from a federal trust fund to help pay for their new responsibilities. The trust fund was to be supported by federal excise taxes. Dollars paid from the trust fund would not be restricted to expenditures on the reassigned programs, however. The funds could, if a state so decided, be allocated to other state programs or to state tax relief. The intention of the president's new federalism was clear: to reduce federal influence on state-local fiscal choice.

What is less clear is how state and local governments will react to this restructuring of our current federalist fiscal system, and whether the present federal government will relinquish control and embrace the new federalism. Yet from the perspective of fiscal policy these are the central issues. Changes in federal grants and federal regulations will assuredly affect state and local governments' budgetary decisions. The relative economic and political attractiveness of the new federalism and of the creative federalism it seeks to replace ultimately rests on the actual allocations of resources. This paper offers some first (but considered) guesses as to what might happen under, and to, Reagan's reforms, and concludes that President Reagan and his supporters are not likely to be disappointed in the economic consequences of the new federalism, though they will be unhappy with its political prospects.

Our Current Federalist Fiscal Structure

In order to predict the future under President Reagan's new federalism, it is important to understand our present federalist fiscal structure and how it evolved. The Reagan proposals are a significant challenge to the historical trend toward federalization of United States fiscal policy. Table 1.1 summarizes the federal and state/local government spending patterns

over the past eighty years. As columns 1–3 show, the federal government share has grown steadily; today each sector is responsible for about half of all domestic public spending.^[2] The major source of the growing federal share was expansion of social insurance programs—"transfers to persons"—from the mid 1930s to 1950 and again from 1960 to 1980 (Table 1.1, columns 7–9). While the federal government's responsibility for transfers was growing, its responsibility for direct provision of nondefense public goods and services was declining (Table 1.1, columns 4–6); the state-local sector is the primary producer of domestic public services. Overall, the major source of the historical growth in total government activity has been transfer programs, however, and these programs are the financial responsibility of the federal government.

A similar move toward the federal level is observed in the historical trends of government receipts measured as taxes plus contributions to social insurance (see Table 1.2). In 1929, all government receipts were $430.80 per capita (in 1972 dollars), of which 33 percent was raised by the federal government. By 1950 the federal share was 74 percent of all receipts. The federal share has declined slightly in recent years, but a comparison of the 1929 and 1983 divisions of the revenue-raising function shows essentially a reversal of roles between the federal and state-local sectors. In 1929, state-local governments were the main source of public dollars; now the federal government has assumed that role.

The federalization of our fiscal structure has manifested itself in other ways as well. The federal government now actively participates in state and local budgetary choices. It does so in two ways: through a carrot called grants-in-aid, and through a stick called regulation. Table 1.3 details the recent growth of federal aid to state and local governments. From a position of very modest absolute and relative importance in 1940, federal aid had grown by 1980 to a sizable real-dollar transfer—a total of $192/capita—and to over 25 percent of state and almost 14 percent of local non-debt revenues. Growth has been particularly dramatic over the last two decades, most noticeably in direct federal-to-local aid. Table 1.4 illustrates the other important features of the aid explosion. The growth has occurred in categorical formula grants and categorical projects grants rather than in block grants. Block grants are federal dollars targeted for broad programmatic missions—e.g., education, employment, community development. Categorical aid is directed at narrow program categories—e.g., school lunches, sewer construction, low-income transfers to fatherless families. Categorical aid can be given to states or localities

according to a well-specified formula (formula grant), or states and localities may submit a specific project proposal for federal funding (project grant). Categorical aid often requires the state or local government to match the federal funds with state or local revenues. Block grants generally do not require such a match; instead, they serve as "free" money to the state-local sector.

Does such aid influence state-local fiscal allocations? The econometric evidence from over twenty years of research is quite unequivocal on the point: it does indeed (see the surveys in Gramlich 1969 and Inman 1979). Federal aid increases state and local spending on the targeted activities, and often on non-targeted activities as well. Categorical matching aid is the greatest stimulus to state-local spending. Block grants and the totally unconstrained, general-revenue-sharing aid induce the least increase in public spending. In the next section I review three recent studies which show that the impact of federal categorical and categorical matching grants on state and local budgets has been sizable. While there has been much political debate about whether these federal aid programs are good or bad for the state-local sector,^[3] no one has denied that these programs have been an important influence on state-local budgetary choices.

The federal government also regulates many activities of the state-local sector, now more than ever. Figure 1.1 summarizes the growth of federal regulatory programs. Four types of regulation have been identified: direct orders , which must be obeyed to avoid civil or criminal penalties (e.g., the Equal Employment Opportunity Act of 1972, which bars job discrimination on the basis of race, color, religion, sex, or national origin); crosscutting regulations , broad federal mandates that apply to all forms of federal assistance (e.g., the Civil Rights Act of 1964); crossover regulations , which require performance on one policy dimension under penalty of loss of federal assistance from another, well-specified program (e.g., the 55mph speed limit required under threat of loss of federal highway assistance); and, finally, partial pre-emptions , in which federal law establishes a policy goal which, if not met by the state-local sector, will allow direct federal provision or enforcement (e.g., the Clean Air Act Amendments of 1970). The 1970s saw by far the greatest growth in federal regulation of the state-local sector. The results have been mixed. The Civil Rights Act is an example of significant accomplishment, but most environmental regulations have produced only modest gains in air, water, and land-use quality, yet have imposed large costs on state and local governments (Advisory Commission on Intergovernment Relations n.d., p. 22). Again, while the relative benefits and costs of federal regulation of the

Figure 1.1
Federal Regulation of State-Local Governments, 1930–1979
Source: Advisory Commission on Intergovernment Relations (n.d.), p. 4.

state-local sector can be debated, there can be little doubt that these regulations have significantly influenced the policies and budgets of state and local governments.

We are now a federalized, federalist system. President Reagan's "new federalism" reforms strike at the heart of this centralization process. First, the two major income-maintenance programs now at the federal level—food stamps and AFDC—are to be returned to the state and local sectors. In 1983, these programs were estimated to cost the federal government $16.4 billion. A dollar shift of this magnitude will reduce the federal share of transfer to persons from 87.2 percent in 1983 (Table 1.1, column 8) to 83 percent.^[4]

Second, the Medicaid program is to become fully a federal responsibility. On the surface, this adjustment appears to run counter to the defederalization intentions of the Reagan policy. The new federal Medicaid outlays, approximately $19 billion, will increase the federal share in transfers to persons, from 83 percent following the decentralization of welfare, back to 87 percent. Despite this upward adjustment in federal

spending, it can be argued that the offer to take on the state's share of Medicaid is a necessary component of the new federalism package. The states and their congressional representatives will not accept the income-maintenance programs without some form of budget relief to compensate for the additional expense of these programs. Direct grant relief might be more of the same—federally funded aid for federally mandated programs. Relief could, however, be offered in the form of a federal takeover of a state responsibility. But which state responsibility? The state share of the Medicaid program is an excellent choice for three reasons: (1) the state Medicaid outlay is large and growing and a bit more than the federal dollars now spent on food stamps and AFDC, i.e., it is a "fair" trade; (2) the states have been doing a very poor job of controlling Medicaid outlays; and (3) there is a growing desire at the federal level to control federal health care costs, and Medicaid, if federalized, can more easily be included within any new federal regulations. The states do not want the responsibility of Medicaid, and for reasons not entirely related to federalism, the Reagan administration is willing to take it on. From the point of view of the new federalism, the swap of the state share of Medicaid for income-maintenance programs seems the best available trade.

Third and finally, Reagan's new federalism proposes to return to the state-local sector sixty-one existing federal-education, social-service, transportation, and community-development categorical aid programs for state administration and funding. To ease the estimated $30.2 billion financial burden imposed by these new programs, the federal government would make available to the states $27.6 billion from a federal trust fund supported by federal excise taxes. (The states' shortfall between program costs and trust fund transfers equals the states' gain from the welfare/Medicaid swap.) The intention of this exchange is to reduce federal control over state and local spending; categorical aid and its particular expenditure restrictions are dropped and replaced with unconstrained federal assistance. This federal-trust-fund aid is to be phased out, however, over four years beginning in 1988. Thus, the third component of the new federalism is intended first to reduce federal aid restrictions and then, beginning in 1988, gradually to reduce federal government spending and taxes.^[5] Overall, the new federalism is: (1) to leave the federal share of transfer outlays largely unaffected but to move to the federal level those redistribution expenditures (Medicaid) that the states have found the most difficult to control; (2) to replace federal categorical aid with lump-sum aid, thereby reducing federal control over state-local spending; and finally, (3) to reduce federal spending and taxes by gradually phasing out

the trust fund and its associated taxes. Or so the proponents of the new federalism hope. Whether the new federalism will in fact have these defederalizing effects is an open question.

Will the "New" Federalism Work?

While the calculations above concerning effects of President Reagan's new federalism on our fiscal system are reasonable first guesses, they are really only that—first guesses. Budget aggregates were simply moved from the federal to the state-local column and back again. Missing from such calculations is any sense that there are governments, and voters, who determine what those budget aggregates will be. Yet an analysis of such a fundamental realignment of fiscal responsibilities as the new federalism proposes must recognize that the state-local sector will react to the reforms. To predict the consequences of new federalism we must first understand the budgetary process of state and local governments.

State-Local Fiscal Choice in a Federalist Economy

Beginning with the early "determinants" studies of state and local spending, economists and political scientists have tried to untangle the causes of decentralized governments' fiscal choices. Analysis has proceeded from simple linear regression models of aggregate state and local spending to sophisticated maximum-likelihood estimation of utility maximization models for decisive voters in individual jurisdictions. But all these models have one element in common: they are economic models, with a central focus on citizen preferences and their budget constraints as causes of governmental expenditures.

Figure 1.2 illustrates the basic, and now familiar, story. A "typical" resident's preferences for public services and private goods are represented by a utility relationship over after-tax private income (y ) and public goods (g ), denoted U(y,g ). The indifference curves in Figure 1.2 rank the relative value of different combinations of y and g as represented by U(g,y ). Combinations of y and g on a higher indifference curve are preferred by the typical resident to those combinations on a lower indifference curve. A budget constraint will restrict the level of services which the typical resident can afford. The constraint is defined by the identity: Î=y +p·g , where Î is the typical resident's "full fiscal income" and p is the resident's "tax price" of local public goods. Full fiscal income is the sum of the resident's before state-local-tax private income (I) and the resident's

Figure 1.2
The "Typical" Voter's Preferred Budget

share of federal to state/local lump-sum aid per resident (z ). The tax price p equals the resident's share of the tax costs of each unit of the public service, net of federal matching aid.^[6]

Given preferences and the budget constraint, the typical resident will wish to buy the level of public goods that maximizes U(y,g ) subject to the constraint. This is point (y*,g* ) in Figure 1.2, where the budget line is just tangent to the highest possible indifference curve. The utility maximization/budget constraint model yields a demand curve for local public goods of the following form:

where the theory predicts that ¶ g/¶ p<0, ¶ g/¶ Î>0. The taste variables (Tastes) are unique to the individual resident.

Economists are now quite comfortable with this specification of state-local fiscal choice and have applied it on numerous occasions to estimate local governments' responsiveness to various federal-government aid and tax policies. (Inman 1979 reviews these studies.) One nagging question

has, however, been left unanswered by virtually every application of this approach. Who is the "typical" resident whose demand curve we are estimating? Answers have relied more on hand-waving than hard work. Empirical analyses of state-local fiscal choice have generally lacked any notion that politics, the process of conflict resolution, affects state-local fiscal allocations. Exactly whose preferences and whose budget constraints dictate the final allocations? Is the final allocation a compromise among several players, or just the preferred outcome of a "chosen" one? Who are these players? What is their standing and what are their rights within the political budgetary process? How are disagreements resolved? What happens if there is no agreement? These are political questions, and their answers require political analysis.

Introducing political considerations into the systematic analysis of budgetary choice is not a simple matter. Once we begin to study allocations involving more than one policy dimension—say, education and welfare spending—we confront a fundamental analytic difficulty. If there are more than two interested voters, such allocations will generally not have a stable equilibrium if budgets are decided by a simple majority-rule process, for there is no identifiable median voter in such cases (see Inman 1984). Yet state-local budgetary choices are often very stable; allocations do not change much from year to year. In an important line of new research, Shepsle (1979) has described and analyzed various legislative extensions of simple majority rule that are sufficient to produce stable fiscal allocations. The final allocations—called structure-induced equilibria —are conditioned by the status quo and the constitutional rules that determine legislative structures. Shepsle identifies three central structural features of legislatures: (1) a committee structure that identifies who is allowed to offer proposals for consideration by the full legislature; (2) a jurisdiction structure that defines which proposals may be considered by the committee and the legislature; and (3) an amendment structure that describes how the committee's proposals to the legislature may be altered. Together these structural features can insure a stable allocation. It is important to emphasize, however, and particularly for our purposes here, that a structure-induced equilibrium is a partial equilibrium, conditioned by the starting or status-quo allocation and the detailed rules of fiscal choice. Exogenous changes in the starting allocation or legislative rules because of new federal regulations or aid programs, such as the new federalism, will alter the observed equilibrium allocations.

The political theory of structure-induced equilibria allows us to specify just how such institutional changes will affect budgetary choices. The

theory predicts that final fiscal allocations will be a simple weighted average of the preferred allocations of each voter. The weights depend on the status quo (denoted as g°, a vector of initial service levels) and the political institutional structure (denoted S, a vector of institutional variables). The allocation for service t preferred by each voter or interest group i is defined by the demand curve g_it = f _it (p_i , Î_i ¦Tastes_i ), as specified in equation (1) above, where now p_i is a vector of public-good tax prices. The resulting model of budgetary choice is therefore

where

and where there are t = 1 . . . m public services and i = 1 . . . n identifiable voter types or interest groups. Variables that might be included in the vector S of political structure include controlling interests (chairmanship, majority) of the legislative committees that set the agenda, jurisdiction and budgetary and bargaining rules on how dollars can be allocated, size of voting blocs within the legislature or community, political allegiance of those with veto power over final allocations (e.g., governor, mayor), and amendment rules that allow proposals to be submitted from at-large interests. The budgetary model outlined here—once it has been estimated—gives us exactly what we need to begin to analyze the effects of changes in federal dollars and fiscal structures on state-local allocations. Three recent studies (Craig and Inman 1982, Gramlich 1982, and Craig and Inman 1984) have examined President Reagan's new federalism from this perspective; their results are instructive as to the likely budgetary consequences of the proposed Reagan reforms.

The Medicaid-Welfare Swap:
What Will Happen to Low Income Assistance?

Gramlich (1982) and Craig and Inman (1984) have examined the likely consequences of the new federalism's proposed exchange of the food stamps and AFDC programs for Medicaid. Both studies reach essentially the same conclusion: the states will appreciably reduce their support of low income maintenance programs despite federal assumption of Medicaid and the availability of $27.6 billion in trust fund aid.

Employing an analytic framework similar to that proposed in Equation 2 above, Gramlich and Craig–Inman estimate the effects of the current

fiscal structure—categorical matching aid for AFDC and Medicaid, federal provision of food stamps, and the availability of general revenue-sharing and other (nearly) unrestricted, lump-sum aid—on the state decision to provide AFDC benefits (Gramlich) and AFDC plus other low-income assistance (Craig–Inman).^[7] Both studies find that dropping federal, categorical matching aid for AFDC and giving states full financial responsibility for the program will reduce spending for AFDC by at least 70 percent and perhaps as much as 95 percent. Gramlich arrives at the larger estimate when he allows for the fact that states are likely to adjust their benefit levels to those of their neighboring, fiscally competitive states. Since high benefit levels are likely to attract low-income families (see Gramlich's paper in this volume) and possibly discourage the location of richer families and firms, no state can afford high benefit levels by itself. Once the federal categorical matching aid, a strong incentive for high benefits, is removed, interstate fiscal competition dominates the allocation process, and AFDC spending is likely to fall dramatically.

Both studies conclude that low-income support from the food stamp program will also be significantly reduced. They find that state officials seem to behave as if food stamp benefits do not exist; in other words, it is a federal program and provides no direct political benefit to state politicians. A program thus ignored in the past will probably be supported only marginally under reform. The net effect of the new federalism's food-stamp transfer will be to reduce this form of assistance drastically, if not totally. Together, the transfer of AFDC and food stamps to the states is estimated to reduce the total spending on these programs by 70 percent or more.

Federal assumption of state Medicaid payments may also reduce the states' contribution to low-income medical assistance. First, as the federal government takes over the program, the previous federal Medicaid matching grant to the states is lost (i.e., the matching rate falls to zero) and state spending for low-income medical assistance declines. Second, the federal government will pay such low-income medical assistance directly to residents within the state, a possible inducement for the states to eliminate any remaining low-income medical-assistance programs. Craig and Inman (1984) estimate that the combined effect of these forces about equals the current level of state expenditures on Medicaid; that is, state welfare spending will fall by the amount of spending assumed by the federal government when it takes over Medicaid (~$19 billion).

The new federalism does offer the states financial assistance, however, and some of this relief may spill over into low income assistance. The

proposed trust-fund account is to pay states $27.6 billion a year initially (less in future years). Craig and Inman estimate that trust fund aid will have no significant effect on low income assistance and may in fact depress welfare spending. Other studies (reviewed in Inman [1979]) also show no major positive effect of lump-sum aid on welfare outlays. It therefore seems safe to assume that trust fund assistance will do little to offset any decline in welfare spending.

What, then, will be the overall effect of the new federalism on state-provided low-income assistance? As we have seen, a conservative estimate is that AFDC and food stamp spending will fall by 70 percent when transferred to the state-local sector. In 1983, the federal and state governments combined spent approximately $24.04 billion on AFDC and food stamps.^[8] The new federalism will reduce this total by $16.83 billion. With the Medicaid transfer, state expenditures on medical assistance for the poor will fall by approximately $19 billion, but federal spending is assumed to rise to fill this gap. There is therefore no net effect on low income assistance from the Medicaid transfer.^[9] The introduction of $27.6 billion in trust fund assistance is estimated to have no effect on state welfare spending. The combined effect of the new federalism reform is therefore an estimated decline of at least $16.83 billion in aid to low-income families provided by all levels of government. From the perspective of a typical low-income family, the new federalism may mean a loss of $2406 annually, approximately a 44 percent decline in average annual benefits.^[10] Clearly, the new federalism is an important fiscal reform affecting low-income households.

The Trust-Fund and Program Turnback:
What Will Happen to Service Provision?

The sixty-one federal categorical-aid programs proposed for transfer to the states include programs in education, social services, transportation, and community development. Also included is the general revenue-sharing program. In their place the federal government will establish a trust fund that will pay to the states a lump-sum grant approximately equal to the costs of transferred categorical-aid programs. This component of the new federalism essentially substitutes unconstrained aid for categorical aid. What will happen to the provision of government services in the affected program areas?

Gramlich's (1982) analysis gives an overview of the likely effects of fiscal reform. He finds that the loss of $1 of federal categorical aid will

lower state-local service expenditures by $0.38; the gain of $1 of federal lump-sum assistance will increase state-local service spending by $0.04. Gramlich's estimates of the state-local sector's response to lump-sum, revenue-sharing assistance is somewhat lower than that obtained by other researchers (see Inman 1979), but it should be pointed out that his sample covers the state-local experience to 1981 and therefore includes those governments' response to the recent pressure for fiscal responsibility and tax relief. Accepting Gramlich's estimates, we can calculate the aggregate effect of the new federalism on service provision. Turning sixty-one federal programs over to the states is equivalent to a loss of $25.6 billion in categorical aid and, since general revenue-sharing is included in the turnback program, a loss of $4.6 billion in lump-sum aid. The loss of $25.6 billion in categorical aid will reduce state-local service provision by $9.7 billion (=0.38 × $25.60). The loss of $4.6 billion in revenue-sharing aid will reduce state-local service provision by $0.2 billion (=0.04 × $4.60). The trust fund, on the other hand, increases federal lump-sum aid by $27.6 billion; the state-local sector's response is to increase state-local provision by $1.1 billion (=0.04 × $27.60). The final net effect of the new federalism is to reduce state-local spending initially by $8.8 billion. This loss is equivalent to a decline of $16 per capita in services (measured in 1972 dollars) or a fall in public-service provision by the state-local sector of 2.1 percent from the 1983 service level of $768.40 per capita (see Table 1.1, column 7).

Craig and Inman's (1984) analysis of the trust-fund exchange shows a similar, though slightly larger, decline in state-local spending. Their equilibrium analysis predicts that the trust fund–program swap will reduce state taxes by 6 percent (due to relaxed matching requirements), lower state education aid and total state welfare spending by 5 percent and 33 percent respectively, and leave other state spending virtually unchanged. The states' human resource budgets seem to bear the full brunt of the spending cuts. Local government spending may rise to offset some of the decline, but the local increase will probably not match the fall in state spending.

A more detailed look at state-local education spending shows that this is exactly what happens. Craig and Inman (1982) have examined the effects of Reagan's new federalism on spending for one of the most important program areas—elementary and secondary education. Their results are suggestive of what may happen to expenditures on the other directly affected services. After estimating an econometric model of state and local school spending, Craig and Inman simulate the effects of the

new federalism in which fourteen categorical education aid programs are consolidated into a single trust-fund, or lump-sum, grant to states. Table 1.5 summarizes their results, where spending in 1977 (the last year of their sample) defines the baseline, or pre-reform, allocation. In the status quo period, the average state allocated $375.58 per public-school enrollee for state aid for education. The average state raised $1,025.86 per enrollee in state taxes, where the difference between taxes and state education aid ($650.28 per enrollee) was allocated to other, non-education state expenditures. Local school districts spent $670.99 per enrollee on education, and raised $286.34 per enrollee in local taxes to support this spending. The remaining $386.65 per enrollee of local spending was financed by state education aid ($375.58) and direct federal-to-local school aid ($9.07 per enrollee). Following new federalism reforms, Craig–Inman estimate that state support for public elementary and secondary education will fall by about 23 percent, to $290.20.

The primary reason for this decline is the same that led to a decline in welfare spending: the new federalism substitutes less constraining lump-sum aid for categorical and categorical matching aid.^[11] Left on their own, states do not want to maintain spending on public education. What happens to the money now freed from the state education budget by the new federalism? The released $85.38 per enrollee (=$375.58 – $290.20) is allocated to tax relief; total state tax relief is $101.44/enrollee (=$1,025.86 – $924.42) supported in part by a fall in other (administrative?) state expenses.^[12] The fall in state-to-local education aid and federal-to-local education aid reduces local school spending by about 8 percent, or $616.35/enrollee. To maintain this level of spending, local taxes must rise by 14 percent, to $326.15/enrollee. The new spending level is supported entirely by local taxes and state aid (616.35 = $326.15 + $290.20). This detailed analysis of the new feder-

alism's effects on education reveals three important conclusions. First, education itself is not a particularly favored activity of state governments; it is federal categorical aid that keeps state funding at its present levels. Second, as the federal and state governments retreat from a fiscal responsibility for education, local school districts fill the void, somewhat; clearly, with the new federalism, financial responsibility for public education moves downward to the local level. Third, as President Reagan might have hoped, the new federalism will alter the mix of national income allocated between the private and public sectors. State taxation is lowered by $101/enrollee, while local school taxes rise by about $40/enrollee. The net result is a $61/enrollee increase in private income.

Summary

It is hoped by the proponents of the new federalism that the proposed reforms will reverse the trend toward federalization of our public economy and to a large degree shift the activity of government back to the state and local levels. The analysis summarized here suggests that they will not be disappointed. The very nature of the reform reallocates responsibilities from the federal to the state-local sector, and there is nothing in the empirical research reviewed here to suggest that reforms will not be decentralizing. The states, for example, have not been held in line by federal programs so that, now unleashed, they can spend unchecked to replace one bureaucracy with another. Quite the contrary, the evidence indicates that the states will retreat from many of these categorical program areas too and, at least in the case of education, will be eager to pass fiscal responsibility down still further, to the local level.

The empirical analyses of the new federalism reveal another important consequence of reform, one which might please President Reagan even more than the move to decentralization. The new federalism will likely shrink the size of government. The exchange of Medicaid for AFDC and food stamps will markedly reduce spending on income maintenance programs and return those released dollars to state taxpayers as tax relief. The new federalism's exchange of the lump-sum trust fund aid for categorical aid also means tax relief. Craig and Inman (1982) find that, in education at least, the resulting lower level of state education spending is transferred to taxpayers as state tax relief. Local education taxes do rise, but not by enough to offset the decline in state taxation. Again, the overall size of government is reduced and taxpayers' private incomes are increased.

From the point of view of those who wish to decentralize fiscal choice and to decrease the size of government, the answer must be yes to the question: Will the new federalism work?

New Federalism as a Federal Issue

If Reagan's new federalism will check the tide of fiscal centralization and shrink government, why then, in this era of apparent fiscal conservatism has the new federalism not been approved by the federal government? The early debates of the reform raised the obvious questions. There was much discussion about the accuracy of the administration's estimates of program costs, and the states were obviously disappointed about the projected 1988–1991 phase-out of the supporting trust fund. Would the dollar trade really be a fair one? But if a balancing of dollars was all that was at issue, we should have resolved the difference long ago. We have not. The matter runs deeper than that. The new federalism is not simply a realignment of programs between the federal and state-local sectors of government, it is a fundamental challenge to how public policy decisions are now made. Our existing, highly federalized fiscal structure did not just happen. It has evolved, I will argue, as a logical outcome of a changing federal political structure and a growing economic pressure on local governments to redistribute resources. Until the political structure changes again—and in a particular way—or until the demand for redistribution subsides, our fiscal structure will remain largely as it now stands: categorical, regulated, and centralized.

The pressure to use government as a means to redistribute resources among the members of the society is endemic to stable economies. Coalitions of potential beneficiaries form around existing, shared institutions capable of transferring dollars in their direction. Producer or consumer groups form about markets; the religious, the poor, and the ill cluster about churches; and all of us look to government for assistance. Small groups can generally organize more quickly and efficiently than large groups; while the fortunes of individual redistributive coalitions may rise or fall, such coalitions rarely die (see Olson 1982). American economic history is rich in examples of the growth and influence of redistributive coalitions (see, e.g., North 1981; Olson 1982).

With the baby boom and the explosion of home ownership and suburbanization following World War II, a new and fertile ground for the growth of such coalitions appeared—state and, most importantly, local governments. Table 1.1, column 6, illustrates the increasingly important

role of the state and local sector in the provision of domestic public goods and services. The growth was particularly explosive from 1945 to 1960 as the share of the state-local sector in public-good provision rose from 69.3 percent to 84.5 percent. During this same period the number of municipal, township, and special district governments grew from 43,440 in 1942 to 56,417 in 1967.^[13] The period from 1945 to 1960 saw the emergence of new redistributive coalitions with their focii almost exclusively on local government. Public employee unionization began in earnest in this period. The National Teachers Association began its slow evolution from a social and professional society to a politically active union, prodded in part by the aggressive, blatantly redistibutive behavior of the newly formed American Federation of Teachers. In our larger cities, downtown business interests began to organize for the development, or redevelopment, of shopping, residential, and business centers. In the suburbs, developers lobbied for new infrastructure (paid for by tax-exempt bonds) and zoning variances. Even the large group of innercity poor had become organized into a politically active redistributive coalition, the National Welfare Rights Organization. Each of these coalitions—public employees, business groups, the poor—turned to local governments for more—more pay, more services, more transfers. Yet the local public sector had only limited resources, and it was severely constrained by the mobility of its tax base. It was only natural that, as the brokers in the redistribution game, local officials should seek more dollars. They organized too, and in the 1960s they went to Washington as the "intergovernmental" lobby.^[14] Would Washington respond?

Washington did respond—with creative federalism. The redistributive coalitions collecting about the local public sector had created a new demand for transfers. The supply of those transfers came from a United States Congress that by the mid 1960s found it particularly attractive to be responsive. Domestic policies began to assume a decidedly regional and local focus. The fiscal structure we now call creative federalism, with its extensive use of categorical aid, project aid, and federal regulations, is one consequence of this change.

Two events converged to produce this important shift. First, American voters were becoming better educated and better informed. The old political party labels were no longer sufficient to insure voter loyalty. Nie, Verba, and Petocik (1979) found the 1960s voter to differ in important ways from the 1950s voter. There were more independents, and even those voters who wore a party label found it easy to desert the party leadership when proposed policies did not meet their demands. Second,

as voters became more independent so did their congressional representatives. For exogenous reasons the strong central leadership of Lyndon Johnson in the Senate and Sam Rayburn in the House gave way in the 1960s to the less autocratic styles of Mansfield and McCormack. The 1964 Democratic landslide brought a new and active group of Democratic liberals into the House, many representing urban local coalitions. The old, fabled "conservative coalition" of Republicans and southern Democrats gave way, lost in a flurry of liberal legislation—including Medicaid, federal grants for education, urban poverty programs, and the creation of the Department of Housing and Urban Development. Much of the legislation was a response of a sympathetic Congress to the memory of a president deeply committed to social legislation.

The period 1967–1970 was one of frustration for the Democratic liberals. A fall in their numbers from the peak in 1965–1966, division within their ranks over the war in Vietnam, the election of a Republican president, and the return to importance of the conservative coalition thwarted the efforts of liberal, urban Democrats to introduce further domestic legislation. These three years proved only a temporary setback in the trend toward a more open, more locally focused Congress. Beginning in 1970, a series of procedural reforms initiated by the liberals were enacted by the Democratic Caucus and by the whole House. The intention of the reforms was to open the key leadership positions of the House—committee and subcommittee chairmanships—to more members, not just the senior, select few (for an interesting description of the process by which these reforms were introduced, see Ornstein 1975). The result was a more diffuse, younger, liberal leadership. The incentives within this more decentralized structure no longer encouraged individual representatives to look to the older leadership for signals on how to form policy. Rather, the structure encouraged offering a favored policy and bargaining with other members for its approval. And how should a representative define a policy agenda? Again, not by turning to the central leadership or to the Democratic or the Republican party. Their approval no longer won local congressional elections. The representative's policy agenda came from the local voters, for if local voters were satisfied, reelection was guaranteed (see Fiorina 1977, esp. chs. 4 and 5). The policies and budgets that emerge from such a decentralized political body will themselves be decentralized in their effects, satisfying the needs of each representative and the redistributive coalitions he or she represents.^[15] When representatives are chosen from geographically prescribed areas, policies will assume a clear regional or local focus. Creative feder-

alism stands as a telling example of what such a political process will produce.

A decentralized congressional policy process seeking to satisfy the demands of local redistributive coalitions will necessarily design policies that will target federal dollars to local areas. Further, these dollars must be targeted to particular redistributive coalitions in a way that will assure that the congressional representative can claim credit for the transfer. Finally, if the federal dollars can be leveraged from other, non-local revenue sources such as the state, so much the better. What type of federal grants policy will achieve these objectives? The answer is categorical project aid and formula aid that uses the states whenever possible as the administrative agent but that monitors state performance closely with "pass-through" and, ideally, matching requirements. To the extent that local redistributive coalitions do not trust the states to administer the grant in their favor, the federal program will bypass the states and give dollars to local governments, or perhaps to private groups directly. Finally, since there are many congressional districts to be satisfied and each district has a different set of local redistributive coalitions, a wide range of federal grants programs are needed, with, ideally, each program capable of allocating dollars to specific local coalitions.

Not surprisingly, exactly this grants system, which we now call creative federalism, emerged over the decade 1963–1973. Table 1.4 shows the proliferation of project-aid and categorical-formula grants. Project-aid programs, the most flexible form of grant and the most susceptible to congressional direction, nearly tripled in number from 1962 to 1976 (Chernick 1979, Arnold 1981, and Plott 1968 present interesting studies of possible congressional control of project aid). The number of categorical formula grants more than doubled during this period; many of these grants employ state "pass-through" requirements and almost all have a state matching provision of some kind.^[16] Further, when the states could not be trusted to target federal aid to local coalitions, direct federal-to-local programs were devised; see Table 1.3. By 1973, direct federal-to-local aid had become nearly 20 percent of all federal categorical assistance. Most of this assistance was for urban areas, bypassing state governments presumably to avoid their rural-suburban bias (see Maxwell and Aronson 1977, tables 3.1, 3.2). Finally, as the analysis predicts, almost all local governments in the country received some categorical aid. A national survey of local governments in 1974 by the Advisory Commission on Intergovernmental Relations (1978) indicated that 73.3 percent of the responding city governments received federal categorical aid, and

80.6 percent of the responding county governments received such assistance. The typical city in the survey obtained money from an average of 9.3 grants; the typical county obtained money from an average of 20.6 grants. By the mid 1970s, we had in place a federalist fiscal structure which stood as the logical outcome of the economics and politics of its time.

One anomaly in the recent history seems to run against this logic—General Revenue Sharing (GRS0. In fact, however, close examination of the GRS history reveals that it was shaped by the same economic pressures and the same political structure that has given us creative federalism. When all the political maneuvering was done, the bill approved in June 1972 looked very little like the first Nixon proposal presented in August 1969. The initial Nixon plan gave 73 percent of the money to states, and those states that paid more federal taxes got more federal GRS aid. The GRS bill that was passed gave most of the aid to local governments (66 percent), aid given to the states contained an implicit match on state taxes, and the final distribution of money across local governments was effectively on the basis of population—i.e., equal funds per capita (see Nathan, Manuel, and Calkin 1975). Lastly, GRS was all new money for local governments; it did not replace any existing categorical aid programs. In effect, the political history of GRS proves to be just another chapter in our story.^[17]

President Reagan's new federalism seeks to put an end to this tale called creative federalism. Will he succeed? The answer, I think, is no . The economic pressures and the political structure that produced creative federalism are still in force. The new federalism is too fundamental a change to be embraced by those who now benefit from our present fiscal structure.

Members of the House, for example, now have a wide range of categorical aid programs which direct federal resources to their favored local coalitions. As Representative John Brademas, the author of a number of categorical aid programs, commented concerning the local activities financed by general revenue sharing, "They don't even ask us to the ribbon-cutting ceremonies" (quoted in Beer 1976, p. 185). It is difficult for a member of Congress to claim credit for tax relief and general trust-fund grants—unlike classrooms, roadways, sewers, or jobs—when running for reelection. Congress prefers particularized programs (see, e.g., Mayhew 1974, pp. 53–57).

State and local officials have also shied away from Reagan's new federalism. The states, of course, were eager to give up Medicaid since it meant

freed dollars, but they did not want to take on AFDC and food stamps in return. And for good reasons. State and local officials prefer redistribution to be handled at the federal level. Only in those few states with no significant low-income population will state officials prefer a decentralized welfare system. Those states can neutralize any political pressure from their small low-income group while receiving the spill-in benefits of being a low-transfer, low-tax state. Elected officials in all other states, however, will feel the pressure of having to do something for the poor at the same time that they try to control taxes on firms and mobile upper-income households. The current federal categorical welfare programs make these hard decisions for them and apply them uniformly across all the states. State officials want more lump-sum aid, but not at the price of fewer welfare aid programs and greater interstate competition (see Rose-Ackerman 1981). Local officials fear the loss of regulated, categorical aid, for the simple reason that they see less money coming their way. As Craig—Inman (1982) show for education aid, when the federal strings are loosened, the level of state-to-local support can fall significantly (see the comments of state and local officials in National Journal 1982).

Finally, voters have not risen to demand reform, and there are two good reasons for this. First, many voters are direct beneficiaries of categorical aid; people do not want to lose the flow of dollars to their redistributive coalition. Some voters, however, do not benefit directly from the present federalized fiscal system. They pay taxes but get no return transfers. Craig—Inman (1982) have shown that the new federalism may mean significant tax relief. Why, then, have taxpayers not formed a coalition in support of the new federalism? The answer is simply that low-spending, tax-relief coalitions are hard to put together. We all benefit a little from the hard work of a few. Yet almost no one—elected local representatives or private citizens—can afford to be one of the hardworking few for tax relief. What is required is a national coalition held together by a commitment to the coalition's central objective, even when it is in everyone's best interest to "free-ride" on the group. Such coalitions may occur once in a while—the FY1982 budget and tax cut is an example—but not often.^[18]

It is not surprising, then, that the new federalism has not yet won congressional approval. Our current fiscal system is what it is because of political and economic forces which will not be easily reversed. Whether a new economic coalition of taxpayers arises and proves that it can compete effectively within a special-interest political system as another special interest remains to be seen. To date, the proponents of the new federalism have not yet succeeded in fashioning such a coalition.

Conclusion

A federalist fiscal structure is an ever-evolving institution, responding, as will all societal institutions, to changing economic and political pressures. Our present structure, creative federalism, is the logical consequence of a rising pressure on local governments to respond to new local coalitions, and of a congressional political system that rewards those who are most responsive to these local demands. The resulting fiscal system channels federal funds to local governments and organizations via many narrowly prescribed categorical grants closely monitored by Washington. While such a fiscal system can be rationalized historically, there are good reasons to doubt its economic logic. Too much money is being spent on too many marginally inefficient programs.^[19] A cutback and a restructuring of our aid program seem in order. Reagan's new federalism is one such proposal and, as we have seen, it is likely to have its intended effects of decentralizing fiscal choice and shrinking the size of government. Gramlich (this volume) has offered a plausible alternative reform package.^[20] Neither reform, however, is likely to win congressional approval in the near future, for the simple reason that what is now in place is Congress's preferred response to our present economic and political environment. Only a substantial change in that environment will lead to substantial change in our fiscal structure.

Acknowledgments

This paper was written while the author was a visiting professor of economics at the University of California, Berkeley. It was financed in part by NSF grant SES-8112001 to the author and by a grant from the Center for Real Estate and Urban Economics, University of California, Berkeley. The financial support of these organizations is appreciated. Ken Shepsle, Art Frank, Dan Rubinfeld, and John Quigley were kind enough to read and comment on an earlier version of this paper, for which I am grateful.

References

Advisory Commission on Intergovernmental Relations. 1978. Categorical Grants: Their Role and Design. No. A-52. Washington, D.C.: U.S. Government Printing Office.

———. 1980. Recent Trends in Federal and State Aid to Local Governments. No. M-118. Washington, D.C.: U.S. Government Printing Office.

———. 1981. The Conditions of Contemporary Federalism: Conflicting Theories and Collapsing Constraints. No. A-78. Washington, D.C.: U.S. Government Printing Office.

———. N.d. Regulatory Federalism: Policy, Process, Impact and Reform. No. B-7. Washington, D.C.

Arnold, D. 1981. "The Local Roots of Domestic Policy." Pp. 250-87 in T.E. Mann and N. J. Ornstein, eds., The New Congress. Washington, D.C.: American Enterprise Institute.

Beer, S. 1976. "The Adoption of General Revenue Sharing: A Case Study in Public Sector Politics." Public Policy 24:127-96.

Bowen, H. 1943. "The Interpretation of Voting in the Allocation of Economic Resources." Quarterly Journal of Economics 58: 27-48.

Break, G. 1980. Financing Government in a Federal System. Washington, D.C.: Brookings Institution.

Bureau of the Census. 1975. Historical Statistics of the United States, Colonial Times to 1970. Washington, D.C.: U.S. Government Printing Office.

———. 1983. Statistical Abstract of the United States, 1983. Washington, D.C.: U.S. Government Printing Office.

Bureau of Economic Analysis. 1981. National Income and Product Accounts, 1929-1976. Washington, D.C.: U.S. Government Printing Office.

———. 1984. Survey of Current Business. Vol. 64. Washington, D.C.: U.S. Government Printing Office.

Chernick, H. 1979. "An Economic Model of the Distribution of Project Grants." Pp. 81-103 in P. Mieszkowski and W. Oakland, eds., Fiscal Federalism and Grants-in-Aid. Washington, D.C.: Urban Institute.

———. 1982. "Block Grants for the Needy: The Case of AFDC." Journal of Policy Analysis and Management 1: 209-222.

Council of Economic Advisors. 1984. Economic Report of the President, February, 1984. Washington, D.C.: U.S. Government Printing Office.

Craig, S., and R. P. Inman. 1982. "Federal Aid and Public Education: An Empirical Look at the New Fiscal Federalism." Review of Economics and Statistics 64: 541-52.

———. Forthcoming. "Education, Welfare, and the New Federalism: State Budgeting in a Federalist Public Economy." In H. Rosen, ed., Studies in State and Local Public Finance. Chicago: Chicago University Press.

Filimon, R., T. Romer, and H. Rosenthal. 1982. "Asymmetric Information and Agenda Control: The Basis of Monopoly Power in Public Spending." Journal of Public Economics 17: 51-70.

Fiorina, M. 1977. Congress: Keystone of the Washington Establishment. New Haven: Yale University Press.

Fiorina, M., and R. Noll. 1978. "Voters, Bureaucrats and Legislators: A Rational Choice Perspective on the Growth of Bureaucracy." Journal of Public Economics 9: 239-54.

Gomez-Ibañez, J. A. 1984. "The Federal Role in Urban Transportation." This volume.

Gramlich, E. 1969. "The Effects of Federal Grants on State and Local Expenditures: A Review of the Econometric Literature." Pp. 569-83 in Proceedings of the 62nd Annual Conference on Taxation. Lexington, Ky.: National Tax Association.

———. 1982. "An Econometric Examination of the New Federalism." Pp. 327-60 in Brookings Papers on Economic Activity .

Hamilton, B. 1983. "The Flypaper Effect and Other Anomalies." Journal of Public Economics 22: 347-62.

Inman, R. P. 1978. "Optional Fiscal Reform of Metropolitan Schools: Some Simulation Results." American Economic Review 68: 107-122.

———. 1979. "Fiscal Performances of Local Governments." Pp. 270-321 in P. Mieszkowski and M. Straszheim, eds., Current Issues in Urban Economics . Baltimore: Johns Hopkins University Press.

———. 1981. "Wages, Pensions, and Employment in the Local Public Sector." Pp. 11-57 in P. Mieszkowski and G. Peterson, eds., Public Sector Labor Markets: COUPE Papers on Public Economics , Vol. 4.

———. Forthcoming. "Markets, Governments and the 'New' Political Economy." In A. Auerbach and M. S. Feldstein, eds., Handbook of Public Economics , Vol. 2. Amsterdam: North-Holland.

Jackman, R., and J. Papdachi. 1981. "Local Authority Education Expenditure in England and Wales." Pp. 47-62 in M. J. Bowman, ed., Collective Choice in Education . The Hague: Martinus Nijhoff.

McGuire, M. 1978. "A Method for Estimating the Effect of a Subsidy on the Receiver's Resource Constraint." Journal of Public Economics 10: 25-44.

Maxwell, J., and J. R. Aronson. 1977. Financing State and Local Governments. Washington, D.C.: Brookings Institution.

Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale University Press.

Moffitt, R. 1984. "The Effects of Grants-in-Aid on State and Local Expenditures: The Case of AFDC." Journal of Public Economics 23: 279-306.

Nathan, R., A. D. Manuel, and S. E. Calkin. 1975. Monitoring Revenue Sharing. Washington, D.C.: Brookings Institution.

National Journal. 1982. "The New Federalism." Vol. 14, 27 February 1982.

Nie, N., S. Verba, and J. Petrocik. 1979. The Changing American Voter. Cambridge, Mass.: Harvard University Press.

Niskanen, W. 1975. "Bureaucrats and Politicians." Journal of Law and Economics 18: 617-43.

North, Douglass. 1981. Structure and Change in Economic History. New York: W. W. Norton.

Olson, M. 1982. The Rise and Decline of Nations. New Haven: Yale University Press.

Ornstein, N. J. 1975. "Causes and Consequences of Congressional Change: Subcommittee Reforms in the House of Representatives, 1970-73." Pp. 88-114 in N. J. Ornstein, ed., Congress in Change . New York: Praeger.

Plott, C. 1968. "Some Organizational Influences on Urban Renewal Decisions." American Economic Review 58: 306-321.

Pommerehne, W. W. 1978. "Institutional Approaches to Public Expenditures: Empirical Evidence from Swiss Municipalities." Journal of Public Economics 9: 255-80.

Romer, T., and H. Rosenthal. 1979. "Bureaucrats vs. Voters: On the Political Economy of Resource Allocation by Direct Democracy." Quarterly Journal of Economics 93: 562-87.

Rose-Ackerman, S. 1981. "Does Federalism Matter? Political Choice in a Federal Republic." Journal of Political Economy 89: 152-65.

Russell, L. 1983. "Medical Care." Pp. 111-14 in J. Pechman, ed., Setting National Priorities: 1984 Budget . Washington, D.C.: Brookings Institution.

Scheiber, H. N. 1966. "The Condition of American Federalism: An Historian's View." Subcommittee on Intergovernmental Relations to the Committee on Government Operations, U.S. Senate, 15 October.

Shepsle, K. 1979. "Institutional Arrangements and Equilibrium in Multidimensional Voting Models." American Journal of Political Science 23: 27-59.

Shepsle, K., and B. R. Weingast. 1981. "Political Preferences for the Pork Barrel: A Generalization." American Journal of Political Science 25: 96-111.

Stockman, D. 1975. "The Social Pork Barrel." The Public Interest no. 39: 3-30.

Weingast, B., K. Shepsle, and C. Johnsen. 1981. "The Political Economy of Benefits and Costs: A Neoclassical Approach to Distributive Politics." Journal of Political Economy 89: 642-64.

Chapter Two—
Reforming U.S. Federal Fiscal Arrangements

Edward M. Gramlich

Economists have written any number of articles calling for tax and expenditure reform. There have been briefs for reforming the income tax, integrating it with the corporate tax, switching to a consumption tax, moving away from or toward a more progressive rate structure. There have been just as many normative treatises about expenditures—arguing for constitutional limits on total expenditures, reforming the social insurance trust funds, replacing certain expenditures with negative income taxes or vouchers. In a discipline that is alleged to emphasize the positive over the normative, public finance economists have certainly bucked the trend.

But not in one area. In contrast to many other developed countries, the United States has a very decentralized system of fiscal relationships. Over $180 billion is given as intergovernmental grants from higher to lower levels of government, and many promising revenue sources are left completely to local governments. These arrangements, loosely termed a fiscal federalism system, have not received much normative, reformist attention from economists. Politicians have certainly become aware of the potential of the federalism issue, and presidents Nixon and Reagan have both advanced well-publicized reform proposals.^[1] The quasi-governmental Advisory Commission on Intergovernmental Relations (ACIR) has noticed the system and has its own reform proposals.^[2] Economists have done a multitude of theoretical and empirical studies on various aspects of federalism, trying to determine optimal governmental arrangements, predicting the effects of grants or taxes, estimating the degree to which

fiscal decisions are "capitalized" into property values, and the like. Unlike studies of other aspects of public finance, however, these analyses have not provided very pointed statements of what is wrong with the present federal arrangements and how they might be changed to further various goals.^[3] In this paper I try to come up with such a statement.

There are obvious risks in such an attempt—one person's item to be reformed may be another person's ideal. And the theoretical basis for many of these supposed improvements is, as always, in doubt. However, a number of aspects of the present United States federal system seem unlikely to appeal either to those economists who worry primarily about efficiency or to those who worry primarily about equity. For all the positive papers analyzing the empirical impact of federalism, most of these features have not gotten the criticism they deserve from economists.

Two apologies are necessary at the outset. One is that a complete discussion of all aspects of federalism in need of major or minor reform would require a lengthy treatise. There have been two very extensive reviews of federal theory and present-day arrangements in the past decade—by Oates (1972) and Break (1980). Obviously, in one paper I cannot cover all the ground covered by these books and by countless shorter articles. I am forced to be selective both in choice of topics and in the treatment of arguments bearing on the topics. Readers desiring a more comprehensive, and undoubtedly more balanced, discussion can refer to these earlier sources. I must also mention that none of the items I single out for reform is original. I have made a stronger case for many of these measures than is typically found in the literature, but I am certainly not the first to use the relevant arguments.

The second apology is for the omission of a topic that should be fundamental to any discussion of fiscal federalism—that of reforming the structure of governments themselves. The United States has an extremely eclectic structure, with strong historical roots. Some states are large and diverse, others are small; some states conduct extensive expenditure operations of their own, other states leave these operations to localities or special districts; in some areas cities and counties overlap, in others they do not; in some areas special districts are organized to conduct functions, in others they are not; in some areas there is freedom for cities to annex suburbs, in other areas there is not. In all areas it is quite difficult to change whatever arrangements do exist. Economists such as Buchanan (1965) have developed some theories for understanding these arrangements, but there is as yet a wide gap between these theories and their practical applications. I do not even try to fill the gap here, but—as will

become apparent—the optimality or lack of optimality of a set of budgetary arrangements among existing governments depends very much on the existing structures. The two questions should ideally be studied simultaneously, not separately.

The Theory of Federalism and the United States System:
The Musgrave Trichotomy

Twenty-five years ago Musgrave (1959) advanced his now-famous trichotomy that divided governmental functions into their allocation, distribution, and stabilization components. The Musgrave trichotomy is not always very helpful in making particular decisions—almost every tax has both allocative and distributional implications, and most expenditures do too, but it serves a useful function as an organizing device in the area of federalism.

Allocation

Two separate traditions apply to public spending decisions within a federal system. Tiebout (1956) proposed a consumer choice model, according to which rational consumers would select a jurisdiction, and its menu of public goods, that would maximize consumer utility. Jurisdictions would then be led to provide the optimal menu; if not, residents would move to other jurisdictions until utility was maximized.

The second tradition follows Breton's (1965) notion of "perfect mapping" of jurisdictions. According to this notion, jurisdictional boundaries would be set to include only that set of individuals who obtain benefits from the relevant public good. In principle there could be as many jurisdictions as public goods, though in later work Breton and Scott (1978) rationalized a lesser number of jurisdictions by taking into account the costs of organizing and coordinating jurisdictions, and the costs to consumers of relocating.

These two traditions have been combined in various ways by various authors. Oates (1972) showed how jurisdiction size can be determined by the balance between two competing forces—the welfare loss from taste differences, which would argue for small jurisdictions, and the welfare gain from benefit spillovers, which would argue for large jurisdictions. His "decentralization theorem" called for public goods to be provided by the jurisdiction covering the smallest area over which benefits are distrib-

uted, so that public goods efficiencies are maximized and the effect of taste differences minimized. Breton and Scott worked out a more general theory of public goods benefits and organizational costs, but they did not formulate any general theorems, on the grounds that it might always be possible to reduce total costs by various kinds of intergovernmental transfers. Atkinson and Stiglitz (1980) built a series of models that included mobility, changes in the marginal product of labor as labor crowds into a jurisdiction, and income differences. The "results" they got were again very agnostic. Sometimes large jurisdictions were appropriate, sometimes small; sometimes there was a stable local public-goods equilibrium, sometimes not.

In light of this theoretical indeterminacy, it is no wonder that little progress has been made in attempting to determine which levels of government should provide what public services for allocation reasons. It is first necessary to adopt what seem to be reasonable simplifying assumptions and then derive the implications of the relevant model. A plausible set of such assumptions might be that organizing any new government is expensive, that mobility is costly, that changes in the marginal product of labor are small, and that income differences can be ignored (so as to focus only on considerations of efficiency). In this case one is led to the pragmatic conclusion that allocation responsibility for providing public services should be meted out to jurisdictions in accordance with Oates' decentralization theorem. But one should recognize that this conclusion is rather specialized and pertains at best only to marginal changes in the administrative structure and the pattern of production.

Turning to the actual numbers, the distribution of expenditures by function and level of government (Table 2.1) seems more or less in accord with the decentralization theorem. Those expenditures that appear to provide benefits over a wide area—national defense and energy—are conducted almost exclusively at the national level. Those that appear to provide benefits over a narrow area—elementary and secondary education and civilian safety—are carried out at the local level.^[4]

The one mystery in this type of analysis involves state governments. These governments make 60 percent of all government purchases at the national or the local level, but it is not clear what public services convey benefits over as large an area as that covered by most states. I will argue below that at least some of the types of expenditure made by state governments—purchases and transfers for income support and health and hospitals—are better left to other levels of government.

Of the remaining state purchases, transportation is probably the one

public service that does have natural statewide benefits, through the geographical linking of road networks. Education, specifically purchases for higher education by state university systems, may also give benefits statewide, but these benefits do not seem as "natural," since they are strongly influenced by the tuition and admissions policies followed by the state universities. Most such universities offer a large tuition reduction and perhaps relaxed admissions standards to in-state students and then find, unsurprisingly, that in-state students attend in very high numbers. One could argue that the benefits are statewide, but that begs the deeper question of whether the tuition reduction should have been granted in the first place.

To make one suggestive test of the degree of intrinsic statewide benefits, I examined data on University of Michigan (UM) freshmen accepted for admission. The results of this test, in the form of a logit regression explaining students' acceptance probabilities, are given in Table 2.2. They show that once the tuition differential has been eliminated, the probability that in-state residents who were accepted for admission will attend UM is no higher than for the accepted out-of-state residents. That sounds like an example of an unnatural statewide benefit.

Distribution

For this governmental function, the basic theoretical analysis was done by Pauly (1973). The Pauly model determines income distribution by the interdependent utilities of individuals—higher living standards for poor transfer recipients make richer taxpayers better off. In most of his cases Pauly arrives at conclusions close to those of the decentralization theorem—that distributional policies should generally be determined by lower levels of government. Two very strong assumptions must be made to arrive at the result, however, and those assumptions are open to question.

The first assumption involves the geographical linking of utilities. The Pauly model assumes that the welfare of donors can be improved only by raising living standards in the donors' own jurisdictions, as if donors are affected by the sight of, and externalities attendant on, poor people. Some survey evidence analyzed by Ladd and Doolittle (1982) sheds doubt on this assumption. Ladd and Doolittle find that an overwhelming majority of respondents to two separate ACIR polls (1981)—respondents who are assumed to be like those who would ordinarily pay for redistribution programs—believe that the national government should retain an impor-

tant role in supporting needy people. Ladd and Doolittle interpret these results as implying that poor people throughout the nation ought to be the beneficiaries of income support programs, as if donors' preference functions contain no state-line distinction.

The other assumption involves the potential migration of beneficiaries. Even if particular states wanted to be generous, they would not be able to be if prospective beneficiaries of transfer programs were highly mobile. Mobility would raise the tax price of redistribution in all states and would prevent states from following the basic redistribution choices of their donors, for fear of attracting hordes of welfare recipients. The prevailing view seems to be that migration is not, practically speaking, a problem, because only 1 or 2 percent of transfer recipients make interstate moves in a year (see Holmer 1975). But this view is belied by transition matrix calculations given in Table 2.3 which indicate that when transfer recipients (most of whom are not working) do move, they are much more likely to move to states with more generous income-support systems. In the long run, even the low degree of mobility pointed to in the prevailing view can lead to major population shifts among the beneficiary population, as is shown in Table 2.3. This evidence provides another argument for retaining some national interest in income redistribution policies.

But it is not obvious how the national interest should be retained. Tresch (1981) views the federal aspect of redistribution policy as an either/or choice: either the national government would determine an income distribution, or a lower government would. For various reasons he favors having lower governments make the determination, and this leads him to advance a hierarchical redistribution plan. Under this plan, the national government would redistribute income among states, the states among localities, and the localities among households. Legislators at any level could vote for as much or as little redistribution as they wanted. Migration of beneficiaries and positive taxpayers alike could be stabilizing in such a system, if generous localities were entitled to greater transfers from higher levels of governments when low-income families immigrated and high-income families emigrated. But if migration were costly, this system would represent the national interest no better than a fully decentralized system, because there would be no way for national legislators representing national preferences to insure that low-income people were taken care of in particular states. Moreover, even if migration were not costly, the outcome might be socially undesirable, because it could lead to extreme differences in state and local incomes.

An alternative view is taken by Boadway and Wildasin (1984). They

do not see the national-subnational question as an either/or choice but, rather, analyze the question as a matter of benefit spillovers, where the spillover represents both the fact that donors may care about recipients from all states and the fact that when one state raises support levels and attracts migrants, other states benefit. In this logic, taxpayers outside the jurisdiction are willing to contribute to support levels in the jurisdiction, and the appropriate policy would be to allow subnational units to set support levels, partly financed by open-ended matching grants from the federal government. That, as it happens, it close to the present arrangement used in the United States for two of the main general programs for redistributing income, Aid to Families with Dependent Children (AFDC) and Medicaid. Both programs contain the added stipulation that the matching rates are more favorable for low-income states. The third main general redistribution program, food stamps, is a national program with minimum nationwide support levels.^[5]

While the structure of income support programs receives extensive criticism, the Boadway—Wildasin analysis suggests that it is not so obviously in need of repair. I disagree. In the Appendix I give an analysis that shows, first, that any inefficiencies caused by migration spillovers are small, and, second, that present matching grants greatly over-correct for these inefficiencies. Yet even with these overly generous matching grants, AFDC support levels are extremely low in the states of residence of slightly more than half of AFDC beneficiaries. The basic reason is that voters in these low-benefit states appear to have little taste for redistribution, as is readily inferred from their low benefit levels in spite of generous federal matching and upward pressure from the higher support levels in other states.

That leads me to a somewhat paternalistic position, carefully spelled out in the Appendix. I would like to see a uniform national minimum standard somewhere near the level that now obtains in the states of residence of slightly less than half of the AFDC beneficiaries, roughly the Health and Human Services poverty living standard. Certainly this minimum standard could be supplemented by states, and perhaps there should even be a slight federal match for supplementation. My reason for desiring minimum standards is the simple one that I am bothered by the fact that support levels in states paying low benefits are so low. Given the numbers, there is no reasonable way to raise these support levels substantially with Boadway—Wildasin-type matching grants, and no reasonable way to justify national standards by resorting to migration inefficiencies.

Stabilization

The prevailing view as of a decade ago was that national governments should try to stabilize the economy by manipulating taxes and expenditures; subnational governments should not attempt to do so. Over the past decade the first statement has come under a series of withering attacks: the criticism of Mundell (1963) and Fleming (1964) that with flexible exchange rates, foreign capital flows will automatically crowd out fiscal changes; the criticism of Lucas (1972) and Sargent and Wallace (1975) that flexible prices and rational expectations render ineffective any systematic macro-policy changes; the criticism of Barro (1974) that households with long horizons expect to pay the cost of government at some point in history, and hence that the actual timing of tax liabilities (and the split between debt and taxes) has no impact on consumption. There are still unreconstructed Keynesians around (like me), but the faith in activist fiscal policy is substantially less than in former times.

Even in the fiscal activist's heyday, Oates (1972) was arguing that subnational fiscal policies were pointless. In part, his argument was based on the belief that national stabilizing fiscal policies presented a realistic alternative to subnational fiscal policy—now not so readily accepted. In part, the argument was based on a belief that the debt of subnational governments was external and that of national governments internal—a distinction now viewed as obsolete. To the extent that debt is floated on a national or worldwide capital market at a predetermined interest rate, bondholders are no better off by virtue of getting a particular interest payment, all debt is effectively external, and there is no differential advantage in having the national government float the debt.^[6]

In part, the lack of faith in subnational fiscal policy was also based on a view that either the mobility of labor or goods in a country was very high. If the mobility of goods in response to spending demand was high, movements in aggregate demand throughout a country would be highly positively correlated and demand stimulation in one area would not cause extraordinary income changes there. If the mobility of labor was also high, whatever differential movements in demand might occur would inspire offsetting by changes in labor supply.

The last two rows of the transition matrix (Table 2.3) try to verify the latter of these critical assumptions, the assumed mobility of labor. Here the topic is the movement of unemployed workers, either short or long term, between states of high and low aggregate unemployment. The message is certainly to downgrade the importance of labor mobility for

any but the longest run. As with the transfer recipients discussed earlier, in the short run very few workers, even when unemployed for as long as half a year, move from a high to a low unemployment state.

The upshot of all these considerations is that perhaps the question of subnational fiscal policy should be reopened. If most demand shocks these days are ultimately due to relative price shifts that benefit some areas of the country and hurt others (see Medoff 1983 for some evidence on the importance of these), if these shocks are largely transitory, if labor is immobile across regions of the country in the short run, and if currency value changes weaken national fiscal policies, then use of subnational fiscal policies may present a sensible way to decentralize responsibilities for this function of government. In another paper (Gramlich 1984) I make this argument in more detail. Most states have constitutional provisions that prevent them from running current-account-budget deficits, but they are not prevented from altering taxes and expenditures in response to income changes in their areas, and it appears to me that they should follow such policies.

Tax Assignment

Two basic questions arise in any examination of the federal structure of taxation. The first involves the levels of taxes raised by national and subnational governments; the second, the types of tax used.

National vs. Subnational Taxes

While the presumption on the expenditure side of the budget is that expenditure programs should be conducted by subnational governments whenever possible, there is an opposite presumption on the tax side. Partly because of a belief that the administrative costs of levying taxes are higher for subnational than national governments, partly because of a fear of tax competition, the standard belief is that tax collection should be centralized whenever possible.

As on the expenditure side, this presumption can at least be said to be specialized, perhaps appropriate in some cases but certainly not in general. For one thing, it totally ignores a point brought out by the new "rent-seeking" literature: that inefficiencies due to lobbying for the grants may dwarf conventional economic inefficiencies. When expenditure programs are decentralized and taxes centralized, large-scale general-purpose transfers (which actually exist in other federal countries such as Canada

and Australia) are needed to balance budgets at both governmental levels. These large transfers place a premium on local politicians who can lobby for grants from the federal government and very little premium on those who are effective managers of their governments—a common complaint in countries that rely heavily on tax-sharing grants (Walsh 1983). The rent-seeking literature should then alert us to a competing principle—that, as a rule, those governments that buy government services should impose their own taxes.

Moreover, the administrative cost argument given in favor of tax centralization seems quite weak. Perhaps in less developed countries it may be true that the national government can administer tax laws more effectively than can subnational governments, but there is no research supporting such a proposition in the United States. And it would be strange if such research could be found, since all states have to do to lower their own administrative costs and the compliance costs of their taxpayers is to use the federal tax base and apply their own rate.^[7]

The question of competition among various governments for desirable tax bases is more complicated. In a Tiebout model with costless migration, competing governments at the same level should strive to eliminate what Buchanan (1950) called fiscal residuals : the difference between taxes paid to a local government and expenditure benefits received from it. The threat of tax competition among subnational governments will then limit the extent to which any of these governments can tax either industry or well-off individuals in their own community, for if these groups are asked to pay extra costs, they will simply leave the community. There is then relatively little scope for assessing redistributive taxes at any but the national level.

This innate limitation on redistributive taxes at the subnational level does not require fiscal transfers from higher to lower levels of government for redistributive spending, as long as the redistribution is done mainly by the federal government, as I have previously argued that it should be. But it could justify such transfers for spending done for allocation reasons. Gordon (1983) points out that horizontal tax competition eliminates an opportunity for a decentralized government to assess completely nondistorting taxes, if there are some factors that are in highly elastic supply to subnational jurisdictions and completely inelastic supply to the nation. In principle Gordon's point is important, but I would be more worried about it if I could determine what such productive factors are—in today's open-economy models, a routine assumption is that capital and perhaps even entrepreneurship are in elastic supply to the whole

country (called the "small country" assumption, for obvious reasons). Pending illumination on this point, my tentative position on this issue is that while it is theoretically possible that principles of tax assignment would call for having certain taxes levied by national governments and accompanied by grants to the subnational governments actually doing the spending, there is no clear evidence that such an arrangement is appropriate for the United States.

Another form of tax competition would also call for more centralization of the revenue-raising function than the spending function. This form is not very common in the United States, though it exists in resource-rich countries such as Australia. It involves the vertical tax competition among all levels of government that could lay claim to taxing profitable resource deposits. Cassing and Hillman (1982) tell a story about rail freight for coal in Queensland, the most resources-rich Australian state. The federal government tries to gain its return from Queensland coal by imposing an export duty. The state of Queensland tries to gain its own return by charging exploitive rail freight rates. As the federal government raises its rates to gain revenue, the profitability of coal is reduced, as is the monopolistic freight rate Queensland can charge. Similarly, by raising its freight rates Queensland can reduce revenues available to the federal government. If the two were to compete, they would tax coal excessively and generate suboptimal tax revenues for both governments. This is one case where it would make sense to centralize taxes and have one government distribute a share of the optimal tax revenue to the other.

Whatever the resolution of these typically complex normative issues, the previously reported data on general-purpose grants suggest that there is in fact broad adherence to the levy-your-own-taxes principle in the United States. Table 2.1 shows that only $5 billion of the $88 billion in federal grants to state or local governments in 1981 were for general purposes, and only $11 billion of the $93 billion in state grants for local governments.

Types of Taxes

The two basic principles for organizing a tax system are the ability-to-pay and the benefit principles. In a federal system we would expect that migration among subnational jurisdictions would be an important factor in the long run (as the tax competition argument and the evidence on AFDC benefits, described above, suggested), which in turn implies that subnational jurisdictions ultimately have only one feasible taxing ar-

rangement, the benefit principle. If they try to make well-off individuals or industry pay taxes that will be spent in the form of programs that benefit others in the jurisdiction, these groups will move out of the jurisdiction.

It follows that in a federal system, most ability-to-pay, or redistributive, taxes will be imposed at the national level, while most benefit taxation will be done at the state or local level. An examination of actual tax data for the United States (Table 2.4) shows this to be generally the case. Income and corporate taxes, the most important ability-to-pay taxes, are imposed mainly at the national level. Contributions for social insurance, used to finance trust funds such as social security, Medicare, unemployment insurance, and the like, are also assessed at the national level. Those state or local contributions that are assessed are for pension systems for the employees of these governments, and should be viewed as a component of the wages of these employees. Those taxes that are mainly benefit taxes, fees and charges and property taxes, are assessed at the local level.^[8] As on the expenditure side, the taxes of state government represent a mixed bag, with some ability-to-pay taxes, some benefit taxes, and the state sales tax, which is hard to classify.

Tax Exportation

One important source of inefficiency in this division of taxing responsibility involves the possibility of exporting taxes. A standard claim on the expenditure side of the budget, routinely advanced as a rationale for categorical grants from the federal to lower governments, is that spending can be too low if some benefits from an expenditure program are realized outside the community. There is a similar, though less commonly heard, argument on the tax side (first made by McLure 1967). If taxes can be exported from a jurisdiction to individuals outside the jurisdiction, without a concomitant transfer of expenditure benefits, local citizens are not internalizing all the costs of public services, and they will spend too much on these public services. Just as we have categorical subsidies for those types of expenditures with benefit spillovers, we should in principle also assign public-service excise taxes for whatever spending is financed by exportable taxes.

The difficulties of matching expenditures and the taxes used to pay for them probably make any formal excise-tax scheme impractical, but there may be other arrangements that should be made to deal with tax exporting. One is to have the federal government assume all tax sources that can easily be exported. It is sometimes argued that this is why the federal government should take over responsibility for the corporate tax (as it largely has), but that view does not accord with prevailing views on the incidence of the corporate tax. That tax is now generally considered to be a tax on a mobile factor, capital, which would drive it, and ultimately labor, out of a jurisdiction, leaving the tax to be paid by the locationally fixed factor, land. Hence in general the corporate tax would not be exported. What would be exported are capital taxes on factors within a jurisdiction owned by outsiders (such as on resources) and excise taxes on travelers with a low price elasticity of demand.

As a practical matter, tax exporting inefficiencies do not seem to be of overriding practical importance in the United States. A recent study by Mutti and Morgan (1983) finds the revenue implications of excise-tax exporting to travelers to be very small, even for states, such as Florida, for which tourism is very important. Beyond that, though the decisions have been made on constitutional grounds, a long series of Supreme Court rulings have effectively prevented the taxation of outsiders and have thereby kept down the distortions that could have arisen from exporting (Hellerstein 1977).

Although the Supreme Court is of course concerned with legal tradi-

tions and not economic inefficiencies, there is one way an alleged constitutional restriction does cause tax exportation. As a result of court rulings in the early 1940s, the federal government now does not tax state and local bond interest payments. This treatment lowers state and local bond interest rates to about 80 percent of the rate for comparable-risk corporate securities, and subsidizes state and local investment in all communities affected by cost-of-capital. There seems to be no economic point to such a subsidy—if there are spillover benefits, matching grants can be used—and its elimination would improve the overall allocation of capital in the United States, as well as improving the equity of the federal income tax. If the subsidy really is rooted in constitutional constraints, this distortion will be with us as long as we have an income tax, though it could still be removed by moving to an expenditure tax, which taxes return to capital only when it is consumed. But there is enough confusion about the ultimate origin of the subsidy that one might also argue for another court ruling on the constitutionality of having the federal government tax state and local interest payments.

While the courts have generally tried to limit tax exporting, congressional actions have generally gone the other way. In one significant case, Congress explicitly encourages a form of tax exporting through the income-tax deduction for state and local taxes paid. This federal deduction lowers the marginal tax price for local public goods for those voters who itemize deductions, and represents exactly the sort of tax exporting that should be prevented from the standpoint of efficiency.^[9]

An examination of the impact of this tax deduction within a state suggests that its effects might be considerably more pernicious than are commonly supposed. A first point is that not all state and local revenues are deductible—in general, fees and charges, the revenues that most closely conform to the benefits principle and hence cause least deadweight loss, are not deductible, while the less efficient income and sales taxes are. A second point is that only 30 percent of all tax returns claim itemized deductions. If the median, or decisive, voter in a community does not itemize, then the deduction does not affect state or local spending but merely represents an unwarranted tax break for the high-income taxpayers who do itemize.

But that is not the end of it. Although only 30 percent of all tax returns claim itemized deductions, it may still be that a high percentage of the tax returns filed by voters itemize deductions. Table 2.5 presents data from a survey of Michigan voters. The bottom row of the table shows that of all 2001 survey respondents, 862, or 43 percent, claim itemized deductions.

The difference between the 30 percent share from overall statistics and this 43 percent share is apparently due to tax returns filed on behalf of minors and others not likely to appear in a voter survey sample. The bottom row also indicates that among voters, the share of itemizers rises to 49 percent. And some of the disaggregated numbers in the table show that among voters in high-income areas, such as the Detroit and Lansing suburbs, the share rises to 60 percent.

Assume for the moment that the median voter model gives a reasonably accurate picture of how public-spending decisions are made in local jurisdictions. The numbers in Table 2.5 indicate the maximum number of voters who would be swayed by the state-local tax deduction to switch their vote in favor of higher expenditures. Using the overall statistics as an example, say that 49 percent of the voters itemize, and that voting turnout decisions are unaffected by the deduction provision. Then up to 49 percent of the voters could have their desired size of public spending increased by the tax deduction. If not too many of these voters had previously favored big governments (in which case they would have already voted for high spending), public spending would be likely to rise in the community. Reasoning in this way, we can see that the effective tax price for public spending is more likely to fall, and public spending to rise, the richer the community.

The tax price for public spending on such social investment services as education is already relatively low in these richer areas because of their high tax base. Now this basic advantage is compounded by the federal

tax deduction. Indeed, a firm believer in migration will also argue that the tax deduction sets up incentives for rich people to live together so that they can export their taxes to others. It is hard to imagine a consciously designed public-policy measure having worse impacts on both efficiency and equity, in the short and the long run, than the federal income-tax deduction for state and local taxes.

Intergovernmental Grants

The other important financing mechanism in a federal system involves intergovernmental grants. As the previous numbers made clear, the United States has a very extensive grant system at both the national and the state level. While most existing grants are categorical, there are separate rationales for categorical grants and for general-purpose grants.

The Rationales for Grants

Grants from higher to lower levels of government can be of a form that alters relative prices facing the recipient government, or that does not. With general-purpose grants the price structure is not altered: these grants affect community income only, and stimulate local consumption of public services according to the income effect. With open-ended categorical grants the price structure is altered: these grants lower relative prices for certain types of expenditures, have both income and substitution (price) effects, and stimulate local consumption of public services according to the uncompensated price elasticity of demand. Since the substitution effect expands spending, if anything, it is easy to establish that open-ended price reduction grants stimulate more spending on the public service in question per dollar of the federal grant.

Whether one type of grant or the other is appropriate depends on the type of problem being corrected by the grant. If there are externalities that imply that social benefits from public services exceed those realized within a community, open-ended price subsidies are appropriate—just the reverse of the tax exportation argument. If the mismatch of expenditure and revenue responsibilities described above is present, general-purpose transfers are appropriate. Indeed, Breton and Scott (1978) point out that there may be any conceivable mismatch of administrative responsibilities for taxes or expenditures at any level of government, making any set of transfers, from higher to lower governments (as most now are), or from lower to higher governments, appropriate.

But the most commonly discussed rationale for general-purpose transfers, and the one that is potentially most relevant in the United States, involves income differences across communities. Should these exist, there will be one of two outcomes. If benefit taxation is not complete, rich people will be net contributors to the public budget and poor people will receive some transfers through the public budget (in Buchanan's [1950] terms, the rich have positive fiscal residuals and the poor negative ones). If benefit taxation is complete, poor people will gain from the higher demand by rich people for public goods.^[10] Either way, the tax price for public services in a community will depend on how many rich people there are or, crudely, on community income.

The argument often stops there, but it should not. Differing tax prices do not necessarily constitute a social ill, though they can if they lead poor communities to under-consume (relative to rich communities) merit public services such as education. One of many ways to eliminate these tax price differences across communities is through general-purpose transfers. But the general-purpose transfers must be compensatory, that is, they must be given in greater per-capita amounts to poor than to rich communities. And it is by no means obvious that general-purpose transfers are the best way to deal with these community income differences. Yinger (forthcoming) points out that two separate notions of equity could be applied in problems such as this: (a) fair compensation, under which all communities would have access to the same bundle of all goods; or (b) categorical equity, under which the expected expenditures on designated public services would be equalized. Under the former notion, general-purpose grants would be appropriate and would have to be given in the amount of income differences among communities. Under the latter, general-purpose grants could be given, but it would be possible to achieve the same end with fewer grant dollars by using Feldstein's (1975) variant of the power equalization approach.

Say that expenditures on public services in the i th community were determined by the linear equation

where E_i is real consumption of public services, Y_i is community income (here used as a proxy for community living standards), G_i is general-purpose grants received by the community, P_i is the dollar size of price-reduction (open-ended matching) grants received by the same community, and the residual u_i refers to all other reasons why spending might differ across communities. With normal preference functions, a₃ >a₂ ; that

is, the price-reduction grants would provide an added impetus to spending. If there exist what are known as "flypaper" effects, whereby general-purpose grants stimulate more spending than income, a₂ >a₁ .^[11] To achieve fair compensation through general-purpose grants, ¶ G_i /¶ Y_i must equal - 1, that is, all income deviations must be compensated dollar for dollar by larger general-purpose grants. To achieve categorical equity, it is merely necessary to insure that expected spending is equal across communities. This is done simply by taking the total derivative of (1) with respect to income, arriving at ¶ G_i /¶ Y_i = - a₁ /a ₂ if equalization is accomplished through general-purpose grants, and ¶ P_i /¶ Y_i = - a₁ /a ₃ if through open-ended matching grants. The first derivative is greater than or equal to - 1; the second is clearly greater than - 1. In these latter cases, because the price sensitivity is relied on to stimulate public spending, and because the standard is not to eliminate all spending deficiencies but only the public-services spending deficiency, less grant money is needed to achieve categorical equity.^[12] Another advantage of this form of matching grants, not shown in the analysis, is that the stimulated public spending could be limited to those public services that really are merit goods.

The upshot of all this is that the usual rationales for general-purpose grants are all quite limited. One rationale could be the assignment-of-responsibilities mismatch described above, though it seems that American subnational governments are able to raise enough revenue to pay for their spending programs without resorting to obviously inefficient taxation. One could be the income-differences argument, though if the categorical-equity standard is used, categorical grants can achieve the same objective with fewer grant dollars.^[13]

Actual Grants

Table 2.6 shows real levels of intergovernmental grants at the federal level over the past decade, disaggregated by type of grant. In principle, a similar analysis of state grants to local governments should be done, but I do not show these data because nobody has developed disaggregations based on the type of grant, and those distinctions are crucial in the analysis.

The table shows that a decade ago, in 1972, there were almost no general-purpose grants, fairly large grants for income support, and even larger categorical grants for other, benefit spillover programs. Over the decade, general-purpose grants rose until 1980 because of the introduction of the Nixon administration's general-revenue-sharing program,

then dropped back in real terms when the Carter administration eliminated general-revenue-sharing for state governments. Categorical grants remained stable until 1980, then dropped back when the Reagan administration killed some of them and converted others to what are known as block grants—grants that are nominally categorical (money has to be spent on a designated function), but effectively for general purposes (the designated functions are so broad, and enforcement so limited, that recipient governments can effectively do what they want with the grant money). Income-support grants have grown from 1972 levels due to the exploding costs of Medicaid, offset by a real drop in AFDC grants.

General-Purpose Grants

The present general-revenue-sharing program goes only to local governments, basically on a per-capita formula. There are provisions in the law that have the effect of giving slightly more funds to poorer areas, but the redistribution is minimal and haphazard.^[14] Even if the redistribution were effective and systematic, however, there would seem to be little cause to retain general revenue sharing. The program violates the levy-your-own-taxes principle described above, in a way that has never been defended in terms of administrative cost saving. And even if redistribution were a more important objective of the program than it apparently is, open-ended matching grants with federal matching rates depending neg-

atively on community income or positively on needs could accomplish the goals of categorical equity with fewer grant dollars.

Block Grants

These grants appear to represent a political compromise: conservatives would like to kill many categorical grant programs altogether but, lacking the muscle to do that, they settle for converting the grant to block form. As said above, this effectively makes the grant into a general-purpose grant. But the allocation of funds for the grant program is based on whatever categorical program just got cashed out—miles of highway, numbers of dilapidated houses, or whatever. Since the grants now are for general purposes, the random elements in the grant distribution formula make the funds allocation even more haphazard than for general revenue sharing. There are also excess administrative costs to maintain the fiction of the block grant. Finally, as argued above, there is no very good argument for general-purpose grants anyway. For all these reasons, these grants should either be terminated or converted to categorical-equity-matching grants for poorer communities. Block grants may provide a useful political compromise, but it is hard to see why an economist who worries about efficiency and equity would ever favor such programs.

Income-Support Grants

I argued above in favor of replacing federal grants for income support with a national program paying basic income-support levels. This should consist of a basic national benefit level (say somewhat above the present average level of AFDC and food stamps). States should be allowed to supplement this level, perhaps with slight (say 25 percent) federal matching support.

In 1983, federal matching grants for all income-support programs totalled $45 billion (see Table 2.6). States spent another $13 billion of their own funds on Medicaid and yet another $6 billion on AFDC. Were the federal government simply to assume these expenditures, federal costs would rise by about $19 billion. Then there should be some reallocations, with most of the $6 billion going to raise AFDC benefits in low-benefit states in the South. The total amount of funds devoted to AFDC can be greater than $6 billion because of the fact that Medicaid is a program much in need of cost-saving reform apart from the federal aspects focused

on here, and various cost-sharing measures should be able to reduce expenditures on it.

Other Categorical Grants

Until the recent introduction of general revenue sharing, conversion to block grants, and rapid growth of Medicaid, the main form of federal intergovernmental transfer has been categorical matching grants in areas such as transportation, education, community services, environmental protection, and hospital construction. These grants appear to have as their rationale benefit spillovers across jurisdictional lines,^[15] but in fact the grants are structured so that they are unlikely to achieve any such objectives.

A first fact about these grants is that legal federal matching shares are very high, averaging 80 percent across the present $29 billion of other categorical grants.^[16] While there may be some benefit spillovers, at the margin the ratio of internal to total benefits for these programs seems to be much higher than 20 percent. This gives states an incentive to overspend, and overspend they probably will. To prevent grant levels from becoming very large, the federal government is forced to impose limits on the size of the grant—overall program limits enforced by the Office of Management and Budget, formula limits for individual governments or groups of governments, and agency limits for application grants. Standard indifference-curve analysis next shows that the price reduction is not effective at the margin, and that the grants have effects much like those of general-purpose grants.^[17] Then budget-cutters such as David Stockman come along and argue that the grant should be either terminated or converted to block form.

This political cycle is designed to end in the termination of the categorical-grant program. Perhaps many of these grants should never have been passed—that is a question for the benefit-cost analysts. But if there is a valid spillover rationale for categorical grants, a better way to improve the grant than by simply converting it to block form and effectively killing it can easily be found. That better way is simply to lower federal matching shares until the ratio of internal to total program costs at the margin equals the ratio of internal to total program benefits at the margin.

In making such a cavalier proposal, I realize that it will often be difficult, if not impossible, to estimate the critical marginal ratio very precisely. But it should not be difficult to come closer than the 20 percent that is now the standard. My own preference would be to assume an internal

share of 80 percent unless it could be shown to be significantly lower. If this is the appropriate share, a very simple demand analysis indicates that such a change should reduce federal categorical grants by about $11 billion, increase expenditures on public services with benefit spillovers, and eliminate a dead-weight loss that appears to be about 1 percent of the level of expenditures.^[18]

The Reform Agenda

Rather than summarize the paper, I will try to maintain its spirit by listing what seem to me to be the major problems with present-day United States fiscal federal arrangements. These problems and the proposed remedies are stated very bluntly, without even trying to list the many compromises and intermediate reforms that could work in the right direction.

1. Given the inability of the national government to stabilize demand shocks, or to stabilize them in different regions simultaneously, states should undertake limited use of stabilization policies. They can do this by creating rainy-day funds, building up these funds (running budget surpluses) in boom years, and running down the funds in recession years.

2. The federal deduction for state and local taxes paid should be abolished. This change will raise federal revenue by an estimated $26 billion (see the summary in Table 2.7), it should affect tax prices and public-spending levels relatively little in rural jurisdictions and central cities, but it should raise marginal tax prices and lower public-spending levels in high-income suburbs. I would have made a similar recommendation, for a similar reason, about the income-tax exclusion for state and local interest, but I am assuming that the provision exists for constitutional reasons, and even my reform proposals do not go that far.

3. Federal grants for income-support programs should be replaced by direct federal income-support programs with uniform national benefit levels and optional state supplementation, perhaps with limited matching support. At today's levels, a reasonable package would raise federal budget expenditures by $20 billion.

4. Federal general-purpose grants and block grants should be replaced by categorical-equity matching grants to poorer communities for merit public services such as education, health, and housing. At

today's levels, $17.5 billion of federal grants would be saved, and perhaps $10 billion could be used for the categorical-equity grants.

5. Federal categorical grants should be altered by lowering federal matching shares to a level that better corresponds to the ratio of internal to total benefits and by eliminating limits on the size of the grant. At today's levels, such a change is likely to reduce federal budget expenditures by about $11 billion.

The measures are not advanced as a package—any one of them could be adopted with or without any of the others. If the whole package were passed, Table 2.7 suggests that all changes combined should reduce the federal budget deficit by about $24 billion—a saving equal to about one-seventh of the enormous present level of the deficit. The short-run impact of the package would be to raise income-support levels greatly for low-income people in the South; the long-run impact should be to raise support levels for all low-income people. People in low-income commu-

nities should benefit further through increased consumption of merit public services. But not everybody will be better off. For a change, high-income itemizers will be made worse off.

Appendix:
Decentralization of Income Distribution Responsibilities

In the text I asserted that the twin forces of beneficiary migration and the diverse levels of AFDC benefits desired by different states argue for a more centralized system of income support. In this appendix I give the exact nature of this argument. The appendix uses real numbers and parameters from Gramlich and Laren (1984).

Suppose we had a country consisting of two states, one that preferred to pay relatively generous AFDC support levels (B₁ ) and one that preferred to pay very low support levels (B₂ ). Each state determines benefits by maximizing the utility function for its decisive voter:

subject to the constraint that

where i indicates the state determining benefits, j indicates the other state, U_i the utility value for the state's decisive voter, Y_i this voter's pre-tax income (for which I will use average per-capita income), T_i the proportional state income tax rate to pay for AFDC, M_i the federal matching rate for the state, and R_i for the level of recipients per capita if state i paid the same benefits as state j . If state i benefit levels exceed those in the other state (B_i >B_j ), recipients would rise according to the migration-sensitivity parameter b . There are cross-state transmission effects whenever b >0.

Determination of benefits in this two-state country is shown in the well-known bargaining diagram Figure 2.1. At the Nash point N each state is maximizing utility under the assumption that all federal matching rates are zero and that benefits in the other state are fixed. At this point we have, for state 1,

where u₁ is the partial derivative of U₁ with respect to B₁ , and where the partial derivative of U₁ with respect to after-tax income in state 1 has been normalized at unity. At the Nash point there is no first-order impact of changing benefits on utility, because the state has already maximized by

Figure 2.1
Determination of Benefits by Bargaining

equating the marginal value of benefit increases (u₁ ) to the marginal cost of income losses (the term in brackets). Hence the indifference curve passing through this point, U₁ , is horizontal there.

But we also know that at the Nash point B₂ increases will raise U₁ , because they attract AFDC recipients into state 2 and raise after-tax income in state 1 for every level of B₁ . This can be seen by differentiating the utility function with respect to B₂ :

This result shows why the U₁ curve is concave from the horizontal axis:

as B₁ is, say, raised above its optimal level at N, increasing rises in B₂ are necessary to keep the state on its indifference level.

The same results are true for state 2. The equation analogous to (3) shows that the U₂ curve is vertical at Nash point N. The equation analogous to (4) shows that benefit increases in state 1 raise utility in state 2. And the curve is concave to the vertical axis because increasing rises in B₁ are necessary to compensate for non-maximizing changes in B₂ .

The fact that indifference curves cross at N implies that utility can be raised simultaneously in both states by bargaining, or by having the central government simply set benefits in both states at some level that raises utility in one state without lowering it in the other. Such a contract curve solution could be found by maximizing the joint utility function

for some arbitrarily specified weight w , assumed to lie between zero and one. But for reasons that will become apparent shortly, that is not going to be my argument for greater centralization. Another proposition is that the Boadway and Wildasin (1984) solution of federal matching could in principle be used to arrive at a contract curve solution, by effectively externalizing some of the cross-state gains of higher benefit levels in a state. Shortly we will also see that if this is the justification for federal matching, actual matching rates turn out to be far too generous to achieve this limited goal.

The diagram gives the area of possible gains from greater coordination as the shaded area. The northeastern limit of this area, denoted by X, can be approximated by reasoning that U₁ becomes vertical when rises in B₁ no longer raise utility in the state, that is, when B₁ is so high that u₁ is zero. Similarly, U₂ becomes horizontal when u₂ is zero. These points can be located with the equations

where the solution is just as before except that the derivatives are used to solve for the levels of B₁ and B₂ where the indifference curves become vertical and horizontal respectively. It can be seen from the graph that the exact crossing point is at a B₁ value less than that at which U₁ is vertical and at a B₂ value less than that at which U₂ is horizontal. I could locate this point exactly if I were willing to assume a specific form for the utility functions, but we will shortly see that such a step is not necessary.

This is all the positive theory that is required; I now try to find the various points with real-life values. The values for B₁ and B₂ will be state

monthly AFDC guarantee levels for a family of four in 1981 dollars, assuming food stamps at a standard national level. Given this level, AFDC guarantee levels of about $600 would have been necessary to keep a family out of poverty status in 1981 (labeled P on the diagram).

The first step is to locate actual benefit levels in the presence of federal matching and migration. Data from Gramlich and Laren (1984) show that in typical high-benefit states, covering nearly half of AFDC recipients, B₁ ranged from $540 to $720 with a mean close to $600. In low-benefit states, B₂ ranged from $115 to $510, with a mean close to $200. These values are indicated on the diagram by point A.

Values at the Nash point N are determined from the equations given in Gramlich and Laren. Equation 9 there shows benefits to be determined in the two states by the relationships

where L(x ) denotes the log of a variable, c is the price elasticity of demand for AFDC benefits, and the Z_i terms represent the influence of all other variables. The Nash point is found by using the actual values listed above and reasonable parameter estimates from the paper (b = 0.65, c = 1.8) to evaluate the Z_i . Then these figures replace Z_i , the M_i are set at zero, and the values at point N are computed: B₁ falls from $600 to $105, B₂ from $200 to $23. The sharp drops are caused by three factors.

1. There is a sharp change in matching ratios, from M₁ = 0.5 and M₂ = 0.75 to zero.

2. The estimated price elasticity is high.

3. The estimated migration effect is high, causing the fall in B₁ to lower B₂ , this to lower B₁ , and so forth as in a multiplier.

Whether these estimates are believable is, of course, not as obvious. The paper finds both estimates to be highly significant, and it does confirm the all-important migration parameter with two different bodies of data.

The next step is to locate point X, the approximation for the northeastern crossing point. Because u₂ is already found to be very low (that is why the group 2 states pay so little AFDC, even when most costs are financed by the federal government), U₂ becomes horizontal when B₂ rises to 30; the same type of calculation makes U₁ vertical when B₁ equals 150. These two values are shown as point X on the graph; the true crossing point is

slightly to the southwest. That these AFDC levels are so low explains why I do not use this lack of coordination argument for greater centralization: the rises in benefits due to improved coordination would not be very high.

Two other points are shown on the graph. Point C, for a closed economy, shows the solution to the model given in Equation 7 when there is no matching and when there is no migration effect. What migration does is to bring benefits in the two states together; when b is set at zero, states diverge to what might be thought of as their true preference benefit levels—close to the actual value for B₁ but close to zero for B₂ . In this sense the present matching-grant system is about right for preserving the closed economy solution for the B₁ states but much more generous than is necessary for preserving the closed economy solution for the B₂ states. Finally, the point labeled M goes the other way, showing the equilibrium values for benefits when migration is infinitely sensitive to benefit disparities, as might be assumed in extreme versions of the Tiebout model. In this case benefits must be equal in the two states, by my calculation at about $40.

Hence the results here are dominated by the low intrinsic desire on the part of the group 2 states to pay benefits. These states now pay benefits of only $200 a month for a family of four, roughly one-third of the poverty line. And this in the presence of a price reduction by the federal government that averages 75 percent, a large estimated price elasticity of demand, a large migration sensitivity, and benefits in other states over three times as high. When the federal matching is eliminated to find the Nash point, benefits in these group 2 states fall to very low levels, and in the presence of the migration sensitivity, this pulls down benefits in the other states. With a Nash point anchored at this low a level, none of the normal federal policy measures aimed at improving coordination will have much effect—even at point X, benefits will be less than $30 in the low-benefit states.

To be candid, then, my policy suggestions are not aimed at preserving tastes in these low-benefit states. My goal is that benefits in all states be set at something like the poverty level of $600 per month, which also happens to be the preferred level in the high-benefit states. The much-maligned present system does bring about this result in these high-benefit states; its defects are highly overrated. But it still does fall short of my goal because of the low benefits in the other states. The easiest way to achieve the goal is simply to have the federal government establish uniform national benefit levels, with optional state supplementation.

As a final matter here, one might ask about two types of sensitivity

tests. First, does the low intrinsic "taste" for AFDC benefits among low-benefit states reflect their low income? The answer is no . The Gramlich–Laren paper also estimates income elasticities as part of the model, and even when these are complicated by the fact that low-income states are likely to have more AFDC recipients, other things being equal, the income elasticities are extremely low. Redoing the Figure 2.1 analysis with all incomes standardized would lead to only trivial changes. AFDC benefits are low in low-benefit states for reasons that are not captured in the income term: many of these states have high incomes.

The other type of sensitivity test refers to the migration parameter b . It is already high enough that the Nash point is geometrically closer to M than to C. But it is also true that if b were underestimated for any of a number of reasons (the opposite of the usual criticism of the Gramlich–Laren paper), point X could stretch to cover point A. Unfortunately, it would take a b value close to 10 (15 times the value now estimated) to get X up to A, and close to 26 (40 times the value now estimated) to get X up to P. Theoretically, more sensitive migration could relieve the strain of the paternalistic argument for centralization: realistically, it cannot.

Acknowledgments

I would like to thank the editors of the volume, along with Henry Aaron, Harvey Brazer, Theodore Bergstrom, Paul Courant, Roger Gordon, Wallace Oates, David Wildasin, and John Yinger for helpful comments. Deborah Laren and Marieka Klaiwitter produced some of the numbers, and some of the work was financed by a grant from the Sloan Foundation.

References

Advisory Commission on Intergovernmental Relations. 1981. "Changing Public Attitudes on Governments and Taxes, 1981." S-10. Washington, D.C.

Atkinson, Anthony B., and Joseph E. Stiglitz. 1980. Lectures on Public Economics . New York: McGraw-Hill.

Barfield, Claude E. 1981. Rethinking Federalism: Block Grants and Federal, State, and Local Responsibilities. Washington, D.C.: American Enterprise Institute.

Barro, Robert J. 1974. "Are Government Bonds Net Wealth?" Journal of Political Economy 82: 1095-1117.

Boadway, Robin W., and David E. Wildasin. 1984. Public Sector Economics . Boston: Little, Brown.

Break, George E. 1980. Financing Government in a Federal System . Washington, D.C.: Brookings Institution.

Breton, Albert. 1965. "A Theory of Government Grants." Canadian Journal of Economics and Political Science 31: 175-87.

Breton, Albert, and Anthony Scott. 1978. The Economic Constitution of Federal States. Toronto: University of Toronto Press.

Buchanan, James M. 1950. "Federalism and Fiscal Equity." American Economic Review 40: 583-99.

—––. 1965. "An Economic Theory of Clubs." Economica 32: 1-14.

Cassing, J. H., and A. L. Hillman. 1982. "State-Federal Resource Tax Rivalry: The Queensland Railway and the Federal Export Tax." Economic Record 58: 235-41.

Courant, Paul N., Edward M. Gramlich, and Daniel L. Rubinfeld. 1980. "Why Voters Support Tax Limitation Amendments: The Michigan Case." National Tax Journal 33: 1-20.

Feldstein, Martin S. 1975. "Wealth Neutrality and Local Choice in Public Education." American Economic Review 65: 75-89.

Fleming, Marcus. 1964. Domestic Financial Plans under Fixed and Flexible Exchange Rates. Washington, D.C.: International Monetary Fund Staff Papers.

Gordon, Roger H. 1983. "An Optimal Taxation Approach to Fiscal Federalism." Quarterly Journal of Economics 98: 567-86.

Gramlich, Edward M. 1982. "An Econometric Examination of the New Federalism." Brookings Papers on Economic Activity 2: 327-60.

—––. 1984. "Subnational Fiscal Policy." Mimeo. Ann Arbor: University of Michigan.

Gramlich, Edward M., and Deborah S. Laren. 1984. "Migration and Income Redistribution Responsibilities." Journal of Human Resources 19: 489-511.

Hellerstein, Jerome R. 1977. "State Tax Discrimination against Out-of-Staters." National Tax Journal 30: 113-34.

Holmer, Martin. 1975. "Economic and Political Causes of the Welfare Crisis." Ph.D. dissertation. Cambridge, Mass.: MIT.

Ladd, Helen F. 1982. "Financing Public Services in the Federal System." Pp. 31-44 in Federalism: Making the System Work . Washington, D.C.: Center for National Policy.

Ladd, Helen F., and Fred C. Doolittle. 1982. "Which Level of Government Should Assist the Poor?" National Tax Journal 35: 323-36.

Levin, David J. 1983. "Receipts and Expenditures of State Governments and of Local Governments, 1968-81." Survey of Current Business 63: 25-38.

Lucas, Robert E. 1972. "Expectations and the Neutrality of Money." Journal of Economic Theory 4: 103-124.

McLure, Charles. 1967. "The Interstate Exporting of State and Local Taxes: Estimates for 1962." National Tax Journal 20: 49-77.

Medoff, James L. 1983. "U.S. Labor Markets, Imbalances, Wage Growth, and Productivity in the 1970s." Brookings Papers on Economic Activity 1: 87-120.

Mieszkowski, Peter, and William H. Oakland. 1979. Fiscal Federalism and Grants-in-Aid . COUPE Papers on Public Economics 1. Washington, D.C.: Urban Institute.

Mundell, Robert A. 1963. "Capital Mobility and Stabilization Policy under Fixed and Flexible Exchange Rates." Canadian Journal of Economics and Political Science 29: 475-85.

Musgrave, Richard A. 1959. The Theory of Public Finance: A Study in Public Economy . New York: McGraw-Hill.

Mutti, John H., and William E. Morgan. 1983. "The Exportation of State and Local Taxes in a Multilateral Framework: The Case of Household Type Taxes." National Tax Journal 36: 459-76.

Nathan, Richard P., Allen D. Manvel, and Susannah E. Calkins. 1975. Monitoring Revenue Sharing. Washington, D.C.: Brookings Institution.

Oates, Wallace E. 1972. Fiscal Federalism . New York: Harcourt, Brace, Jovanovich.

Pauly, Mark V. 1973. "Income Redistribution as a Local Public Good." Journal of Public Economics 2: 35-58.

Sargent, Thomas J., and Neil Wallace. 1975. "Rational Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule." Journal of Political Economy 83: 241-54.

Schultze, Charles L. 1974. "Sorting Out the Social Grant Programs: An Economist's Criteria." American Economic Review 64: 181-89.

Tiebout, Charles M. 1956. "A Pure Theory of Local Expenditures." Journal of Political Economy 64: 416-24.

Tresch, Richard W. 1981. Public Finance: A Normative Theory. Plano, Texas: Business Publications.

U.S. Department of Commerce. Survey of Current Business , various issues.

U.S. Office of Management and Budget. Budget of the United States Government , various issues.

Walsh, Cliff. 1983. "Reforming Federal Financial Relations: Some Radical (Or Are They Conservative?) Proposals." Paper presented at the Federal Finances Symposium, Hobart, Australia.

Wilde, James A. 1971. "Grants-in-Aid: The Analytics of Design and Response." National Tax Journal 24: 143-56.

Yinger, John. Forthcoming. "On Fiscal Disparities across Cities." Journal of Urban Economics.

Chapter Three—
Commentary

Henry J. Aaron

Gramlich's and Inman's articles both discuss problems of federalism, but they share only a few elements. Accordingly, I shall comment on them separately.

Gramlich's article is meaty, thought-provoking, and highly professional. No economist I can think of does a better job of integrating economic theory, statistics, and institutional realities to say something useful about practical problems. This encomium is not the prelude to devastating criticisms, but I do have some criticisms.

In the discussion of the various motives for grants, I believe that one motive (to which Inman alludes briefly) is omitted. Higher-level governments often employ lower-level governments as agents, using grants as a means of virtually drafting their administrative services. The responsibility appears to be joint, but the activity is controlled in all essential respects by the higher-level government. Such grants are best explained and appraised, in terms not of spillovers or interregional income redistribution, but of their effects on administrative efficiency. From this standpoint, narrow categories and high matching percentages are entirely rational. Better grounds for criticism would be whether the federal government should be involved in such functions, or whether administration by lower-level governments is truly advantageous.

Economists have almost completely neglected yet another motive for grants. I believe that the political leaders who fashioned the large increase in federal grants during the 1960s and 1970s thought that by offering money to hesitant states and localities on irresistible terms they could

develop interest groups for activities that had previously had none and that by closely supervising how money was spent they could improve administrative processes in backward lower-level governments. In the jargon of our profession, they thought that both the preference functions and the production functions of the lower-level governments were malleable.

Economists have tackled endogenous production functions in models of learning-by-doing, but they have not carried this insight to governmental behavior. And the possibility that utility functions are endogenous to consumption has played little role except in speculations about the role of advertising and in Thomas Schelling's brilliant essays on self-control. I suggest that the history of grants-in-aid in the last quarter of a century is incomplete without both concepts. I suggest also that the endogeneity of state and local production functions and tastes for public services would go a long way toward explaining the fly-paper effect. What I have in mind is the process by which an increase in grants for a particular spending program will alter the tastes of the local politicians toward favoring more spending on that program.

My remaining comments on Gramlich's paper concern specific points. He states that there is little evidence from the distribution of expenditures among levels of government to show that there are gross misassignments of functions among levels of government. That statement is logically correct, but so is the converse—there is little evidence that functions, other than national defense, energy, and a few others, are correctly assigned. How bad are educational spillovers? How many people reside in the school district in which they received their precollege education? Where counties or states finance education, the same question applies with two words changed. Does the particular split of responsibilities for health and hospitals shown in Table 2.2 make economic sense? I do not even know how to go about answering that question.

Gramlich labels personal and corporate income taxes as the most important ability-to-pay taxes; property taxes are listed as benefit taxes; sales taxes are left unclassified, but they must be either benefit taxes or unimportant ability-to-pay taxes. Behind any such taxonomy lie implicit judgments about the value of benefits from public services that accrue to different households. Many years ago Martin McGuire and I showed that you could get virtually any result you wanted regarding the distribution of benefits from public services by selecting among plausible assumptions regarding the income elasticity of utility. For some assumed values, federal taxes were well-matched benefit taxes. For this reason I cannot assign

much meaning to Gramlich's suggestion that ability-to-pay taxes should be levied at the federal level.

Gramlich suggests another principle for assigning taxes, that the federal government should take over easily exported taxes. I am not sure which taxes this rule would exclude, since even taxes borne entirely by the residents of one state may produce migration effects. Given such effects, state income taxes, for example, would have to be stricken from the list of revenue sources appropriate for state administration. The point is that all taxes place states at some disadvantage in competing for labor and capital. The degree of disadvantage depends not only on the type of tax but also on its level and structure. A poorly designed sales tax at a high rate may be more of a deterrent to business than a well-designed and well-administered corporation tax.

Gramlich's jeremiad against federal income tax deductibility of state and local taxes of all kinds is eloquent and justified. He bases his criticism on the damage such deductibility does to rational decisionmaking at the state and local levels. He might also have mentioned that deductibility, along with countless other base-narrowing provisions, raises marginal rates and increases deadweight losses to raise a given amount of revenue.

I am even more pessimistic than Gramlich on the subject of the exclusion of state and local bond interest from the federal tax base. While it is true that the interest differential between taxable and nontaxable bonds of given riskiness and maturity is only about 25 percent, the differential on short-term securities is considerably larger, in substantial measure because banks continue to hold short-term securities, as a result of their financial needs and the quaint and curious way we tax financial institutions. Furthermore, while the spread between taxable and nontaxable bonds has decreased, the volume of tax-exempt issues has mushroomed as they have become available for industrial and housing bonds. The scope for tax exemption to distort investment is thus vastly expanded. It seems safe to say that the total excess burden from the exemption of interest on state and local securities is at an all-time high.

Finally, Gramlich offers a set of recommendations regarding welfare. I found the appendix summarizing the analysis and results of another paper fascinating. The use of a bargaining model to handle migration effects of welfare payments is very clever. The most striking aspect of this analysis is that when he comes to policy recommendations, Gramlich drops the analytical framework and says that he would impose poverty-income-threshold transfer levels on both high- and low-benefit states. I cannot quarrel with the recommendation, although I do think that other

aspects of the welfare program, such as work requirements, job creation, and other administrative rules are so important that benefit levels do not adequately characterize any welfare program. My point is that one can reach Gramlich's recommendation only if in Equation 5, (1 – w ), the weight attached to utility of residents in the low-benefit state in the social welfare function, is zero. In short, in making policy recommendations he drops his economic framework and turns—as I think he should—to value judgments.

That reminds me of two waggish laws of economics. The first states that nothing of importance in economics was ever settled by empirical analysis. The second states that nothing of importance was ever settled by theory either. I hate to think of the corollary.

I now turn the the Inman paper. He gets the big points right. The New Federalism has not been adopted and is not likely to be unless the November 1984 election is to the November 1980 election as the 1980 election was to the 1976 election, a truly awe-inspiring prospect. If it were to be adopted, the New Federalism would result in large but unevenly distributed reductions in spending by state and local governments. I do not trust the precision of his estimates (neither, I think, does Inman), because they derive from data generated by gradual year-to-year changes or from cross-sectional estimates where tastes may not have been adequately controlled. But the qualitative characterizations of those effects on outlays, i.e., big and negative, seem indisputable. The history is incomplete, however, without taking note of the effects of President Reagan's great budgetary leap backward. Even without the New Federalism, things have changed a lot. A fair amount of grant consolidation and budget reduction has occurred.

Inman's results suggest a ranking of motives on the part of the proponents of the New Federalism rather different from those he derives. He claims that the primary motives were to scale back the size of government and of federal intrusiveness, and that the achievement of these objectives was foiled by the political facts of life. States really want matching grants, not unrestricted funds, at least for welfare and other programs that help the poor. That way they can meet the requests of the teeming masses yearning to receive aid at low net marginal cost to the resident middle class. Members of Congress too were reluctant to surrender categorical grants, because each representative had one or more grants that served him or her and only a few others; logrolling did the rest.

Relegated to bit parts in this drama are two other factors to which participants in the negotiations between federal and state officials over

the New Federalism would, I think, assign primary importance. The first is that, quite apart from any incentive effects, the New Federalism was a raw deal for the states. They lost money right away and were promised that after a few years they would lose a lot more. If Reaganauts had been primarily interested in reducing the intrusiveness of government, they could have upped the ante sufficiently to bring state officials on board. Even if federal grants were greatly increased, it appears from both Gramlich's and Inman's calculations that total government spending could have been cut sharply. The necessary offer was not made, because the primary objective was cutting domestic federal spending. That the goal of reducing federal intrusiveness may have been a rhetorical sham is suggested by the Reagan administration's proposals for a massive increase in federal intrusiveness through regulation of state and private behavior with respect to abortion and school prayer and by overriding clinical and parental freedom in the treatment of infants. The fact is that the New Federalism was a federal budget-cutting initiative, first and foremost.

The second factor is best summarized by an ironic comment that Michael Stern, then staff director of the Senate Finance Committee, made to me in 1977. "When it comes to welfare," Stern said, "all members of Congress are statesmen. The welfare lobbies are so weak, most members can and do vote their consciences." I realize that conviction as a motivation for voting has played little role in formal models, but I would suggest that welfare is different in degree, not kind, from many other programs included in the New Federalism proposals. Although they may not understand the Inman—Craig or Gramlich models, many members of Congress do understand that giving less money with fewer incentives will produce fewer effects than does more money with more incentives. Again, I think that the Reagan administration could have swayed enough members of Congress to accept the proposed restructuring of federalism if more money had been put into the deal. But the desire to cut federal spending was overmastering.

Commentary

George F. Break

Gramlich and Inman have both written excellent articles, providing much stimulus to thought and leaving little to criticize. I do have one major disagreement with Gramlich concerning the role played by federal deductibility for state and local taxes. He contends that "it is hard to imagine a consciously designed public policy measure having worse effects on both efficiency and equity counts than the federal income tax deduction for state and local taxes." In my view that feature of the federal individual income tax is both sound in principle and equity-enhancing in practice. Its effects on economic efficiency may well be generally unfavorable, but their quantitative significance has yet to be firmly established. Policy evaluation of federal deductibility is, I believe, much more complicated than Gramlich maintains, involving important, but uncertain, tradeoffs between equity gains and efficiency losses.

My argument begins with the following basic proposition concerning the role of taxes in a federal system of government: when two independent levels of government both make use of ability-to-pay taxes, the proper income tax base for each level is personal income minus the ability-based taxes of the other. This is simply a specific application of the general principle that ability to pay taxes is a function of each person's discretionary income—namely, that part of total income remaining after provision has been made for minimum survival consumption expenditures, unavoidable casualty losses, and so forth. Ability-based taxes, by definition, are compulsory transfers to the government in return for

which no tangible benefits are received. Payment of such taxes to one level of government reduces ability to pay at other levels.

This is not the place to debate the relative philosophical merits of this basic proposition. Instead, let us accept it as a general principle of federal finance and discuss some of its important policy implications. Consider first a dual system of proportional income taxation, the federal rate being f and the state rate s . With no deductibility, the total nominal tax rate is t = f + s , and each component understates the tax rate each level of government is in fact imposing. The effective federal rate is f * = f /(l - s ), and the effective state rate is s * = s /(l - f ). With unilateral deductibility by either level, the total nominal tax rate is

For federal deductibility, which permits deduction of the state income tax from the federal base, it is convenient to concentrate on the second two terms. Taken together, s (l - f ) measures the net cost to state taxpayers arising from the state income tax. Taken separately, s indicates the gross rate of tax received by the state, part of which is provided by its own taxpayers [s (l - f )], and part by the federal government (sf ). In effect, the interaction term, sf , represents an implicit, unrestricted federal grant channeled to the state government through its resident taxpayers.

For state deductibility, on the other hand, it is convenient to concentrate on the first two terms of Equation 1. The term f (l - s ) now measures the net tax rate imposed on federal taxpayers, and fs is an implicit unrestricted state grant to the federal government. Double deductibility, which need not be discussed here, simply combines these two separate patterns.

Suppose, next, that each level of government wishes to operate Musgrave's Distribution Branch by levying progressive ability-based income taxes. In this joint venture the federal government is the senior partner, and states participate because there is no national consensus about the optimal degree of income redistribution to be achieved through the tax system. One would therefore expect to find considerable diversity in state Distribution Branch policies, some wishing to augment and others to offset the redistributive effects of the federal tax system. In such a fiscal environment the tax parameter of major interest to state policymakers is what I have called the incremental burden ratio, that is, the ratio of incremental state tax burdens imposed at each level of personal income to the taxpayers' after-federal-tax incomes.^[1] The trouble with that ratio is its low visibility, not only to the general public but also to many state

policymakers. In public discussions of tax policy it is the nominal state rate, or rates, that are the focus of attention. That being the case, the important question is how good, or bad, an indicator of the redistributive effects of state taxes those nominal rates are.

If there is no deductibility of one governmental level's income tax from the income tax base of the other level, neither tax is being imposed on a correctly defined base, and the nominal state tax rate is a seriously flawed indicator of the true redistributive effects of state taxes. This may be readily seen for the simple case of a proportional state income tax, levied at rate s . The incremental burden ratio at any income level i is then

where f_i is the average federal tax rate at income level i

Since f_i is higher at higher income levels, the IBR structure will be progressive. Combined with the average federal income tax rates shown in Table 3.1, a flat-rate state tax of 10 percent, for example, would produce incremental burden ratios that increased from 10 1/2 percent at the $10,000 AGI level to 14 percent at the $75,000 level. With no deductibility, in short, a progressive state income tax would be more progressive than its nominal rate structure indicates, a nominally proportional tax would be progressive, and even a nominally regressive tax might be progressive.

Federal deductibility alone, which is the policy at issue here, would improve both the equity of the federal tax and the usefulness of the state nominal tax rate structure as an indicator of state redistributive policies. In this case, the incremental burden ratio is

where f'_i is the marginal federal tax rate at income level i

It will be recognized that this is simply the nominal state tax rate multiplied by one of the standard measures of income tax progression, namely, the elasticity of after-tax income with respect to before-tax income, commonly called residual income progression .^[2]

For federal deductibility the relation between the nominal state tax rate structure and its incremental burden ratio structure may be summarized as follows:

1. s_i > IBR_i , since (l - f'_i )/(l - f _i ) < 1 for a progressive federal income tax.

2. Over any income range for which federal residual income tax progression is constant, the nominal state tax rate structure will be an accurate indicator of the incremental burden ratio structure. That is, the incremental burden ratio structure will be proportional when the nominal rate structure is proportional, progressive when it is progressive, and regressive when it is regressive.

3. Over an income range where residual income progression is falling—(l - f'_i )/(l - f_i ) is rising—nominal state rates understate the progressivity of state tax policies, and vice versa.

The incremental burden ratio structure for a proportional state income tax of 10 percent, for example, may be found in the last column of Table 3.1 by moving the decimal point one place to the left.

On equity grounds alone, then, federal deductibility for state and local ability-based taxes has much to recommend it. Its efficiency rating is both more mixed and less favorable. Three main sets of effects need to be considered.

1. By lowering the net cost to federal-income-tax itemizers of state and local ability-based taxes, federal deductibility may stimulate state-local spending beyond optimal levels. Given the loose linkage between ability-based taxation and spending-level decisions and given the many ambiguities surrounding the concept of optimal government spending, this is a very difficult set of effects to evaluate. In general, it appears that policymakers face a troublesome tradeoff between federal deductibility, which may distort state-local Allocation Branch operations, and no deductibility, which, for the reasons

already given, would be likely to distort state-local Distribution Branch operations.

2. Federal deductibility biases the choice of alternative state-local financing instruments away from those that are not deductible from the federal individual-income-tax base. How strong these effects are is not clear. The biases in question could be either eliminated by repealing federal deductibility or moderated by establishing a set of matching, unrestricted federal grants based on state-local use of nondeductible taxes and user charges.

3. Federal deductibility may moderate the distorting effects that high state-local ability-based taxes have on decisions about business and household location. For federal tax itemizers, high-tax states are less unattractive than they otherwise would be, and businesses locating there can, as a consequence, attract such people with offers of lower salaries than might otherwise be needed.

Federal deductibility of state-local ability-based taxes, like most important tax issues, presents policymakers with some very hard choices. Restricting deductibility to fewer taxes (the retail sales tax has in recent years frequently been offered as a candidate for removal from deductibility status) would run the risk of seriously distorting state and local government choices among alternative ability-based taxes. Eliminating deductibility altogether would risk distorting both the operation of state-local redistributive tax policies and the location among states of households and businesses. Keeping federal deductibility intact, because it is regarded as an important structural feature of the federal income tax in a federal system of government, would risk continuation of the other distortions noted above. If deductibility is kept, its significance and equity could be improved by removing it from the official list of federal tax expenditures, thereby recognizing its basic structural function, and by moving it "above the line" into the category of "adjustments to income," thereby making it available to all federal income taxpayers.

Commentary

Julius Margolis

The analysis of fiscal federalism within the framework of comparative statistics has great merit. However, it tells less than the full story. Five mjaor historical trends have affected the multi-level government structure of the United States: the growth of government; technological change; the decline of political parties; the revolution in the household economy, including the women's revolution; and the growth in democratization. These five trends have combined to effect the relative growth of the national government with the expansion of the locals as its agents.

The Phenomenal Growth of Government

All growing governments, whether they are unitary governments, like that of France, or highly centralized, like the Soviet Union's, are confronted by the huge bureaucratic diseconomies of trying to implement policy at a central point. France, the Soviet Union, the United States and, I would guess, almost all other nations, have set up extensive decentralized administrative structures. In the United States the degree of administrative decentralization is cloaked by the federal structure. Instead of regional divisions of federal agencies, the peripheral agencies have been the state and local governments, which are constitutionally independent of the central authorities. We find a modest degree of administrative decentralization of federal agencies and a number of regional organizations which may be created by the states but whose revenues and behav-

ioral rules derive wholly from the federal government. More important is the way the "normal" state and local governments alter their structures and behavior so that they conform to the specifications of the feds. And of course many special districts and authorities were established with taxing powers, but their true role is to contract with the federal government, under terms specified by federal legislation, to perform services demanded by the federal government.

From a fiscal perspective, the federal government is primarily a financial intermediary, levying taxes and then transferring funds to local governments to support services performed by locally elected and controlled governments that have their own taxing and police powers. From a public service perspective, the state and local governments, when receiving federal grants or contracts, appear to be federal agents implementing federal objectives. A principal-agent model would select the federal agency rather than the local electorate as the principal in many of these operations.

When we recognize that a very large part of fiscal federalism is simply a way to adjust administrative decentralization to an American governmental structure, it becomes clear why the matching-grant formula bears no relationship to externalities. Federal loss of interest in a specific categorical program is likely to result in a budget cut rather than a search for a more efficient price.

Increasing government size fostered expertise in formulating and implementing public programs. The federal bureaucracy took the lead in sophistication of techniques and in hiring technically competent personnel. In time, the regional, state, and local bureaucracies hired their own professionals. In formal terms, programs were initiated at the federal and implemented at the local level, but in less formal terms we find a coalition of bureaucrats at all levels formed to administer these programs. Two aspects of this process are of special interest to us: increasing independence from the political process, and increasing reliance on the federal government as financier.

The coalition of bureaucracies became allied with a body of suppliers, much like the military-industrial complex, e.g., the health bureaucracies have allies in the hospitals, nursing homes, medical societies, drug companys, and instruments manufacturers, among others. And just as the military-industrial complex seems to have achieved a charmed independence from the normal political checks and balances, so has the coalition of expanding bureaus. The technical bureaucracies have come to dominate the legislative branch; at the local level they were even more influen-

tial, since they brought their own funds. The organizational power of the bureaucracies was given a powerful political boost by support from the suppliers.

The extent of these complexes of self-interested groups, dedicated professionals, and even some beneficiaries of the services is dependent on the costs of the programs, not the benefits. Herein lies an interesting observation about federalism. The set of input providers are professionalized and organized; the beneficiaries and citizenry are much less so. Of course, the political power of one coalition is not only measured in terms of its influence against the general citizenry and the specific program beneficiaries but is subject to the countervailing power of all the other coalitions competing for the limited resources extracted from the citizenry. Despite the competition among bureaus, the power in the public sector is heavily concentrated in the supply side. Although the elected leaders must eventually act to endorse or check the ambitions of the competitive bureaucracies, this model of federalism is a far cry from response to the median voters.

A second important feature of most public programs is the difference in distribution of cost versus benefits. Even for a public good, such as defense, which is spread throughout the nation, the production of weapons and training of personnel are restricted to a few production centers. Many public services such as health, education, or transportation have a very wide distribution of benefits but the production of the capital equipment is spatially concentrated. If we devised a political benefits-costs ratio, where benefits include gains to input suppliers as well as willingness to pay for outputs, the ratio for a given program would rise as we moved up the hierarchy of governments from local to national. This does not mean that local bureaus want to surrender their autonomy in exchange for a higher budget. On the contrary, there is much evidence that bureaus will resist growth if it means a dilution of independence. However, if the programs have grown due to federal initiative, coalitions already formed will resist returning funding to local levels. However, if this does happen the coalition may be able to keep a program going in the absence of direct federal support—which suggests that programs are overfunded at the federal level because of the political benefits-costs ratio and that a return of allocational decisions to the local or even the state level might bring the public decision makers closer to a social benefits-costs ratio. Another way of expressing it is that it might be of social advantage to "externalize" the costs of a public program: if the inputs of a program could be located outside the jurisdiction, then the only gain offsetting the tax costs of a

program would be the value of the outputs, not the employment gains or the profits on the inputs. It seems paradoxical that even though private market failures may arise from externalities, it might be necessary to create another set of externalities to overcome public failures.

The conditions of growth in the public sector led to an excessive response to the changing trends discussed below. The coalition of suppliers greatly inflated the benefits of a program, and decentralized administration led to excessive growth at the local level. A counter-revolution should not be unexpected.

Technological Change

The rapid technological changes of the past century have had a highly differential effect on local governments, since their boundaries remained relatively fixed while the pattern of land use changed drastically. Some locals became industrial enclaves, others became bedroom suburbs, and the most productive areas of the central cities turned into obsolete and high-cost producers of services for residential or business use. Thriving urban centers with great expectations were abandoned; others were depressed. This is a familiar story, but it is not clear that it had to result in the current huge fiscal disparities.

In the private market, capital equipment, labor skills, even entire industries have become obsolete. When a plant becomes obsolete it remains in productive use in the firm so long as its marginal operating costs are less than the price of the product. The capital declines in value in response to the loss of quasi-rents. Losses are taken by the owners of the capital. In time the obsolescent capital will be replaced by technologically efficient equipment, with the full faith and credit of the firm behind the credit used to finance the new equipment. The process in the public sector is very different. The city (the public plant) is expected to levy taxes sufficient to pay competitive public services (charge the same tax prices) even though the inefficient public facilities cannot attract or hold their old residential or business users. Attempting to charge the same prices (keep the high taxes) leads to abandonment and further decay. It is not surprising that, as in the case of the private firm, a higher level of organization imposes an umbrella of support over the "victims" of technological change. The (inefficient) city's costs (local taxes) are reduced through federal grants—a process similar to a firm's expecting lower quasi-rents from the older plants.

An alternative approach to shifting the problem of support of renewal

of capital and labor to the central government might have been to expand the boundaries of the city containing the obsolescent resources so that the city will contain the new facilities and labor skills. However, the newer suburbs strongly resist such an expansion. Many of them might have been forced to accept annexation since they did not have the resources to support facilities, but federal support programs for new infrastructure construction obviated the need for this. One political response to the historical trend toward democratization represented by the federal support of the suburbs was a huge investment by the federal government in the resources of the central city. Boundary changes and the reassignment of responsibilities within the local units of a "regional government" might have reduced the need for federal fiscal intervention, but the drive to experiment in regional government was frustrated by problems of racial, ethnic, class, and economic differences in the urban population.

The Decline of Political Parties

One major instrument of political decisionmaking is the political party. If a party seeks to rule, then it must be part of a majority coalition, able to influence the agenda of the governing bodies of the central cities and the suburbs. It is the political party that is peculiarly adept at conflict resolution, at compromising among disparate interests. I would argue that we do not have real political parties in the United States at the national level; at the local level, the shadow of a party exists only in the older sections of the country. The local elections are contested by nonpartisan candidates; typically, they run in nonpresidential election years to avoid the taint of party. The ballots are structured to discourage any form of party voting. Therefore, the political capability for boundary changes and functional reassignment of tasks and fiscal resources is very limited.

The New Household Economy

The growth of government is associated with the assumption of power—usually, by the central authority. Analysts have viewed this growth with alarm, especially when it is paired with the growth of the large and now gigantic firms. Opposing this trend has been a surprisingly large burst in economic activity in the household sector—that is, the most decentralized of all entities, the households, have been growing at a more rapid rate than the highly centralized ones. In the quarter-decade after

1950, business reproducible wealth increased by 130 percent and government reproducible wealth increased by 132 percent—while the lowly households' reproducible wealth increased by 220 percent.

A second major resource shift has been the huge decline in the number of hours of paid work over a lifetime and the dramatic increase in labor-force participation by women. The decline in paid work was a combination of a higher average age for leaving school, earlier retirement, and home production using leisure time and household capital instead of market employment. Even more sudden has been the drop in the size of the family, the reduction of the household to the nuclear family, and the woman's role shift from homemaker to fulltime member of the labor force.

Government programs have been an integral part of these changes. The first decades of life are spent in public educational institutions; the last decades of life are spent in publicly supported retirement status, with huge transfer programs to support the aged (and the younger disabled). In the postwar decades, public infrastructure was built to support private investment in suburban housing and cars. In recent decades household investment has shifted away from the house, and less complementary government investment is required. More important, especially for the debate about the new federalism, has been the transformation of women to members of the work force. Now the family pays taxes on this increased income, and the government manages the transfers. The transfer sector of the fiscal government has grown at twice the rate of collective consumption.

Democratization and Egalitarianism

The turbulent decade of the sixties brought into the political system neglected minorities such as the blacks, and broadened of the definition of civil rights. Equality before the law was extended to equality of public services. This revolution broke at the same time as the household revolution that surrendered to the government the role of transfers among generations, and the technological revolution that encouraged firms to locate in the suburbs near a highly skilled work force. The latter weakened the central cities and their capacity to support the revolution. The metropolitan areas had the resources to support the fiscal effects of these changes, but not the organizational structure nor political will. Local governments found it difficult to respond because of tax competition,

boundaries which made it simple to avoid taxes, and a political structure biased toward property interests. We all know the history of the federal government's repeated attempts to resolve the fiscal problems of the cities.

Summary

Our society has been subjected to a set of revolutions that have had a major effect on government at large and on the structure of local governments and the federal system in particular. The burden of change fell on the federal government, which relied heavily on local governments to administer programs designed and financed at the federal level. Each president from Nixon on has sought to reduce the federal role. Revenue-sharing was introduced, but the underlying structure of the relationship was not changed until the Reagan administration. If this administration succeeds in changing the balance, what will happen?

Our explanation implies that the programs will wither. The public has accepted the revolutionary changes and is willing to adopt complementary government programs. However, to continue these programs out of local resources would involve a degree of reorganization of local government that is highly unlikely, for there is no serious movement for regional governments. The federal government may well rescue the programs after the locals have failed to be responsive. There is another option: the states might continue the programs. The states seem to be the favored agency of the Reagan administration. They have been neglected by economic analysts (possibly it has been too difficult to go beyond a two-sector model) but they may be the right compromise between the fractionated locals and the overzealous federal government. The states have a stronger fiscal base, and their bureaucracies have been improved. Possibly they can prevent the counterrevolution in government from going to excessive lengths.

PART TWO—
DISTRIBUTIONAL PROGRAMS: EDUCATION AND ANTIPOVERTY

Budgetary programs that have substantial distributional implications are the focus of Part Two. Sheldon Danziger and Daniel Feaster review the recent history of income transfer programs in the federal budget, showing the nature and extent of the Reagan administration cuts. Richard Murnane evaluates public spending on primary and secondary education, considering such diverse issues as the trend in test scores, how to attract and keep good teachers in the public schools, and the reform of school finance. In the Commentary, Paul Courant raises methodological questions about the Danziger and Feaster evaluation of AFDC program trends and reflects briefly on the difficulties of achieving innovation in the provision of education. Eric Hanushek is critical of Danziger and Feaster's statistical methodology and their failure to account adequately for non-cash transfers; he also argues that we ought to reward good performance whether or not we can measure the characteristics that make for effective teaching. Focusing on the relationship between educational attainment and earning capacity, David Stern claims that recent policy has involved a shift away from reducing inequality in educational attainment toward achieving educational excellence. Finally, Jennifer Wolch cites the case of the service-dependent poverty population, who have become substantially worse off during the past few years. She suggests that to evaluate transfer programs properly it may be necessary to disaggregate the poverty population.

Chapter Four—
Income Transfers and Poverty in the 1980s

Sheldon Danziger and Daniel Feaster

Income Transfers and Antipoverty Policy

The primary intention of the War on Poverty was to promote employment opportunities and higher wages. The poor could then escape poverty in the same manner as the nonpoor, that is, through the private labor market, and not because of government transfer payments. Despite these hopes, income-maintenance expenditures grew rapidly. By the mid 1970s, such spending cost about three times as much in real terms as in the mid 1960s, owing to introduction of new programs and to increases in both the number of beneficiaries and the size of income-maintenance payments in existing programs. However, real transfer growth slowed in the late 1970s and then became negative in some programs as a result of the Reagan administration's budget cuts.

Measuring Poverty and the Impact of Transfers

An analysis of income poverty requires the specification of both a poverty threshold and an income concept. A household is considered poor if its income falls below the poverty threshold. Different poverty thresholds and income concepts convey different information about the nature and magnitude of the poverty problem. While there are a variety

of alternative thresholds and income concepts (Danziger and Gottschalk 1983), we focus here on the official poverty threshold and on two income concepts: pretransfer income and census money (post-transfer) income.

The federal government's official measure of poverty provides a set of income cutoffs adjusted for household size, the age of the head of the household, and the number of children under age 18. (Until 1981, sex of the head and farm/nonfarm residence were other distinctions.) The cutoffs provide an absolute measure of poverty that specifies in dollar terms minimally decent levels of consumption. The official income concept—current money income received during the calendar year—is defined as the sum of money wages and salaries, net income from self-employment, social security income and cash transfers from other government programs, property income (e.g., interest, dividends, net rental income), and other forms of cash income (e.g., private pensions, alimony). Current money income does not include capital gains, imputed rents, government or private benefits-in-kind (e.g., food stamps, Medicare benefits, employer-provided health insurance), nor does it subtract taxes, although all of these affect a household's level of consumption.^[1]

So that they represent the same purchasing power each year, the official poverty thresholds are updated yearly by an amount corresponding to the change in the Consumer Price Index. For 1982, the poverty lines ranged from $4626 for a single aged person to $19,698 for a household of nine or more persons. The average poverty threshold for a family of four was $9862. According to this absolute standard, poverty will be eliminated when the incomes of all households exceed the poverty lines, regardless of what is happening to average household income.

Census income data do not distinguish between income derived from market and private transfer sources (e.g., wages, dividends, alimony) and that derived from government transfers (e.g., social security, public assistance). As such, census figures fail to separate the private economy's antipoverty performance from the performance of government cash-transfer programs. Households that do not receive enough money income from private sources to raise them over the poverty line constitute the pretransfer poor (a more exact title would be pre-government-transfer poor). Pretransfer poverty has received little attention, yet it reveals the magnitude of the problem faced by the public sector after the market economy and private transfer system (e.g., private pensions, interfamily transfers) have distributed their rewards.

The antipoverty effect of transfers is measured in this paper by a comparison of pretransfer and post-transfer poverty. Cash transfers include

social security, railroad retirement, Aid to Families with Dependent Children, Supplemental Security Income, general assistance, unemployment insurance, workers' compensation, government employee pensions, and veterans' pensions and compensation.^[2] Pretransfer income is determined by subtracting government transfers from post-transfer income. This definition assumes that transfers elicit no behavioral responses that would cause income without transfers to deviate from observed pretransfer income. However, transfers do induce labor-supply reductions, so recipients' net incomes are not increased by the full amount of the transfer—true pretransfer income is likely to be higher than measured pretransfer income. Pre/post comparisons, therefore, like the ones made here, are likely to provide upper-bound estimates of antipoverty effects.^[3]

Geographical Variations in Poverty

Table 4.1 shows the incidence of post-transfer poverty for 1967, 1978, and 1982. We use 1967 and 1982 data because they are reported on the earliest and latest available computer tapes from the Census Bureau's annual March Current Population Survey (CPS). We chose 1978 because real cash transfers per household peaked in the late 1970s and because poverty has increased in every year since 1978.^[4] The residence division of residence not identified is not included in the 1967 classification; it was added to preserve confidentiality after the CPS began to identify each household's state of residence.

Poverty as officially measured declined from 14.3 percent of all persons in 1967 to 11.4 percent in 1978, but increased to 15.0 percent in 1982.^[5] During the 1967–1978 period, by far the largest decline was for persons living outside metropolitan areas, who had the highest rates in 1967. Poverty rates were relatively constant inside metropolitan areas during this period. Since 1978, rates have increased sharply in all locations, so that the 1982 rate is below the 1967 rate only for those living outside metropolitan areas. Central city residents now have the highest poverty rate and comprise the largest group among the poor.

As Table 4.2 shows, in the nation and in each region except the South, pretransfer and post-transfer poverty rates in 1982 were higher than they were in 1967. In the 1967–1978 period, poverty rates in the South declined more than those in the other regions; between 1978 and 1982, they rose less quickly. There is now much less variation in poverty rates across regions than there was in 1967. For example, poverty rates in the Northeast, the lowest in both 1967 and 1982, increased from 71 to 87 percent

of the United States average, while those in the South, the highest in both years, decreased from 155 to 121 percent of the United States average.

The last column of Table 4.2 shows the growing antipoverty effect of increased transfers over the 1967–1978 period and their declining effectiveness since 1978: the percentages of all pretransfer poor persons removed from poverty by cash transfers increased from 26.3 to 43.6 percent, and then declined to 37.5 percent. The pattern in each of the regions is the same. Transfers are least effective in the South in each year. Transfer benefits are much lower in the South than in other regions and some programs, notably Aid to Families with Dependent Children for unemployed parents and Medicaid for the medically indigent, are not even

operated by most southern states. However, as with poverty rates, the southern region's antipoverty impact of transfers converges toward the United States average.

The Reagan Policy toward Income Transfers

From 1967 to 1978, real cash transfers per household increased by 67 percent and real GNP per household increased by 9 percent, whereas from 1978 to 1982 real transfers declined by 1 percent and real GNP by 7 percent (Gottschalk and Danziger 1984). How much of the recent increase in poverty is due to the retrenchment in income transfer programs initiated by the Reagan administration?

First let us review Reagan's changes in federal spending, which were designed both to reduce expenditures and to alter their composition. The first two columns of Table 4.3 demonstrate a growth in the ratio of federal spending to GNP from 18 to 23 percent and show the changes in budget shares that took place between 1965 and 1981 (the last pre-Reagan budget). The swing away from defense and toward income security, education and training, and especially health is apparent, and its reversal was the focus of the Reagan administration's reallocation of the budget. Column 3 indicates, by projecting budget shares to 1986, when all proposed changes would have been in place, the planned changes in priorities that Reagan put forward during the winter of 1981. Defense was to be expanded toward percentages prevailing in the early 1970s while education and training were to be rolled back to the lower levels of that same period. The income-security expenditure share was to be cut 10 percent and the expansion in health expenditures was to continue, but at a much more modest pace.

What actually has been enacted by Congress is somewhat different, as can be seen by comparing the third and fourth columns of the table. Column 4 presents February 1984 estimates of the 1986 budget size and shares, assuming that current laws and policies, including those proposed by Reagan and already enacted by Congress, remain unchanged. Obviously, he has not yet gained all his expected defense increases. The 1983 Social Security Amendments provided somewhat larger cuts in income security than was originally anticipated. This outcome was not reflected in the 1981 budget projections. The health share is projected to rise above expectations, because a proposed health-cost-containment package has not yet been fully enacted. Training costs will fall less than expected, because of a new program introduced to combat the long recession. And

the ratio of federal spending to GNP is even higher than it was in 1981, partly because not all the planned spending cuts were enacted, and partly because of the deep and prolonged recession.

On the expenditure side, then, Reagan and Congress have slowed the expansion of the welfare state and stepped up military expenditures. In dollar terms, or even in terms of shares of the budget, as Table 4.3 indicates, expenditure changes other than defense have not been large. However, it was in spending for the means-tested welfare programs that the Reagan administration differed so much from previous administrations and there that the cuts were the largest.

Table 4.4 shows the proposed and enacted percentage changes in outlays for each of the cash transfer programs included in the CPS data used in this paper. Since social security is by far the largest program, its proposed and enacted changes dominate the totals. Cash transfers in the CPS

data are about 6 percent lower than they would have been had pre-Reagan policies remained constant.

President Nixon's Family Assistance Plan (FAP) and President Carter's Program for Better Jobs and Income (PBJI) were both intended to establish a national minimum-income guarantee, to extend benefits to persons who were categorically ineligible under existing programs, and to promote work incentives by keeping marginal benefit reduction rates on earnings well below 100 percent. As such they would both have raised the safety net and filled in some of its gaps, particularly in regional differences in eligibility requirements and benefit levels. But both generated fatal congressional opposition and harsh criticism from policy analysts who pointed out that these reforms and the goal of controlling social spending were incompatible.

Unlike his predecessors, Presidet Reagan succeeded in reforming welfare. The Omnibus Budget Reconciliation Act of 1981 (OBRA) reduced costs and caseloads by raising the tax rate on welfare recipients' earnings and by establishing more restrictive gross-income limits. It did not, however, lower the safety net for those who do not work. The philosophy behind the cuts was to transform certain income transfer programs from a general support system encouraging simultaneous receipt of wages and welfare to a safety net that forced a choice between work and welfare.

The administration argued that the break-even level for welfare benefits was so high in some states because of the work-incentive provisions it eliminated (the $30-and-one-third rule); many who were not "needy" were receiving welfare. Further, work incentives served more to keep families dependent on welfare than to encourage work. Welfare had become an income supplement and previous welfare reforms were, according to a Reagan advisor, attempts "by a largely liberal, intellectual elite . . . to foist on an unsuspecting public . . . a guaranteed income" (Anderson 1984, p. 25). Work effort was best secured by work requirements (proposed but not enacted), not work incentives.

Generally speaking, those who did not work lost only a small portion of their benefits. The biggest losers were the "working poor" and not the "poorest of the poor," who are not in the labor force. For example, the U.S. House of Representatives, Committee on Ways and Means (1984b) reports that in 1980, forty-six states would have provided AFDC benefits to a woman with two children and earnings at 50 percent of the poverty line; by 1984, only twenty-four would have provided benefits. For a woman with wages equal to three-quarters of the poverty line, the number of states paying AFDC dropped from 37 to 6. In addition, the latter

woman had a negative federal tax burden in 1980 (payments from the Earned Income Tax Credit exceeded the sum of the social security and personal income taxes), but a positive burden in 1984. In 1980, this woman's disposable income would have been 108 percent of the poverty line, while by 1984 it would have dropped to 92 percent.

Robert Lampman (1974) has argued that the declaration of the War on Poverty had an immediate and far-reaching effect: it required all existing programs and proposals for policy changes to address the question, "What does it do for the poor?" The Reagan economic program asks instead, "What does it do for the incentives to work and to save?" As a result, we share Lampman's (1983) judgment that the declining rate of growth of federal revenues, the reordering of domestic versus military priorities, and the vast projected budget deficit fundamentally mean that "it is extraordinarily difficult to initiate new social spending measures in the field which the President has set. In that sense, the President's design for calling a halt to the growth of welfare statism seems to have won the day" (p. 381). If this is the case, then the comparisons of 1978 and 1982, which we make below, show too small an effect of the recent changes—they compare "what is" with "what was," rather than with "what would have been had there been no cuts."

Accepting the premise that these changes in the various programs differentially affect demographic groups, we examine changes in the antipoverty impact of transfers for households headed by nonaged persons (less than 65 years of age) by sex of head. We disaggregate by region because of the long-standing regional differences in benefit levels and eligibility requirements in transfer programs, and we use household rates because transfer benefits generally vary with household size and income. The household measure treats all households equally, regardless of the number of persons.

Changes in Transfers and the Trend in Poverty

Table 4.5 shows the pretransfer and post-transfer poverty rates and their percentage changes from 1978 to 1982 and the antipoverty impact of transfers in those two years. While the regional variations noted above are evident in these data, they are much smaller than the male-female differences. In each region, households headed by women are four to five times more likely to be poor than those headed by men. In this period, however, the percentage increase in both pretransfer and post-transfer

poverty rates was larger for men. Lower wage rates and lower labor-force-participation rates for women make women's poverty rates relatively immune to macroeconomic conditions—they are high at all stages of the business cycle.

For both men and women, the most rapid increase in the pretransfer poverty rate was in the North Central region and the least rapid was in the South. This reflects regional differences in the severity of the recent recession. Between 1978 and 1982, the unemployment rate increased from 6.1 to 9.7 percent in the United States as a whole, but from 5.3 to 11.1 percent in the North Central region and from 5.6 to 8.9 percent in the South.

Post-transfer poverty for all the groups shown increased more rapidly than pretransfer poverty (compare column 6 to column 3), which indicates that the total rise in poverty cannot be attributed to the recession. Transfers removed a smaller percentage of the pretransfer poor in 1982 than in 1978 for all groups. The antipoverty impacts shown here for the nonaged are smaller than those shown in Table 4.2 for all persons because by far the largest antipoverty impacts occur among the aged. The impacts are larger for households headed by men than for those headed by women, because men generally receive social insurance transfers, which are based on prior earnings and have higher maximum levels than the welfare transfers generally received by women.

Table 4.6 shows in greater detail some components of the antipoverty impact of transfers. Consider first the percentage of pretransfer poor households receiving transfers (columns 1 and 2) and among those households, the percentage receiving enough to raise them above the poverty line (columns 3 and 4). These two components provide additional insight into the male-female differences in the antipoverty impact of transfers. In each year and each region, poor women are more likely to receive transfers than men, but much less likely to be taken out of poverty. For example, in 1982 three-quarters of households headed by women received transfers, while only 56 percent of those headed by men received them. But only about one-fifth of the former received enough transfers to raise them above the poverty line, while about 40 percent of male-headed households were so raised.

Women are less likely to be taken out of poverty than men both because their mean poverty gaps are higher and because their average transfers received are smaller. In 1982, the mean pretransfer poverty gap for all female-headed households was $6615 and the average transfer amount for those receiving transfers was $4377; for male-headed households the

corresponding figures were $5477 and $5141. In 1978, the gaps were $6489 and $5445 (in 1983 dollars) for women and men and the transfers received were $5124 and $7097, respectively. Thus, the poverty gap for women was about 20 percent higher than that for men in both years, while transfers were about 28 percent lower in 1978 and about 15 percent lower in 1982.

The probability of transfer receipt was relatively constant in all regions and for men and women over the 1978–1982 period, but the probability of escape after receipt declined for all groups. This result follows from the way the budget cuts were structured: poor households were more likely to receive reduced transfers than to be totally removed from the programs. Note that these conditional probabilities of escape after transfer receipt, already rather low, do not reflect the fact that almost half of poor male-headed households and a quarter of poor female-headed units receive no transfers at all. Although the rapid growth of transfers was a prime motivator of the Reagan budget cuts, these significant gaps in coverage and inadequate benefit levels have received little attention.

Columns 5 and 6 of Table 4.6 show that the decreased probability of escape after transfer receipt can be roughly attributed to the declining real value of transfers and not to macroeconomic conditions. The real pretransfer poverty gap increased by less than 1 percent for all poor men and by less than 2 percent for all poor women. The gap for male transfer recipients actually declined by 3 percent; for female recipients, it increased by 2 percent. Some portion of the increased gap is due to the elimination of public employment positions previously funded through the Comprehensive Employment and Training Act (CETA) and might thus be attributed to program changes. The reduced escape probabilities therefore reflect the decline in real value of cash transfers of almost 30 percent for men and about 17 percent for women. This decline is due both to the administration's budget cuts and to the fact that benefits in programs that are not indexed to consumer prices by law (e.g., AFDC, unemployment insurance) have been falling in real terms since the mid 1970s.

The final two columns of Table 4.6 present the results of a simple statistical exercise that computes how much of the observed increase in the number of post-transfer poor households between 1978 and 1982 can be attributed to the observed declines in the probability of transfer receipt and the probability of escape after receipt. We begin by assuming that the program changes did not affect the demographic composition of households and pretransfer poverty.^[6] We then multiply the number of

pretransfer poor households in each group in 1982 by the corresponding 1978 probabilities shown in columns 1 and 3. This yields the number of poor households in 1982 who would have been removed from poverty by transfers if the 1978 transfer probabilities had remained constant. Subtracting from this simulated number the actual number of households removed from poverty in 1982 yields the increase in the number of poor households due to transfer program changes.

In column 7 this number is expressed as a percentage of the total number of post-transfer poor households in 1982; in column 8, as a percentage of the increase in the number of post-transfer poor households between 1978 and 1982. We find that the number of poor households in 1982 was about 6 percent higher both among those headed by men and by women than it would have been if the two probabilities had not changed.

Since the percentage increase in the number of poor households between 1978 and 1982 was so much higher among families headed by men (Table 4.5), the additional 6 percent of poor households accounts for about 15 percent of the increased male-headed poor families and about 25 percent of the female-headed poor families.^[7] This differential is not surprising, since many female household heads either lost eligibility for AFDC or had their benefits significantly reduced by the new program rules. Our results are in general agreement with a recent microsimulation study released by the U.S. House of Representatives, Committee on Ways and Means (1984a), which attributed a larger percentage of the observed poverty increase among female household heads to program changes and a larger percentage among male household heads to the recession.

An Econometric Model of the Antipoverty Impacts of Transfers

The descriptive data presented so far have not controlled for varying personal characteristics of the poor or differences in the degree of their poverty. We have thus ignored the fact that the probability that a poor household will receive a transfer or escape from poverty will depend to a great extent on household characteristics. For example, a nonworking widow with children will be eligible for social security benefits that are not affected by her other sources of income, while a nonworking divorced woman's AFDC benefits will be affected, and a poor household with income just below the poverty line is likely to receive different treatment by transfer programs than a household with little pretransfer income. In

order to examine more closely the antipoverty impacts of transfers, we have estimated a two-equation econometric model of the antipoverty impact of cash transfers.

We do not attempt to model the macroeconomy or the labor market here. Rather, just as in our descriptive analysis, a household's pretransfer income is assumed to be exogenous. The statistical model we use can be termed a probit equation with sample selection. At a given time (in this case 1978 or 1982), a pretransfer poor household has the potential to receive a transfer, and, if one is received, to escape poverty. The following functions describe these potentials:

where y₁ is the potential to receive a transfer and y₂ is the potential to escape poverty.

The specification and estimation of this model are discussed in the Appendix. The first regression included all households with pretransfer incomes below the poverty line; the second, only those households which were pretransfer poor and which receive any cash transfers.

The regression coefficients derived from this model allow us to control for changes in personal characteristics since 1978, changes that could not be incorporated in the decomposition analysis of Table 4.6. Table 4.7, columns 1 and 2, shows several series of predicted probabilities of transfer receipt for pretransfer poor nonaged female household heads for 1978 and 1982. Columns 3 and 4 show the predicted conditional probabilities of escape from poverty given receipt. These probabilities are computed by evaluating the regression coefficients at the 1982 weighted sample means for all the independent variables except the one that is allowed to vary in each panel of the table. Columns 5 and 6 are the unconditional probabilities of escape, the products of the probabilities shown in columns 1 and 3 and columns 2 and 4, respectively. The final column shows the 1978–1982 change in the number of poor households per 100 poor households who escape poverty. It is merely the difference between columns 6 and 5 multiplied by 100. Since we have held personal characteristics at their 1982 means and allowed the coefficients to vary, we now have a "true" measure of the change in the number of poor households due to changes in the antipoverty impact of transfer programs.

For example, the first panel shows that in 1982, 76.4 percent of all white female heads received transfers and 18.4 percent of the recipients

escaped poverty. If the transfer system had not changed since 1978, then 78.3 percent of the 1982 population of female heads would have received them and 23.8 percent would have escaped. The probabilities of receipt for blacks and Hispanics are higher than for whites in each year, but their probabilities of escape are much lower. This is primarily because whites are more likely to receive social insurance transfers, which are on average higher than welfare transfers. For whites and Hispanics, the declines in the escape probabilities are the bigger of the changes in the two components. The last column shows that if the transfer system were as effective in 1982 as it was in 1978, there would be 4.5 percent fewer white, 1.4 percent fewer black, and 9 percent fewer Hispanic households headed by women in poverty. What is most striking is how low the probabilities of escape are for these women in each year.

The second panel shows that transfers have a much greater antipoverty impact for widows than for other female heads; they are more likely to receive transfers and much more likely to escape poverty. This is due to their receipt of social security survivor's benefits, which are on average much higher than other transfers.

The variation by region shows the familiar pattern: the lowest probabilities of receipt and escape in the South, but regional convergence. The variation by distance from the poverty line is as expected: the poorest have a greater probability of receipt but a smaller probability of escape. The percentage decline in the number who escape is not affected by distance from the poverty line.

These results are consistent with those of our statistical decomposition. Since 1978, changes in transfer programs have reduced their antipoverty effectiveness. These declines, however, are relatively small in contrast to the increased antipoverty impacts of the 1967–1978 period (see Table 4.2).

Implications for Antipoverty Policy During the Mid 1980s

Gottschalk and Danziger (1984) have projected that the aggregate poverty rate will remain above the 1978 rate through the late 1980s even if the economy continues to grow as fast as the Reagan administration expects. A return to "full" employment would substantially reduce poverty rates for nonaged men but would have a much smaller impact on the aggregate rate, because this group represents an increasingly smaller portion of the total poverty population. Thus, if poverty is to fall, the

economy must grow and there must be a renewed antipoverty initiative directed at curbing the "feminization of poverty." About half of these households headed by women with children remain poor (Table 4.5), even though about three-quarters of such households receive income transfers (Table 4.6). And both our statistical decomposition and our econometric analysis show that these households have been adversely affected by the recent program changes.

It is beyond the scope of this paper to present a comprehensive strategy to lower poverty among families headed by single women. But we can offer a few suggestions. First, AFDC guarantees, which have fallen on average by more than a third in real terms since 1970, should be indexed to the same price index used to update the poverty threshold, and a national minimum AFDC benefit should be introduced. The indexation of social security benefits and the introduction of a national minimum benefit in the Supplemental Security Income program are the key reasons that poverty for the aged declined throughout the 1967–1982 period. These changes would have a particularly large impact on poverty in the South, where benefits were initially low and where they have eroded the most in recent years.

Second, a targeted employment program that allows recipients to mix work and welfare should be introduced. The Supported Work Demonstration project (Manpower Demonstration Research Corporation 1980) provides an example of the antipoverty possibilities of such a program. Danziger and Jakubson (1982) used Supported Work data to simulate the national effects of implementing such a program. They found that over 80 percent of the AFDC participants would have been poor if they merely had access to current transfer programs, whereas only 35 percent would have been poor if they also had access to the jobs program. Unlike a negative income tax, this type of employment program both increases work effort and reduces poverty.

These reforms obviously increase public spending. In contrast, a third reform is a social child-support program that attempts to minimize the need for additional public funds. Under that program, every adult who cared for a child and did not live with the child's other parent would be eligible for a support payment that would be financed by a percentage-of-earnings tax on the absent parent. If the tax on the absent parent fell below a fixed minimum level, because the parent's earnings were too low, the support payment would be raised to that level by government funds. Even if total government AFDC expenditures were maintained at current levels, the program could reduce poverty because of the additional revenue raised from absent parents (Oellerich and Garfinkel 1983).

What are the prospects for new antipoverty initiatives? While, by electing Reagan, the voters clearly called for a retrenchment of the welfare state, they have shown no enthusiasm for the recent large increases in poverty. And even though the Reagan administration has reduced the scope of transfer programs, it has not abandoned their primary goals—the provision of minimum levels of cash, nutritional, medical, housing, and educational assistance for a substantial portion of the population.

During its first year in office, the administration proposed drastic cutbacks in most social programs. But because so many households are direct beneficiaries of at least one program, widespread voter and congressional opposition developed. This was particularly true for programs with the broadest range of recipiency across the income distribution (e.g., social security, educational financing program for college students). As a result, only a small portion of the cuts were in programs whose benefits were widely distributed. The large cuts were in income-tested welfare programs that provide benefits only for those toward the bottom of the income distribution. But because they serve fewer recipients, and because the benefits of the poorest were least affected, budgetary savings as a percentage of the total budget were small.

These cuts came, however, on top of a deep recession, and those events together raised poverty rates back to the levels of the late 1960s. This increase now limits the administration's ability to obtain enactment of additional cuts in transfer programs.^[8] Indeed, it is now clear to the administration that the technical and political problems that make large distributional changes difficult are synergistic. In its defense of the fiscal year 1984 budget proposals, the U.S. Office of Management and Budget (1983) stated that "entitlement programs develop vast networks of dependency that cannot be precipitously altered without unacceptable social and human costs. As a consequence, their claim on the budget and national economy . . . can be reduced only slowly . . ." (ch. 3, p. 9). Unfortunately, there remains a large gap between a decision to refrain from asking for further transfer cuts and one to propose increased antipoverty expenditures.

Appendix:
Specification of Econometric Model

Let y₁ be the potential to receive a transfer and y₂ be the potential to escape poverty. Then

However, only y*₁ and y *₂ , indicators of y₁ and y ₂ , are observed

and y *₂ is observed only if y *₁ = 1.

There is some ambiguity surrounding the domain of e₂ . In what is termed the "sequential-decision" model by Maddala (1983), e₂ is only defined if y *₁ = 1 (or e₁ > - Xb ). The likelihood function, £ (given that the e_i are normally distributed), is

where F is the standard normal cumulative distribution function.

The estimation of this model can be accomplished in two steps. First, to get estimates of b , estimate a univariate probit on the indicator y *₁ for those who are pretransfer poor. Then to get estimates of d , estimate a univariate probit on the indicator y *₂ only on those pretransfer poor who received a transfer. This procedure allows conditional inference on the probability of escaping poverty, but it does not allow the likelihood of escaping poverty to affect the probability of receiving transfers. If there are unobserved characteristics that affect both probabilities, then the estimates will be biased. Consider, for example, the unobserved variable "knowledge of how the system operates." A household head who scores high on this variable will be both more likely than one who scores low to receive a transfer at all, and more likely to receive a larger transfer and hence to escape poverty.

This correlation can be incorporated by changing the assumption on the domain of e₂ . While the situation is still one of sequential decision, e₂ is defined for the entire population but is only observed when e₁ > - Xb . This is a censoring problem which can easily be handled with the additional assumption that

The estimation can no longer be done in two steps. The likelihood function is now

where F is as before and F is the bivariate normal cumulative distribution function.

Another way to think of the problem is to examine the log likelihood and define three mutually exclusive subsets of the sample. Let A be those for whom y*₁ = 0, that is, those who do not receive a transfer. Let B be those for whom y*₁ = 1 and y*₂ = 0, those who receive a transfer but are not removed from poverty. And let C be those for whom y*₁ = 1 and y*₂ = 1, those who both receive a transfer and escape poverty. The log-likelihood function, L, can then be expressed as

The preceding models analyze only whether transfers remove a poor household from poverty. Since transfers are continuous, the model could be rewritten to focus on the extent to which transfers alleviate poverty. Let T equal the amount of transfers, and rewrite Equations 1 and 2 as

where s is a scale parameter.

The likelihood function would then be

We have estimated weighted regressions for both the univariate (Equation 5) and bivariate (Equation 7) models for nonaged female household heads with children. In the second specification, there was strong evidence against the hypothesis that r = 0. We report those results in Table 4.8. (The other results are available on request. Because of high computational costs we did not estimate these models for male-headed households or any models reflecting the specification of Equations 8 and 11.) The first stage of Equation 7 included all households with pretransfer

incomes below the poverty line. The dependent variable took the value of 1 if the household received any cash transfers, 0 otherwise. In the second stage, we included only those households that were pretransfer poor and received cash transfers. The dependent variable took the value of 1 if the household was removed from poverty by the transfers, 0 otherwise.

We weighted the log-likelihood function by the CPS sample weights so that the estimated equation would accurately reflect the published (weighted) variable means. The weights were adjusted so that the number of observations shown at the end of Table 4.8 is the number of unweighted observations. The procedure is the same as Manski and Lerman's (1977) correction for choice-based sampling. The fact that weighted and unweighted means for the dependent variables differ raises the possibility that the CPS sample stratification is in some way correlated with transfer receipt or low income. Examining this possibility is beyond the scope of this paper.

The independent variables were sets of dummies for the household head's race or Hispanic origin, household size, regional and metropolitan residence, age, education, and marital, disability, and student status. Also included was the ratio of pretransfer income to the poverty line. The probability-of-receipt equation included all the variables listed. Tunali (1983) shows that at least one restriction is necessary for identification of the model. Therefore, we omitted the residence and education dummies from the second equation. Education and residence are expected to affect one's knowledge of and access to transfer programs and thus to influence participation, but not to affect the amount of the transfer received and hence the probability of escape from poverty.

The general pattern of results—though not the relative magnitudes of the coefficients—is similar for each year. For example, ceteris paribus, the probability of receiving transfers increases with family size, is lowest in the South, higher for those with a disability, highest for widows, and lowest for Hispanics. Not surprisingly, those among the poor who are closest to the poverty line are least likely to receive transfers, but if they do receive transfers they are much more likely to escape poverty. The probability of receiving enough transfers to escape poverty is lowest in the South, and highest for widows and whites.

Acknowledgments

This research was supported in part by the Graduate School Research Committee at the University of Wisconsin–Madison. Christine Ross pro-

vided valuable research assistance. The authors benefited from the comments of the editors and the conference discussants and participants.

References

Anderson, Martin. 1984. "The Objectives of the Reagan Administration's Social Welfare Policy." Pp. 15-27 in D. Lee Bawden, ed., The Social Contract Revisited: Aims and Outcomes of President Reagan's Social Welfare Policy . Washington, D.C.: Urban Institute Press.

Danziger, Sheldon, and Peter Gottschalk. 1983. "The Measurement of Poverty: Implications for Antipoverty Policy." American Behavioral Scientist 26: 739-56.

Danziger, Sheldon, and George Jakubson. 1982. "The Distributional Impact of Targeted Public Employment Programs." Pp. 210-26 in Robert Haveman, ed., Public Finance and Public Employment . Detroit: Wayne State University Press.

Gottschalk, Peter, and Sheldon Danziger. 1984. "Macroeconomic Conditions, Income Transfers and the Trend in Poverty." Pp. 185-215 in D. Lee Bawden, ed., The Social Contract Revisited: Aims and Outcomes of President Reagan's Social Welfare Policy . Washington, D.C.: Urban Institute Press.

Lampman, Robert. 1974. "What Does It Do for the Poor? A New Test for National Policy." Public Interest 34:66-82.

———. 1983. "Review Article: Economic and Social Policies under the Reagan Administration." Journal of Social Policy 12: 378-86.

Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics. Cambridge: University of Cambridge Press.

Manpower Demonstration Research Corporation. 1980. Summary and Findings of the Supported Work Demonstration. Cambridge, Mass.: Ballinger.

Manski, Charles, and Steven Lerman. 1977. "The Estimation of Choice Probabilities from Choice-Based Samples." Econometrica 45: 1977-88.

Oellerich, Donald, and Irwin Garfinkel. 1983. "Distributional Impacts of Existing and Alternative Child Support Systems." Policy Studies Journal 12: 119-30.

Palmer, John, and Isabel Sawhill, eds. 1984. The Reagan Record. Cambridge, Mass.: Ballinger.

Plotnick, Robert. 1984. "The Redistributive Impact of Cash Transfers." Public Finance Quarterly 12: 27-50.

Plotnick, Robert, and Timothy Smeeding. 1979. "Poverty and Income Transfers: Past Trends and Future Prospects." Public Policy 27: 255-72.

Tunali, Insan. 1983. "A Common Structure for Models of Double Selection." Social Systems Research Institute Paper 8304. Madison: University of Wisconsin.

U.S. Bureau of the Census. 1984. Estimates of Poverty including the Value of Noncash Benefits, 1979 to 1982. Washington, D.C.: U.S. Government Printing Office.

U.S. House of Representatives, Committee on Ways and Means. 1984a. Effects of the Omnibus Budget Reconciliation Act of 1981 (OBRA) Welfare Changes and the Recession on Poverty. WMCP 98-33, 25 July 1984. Washington, D.C.: U.S. Government Printing Office.

———. 1984b. Families in Poverty: Changes in the "Safety Net." 98th Congress, 2nd Session, Committee Print 98-37. Washington, D.C.: U.S. Government Printing Office.

———. 1984c. "Tax Policy under Reagan Increases Number of Poor." News release, 31 July 1984.

U.S. Office of Management and Budget. 1983. Budget of the United States Government, Fiscal Year 1984. Washington, D.C.: U.S. Government Printing Office.

Chapter Five—
An Economist's Look at Federal and State Education Policies

Richard J. Murnane

Schooling in the United States is an extraordinarily decentralized activity. Over 15,000 school districts make decisions about what salaries and benefits to offer teachers and how to allocate students and teachers to schools and classrooms. Over two million teachers decide where to apply for teaching positions, whether to request transfers, and whether to leave the profession. Over 40 million students decide how to respond to the varied academic demands placed on them. All in all, there are a tremendously large number of actors in the schooling sector who can and do respond to the incentives they face and whose responses influence the amount of learning that takes place in our schools. A central theme of this chapter is that in analyzing federal and state education policies, it is critical to examine how policies alter these incentives and to review carefully what we know about these actors' responses to changes in incentives.

One consequence of the decentralization of United States schooling is enormous variation in the quality of education provided in different school districts and, for that matter, among schools within a district. A second theme of this chapter is that evaluations of public policies should focus on how they affect the entire distribution of educational achievement, not only on how the policies affect "average achievement." In particular, it is important to consider how policies affect the education of children who in the past left the schools not only with the preexisting characteristics of minority group membership or low family income, but also without cognitive skills.

The strategy I employ in developing a framework for analyzing govern-

ment education policies has three steps: first, I assess trends in educational outcomes; second, I summarize what is known about the causes of these trends; and third, I attempt to map the causes to policies. I encountered a number of problems in carrying out this strategy. First, the only systematic outcome data available are students' scores on tests of cognitive skills. These test scores do provide significant information about the skills of students in our schools. However, the linkages between these scores and life outcomes such as income are not extremely tight in the United States (cf. Jencks 1972).

Second, the arrows connecting trends to causes in many cases are very faint, if not invisible. One reason is that the design of the National Assessment of Educational Progress (NAEP), the testing program that provides the best test-score data, precludes identification of the school district or even the state in which any individual student lives. As a result, the NAEP data cannot be used to evaluate whether specific policies or educational programs affect children's test scores.

Third, linking causes to policies is very hazardous in an environment in which so many factors influence children's achievement, and the same policy (e.g., federally sponsored compensatory education) results in very different educational programs at different sites.

With these caveats explicit, I proceed.

Testscore Trends and Explanations

SAT Scores

Over the period 1963–1980 the average score of high school seniors taking the College Board's Scholastic Aptitude Test (SAT) of verbal ability fell in an almost unbroken trend from 478 to 424. The average SAT math score fell over this period from 502 to 466 (O'Neill 1984). While the declines up to 1970 are explained in large part by changes in the composition of the test-taking population, these changes do not explain the declines during the 1970s (On Further Examination 1977). These trends, a cause for real concern, have been emphasized in many of the recent reports criticizing the quality of United States education (e.g,, NCEE 1983).

In assessing explanations for the decline, it is important to keep in mind that the scores measure the skills of students who have completed twelve years of education. Consequently, any set of SAT scores tells us more about what was happening during the previous twelve years than it tells us about the factors affecting student skills in the year the tests were

taken. Thus, to explain the declines in scores during the 1970s, we must look for changes that took place primarily during the 1960s.

One category of explanation for the SAT score declines focuses on changes in the world outside the schools, including a sense of disillusion arising from the Vietnam War, an increase in TV watching, and a higher rate of marital dissolutions (On Further Examination 1977). A second category includes changes directly related to formal schooling. The best-documented schooling change contributing to the score decline is a reduction in the number of academic courses students take. This decline was the result, in part, of an attempt during the 1960s and early 1970s to tailor high school education to the needs and preferences of individual students. Subsequent research (Bryk et al. 1984; Coleman et al. 1982; Harnischfeger and Wiley 1984) supports the link between the number of academic courses students take and their scores on standardized tests.

In the four years since the SAT scores bottomed out in 1980, the average SAT scores have risen slightly (from 424 to 426 on the verbal section and from 466 to 471 on the math section: Ranbom 1984). Also, during the period 1976–1983, the gap between the average score of black students and that of white students on both the SAT verbal and math tests declined slightly. In 1976, the ratio of the average score of black students to that of white students on the SAT verbal tests was 0.736; in 1983, it was 0.765. The comparable numbers for the SAT math test are 0.718 and 0.762 (Wainer 1984). Why have the SAT scores started to rise again—albeit slowly? Why has the black-white gap closed somewhat? To answer these questions, we need to learn about the early schooling experiences and the achievement of the students who took the SAT tests in the early 1980s.

National Assessment of Educational Progress

The National Assessment of Educational Progress (NAEP) is a federally sponsored testing program under which large samples of American children, aged 9, 13, and 17, are tested in a range of subject areas. The first tests were administered in 1970, and a new round of tests is given every three to four years, depending on the subject area. These test scores provide the best available evidence on the skills of students of different ages attending United States schools.

Reading

The reading skills of American children were assessed by NAEP in the school years ending in 1971, 1975, and 1980. The trends in the scores

differ by age group. The scores of 17-year-olds show a reduction over the decade in inferential comprehension skills (making inferences from material presented in short passages) and stability in literal comprehension skills (locating information contained in the text) (NAEP 1981b). Since ability to make inferences is tested in several parts of the SAT, the decline in the inferential comprehension scores on the NAEP tests lends support to the conclusion that the SAT score decline over the 1970s reflected a real reduction in certain skills in the test-taking population and was not simply an artifact of a changing test-taking population.

The trend in the reading scores of 9-year-olds during the 1970s is very different from the trend for 17-year-olds. In fact, the trend for younger children provides quite encouraging data on recent changes in United States education. Over the decade the test scores of 9-year-olds increased significantly. Moreover, black children experienced the greatest gains, resulting in a closing of the black-white achievement gap. The scores of 13-year-olds increased in one skill area (literal comprehension) and remained stable in the other skill areas. As a group, the 13-year-old black children made greater gains than the white children (NAEP 1981b).

One explanation proposed for the gains in the skill levels of younger children, especially black children, is compensatory education. During the 1970s, Title I of the Elementary and Secondary Education Act of 1965 provided federal funds to local school districts to improve the basic skills of educationally disadvantaged children. It took local school districts several years to learn how to use the new categorical funds and to direct them to targeted children. However, by the mid 1970s, most local school districts complied with the regulations mandating that the funds be used to provide extra services to educationally disadvantaged children (Goettel et al. 1977).

The educational effectiveness of Title I programs has proved difficult to evaluate, since local districts design and implement quite varied programs. However, recent evidence suggests that the program has been effective in improving the skills of the target group—primarily low-achieving children in the primary grades (Anderson 1982; NAEP 1981a).

Mathematics

NAEP assessed students' skill levels in mathematics in the school years ending in 1973, 1978, and 1982. Over this period the average math skills of 9-year-old children were stable. The scores of 13-year-olds fell between the first two assessments and rose between the second and third assessment. As a group, black children gained more than white children. The

scores of 17-year-old children fell between the first two assessments and were stable between the second and third assessments (NAEP 1983).

Why have students' mathematics skills remained stable or declined during the 1970s, while skill levels in reading increased during this period? One explanation is that a shortage of qualified mathematics teachers during the decade hindered student learning (more on teacher shortages below).

Science

Students' skill levels in physical and biological sciences were assessed in the school years ending in 1970, 1973, and 1977. Scores in physical sciences declined over the period between the first and third assessments for all age groups. Seventeen-year-olds also experienced a decline in biology test scores over this period. The biology scores for 9-and 13-year-old children declined between the first two assessments and rose between the second and third assessments (NAEP 1978).

Why the dismal record in science achievement in the 1970s? One proposed explanation is a shortage of qualified science teachers. A second is reduced emphasis on the teaching of science, brought about in part by the increased emphasis placed on the teaching of reading and math—the basics (NAEP 1978, p. 17).

Summary of Trends and the Roles of School Factors

There have been significant improvements over the 1970s in the reading skills of younger children, especially those of black children. We do not yet know whether the gains of younger children achieved over the 1970s will be retained by these children during the 1980s (a cohort model that bodes well for the future) or whether the evidence should be interpreted as simply that our educational system as currently constituted is more successful in raising the skill levels of younger children than those of older children (a life-cycle model). There is limited evidence to support each hypothesis.

Although compensatory education has apparently been effective in raising the basic reading skills of disadvantaged children in the primary grades, it has not led to improved reading scores for 17-year-olds. There are two reasons for this. First, most compensatory education programs are directed at younger children. Second, most children who benefited

from compensatory education in their early years in school were not yet 17 in 1980 and so were not included in the NAEP samples.

Overall, the evidence indicates that compensatory education cannot be understood as analogous to a vaccination—something given to young children that sustains them for an extended period of time without further treatment. A better metaphor appears to be a vitamin pill—a treatment that is effective but must be continued.

The most compelling explanations for the test-score declines in science deal with teacher labor-market conditions, which have changed so markedly over the last fifteen years that even a summary of the changes requires a section of its own.

Status of the Teacher Labor Force

Changes in Labor-Market Conditions During the 1970s

During the 1970s, a variety of conditions combined to make it more difficult for local school districts to attract and retain talented teachers. The first of these conditions was a decline in student enrollments. Between 1970 and 1980 the number of students attending United States public elementary and secondary schools fell from 46.1 million to 40.2 million, a drop of 13 percent (The Condition of Education 1983, p. 14). This enrollment decline drastically altered career opportunities for teachers. Many teachers were "riffed" during the 1970s, to use the term common in teacher circles for layoffs resulting from reductions in force. Opportunities for transfers to more desirable schools, a traditional method of gaining improved working conditions, disappeared (Murnane 1981). Involuntary transfers necessitated by school closings increased, often causing disruption of school programs.

For many teachers and for college students training to become teachers, the layoffs, transfers, and lack of job opportunities were sources of great disillusionment and reassessment. Job security was one attribute the teaching profession had in the past always offered, and it had compensated to some extent for low salaries. Now even this had disappeared.

A second, related factor affecting teacher labor markets during the 1970s was a decline in the real wages of teachers. Between 1971 and 1981, the average salary of public school teachers rose from $9269 to $17,602. However, real salaries (in 1981 dollars) fell from $20,212 to $17,602 during this period, a reduction of 13 percent. In part, the decline in real wages was the result of teachers' unions trading salary increases

for continued employment of their members. Their success in this regard is reflected in the decrease in the average student-teacher ratio from 22.5 to 18.8 over the years 1970–1980 (Digest of Education Statistics 1983 , as reported in O'Neill 1984).

The lack of jobs and the decline in real wages provided strong signals to college students making career decisions. One response to these signals was a reduction in the number of students in United States colleges and universities graduating with education degrees, from 177,000 in 1971 to 101,000 in 1982. The figures on the number of graduating students prepared to teach—that is, fulfilling minimum requirements for certification—are even more dramatic, with the number falling from a peak of 317,254 in 1971–1972 to 140,639 in 1980–1981 (Teacher Supply and Demand 1983). Between 1970 and 1982, the percentage of full-time freshmen indicating that they planned to choose elementary or secondary school teaching as a probable occupation fell from 19.3 percent to 4.7 percent (Condition of Education 1983, p. 218).

There is some evidence to suggest that another part of the response was a reduction in the quality of the pool of graduates planning to enter the teaching profession (Schlechty and Vance 1981). In part, this may have been the result of more able students choosing more financially attractive occupations. The increased career opportunities for women during this period may have led many women, who in earlier decades would have taken a low-paying teaching position, to take advantage of new opportunities in higher-paying occupations. The quality of the pool of applicants for teaching positions may also have fallen as schools of education, attempting to retain enrollments during a period of declining demand for their services, lowered admission standards (Weaver 1978).

In one respect the responses of students during the 1970s to the decline in the attractiveness of the teaching profession are not only what one would expect to happen in a market economy in which individuals are free to choose occupations; they are also desirable in the sense that these responses to market signals are the mechanisms through which labor is redistributed from declining industries to growing industries. There are three reasons, however, for concern. The first is the issue of quality, mentioned above. A second is that public school enrollments are expected to rise again, beginning in 1985, as the offspring of the post–World War II baby boom generation enter the nation's schools (Frankel and Gerald 1981, p. 15). This will result in an increased demand for teachers. Given the lag in the supply response, the teacher glut of the 1970s may be replaced by a widespread teacher shortage in the late 1980s. A third

concern deals with variation in demand and supply conditions across subject areas—in particular, with the shortage of qualified teachers of mathematics and science.

The Shortage of Math and Science Teachers

Teaching has never been a financially attractive occupation for college graduates trained in mathematics or science. For example, in 1974, graduates trained in mathematics who entered business or industry earned starting salaries that were on average 36 percent higher than beginning teacher salaries. By comparison, graduates trained in the humanities who entered business or industry in 1974 earned salaries that were 7 percent higher than beginning teacher salaries. Thus for graduates trained in the humanities, salaries offered in teaching in 1974 were almost competitive with alternatives—if a teaching position was available. This was almost equally true in 1981, when the average salary differential between business and industry, on the one hand, and teaching, on the other, was 12 percent for humanities majors. But for graduates trained in mathematics, the attractiveness of teaching relative to industry or business fell drastically during the 1970s. In 1981, the average salary differential was 61 percent (Bacharach et al. 1984, p. 66).

One consequence of this trend is that while there was a surplus of teachers overall during the 1970s, many areas of the country experienced a shortage of math and science teachers. Moreover, this shortage has become increasingly acute. Local school districts deal with the shortage of math and science teachers in a variety of ways, in part due to variation in state licensing regulations. However, a recent survey (Impact of Teacher Shortage 1983) identified the following responses, with the most prevalent responses listed first.

• Arrange for persons to get emergency certification.

• Use teachers certified in other subject areas.

• Increase class size.

• Cancel courses.

These responses "solve" the shortage problem by lowering the quality and quantity of math and science instruction, and thus probably contributed to the decline in the NAEP math and science scores between 1970 and 1980.

Salary trends may also help to explain why NAEP test scores in biology

were more stable over the 1970s than the physical science scores were. For biology majors entering business or industry, average starting salaries were 12 percent higher than starting teacher salaries in 1974, and 31 percent higher in 1981. For graduates trained in other natural sciences, the comparable numbers are 39 percent and 86 percent (Bacharach et al. 1984, p. 66). Thus, school districts may have found it easier to find qualified biology teachers than qualified physical science teachers during the 1970s.

What should be done about the math and science teacher shortage? A central issue in considering policy alternatives is the role of certification requirements. To obtain a teaching certificate (a license to teach) in most states, college students must devote a significant part of their course work during their junior and senior years to education courses. One cost of this requirement is a reduction in the number of subject-area courses taken. For students who would like to teach math or science for a few years but who want to retain the option of moving on to a higher-paying position in business or industry at a later time, this is a heavy loss. With less preparation in math or science than other students majoring in these areas, they may be at a significant competitive disadvantage if they do decide to leave teaching.

Seen from this perspective, raising the salaries of math and science teachers above those of other teachers—a proposal that has been debated for many years (cf. Kershaw and McKean 1962)—may be less effective than policies that would make it easier for graduates who have majored in math or science to try teaching. I do not intend to imply by this argument that nothing is learned in education courses or that no special pedagogical skills are needed to teach effectively. What I do mean is that there are many alternative ways to acquire these skills, including supervised on-the-job training complemented perhaps by formal in-service instruction. Certification requirements that effectively make college students choose between preparation for teaching or intensive preparation in math or science may keep many talented potential teachers out of the classroom.

Summary

When we look to explicit public policies to explain the educational problems of the 1980s, it is easy to neglect the importance of two implicit policies of the past:

• Letting teachers' salaries deteriorate relative to alternatives during a period in which alternative employment opportunities for women improved markedly.

• Requiring that aspirants for teaching positions specialize in education during college to the extent of reducing alternative career options, especially in technical areas.

Improving the quality of United States education in the 1980s will require reversal of these implicit policies.

Federal Education Policies

Policies Prior to the Reagan Administration

The federal role in funding elementary and secondary education in the United States has always been small. In 1950, the federal government contributed only 2.9 percent of the funds spent on United States public elementary and secondary education. Over the next thirty years the federal role did grow; by 1980 the federal contribution had increased to 9.3 percent of total expenditures (Condition of Education 1983, p. 42).

The major event marking a more active federal role was passage of the Elementary and Secondary Education Act of 1965 (ESEA). Title I of ESEA provided $1 billion to local school districts for compensatory education. By 1980, Title I funding had increased to $3.4 billion (O'Neill and Simms 1982). Correcting for inflation, the 1980 funding level provided a 30 percent increase in real expenditures over the 1966 funding level.

The regulations governing use of Title I funds mandate that recipient districts spend as many non—Title I dollars on low-achieving children participating in Title I programs as they spend on other children. In addition, Title I funds must be used exclusively to supplement the educational resources made available to Title I—eligible children; the funds may not be spread across all children. Gaining compliance with these regulations was a significant problem in the first years of Title I, when many districts treated Title I funds as general aid. By the mid 1970s, however, most of the 14,000 school districts receiving Title I funds complied with the regulations (Goettel et al. 1977). As summarized above, there is also limited evidence that Title I has been effective in improving the education of the target population and in reducing the achievement gap between black and white children.

Title I was the largest federal categorical grant program, but it was not

the only one. In fact, the number of categorical grant programs directed to disadvantaged children grew markedly over the 1970s. Since children eligible for participation in the federal programs are concentrated in certain geographical regions of the country, especially in urban areas and in the South, one consequence of the increase in the number and size of these programs was that federal aid also became concentrated. For example, in 1980 Mississippi received federal funds amounting to $380 per public school student, while Nevada received $134 per student (Condition of Education 1983, pp. 16, 42). A second consequence of the increase in the number of categorical grant programs, each of which had its own compliance regulations, was that fulfilling the varied and sometimes conflicting requirements created significant administrative problems for some school districts (Hill 1979).

Federal Education Programs under the Reagan Administration

The theme underlying President Reagan's treatment of education during his first two years in office is that education is a responsibility of state and local governments and federal involvement should be reduced. This theme was manifested in the Educational Consolidation and Improvement Act of 1981 (ECIA), significant provisions of which included:

• Chapter II of ECIA consolidated twenty-nine categorical programs into one block grant program. (The largest of the consolidated programs was the Emergency School Aid program, which provided funds for districts coping with desegregation-related problems.)

• The total budget authority for Chapter II was set at $456 million, approximately 24 percent less than the funds provided under the twenty-nine categorical programs. Chapter II funds are distributed to states on a per-student basis, with all students having equal weight in the formula.

• The regulations governing use of Title I (now Chapter I) funds were reduced, as were reporting requirements.

It is still too early to determine how the Reagan administration's policy changes will influence the distribution of student achievement. Among the fears of advocates for the disadvantaged is that the program changes will result in a significant reduction in the quality of education provided to minority group children and children from low-income families. They fear that the gains reported by NAEP in the achievement of black chil-

dren, especially in the primary grades, where Title I programs concentrate, will not continue in the years ahead (A Children's Defense Budget 1984). The reason this may happen, they argue, is that the elimination of the categorical programs and the reduction in the reporting requirements for compensatory education programs will result in a transfer of funds once devoted to programs for the disadvantaged to general educational use, to other public services, or to tax relief. They cite the experience of the early years of Title I as indicating the need for strong regulations to assure that resources intended for the disadvantaged actually reach them. I find these arguments compelling because they are consistent not only with the early Title I experiences but also with the evidence on local school districts' responses to state aid. As discussed below, the evidence is that local districts tend to use unrestricted state aid to substitute for locally raised revenues, thereby permitting tax relief, instead of using the aid to increase school services (Carroll 1979).

Although we do not yet have evidence on the consequences for student achievement of the Reagan administration's education policies, we can predict that the consequences of ECIA will vary from state to state. One reason is that some states (e.g., California) have significant compensatory education programs with their own accountability and reporting requirements. In these states relaxation of the federal reporting requirements will probably not lead to resource reallocation. Other states appear more ready to let local districts determine the use of Chapter I funds, and there are strong political pressures to treat such funds as general aid. A second reason is that the distribution of political power in big cities (the primary recipients of Title I funds) varies enormously across the country. Cities with leaders elected with strong support from minorities will use Chapter I funds differently from cities whose officials are dependent on other constituencies. Thus, one general consequence of the Reagan administration's policy changes is to make the quality of education received by disadvantaged children more sensitive to where the children live and go to school.

Proposed Changes in Federal Education Policies

An underlying theme of this chapter is that a paramount goal of federal and state education policies should be to provide all children in the United States, no matter where they live, with strong basic cognitive skills. The question is how federal education policies can further this goal, while remaining sensitive to the following realities.

• Local school districts tend to treat education block grants as general aid to be spread across all students or to permit tax relief.

• School districts that participated in a variety of categorical grant programs often found the regulations conflicting and the burden of compliance onerous.

• Local school districts did learn to comply with the regulations mandating that Title I compensatory education funds be targeted to low-achieving children.

• Title I appears to be at least partly responsible for the improvement over the 1970s in the reading skills of minority group children in the primary grades.

• Concern about the size of the federal deficit prevents a dramatic increase in federal education funding.

I suggest the following policy changes. First, the reporting and accountability requirements of Title I that were relaxed under Chapter I of ECIA should be reinstated. The first ten years of experience with Title I demonstrated both the difficulty of targeting aid to disadvantaged children and the success of the regulations in achieving this goal.

Second, a significant part of the funds currently provided to states as block grants under Chapter II of ECIA (perhaps 50 percent of the FY1983 funding level of $451 million) should be directed to low-achieving children by adding the funds to the Chapter I compensatory education program. These additional funds could permit the program to serve more of the primary-school-aged eligible children and possibly permit expansion in some districts to middle-school-aged eligible children. Such an expansion of an existing program would not create new compliance burdens.

Reallocating funds from Chapter II to Chapter I does redistribute federal funds to districts with large concentrations of low-income children. Given the evidence supporting the effectiveness of Title I, this reallocation is consistent with a federal goal of providing all American children, no matter where they live, with the basic cognitive skills they will need to fulfill their potential.

Changes in the Reagan Administration's Education Policies

When President Reagan appointed Terrell Bell to be Secretary of the Department of Education, the appointment was accompanied by a man-

date to eliminate the department. This was consistent with the philosophy that the federal government should minimize its involvement in elementary and secondary education. During the second half of Reagan's first term, this attitude changed as education became a campaign issue. Secretary Bell became one of the most visible cabinet members. He commissioned a panel to study United States education and publicized the panel's report, A Nation at Risk , published in April 1983. This report was highly critical of United States education, as is reflected in the oft-cited statement on its first page: "The educational foundations of our society are presently being eroded by a rising tide of mediocrity." In May 1984 the Department of Education published The Nation Responds , which summarizes efforts of state and local governments and the private sector to improve educational quality. The tone of the document is one of the federal government exercising leadership.

Two sets of facts are necessary to set the record straight. First, the education policies that President Reagan advocates, school prayer and tuition tax credits, are very different from the recommendations made in A Nation at Risk . Second, most of the "responses" by states cited in The Nation Responds were started several years before the publication of A Nation at Risk . Thus, to understand the governmental policies that have been most instrumental in shaping United States education, we must redirect our attention from Washington to the states.

State School Finance Policies

In contrast to the Reagan administration's philosophy that education decisions should be left to units of government closer to the people, state governments in recent years have increased their role, relative to local governments, both in funding public education and in regulating it.

State Aid Policies

In the five years between the 1977–1978 and 1982–1983 school years, state support for public education increased by 62 percent (14 percent in real terms), resulting in an average state expenditure of $1611 per pupil in the 1982–1983 school year. (This average conceals extreme variation from state to state: per pupil state-aid levels of $189 in New Hampshire, $3081 in Delaware, and $5920 in Alaska [Augenblick and Van de Water 1983]). Among the reasons for the increases in state aid were changes in state aid formulas brought about by court challenges.

Until the early 1970s, most states distributed aid to local school districts on something close to a per-student basis. As a result, state aid policies did little to alter the variation in per-pupil spending levels caused in part by differences in community wealth, defined both in terms of size of the property tax base and the personal income of residents. This changed during the 1970s as existing systems of school finance in many states were successfully challenged in court cases, the first and best-known of which was the Serrano case in California. In most states in which challenges were successful, the courts mandated that the state legislature devise a state aid plan that reduced the relationship between community wealth and per-pupil spending.

The most popular of the reform plans was labeled the Guaranteed Tax Base (GTB) plan or District Power Equalizing. Under a GTB plan, a property-poor community's per-pupil expenditure level depends only on the property tax rate that the community sets, not on the amount of revenue raised. For communities in which the per-pupil property tax base is smaller than the guaranteed tax base, the state provides aid equal to the difference between the revenue raised and the revenue that would have been raised had the community's chosen tax rate been applied to the GTB.

The attraction of the GTB concept is that it has the potential to equalize "opportunities" while retaining, to a significant degree, local control of education spending. Communities are free to choose a tax rate. Communities valuing education highly can retain a high level of education spending by choosing a high rate.

In essence the GTB plan reduces the price of education for communities with low per-pupil tax bases. Thus, the effect of a GTB plan on the distribution of per-pupil spending across districts in a state depends on the sensitivity of spending levels to price. GTB plans equalize spending if property-poor communities respond to the lowering of price by purchasing more education.

To date, the evidence from the states that have adopted GTB plans is that the plans have not resulted in a significant equalization of per-pupil spending across districts. The reason is that property-poor communities responded to the GTB incentives, not by purchasing more education at the now lower price but, rather, by continuing to purchase the same amount of education and using the increased state aid to lower property taxes. Thus, the net effect of GTB plans has been drastically to increase state aid levels and to equalize property tax rates across communities, but not significantly to equalize education spending levels across districts (Carroll 1979).

The final word has not been spoken on the effects of GTB plans. The incentives present in the plans actually implemented by states differ in important respects from the incentives embodied in the model plan.

• Hold-harmless clauses provide continued aid to districts with high tax bases.

• Underfunding has led some states to put a cap on the amount of aid they will provide to property-poor districts that set high tax rates.

• Cost-of-education adjustments in some states benefit districts that do not have high costs but, rather, choose to pay high teacher salaries in order to attract high-quality teachers (Carroll 1979; Brazer et al. 1982).

Despite these qualifications, it seems clear that the main lesson from the first ten years of school finance is that GTB finance plans which lower the price of education to property-poor communities, but leave the communities free to choose between more spending on education or lower tax rates, will not produce an equalization of per-pupil spending levels across school districts and will not result in all districts spending enough to provide their students with a strong basic academic program.

In my view, states should concentrate on designing policies that provide all districts with the resources to develop strong basic academic programs. A corollary to this emphasis is that states should deemphasize concern with equalizing expenditure levels across districts, even if this means that the quality of extracurricular activities varies across districts. The primary reason for this suggestion is that affluent districts can and do subvert equalization policies in countless ways. For example, if the state places a cap on per-pupil spending in local districts (a necessary condition for full equalization), affluent districts form non-profit civic foundations to support extracurricular activities so that the entire school budget can be used to pay for core activities. This means that the quality of the educational services is superior in the affluent districts even though the levels of per-pupil spending listed on the books are no higher than in other districts.

A foundation formula can achieve the goal of providing all districts with the resources to develop a strong basic academic program. Under this type of formula the state mandates that every district set a tax rate at least equal to a state-specified minimum. In return, the state guarantees that every district will have the funds to spend a specified amount per student on education (the foundation level). If the local revenues raised

by applying the required minimum tax rate to the local tax base do not generate sufficient funds to support the foundation level of spending, the state provides a grant in the amount of the deficit. If the foundation level is set suitably high—and this is the crucial step—this type of state aid formula can achieve the goal of assuring that all school districts have the resources needed to develop a strong instructional program. One way to keep the foundation level high is to mandate that it be set equal to a given percentage (perhaps 90 percent) of average per-pupil spending on education in the state.

State Policies to Improve Teaching

By the late 1970s, many states became concerned when they realized that although state spending for education was rising rapidly, performance did not appear to be improving. This led to a variety of state efforts to tie education aid to new regulations aimed at improving the quality of instruction provided in schools and increasing the effort levels of students and teachers.

Tougher Certification Requirements

Many states have recently increased the requirements for obtaining a teaching certificate (a license to teach in a public school). One common change has been an increase in the amount of formal training a teacher must complete before being eligible for certification. A second change is that applicants for teaching positions must score above a minimum grade on a standardized exam —seven states use the National Teachers Exam for this purpose (Condition of Education 1983, p. 66). A third type of new requirement is that teachers undergo on-the-job observation and evaluation before receiving certification. Such "performance-based certification" requirements have been introduced in Florida, Georgia, New Mexico, and Oklahoma (Ordovensky 1983).

In attempting to understand the effects of increasing certification requirements, it is useful to ask what model of local-school-district behavior would justify such state actions. One model that is not consistent with such actions is that local school districts are motivated to hire effective teachers, are competent to choose such teachers from the pool of applicants, offer high enough salaries to attract a talented pool, and can and will dismiss an incompetent teacher if they make a hiring mistake. If all districts in a state fit this description, then no state teacher-certification

requirements would be necessary. Such school districts could examine applicants' college transcripts and evaluate the quality of their academic preparation. They could judge whether the applicants' test scores are high enough to offer promise of intellectual skills. They could evaluate the performance of the teachers they hire and dismiss those who are not able to improve their performance sufficiently to meet minimum standards.

Among the alternative models of local school district behavior that could justify the regulations are that local districts are motivated to hire teachers on the basis of criteria other than potential teaching effectiveness, such as political connections. Or district administrators could be incompetent in making hiring decisions. Or district administrators trying to hire effective teachers could be hampered by salaries so low that competent teachers do not apply.

Tougher teacher certification requirements may improve the quality of teachers employed by such districts by preventing the hiring of incompetent teachers with political connections. They may aid local school officials in gaining taxpayer support for higher teacher salaries by highlighting the inability of the district to attract competent teachers at existing salaries. These accomplishments could justify tough certification requirements if a large percentage of school districts in a state fitted the description of districts in need of help or if the requirements had no adverse affects on other districts. It is not possible to evaluate how large a percentage of the school districts in a state need state certification requirements to influence their hiring practices.

Tougher teacher certification requirements do, however, adversely affect school districts that want to hire the most effective teachers they can find, have administrators competent to choose among applicants, and pay reasonable salaries (cf. Benham 1980). Consider state-mandated increases in the formal training college students must complete before being eligible for teacher certification. These increased requirements raise the cost of becoming a teacher. In some cases, the cost increase may result from the student having to stay in school longer. In other cases, the cost may be in terms of narrowing the set of colleges the student may attend. For example, many liberal arts colleges, attended by only a few students who want to become teachers, find it unfeasible financially to offer all the courses needed for state accreditation. Thus, the student who wants to try public school teaching may not be able to attend a college that in other respects offers an attractive academic program. In still other cases, the cost may be that increased course requirements for certification may reduce opportunities to take other courses. These foregone courses might

have enhanced students' competitive positions in applying for jobs in other industries if they decide to leave teaching, or have difficulty finding a desirable teaching position, due to enrollment declines and layoffs. Such an argument may seem farfetched, for most discussions of teaching emphasize the need to attract talented college graduates to a teaching career. High attrition rates during the first years on the job are a reality among public school teachers in the United States, however—in fact, they have always been a reality (Charters 1970). Consequently, attempts to improve teaching would be more effective if they focused less on how to attract talented college graduates for a lifetime and more on how to attract talented graduates for a few years. (Indeed, many high-quality private schools recruit teachers with exactly this model in mind.) Seen from this perspective, increased course requirements for certification may significantly reduce the pool of applicants for teaching positions and make it more difficult for school districts to find talented teachers.

Strict state requirements for teacher training may also reduce the quality of teaching in public schools by hindering innovation and diversity in training programs. In most states, the first step in the teacher certification process is for the applicant to complete a prescribed course of study in a teacher training program accredited by the state. For a program to be accredited, it must offer the courses that satisfy the state requirements. This would not be a serious limitation, perhaps, if there were strong evidence that a particular set of courses provided the best training for prospective teachers. The evidence is of a quite different nature, however. Studies of the determinants of teacher effectiveness show no relation between preparation and teaching effectiveness, as measured by student test-score gains. For example, there is no evidence that teachers majoring in education are more effective than teachers majoring in other subject areas (Murnane 1983). This should not be interpreted as indicating that preparation does not matter; rather, researchers' ability to identify effective training programs is limited by the similarity among programs mandated by state regulations. It does imply, however, that state policy should encourage innovation in teacher training, rather than demand compliance with a tightly defined set of standards.

How might states promote innovation in teacher training and encourage talented college students to try teaching, while still monitoring the hiring actions of troubled school districts? I would like to offer the following immodest, three-part proposal.

First, the state should require that all college students interested in becoming teachers pass an examination testing basic literacy skills and

minimum subject-matter competence. Students should be allowed to take this test at any time during their college years. The purpose of the test would be to provide information to students about minimum cognitive skills needed for teaching. Those students who did not possess these skills could either work to acquire them while in college or choose an alternative occupation rather than devoting their time and their institution's resources to teacher training. I want to emphasize that this exam is only intended to test potential teachers' basic cognitive skills. It is not intended to distinguish between competent and incompetent classroom teachers. In fact, a recent study that found a statistically significant negative relationship between teachers' scores on the National Teachers Exam and their students' test-score gains serves to highlight the limitations of written tests as indicators of teaching effectiveness (Summers and Wolfe 1977).

Second, all colleges and universities in the state should be allowed to design their own teacher training programs and attempt to recruit students and to place their graduates in teaching jobs. Under such a system, school district personnel officers would find it worthwhile to keep track of the performance of teachers trained in specific programs and to recruit more teachers graduating from programs that had provided effective teachers in the past. Placement officers in teacher training programs would have incentives to give honest appraisals of their graduates' potential, given that school districts would turn away from programs whose graduates turned out to be less capable than was advertised.

Third, the state should observe and evaluate teachers' performance in the classroom during their second year of teaching. (Teachers need one year on the job to overcome initial fears and frustrations.) The teacher must pass the performance-based evaluation test to gain certification. The evaluation system could be similar to that recently introduced in Georgia, where each rating team consists of one teacher and one administrator from the applicant's district, and one state official, all trained in evaluation.

I do not underestimate the well-documented problems involved in any system of evaluating teachers' performance (cf. Wise et al. 1984). Such a performance-based certification system would not assure that all school districts employ only effective teachers. There is not sufficient consensus on the definition of effective teaching to achieve that result, even with the best of evaluators. However, such a system could provide as much assurance as existing certification requirements that troubled districts do not employ incompetent teachers. Moreover, this proposal, by encouraging

innovation in training and by making it easier for college students to try teaching, would increase school districts' ability to find competent teachers.

What would happen if such a market for teacher training were allowed to develop? All districts would be able to recruit better teachers as the quantity and quality of applicants increased. A second, less agreeable consequence is that districts with high salary scales and good working conditions would attract the most talented graduates from the best programs, while districts with low salaries and difficult working conditions would be able to attract only less able students and those graduating from weaker programs. Would the resulting distribution of teachers be a reason to scrap the proposal? I think not, because the pattern is the same under the existing regulatory system. The only real difference would be that the differences in the quality of the teaching staffs of different school districts would be more noticeable, because the range of quality in the overall applicant pool would be greater. Viewed from this perspective, the problem is how to make teaching more attractive in districts that are not able to attract and retain high-quality teachers. Licensing requirements that contract rather than expand the pool of applicants for positions in such districts are not the answer.

One other consequence of such a market system would be that some existing teacher-training programs would lose students. They would resist the changes, and possibly prevent deregulation of the certification requirements. I do not believe, however, that deregulation of teacher training would inevitably lead to the demise of schools of education, although it would lead to changes in what they do. An optimistic, but I believe plausible, scenario is that many education schools would swallow hard and then transform themselves into lean, mean competitors with liberal arts programs. They would concentrate on substance and on developing leadership, supervision, and other people-managing skills, and would claim (correctly) that a few years in teaching is superb background for many other careers. They would create a new market niche between the "major in business and commerce" and the liberal arts degree.

One criticism that might be made of my three-part deregulation proposal is that by expanding the number of applicants competing for teaching positions, it would reduce pressure for salary increases and thereby damage the profession. I do not believe that this criticism is valid. In the long run, the salaries of public school teachers and the prestige of the teaching profession will increase markedly only if taxpayers and employers believe that teachers provide high-quality education. My proposal, by

making it easier for talented college graduates to enter teaching and for school districts to recruit able teachers, will improve the quality of teaching in our schools and thereby will ultimately benefit the teaching profession.

Compensation for Teachers

Merit Pay for Teachers

Several states are providing incentives to local school districts to institute merit pay programs. For example, the state of Florida has appropriated $80 million for 1984–1985 to support merit pay for teachers (The Nation Responds 1984, p. 45). A large part of the attraction of merit pay is that it appears to offer a means of increasing the compensation of the best teachers without incurring the large cost of raising the salaries of all teachers.

The apparent attractiveness of merit pay is drawn into question by knowledge of its history, however. Merit pay is an old idea, dating back to 1908 in the United States. The history of merit pay plans is that they do not last very long: the typical life span is three or four years (Merit Pay for Teachers 1979). Moreover, the temporal and geographical nature of the evidence prohibits blaming union resistance for this brevity. Case studies indicate that the reasons plans were terminated include negative effects on teacher morale and cost of administration. Even teachers picked by their peers as likely recipients of merit pay because of their superior teaching oppose the idea (Jackson 1968).

Insight into the problems of merit pay can be gained by examining the incentives in a system in which teachers are paid according to their students' average performance on specified tests. (This is one of the criteria for teacher merit approved by the state of Florida [The Nation Responds 1984, p. 45]). Such a system provides incentives for teachers to devote instructional time to the skills needed to do well on the test, and to minimize time spent teaching other skills. Many teachers feel that the types of skill that can be measured on tests are only a small subset of the skills that constitute a good education. As a result, they oppose an incentive structure that fosters time allocations they see as diminishing their effectiveness as teachers.

A related problem concerns the incentives for teachers to allocate their time among children according to a strategy that maximizes their salary. To the extent that teaching time is a private good—time spent with one

child decreases the amount of time available for another child—the goal of maximizing the average test-score gain of students in a class provides incentives to allocate time so that the marginal test-score gain resulting from the last unit of time spent with any child is equal across children. Another way to say this is that this compensation system provides incentives to minimize time spent with children who do not reward increased teacher attention by achieving test-score gains. Again, many teachers feel that these incentives conflict with their desire to help all children in their class to learn.

In an ongoing study, a colleague and I are attempting to learn more about the potential for merit pay by examining how it works in five districts that have used merit pay for a number of years. (Of course, the longevity of merit pay in these districts makes them quite atypical.) The preliminary evidence suggests the following conclusions.

1. The districts in which merit pay survives for an extended period of time are districts in which the uniform salary scales are above the regional average and working conditions are good. In other words, merit pay is used to supplement the pay of teachers in already high-paying districts.

2. These districts do not use student test-score gains as criteria for merit pay.

3. These districts weight teachers' activities outside the classroom heavily in determining who gets merit pay. For example, taking on heavy committee assignments, taking additional courses, and providing leadership in community activities are important in determining who gets merit pay. In other words, merit pay in these districts is more like extra pay for extra work than it is rewarding the most effective classroom teachers (Murnane, forthcoming).

Overall, the evidence from our ongoing work indicates that merit pay may be a small part of an attractive compensation package in well-running school districts with high uniform salary scales. The evidence does not suggest that it is an effective strategy for improving the quality of teaching in troubled school districts.

State-Sponsored Mini-Grants

State-sponsored mini-grant programs for teachers provide a means of stimulating teachers' effort levels and encouraging the development of

new teaching ideas without the perverse effects present in most merit pay programs. The assumptions underlying the mini-grant idea are two. First, good teachers have many ideas for innovative projects but lack of funds often frustrates their attempts to implement these ideas. Second, teachers, like everyone else, desire and need recognition for excellent work. Thoughtfully designed mini-grant programs have the potential for stimulating the development of innovative teaching projects, by providing financial support. They can also provide teachers who have good ideas and an entrepreneurial spirit with recognition for their efforts.

Such programs might work as follows. The state announces the availability of funds to support small grants (e.g., $500–$3000) to support the development or execution of any idea that holds promise for improving teaching or enhancing students' experiences. Individual teachers or groups of teachers submit short proposals explaining and defending their ideas. The proposals are evaluated by a committee consisting of teachers and state officials. Awards are announced publicly. At the end of the award year, recipients submit a short summary explaining what they did with the grant money and what they accomplished with the grant. Among the important design provisions of the grant program are that there be sufficient funds to support a large percentage of proposals that hold promise and that there be no quota on the number of teachers in any school or school district who could be awarded grants.

Grant programs such as that described above have been tried in a number of states (e.g., Massachusetts), in many cases supported by funds from Title IV-C of ESEA. Evaluations of these programs are quite encouraging. They are relatively simple to administer. They have, at relatively low cost, provided support for a range of interesting activities. Moreover, they have provided recognition and support to a large number of teachers (McDonnell and McLaughlin 1980).

Salary Increases

A number of states are taking action to raise teachers' salaries. While this alone will probably not improve the quality of teaching in our schools, evidence on teacher labor markets (cited above) suggests that salary increases will be necessary to improve the quality of education in many school districts.

One critical question states face is what strategy to employ to help districts attract effective teachers. I would argue that, as stated above, a state-aid formula based on the foundation concept with a high founda-

tion support level is an effective means of insuring that all districts have the resources to hire effective teachers.

An alternative strategy adopted by some states is to mandate statewide minimum salary levels for teachers. I do not believe that this is a good policy. Labor-market conditions for teachers vary widely among the school districts in any state; consequently, the salary that a district must offer to attract effective teachers varies widely. Districts with good working conditions that are located in or near desirable residential locations can attract talented teachers at lower salaries than other districts. Mandating that these districts offer high salaries reduces the districts' flexibility in using resources in a manner that most efficaciously builds a solid instructional program. Similarly, in districts with poor working conditions, using scarce resources to improve the conditions under which teachers work may be a more effective strategy for attracting talented teachers than raising salaries would be. (Many private schools successfully employ this strategy.) In summary, state policy should concentrate on insuring that all school districts have the resources to attract effective teachers; it should not mandate how much of these resources should be devoted to teachers' salaries.

State Policies to Stimulate Students' Efforts

In recent years, states have undertaken a number of actions aimed at stimulating students' effort levels and enhancing their achievement. A number of states have increased course requirements, especially the number of math and science courses required for high school graduation. For example, in 1982 California passed legislation requiring that, effective in the 1986–1987 school year, high school students must complete two years of mathematics and two years of science before graduating. These new requirements are much less stringent than those recommended by A Nation at Risk. Their significance is indicated, however, by the fact that only 40 percent of the students graduating from California high schools in 1983 satisfied these new requirements (Harnischfeger and Wiley, 1984). Other states have passed more stringent requirements. For example, effective in the 1986–1987 school year, Florida high school students must complete three years of science and three years of mathematics before graduating (The Nation Responds 1984, p. 43). Another common, and more controversial, state initiative is the introduction of minimum competency examinations, which students must pass to graduate from high school. Seventeen states have introduced such exam requirements (Condition of Education 1983, p. 64).

Presumably, the impetus for these new state policies is concern about declining test scores and the correlation between test scores and the number of academic courses students take. Certainly, the hope is that the requirements of more course work and the successful completion of a basic skills exam will stimulate students (and teachers) to work harder and achieve more. The danger is that the new requirements will elicit other, less desirable responses, such as students dropping out of school and teachers teaching tested skills at the expense of skill development in areas not tested on the minimum competency exams (Madaus 1982).

To date, there is almost no evidence on the effects of the new requirements. We simply do not yet know the responses the new requirements elicit from students and teachers, or whether these responses increase achievement. It does seem safe to conjecture, however, that the responses will depend on the extent to which students find that meeting the stiffer requirements results in rewards commensurate with the additional work. One factor that will influence this judgment, and consequently students' responses, are graduates' labor-market experiences. It seems unlikely that students will increase effort levels to earn a now harder-to-get high school diploma unless they find that labor markets reward the extra effort with improved employment opportunities. It is not possible to predict these labor-market responses; alternative models of the operation of labor markets yield different predictions (compare Smith and Welch 1977 with Thurow 1975).

A second factor that will influence students' responses to new academic requirements is the quality of teaching in the new required courses. If competent teachers are not found to teach the new required math and science courses, adolescents may well discover responses less socially desirable than working harder.

A third factor that will influence responses is the treatment of students who fail the graduation exam. Introduction of an exam requirement will be more likely to enhance the effort levels of low-achieving students, instead of inducing them to drop out, if failure leads to significant remedial instruction and additional opportunities to take the exam. These conditions are more readily met if the exam is administered early in a student's high school years, perhaps in grade 10, instead of in grade 12.

Specific Questions to Ask

This chapter has evaluated a wide range of federal and state education policies and has made a number of recommendations for policy changes. I stand behind these recommendations and argue that they are the right directions to move in attempting to provide all American children with a

good education. It is important to keep in mind, however, that the consequences of any policy will depend on the details of the policy and on the composition of the package of which the particular policy is a part. It is not possible in such a broad-brush essay to analyze the significance of particular policy details and packaging, but I would like to suggest that in analyzing any particular state or federal policy, it is useful to ask the following three sets of questions.

1. What model of local-school-district behavior underlies the policy? Does the policy assume that local districts are incompetent or constrained by local politics? How will the policy affect the options of districts that are competent and have taxpayer support for developing a high-quality education program?

2. Does the policy make teaching more, or less, attractive as an occupation? What responses from teachers might the policy evoke? Does the policy make it more, or less, difficult for college students to prepare to teach, at least for a few years?

3. Does the policy offer rewards to students commensurate with the increased demands made upon them? What happens to students who do not meet the demands?

Acknowledgments

I appreciate the first-rate research assistance of Marc Moss and Patrick J. Murnane. I would also like to acknowledge helpful conversations with Jeanne Chall, Mary E. Curtis, Harold Howe, Francis Keppel, Marvin Lazerson, Jerome Murphy, Richard Nelson, Barbara Neufeld, and Carol Weiss. I would particularly like to thank Edward Pauly, who read several drafts of this essay and provided many important ideas.

References

Anderson, Judith I. 1982. "Measuring the Effectiveness of Title I: A Summary of the 1979-80 State Title I Evaluation Reports." Paper presented at the annual meeting of the American Educational Research Association, New York, March.

Augenblick, John, and Gordon Van de Water. 1983. State Support for Education, 1982-83. Denver: Education Policy/Planning Service.

Bacharach, Samuel B., David B. Lipsky, and Joseph B. Shedd. 1984. Paying for Better Teaching: Merit Pay and Its Alternatives. Ithaca, N.Y.: Organizational Analysis and Practice.

Benham, L. 1980. "The Demand for Occupational Licensure." Pp. 13-25 in S. Rottenberg, ed., Occupational Licensing and Regulation . Washington, D.C.: American Enterprise Institute.

Brazer, Harvey E., Deborah S. Laren, and Frank Yu-Hsieh Sung. 1982. "Elementary and Secondary School Financing." Pp. 411-46 in H. Brazer, ed., Michigan's Fiscal and Economic Structure . Ann Arbor: University of Michigan.

Bryk, Anthony S., Peter B. Holland, Valerie E. Lee, and Ruben A. Carriedo. 1984. Effective Catholic Schools: An Exploration. Washington, D.C.: National Catholic Education Association.

Carroll, Stephen J. 1979. The Search for Equity in School Finance: Summary and Conclusions. Santa Monica: Rand Corp.

Charters, W. W. 1970. "Some Factors Affecting Teacher Survival in School Districts." American Educational Research Journal 7: 1-27.

A Children's Defense Budget: An Analysis of the President's FY 1984 Budget and Children. 1984. Washington, D.C.: Children's Defense Fund.

Coleman, James S., Thomas Hoffer, and Sally Kilgore. 1982. High School Achievement. New York: Basic Books.

The Condition of Education 1983. 1983. Washington, D.C.: U.S. Department of Education.

Darling-Hammond, Linda, and Arthur E. Wise. Forthcoming. "Beyond Standardization: State Standards and School Improvement." Elementary School Journal.

Digest of Education Statistics 1983. 1983. Washington, D.C.: U.S. Department of Education.

Frankel, Martin M., and Debra E. Gerald. 1981. The Projections of Education Statistics to 1990-91 , Vol. 1. Washington, D.C.: National Center for Education Statistics.

Goettel, Robert J., Bernard A. Kaplan, Martin E. Orland, Pascal D. Forgione, Jr., and Sheila D. Huff. 1977. A Comparative Analysis of ESEA, Title I Administration in Eight States. Syracuse, N.Y.: Syracuse Research Corp.

Harnischfeger, Annegret, and David E. Wiley. 1975. Achievement Test Score Decline: Do We Need to Worry? Chicago: Cemrel.

—––. 1984. Time and Learning in California Schools. Sacramento: California Testing Program.

Hill, Paul. 1979. Do Federal Programs Interfere with One Another? Santa Monica: Rand Corp.

The Impact of Teacher Shortage and Surplus on Quality Issues in Teacher Education. 1983. Washington, D.C.: American Association of Colleges for Teacher Education.

Jackson, Philip W. 1968. Life in Classrooms. New York: Holt, Rinehart and Winston.

Jencks, Christopher, Marshall Smith, Henry Akland, Mary J. Bane, David Cohen, Herbert Gintis, Barbara Heyns, and Stephen Michelson. 1972. Inequality: A Reassessment of the Effect of Family and Schooling in America. New York: Basic Books.

Kershaw, Joseph A., and Roland N. McKean. 1962. Teacher Shortages and Salary Schedules. New York: McGraw-Hill.

McDonnell, Lorraine M., and Milbrey W. McLaughlin. 1980. Program Consolidation and the State Role in ESEA Title IV. Santa Monica: Rand Corp.

Madaus, George F. 1982. "Competency Testing: State and Local Level Responsibilities." Educational Measurement: Issues and Practice , Winter.

Merit Pay for Teachers. 1979. Arlington, Va.: Educational Research Service.

Murnane, Richard J. 1981. "Teacher Mobility Revisited." Journal of Human Resources 16: 3-19.

—––. 1983. "Understanding the Sources of Teaching Competence: Choices, Skills, and the Limits of Training." Teachers College Record 84: 564-69.

—––. Forthcoming. "The Rhetoric and Reality of Merit Pay: Why Are They Different?" In Henry C. Johnson, ed., Merit, Money and Teachers' Careers . Lanham, Md.: University Press of America.

National Assessment of Educational Progress (NAEP). 1978. Three National Assessments of Science: Changes in Achievement, 1969-77. Denver: Education Commission of the States.

—––. 1981a. Has Title I Improved Education for Disadvantaged Students? Evidence from Three National Assessments of Reading. Denver: Education Commission of the States.

—––. 1981b. Three Assessments of Reading: Changes in Performance, 1970-80. Denver: Education Commission of the States.

—––. 1982. Reading, Science and Mathematics Trends: A Closer Look. Denver: Education Commission of the States.

—––. 1983. The Third National Mathematics Assessment: Results, Trends and Issues. Denver: Education Commission of the States.

National Commission on Excellence in Education (NCEE). 1983. A Nation at Risk: The Imperative for Educational Reform. Washington, D.C.: U.S. Department of Education.

The Nation Responds: Recent Efforts to Improve Education. 1984. Washington, D.C.: U.S. Department of Education.

O'Neill, Dave A. 1984. Education in the United States, 1940-1983: A Survey of Trends and Current Concerns. Special demographic analysis. Washington, D.C.: U.S. Bureau of the Census.

O'Neill, June A., and Margaret C. Simms. 1982. "Education." Chapter 11 in J. L. Palmer and I. S. Sawhill, eds., The Reagan Experiment . Washington, D.C.: Urban Institute.

On Further Examination: Report of the Advisory Panel on the Scholastic Test Score Decline. 1977. New York: College Entrance Examination Board.

Ordovensky, Pat. 1983. "Evaluations Sort out Good, Bad Teachers." USA Today , 8 June, p. 3A.

Ranbom, Sheppard. 1984. "S.A.T. Scores Up 4 Points, Biggest Jump in 21 Years." Education Week , 26 September, p. 1.

The Reagan FY1985 Budget Series: Education. 1984. Washington, D.C.: Federal Budget Report.

Schlechty, P. C., and V. S. Vance. 1981. "Do Academically Able Teachers Leave Education? The North Carolina Case." Phi Delta Kappan 63: 106-112.

Smith, James P., and Finis Welch. 1977. "Black-White Wage Ratios, 1960-1970." American Economic Review 67: 323-38.

Summers, Anita A., and Barbara L. Wolfe. 1977. "Do Schools Make a Difference?" American Economic Review 67: 639-52.

Teacher Supply and Demand in Public Schools, 1981-82. 1983. Washington, D.C.: National Education Association.

Thurow, Lester C. 1975. Generating Inequality. New York: Basic Books.

Wainer, Howard. 1984. "An Exploratory Analysis of Performance on the SAT." Journal of Educational Measurement 21(2): 81-91.

Weaver, W. T. "Educators in Supply and Demand: Effects on Quality." School Review 86: 552-93.

Wise, Arthur E., Linda Darling-Hammond, Milbrey W. McLaughlin, and Harriet T. Bernstein. 1984. Teacher Evaluation: A Study of Effective Practices. Santa Monica: Rand Corp.

Chapter Six—
Commentary

Paul N. Courant

The methodology of economics is powerful at showing where there is no role for government; where it shows that there is a role to be played, the methodology is not very useful as a guide to designing programs to permit that role to be performed effectively. Thus, when we consider issues of social concern, we have a lot to say about what should not be done, and we have something to say about where things should be done, but we say very little about how things should be done, except that perhaps they should be done some other way. It seems to me that if we are to present useful prescriptions for social policy, we need methods that can be used to tell us what the governmental role might be and how to play it. I suggest that we devote more attention to figuring out what rules might meet the ends of "rough justice" and, if I may, "rough efficiency," and less time bemoaning the fact that what we see is not defensible under rigorous rules of evaluation. Put another way, I think that our method often biases us toward getting the right answer to the wrong problem. And our answer is usually "no."

The paper by Danziger and Feaster is relatively straightforward: reasonably good data are used to examine a well-defined question. We come away with some new knowledge about how recent policy has affected poverty, according to the official definition of poverty. The results do not change the way we think about policy, but they add to our understanding.

The article by Murnane is very different in character. While its title implies a focus on recent policy changes, its coverage is much broader. It starts with the now familiar, depressing statistics on SAT scores over the

past two decades, and attempts to focus the diverse literature on elementary and secondary education on the question of why those scores fell and what might be done about it. I learned a great deal about the literature from Murnane's article, but this literature is remarkably unhelpful in providing guidance as to what should be done. I must also confess to some confusion regarding the objectives of public education. Thus, much of what I have to say about this paper will be quite speculative. I note here, however, that in my opinion, Murnane often draws too much from the literature (e.g., inferring that Title I is effective because over a decade the ratio of black to white test scores has improved somewhat) and sometimes too little (e.g., if Murnane's own work shows that teacher training has no effect on teaching effectiveness, why is he so careful not to conclude that education courses are a waste of time?).

Antipoverty Effects of Transfer Programs

Danziger and Feaster measure the effectiveness of antipoverty programs by comparing pretransfer income and post-transfer income. As the authors know well, this approach is strictly accurate only if there are no behavioral responses to the programs themselves. Thus, the measure is an upper bound on the effectiveness of government programs. They cite a study suggestion that the upper-bound estimate of the effect of transfer programs is indeed an upper bound, but a very poor estimate. According to Danziger and Feaster, their method of calculation overstates the reductions in poverty attributable to transfers by a factor of more than two (for 1974 data). If antipoverty programs cause disincentives of this magnitude, it may be worth pursuing some other approach.

A second difficulty with the Danziger and Feaster paper is that they use the census definition of income, under which non-cash transfers are not included. Perhaps their results would not change using a more comprehensive income definition, but the details, especially with regard to sex, might change—given that AFDC eligibility often implies Medicaid eligibility and that it is relatively difficult for male-headed households to receive such assistance in many states.

Another cause of some disquiet arises from the years chosen for comparison—1978 and 1982, years that are at radically different positions in the business cycle. The year 1978 was unusually good, and 1982 unusually bad. A comparison of policy effects should not depend so heavily on statistical controls for the business cycle. The business cycle also varies by region. The North Central region has always been very cyclical, and so, no doubt, have been its poverty numbers. A sample of cycles, rather than just one peak to trough, would be desirable.

Perhaps more important is the interpretation of the descriptive statistics and associated analysis. The authors first state that the probability of escape from poverty, given receipt of transfers, was essentially unaffected by the recession. Later in the discussion, they show that 15 percent of the increase in the number of persons poor in male-headed households (and 24 percent of the increase for female-headed households) can be attributed to program changes. Do we then infer that the remaining 85 percent and 76 percent are due to the recession? If so, their first statement is misleading.

The most striking results that emerge from the analysis are: (1) women are more likely to receive transfers, and less likely to be removed from poverty by them; (2) households with incomes far below the poverty line are more likely to receive transfers and less likely to be removed from poverty by them. Blacks are less likely to be removed from poverty than whites, conditional on receipt of transfers. The authors attribute this finding to the fact that whites are more likely to receive social insurance payments, which are more generous than welfare payments. Might it also be the case that whites are more likely to be near the poverty line in the first place?

I think the authors' assumption that 1978 policies would have stayed in effect, or even become more generous, had Mr. Reagan not been elected is probably wrong. My own experience in and around the White House in 1980 indicates that even the moderate welfare reform that was official administration policy at the time was not pursued seriously.

Danziger and Feaster say that the primary goals of transfer programs are the "provision of minimum levels of cash, nutritional, medical, housing, and educational assistance for a substantial portion of the population." Unfortunately, the analysis is really only about cash programs and about poverty lines, rather than about "minimum levels of assistance." This article simply does not tell us how much the Reagan changes have affected the living standards of the poor in general, or even of those poor who receive transfers. Instead, we know a tremendous amount about the effect on their cash incomes.

I find little to quarrel with in the authors' policy prescriptions. Surely if we have defined a minimum standard, we ought to index it. If targeted employment programs are cost-effective (all that is indicated in the text is that they could eliminate a lot of poverty, and costs are not discussed), they may be worth pursuing. The Garfinkel plan, essentially a national

tax-transfer scheme from absent fathers to present mothers, has always appealed to me, in part because it is a program with antipoverty impact that does not require poverty as a condition for either receipt or contribution.

If it is really true that pretransfer incomes would be as much higher, in the absence of programs, as is implied by Plotnick's results, there should be large potential efficiency gains in the design of antipoverty programs. If over half the effect of antipoverty programs now is to reduce pretransfer incomes, surely we should start thinking about radical changes in program design, especially if current programs are not well adapted to respond to the feminization of poverty. Are there specialized training or workplace rules that affect women differently from men, and that involve interventions in labor markets so that women will be helped by economic growth more than they have in the past?

Elementary and Secondary Education Policy

If our measures of the effectiveness of income redistribution programs suffer because Consumer Population Survey data are incomplete and the poverty line is not a perfect welfare index, the situation for evaluation of education policy is far worse. Nothing in Murnane's paper vitiates Hanushek's conclusion that we do not know much about the production function for education.^[1] Even worse, we do not have social agreement on what we want, beyond vague phrases like "excellence" and "equality of opportunity," goals that may often be in conflict. We do have a sense that the schools are not doing especially well at promoting either goal and that there is some problem with incentive structures at the level of school and classroom. "Merit pay," the economist's natural solution to the problem, needs to become far more than a slogan if it is to have any effect, and Murnane's own work indicates that schools' organizational structure limits its potential. So we have a good that we cannot measure, produced by a process that we do not understand, coupled with a general sense that things are not going well. What's a policymaker to do?

Murnane's major policy proposals are in the area of teacher certification. He argues that current certification requirements in many states raise the opportunity cost of becoming a teacher by preventing the prospective teacher from acquiring training that can be used for anything else. I would add, as a sometime counselor of undergraduates, that another reason why some students stay away from education courses is that the courses themselves are not very interesting. Why would a student with

relatively good scientific aptitude choose to enter an education curriculum as a college junior when by studying chemistry she or he will have a better time in school and have more opportunities, at higher pay, two years hence?

Murnane's suggestion that states experiment with certification requirements is very promising. Indeed, in light of the econometric evidence that education courses have nothing to do with classroom performance, the case for trying this kind of approach seems overwhelming. The problems are essentially political. It is not news that education courses accomplish little, yet they have been around for a long time, and both teacher-training schools and teachers' unions seem to be natural constituencies for their maintenance. When essentially all of the institutional experts in an area have a clear interest in maintaining a policy, it is not going to be easy to abolish it, even if the social interest in doing so is large.

The preceding example is one where the preferred policy involves reducing the state's role in education policy. There are many such examples, leading us to the question of just what we believe the state ought to be doing in this area. Murnane advances a number of reasons for state regulation of local boards, and in all cases that bear on personnel policy the reasons cited make sense only if local boards are incompetent. Moreover, to the extent that education is a private good, it is hard to find a role for state government, much less for federal. Indeed, it may be hard to find a role for government at any level.

Such a conclusion seems to miss the point. Traditionally, it has been argued that there is a public-good aspect to elementary and secondary education, that it is important that there be some common core of knowledge and literacy, and (perhaps) that there be some socialization toward a common set of values. Moreover, there may be some information economies in the choice of curriculum ("experts" can help to guide families regarding the returns to the acquisition of specific kinds of training). Surely a set of school curricula can be established which are likely to be beneficial in most contingencies (arithmetic and literacy in English are surely two elements of such a set). In any of these cases, it would seem that the state does have a role to play in defining a required curriculum and in requiring some minimum performance level in that curriculum. But the state's role here would be quite removed from the details of day-to-day school operations or personnel policy. Promulgation of output standards, coupled with some incentive system to induce schools to meet those standards as they saw fit, would seem to be appropriate.

There is some scope for regulation of a more detailed kind. However,

the system is currently overregulated from the top and leaves little room for innovation. The "best and the brightest" face real barriers to entering teaching; in addition, once people become established as teachers, union rules make it virtually impossible to fire the bad ones and reward the good ones. While the latter problem will always exist, and the internal dynamics of any collective enterprise are such that strict adherence to the marginal-productivity doctrine of wages, individual by individual, probably causes serious morale problems, it should be possible to introduce considerably more flexibility into the training and promotion of teachers than is currently the norm.

A tennis coach once told me, "Always change a losing game." He was a public school teacher, and a good one. I suggest we apply his advice to public schools themselves.

Commentary

Eric A. Hanushek

Poverty and Welfare

The article by Danziger and Feaster is excellent. It attempts to disentangle the various factors affecting changes in poverty over the recent past and, most important, to consider separate demographic and geographic groupings of the poor. In order to separate technique from conclusions, their analysis is done in several ways.

From the analytic viewpoint, I think two related issues are particularly worth considering. These involve the concentration on poverty thresholds and, relatedly, the definition of poverty. The development of an agreed-upon measure of poverty has been very important in the history of social policy. The definition evolved from the early work of Mollie Orshansky and was refined somewhat over time. No one believes the current definition is a perfect measure of the welfare (or misery) of families, but it has provided a yardstick against which intertemporal as well as cross-sectional changes can be identified, quantified, and analyzed. The Census Bureau provides data on poverty regularly; programs use poverty definitions to determine eligibility; and social scientists make a living interpreting the incidence of poverty. Nevertheless, it is important to reflect on the usefulness of poverty measures in the Danziger/Feaster analysis. One interpretation of their approach is that they have taken an inherently continuous variable, that of income, and introduced errors into it to arrive at the dichotomous variable of poverty. To add to this, it is now well known that linear regression models have certain conceptual difficulties when applied to dichotomous dependent variables. Therefore,

we frequently turn to some nonlinear method of estimation such as probit analysis, which can be thought of as a device for inferring what the underlying continuous probability distribution of a variable looks like given the realized values of the probability process—that is, the dichotomous observed values. Unfortunately, the estimation theory surrounding probit analysis is not as well developed as that for linear regression analysis. The estimation appears to be quite sensitive to data problems such as correlations of the exogenous variables or errors-in-variables. Moreover, the method is quite expensive, thus inhibiting some lines of inquiry.

The Danziger/Feaster analysis has pursued such a strategy: it takes a continuous variable (income), transforms it into a discontinuous variable (poverty), and then estimates models of its incidence and the effects of different variables on poverty within a probit context. Moreover, the focus of their analysis is the antipoverty impact of governmental transfers and how this might have changed between 1978 and 1982. Perhaps for symmetry, they have measured transfers simply in terms of receipt; the magnitude of such transfers is ignored.

While their concentration on poverty status conforms to the now-standard poverty data presented by the Bureau of the Census and while this makes for a more "state of the art" article that can introduce interesting likelihood functions, it makes little analytical sense to me. The obvious alternative is to analyze the income position of families and how transfers of different amounts affect that. Such an effort might still require estimating a probit model of the probability of receiving a transfer. That, however, would be a simple probit instead of their simultaneous probit. The one concession the authors have made to the continuous nature of the problem is to include a measure of distance below the poverty line in their models of the probability of escape from poverty. Not surprisingly, this is an important predictor. They never discuss the alternative of using the natural measure of income, even though this would greatly simplify their estimation and might—although it is difficult to know a priori—affect the results of their analysis.

A related point about interpretation involves noncash transfers. While this has recently become a very heated political issue, the underlying conceptual issue is quite straightforward. Many transfers to the poor are not in the form of cash but instead come as food stamps, subsidized housing, or medical insurance. Although it is hard to tell the difference between cash and food stamps—the Federal Reserve shreds food stamps just as it does worn-out dollar bills—poverty measures take no account of noncash transfers.

In truth, this is just one part of a much larger measurement question. Presumably we want to measure the welfare of families; thus, to the extent that food stamps, housing programs, and the like contribute to their welfare, they should be counted. Conversely, to the extent that taxes or high living costs of a particular area detract from welfare, they should also be included in the analysis. To be sure, the measurement issues are large, particularly in the case of in-kind provision of items that are not tradable and are not close to cash. For example, the value of Medicare in improving the welfare of families is difficult to estimate. Some effort, however, should be made to incorporate noncash factors, since these have become an increasingly important component of our income support system. Major means-tested noncash benefit programs rose from $5 billion in 1965 to $47 billion in 1982 (in constant 1982 dollars). The aggregate programmatic changes should make it obvious why it is so important to include noncash items in any intertemporal analysis of poverty.

The desirability of including such items is perhaps even greater in a cross-sectional analysis. Two families with equal cash income are obviously not equally well off if one family pays taxes and the other does not, particularly when pretax income is the standard, as it is in poverty estimates. As social security taxes have grown and as inflation has reduced the effective beginning point of the income tax schedule, taxes have become increasingly important to the working poor. (By recent estimates, the personal-income-tax threshold for 1984 falls 17 percent above the poverty cutoff for a family of four.) Thus, neglecting taxes introduces a systematic distortion between working and nonworking poor. Similarly, there is a distortion between those in subsidized housing and those not, between those receiving Medicaid and those not, and so forth.

Much of the recent (and heated) discussion about whether to include noncash benefits in the definition of poverty has revolved around its potential effect on the perception of poverty and the transfer programs that explicitly mention poverty in eligibility rules. There is no denying, however, that these noncash items are valuable in improving individual welfare and, therefore, that they should be included to the extent possible in analyses of the low-income population.

A related issue that has received less public attention is the potential effect of geographical differences in the cost of living. As one who regularly commutes between Rochester, New York, and Washington, D.C., I can testify that prices do differ around the country. This is obviously more important for stationary items—say, houses—than for mobile ones—

say, hamburgers. (There are, correspondingly, benefits that exist in one area but not in another—for example, the lack of good Thai restaurants in Rochester). These differences surely affect true welfare, but little attention is ever given to them. While this may be a case where the perennial "lack of data" excuse comes into play, I basically believe a crude correction is better than none at all. For example, the now-defunct Bureau of Labor Statistics (BLS) budget estimates could provide a rough standardization for 1978 and 1982.

Finally, with respect to the analytical issues, there is always interest in the conventional poverty statistics. For this reason, it might be useful to provide the aggregate summary of the results in poverty terms, after doing the analysis in income or adjusted-income terms.

The reaction to the Danziger/Feaster article from a Washington and a policy perspective is somewhat clearer. The proposals to provide cost-of-living adjustments for transfer programs do not seem to lead anywhere. First, though motivated by a current poverty problem, this policy would only affect the rate of change of future poverty; inflation would not in the future lead to implicit cuts in benefit levels of transfers, and thus would not have the direct antipoverty bias of a nonindexed transfer system. Unless the policy were applied retroactively, anyone poor today would remain poor in the future as well, regardless of the inflation rate. Second, if in fact inflation is controlled—a primary goal of recent policies—the change to indexing would have no effect. Finally, cutting cost of living adjustments (COLAs) is one of the more popular deficit reduction proposals, in part because it is a way of identifying large expenditures. Therefore, while not impossible, new COLA plans seem somewhat unlikely.

Establishing national minimums for AFDC levels has some appeal in view of regional poverty statistics and locational distortions introduced by different welfare levels. Nevertheless, the case for a national minimum is not really developed, leaving one to wonder whether the argument is entirely based on regional paternalism. Obviously, it has some large budgetary implications.

Education

Murnane has put together a comprehensive picture of educational policies along with some provocative ideas for thought and policy. Everyone from the serious educational researcher to the educational decision-maker to the casual reader is bound to get something from this paper.

Having said that, I must point out the most startling feature of educational research and policy today: almost any assertion or policy suggestion is permissible. There is little requirement to present evidence to support a position; indeed, little systematic evidence is available. This lack of evidence is convenient because there is little chance of any assertion being refuted.

I point this out in part because Murnane is guilty of making unsupported assertions; in fact, he is forced to do so if he wishes to make a broad assessment in the area of education. Perhaps because we share many of the same biases, however, I find Murnane's assertions frequently plausible and always sensible. This stands somewhat in contrast to a variety of major studies and commission reports that have recently been issued. They have provided us with assertions, recommendations, and pronouncements that represent individual opinions. With a few exceptions, they do not represent new evidence about the character of the educational process; about how the various components of our educational sector interact; or about how regulations, financial incentives, and schools interact.

The general perspective that Murnane brings to analyzing school policies might be characterized as an economist's standard view of deregulation. He focuses particularly on the certification requirements imposed by all states. I believe that his arrow is aimed at a good spot. Indeed, this is perhaps a classic case of regulation and a classic example of why economists often recommend deregulation.

Our knowledge about the educational production process is quite imprecise. States have quite generally imposed certification requirements with absolutely no knowledge of what characteristics make for effective teachers or classrooms. A major effect of these certification requirements, which are typically phrased in terms of specific teacher-training courses, has been to impose constraints on entering the teaching profession. In particular, individuals who take a standard liberal-arts or science curriculum in college will not have the required education courses and thus will be ineligible for teaching. (A secondary effect has been the artificial support of teacher-training institutions that, in the absence of such requirements, might find it difficult to compete for students.) This may be an important factor in preventing high-caliber people from entering teaching. It may also have a differential effect by school subject matter (because of varying opportunity costs).

Murnane advocates ways of circumventing the teacher-training requirements. I find his notions appealing, but I think they are too limited.

A plethora of restrictions emanate from state laws, state regulations, local regulations, contract provisions and, to a lesser extent, federal laws and regulations. These restrictions have their supporters, but support is seldom based on evidence of performance in schools. There is little reason, therefore, for Murnane to draw a box around certification and to leave the larger matter out of consideration. If the argument is built entirely on some idea of political feasibility, I suspect that there is no cost in broadening the scope. All these regulations have their own constituency built on the strongest of all bonds—individual labor-market advantages created by regulations.

It is surprising that Murnane makes no mention of labor unions in this discussion. Unions are, of course, key actors in many of the discussions. They are direct participants in a majority of local contracts, where many binding regulations concerning personnel and school organizations are imbedded. They are indirect actors in state regulations, in schools of education, and so forth. The policies that Murnane suggests would very likely lead to resistance from unions, and the natural extensions would almost certainly be ones of considerable conflict with unions.

One reason educational policy is in its current (perpetual) state of chaos, I assert, is that attention has been focused on the wrong set of questions. For the most part, people have been concerned with what characteristics of teachers determine, or simply indicate, good teaching ability and—having discovered that—how these characteristics can be developed. These are appealing questions; if they could be answered, we could undoubtedly make much better educational decisions. The answers might in particular provide a basis for teacher certification requirements and other regulatory devices. However, it appears that answering these questions is beyond our current capability. This does not imply that we cannot improve matters. A more natural approach would be to reward good performance itself, instead of factors that tend to be correlated in the population with good performance. On this score, Murnane's paper is particularly interesting. He provides some insights into merit pay and what can be gleaned from past experience. This indeed seems to be a productive way to proceed. Unfortunately, we currently lack definitive studies of merit pay, so this area again comes down to listing hypotheses and the few shreds of available evidence.

Poverty and education have been the focus of consistent attention for the past twenty years. They have, since the War on Poverty, been the subject of governmental policy attention. Two things seem clear at this

time. First, both issues remain unresolved. Second, with current skepticism about the effectiveness of many governmental programs and with current pressures against the expansion of governmental outlays, new ways of looking at problems and dealing with them are called for. Both of these articles do implicitly contain some such ideas, although evidence of the effectiveness of the specific suggestions of each is scant.

Commentary

David Stern

Direct transfers in cash or kind are the most cost-effective way for governments to provide an adequate standard of living for households that lack sufficient income from their own labor, capital, or private transfer payments. But government transfers to the poor, especially in cash, are politically unpopular in this country. Danziger and Feaster found that even in 1978, when transfer payments by all levels of government had risen to 10.1 percent of the GNP—up from 5.7 percent in 1960—more than three out of four nonaged poor households were not receiving enough cash from governments to raise them above the poverty line. They found that under the Reagan administration, cuts in transfer programs reduced the antipoverty impact to the point where, in 1982, cash transfers failed to lift more than four-fifths of the nonaged poor households over the poverty line.

Giving "a hand, not a handout" is in this country a more popular strategy for helping the poor. Governments at all levels sponsor programs to increase the earning capacity of poor people or of their children. If these programs reduce the number of people whose pretransfer earnings are insufficient, then, obviously, they reduce the need for transfer payments. Examples of such programs are insurance on loans to businesses owned by members of disadvantaged minority groups, employment training for low-income people, agricultural extension services for small farmers, and educational assistance for students from poor families.

Of the various programs that affect the distribution of earning capacity, education is the largest in dollar outlays. I focus here on education as a

means of enabling more people to earn adequate incomes. I will review some evidence that policy in the past two decades has had some effect in reducing inequality of educational attainment. However, since 1980, policy has shifted away from concern for economic inequality toward an emphasis on educational "excellence." To achieve higher standards of academic performance, central and local policymakers are changing the incentive structure for teachers and students. I suggest here how new incentives can motivate efforts to improve educational performance without abandoning the goal of reducing economic inequality.

Education as an Equalizer:
The 1960s and 1970s

If putting more money into the pockets of poor people were the sole goal, education would not be an efficient means. Ribich showed in 1968 that a dollar spent on compensatory education at the elementary or secondary level produces less than a dollar increase in the present value of lifetime earnings of children from low-income households. Jencks and his colleagues in 1972 assembled a great deal of evidence to argue that the link between education and income is too weak for schooling to be useful as a means of altering the income distribution. On the other hand, economists of the human-capital school (Mincer 1974; Chiswick 1974) have demonstrated that more of the variance in earnings can be explained by a model that includes years of work experience in addition to years of schooling, and their findings suggest that schooling may be complementary to learning on the job. Even so, most of the variance in individual earnings cannot be attributed to differences in educational attainment.

Furthermore, public policy cannot control educational attainment directly. Dollars for schools must first be converted into instructional programs, which may or may not increase students' learning (see Hanushek 1981). Any gain in students' learning then may or may not lead to higher educational attainment and earnings. It would be surprising if the eventual increase in earnings produced by an additional dollar of spending on schools were as much as half a dollar in present value. This cost-ineffectiveness is simply the price we pay for trying to change the distribution of pretransfer income rather than using cash transfers directly.

Of course, cash transfers themselves produce less than a dollar of net income gain per dollar of transfer payment, because they induce recipients to spend less time in paid work. The Seattle and Denver income-maintenance experiments found that a thousand-dollar lump-sum grant would cause husbands to reduce their paid work by 47 hours a year

(Keeley et al. 1978, table 4). At the then-minimum wage of $2.00 an hour, this translates to an earnings reduction that is less than 10 percent of the thousand-dollar transfer. For wives, the earnings reduction is less than 40 percent of the transfer, and for female heads of households it is 23 percent. These reductions are not negligible, but they still leave much more than half of the transfer as a net gain in recipients' income.

The rationale for using education rather than cash grants to combat poverty is largely moral and political, not economic. Some low-income adults may be seen as "undeserving" of public aid, but children cannot be held responsible for their families' lack of money. In 1965 this kind of moral argument, combined with the self-interest of the education lobby and representatives of low-income communities, helped to produce a winning political coalition in support of federal aid for schools in low-income neighborhoods (see Timpane 1978). The result was Title I of the 1965 Elementary and Secondary Education Act. This was, and still is (as Chapter I of the 1981 Educational Consolidation and Improvement Act), the largest item in the federal budget for elementary and secondary education. It is also the biggest single source of money for compensatory education of low-income children.

The moral and political forces that produced Title I of ESEA also directed the attention and resources of states, school districts, and individual teachers toward the educational needs of low-income students during the 1960s and 1970s. Sixteen states provided categorical funds for disadvantaged students, and another eight gave extra weight to disadvantaged students in allocating general-purpose aid to local districts (Simms 1984). Many high schools created new curricular options, such as granting academic credit for work experience, to encourage attendance by students who otherwise would have been likely to drop out (Abramowitz and Tenenbaum 1978). Even within individual classrooms, there is evidence from this period that teachers tended to give extra attention to slower students (Brown and Saks 1981). An equalizing tendency was evident from the center to the periphery of the educational system.

There is also evidence that a certain degree of equalization was actually achieved. The latest, and largest, evaluation of the federal Title I program itself found evidence of definite gains by participating students, relative to other low-income students (Carter 1984). More generally, effects of efforts toward equalization at various levels are evident in data from the National Assessment of Educational Progress, which tested students at ages 9, 13, and 17. Achievement in mathematics was tested during the 1977–1978 and 1981–1982 school years; reading was tested in 1974-

1975 and 1979–1980. Changes in average achievement were more positive for blacks and Hispanics than for whites. Changes were also more positive for students whose parents lacked high school diplomas than for students whose parents had some schooling beyond high school. These differences occurred in both reading and mathematics, and in all three age groups.

Similarly, data from the Current Population Survey in 1959, 1969, and 1979 show that illiteracy in the population aged 14 and over declined more steeply for blacks than whites. Among whites aged 14 to 24, the percentage of illiterate fell from 0.5 in 1959 to 0.2 in 1979. Among blacks aged 14 to 24, illiteracy fell from 1.2 percent in 1959 to 0.2 percent in 1979.

High school dropout rates among blacks also fell dramatically: from 30.0 percent of the 14-to-34 age group in 1970 (31.2 percent of the 18-and 19-year-olds) to 17.9 percent in the 14-to-34-year-olds (19.3 percent of the 18- and 19-year-olds) in 1981. In comparison, between 1970 and 1981 white dropout rates in the 14-to-34 age group declined from 15.2 only to 12.1 percent. And among 18- and 19-year-old whites the dropout rate actually rose slightly, from 14.1 to 15.5 percent (Grant and Eiden 1980; Grant and Snyder 1983).

These numbers indicate movement toward more equal educational attainment between blacks and whites, between students whose parents had different amounts of schooling, and between students whose parents had different levels of income. This does not necessarily mean that the distribution of income among these students will be any more equal than it was among their parents, but it does mean that education was doing more to reduce the likelihood that children of poor parents would grow up to be poor themselves.

Enter "Excellence" (Again)

Schools in the 1960s and 1970s were not always champions of the poor. In many communities, new programs for low-income children were born out of struggle and acrimony (Ravitch 1982; Stern and Timar 1982). But whether schools resisted, accommodated, or actively promoted change, they were part of it.

Since 1980, however, the same political and moral sea change that allowed economic policy to produce a 10 percent unemployment rate, make the income tax less progressive, and reduce cash transfers for the poor has also affected educational policy. Federal budget cuts have re-

duced the number of economically disadvantaged children receiving compensatory services by an estimated 500,000 to 900,000 since 1981 (Simms 1984). Instead of reducing inequality, achieving "excellence" is now the top priority.

In 1982 and 1983 dozens of reports viewed with alarm the performance of American students and educators. The first and clearest of these reports was A Nation at Risk (National Commission on Excellence in Education 1983). It complained that "the ideal of academic excellence as the primary goal of schooling seems to be fading across the board in American education" (p. 14). After reviewing evidence of declining educational performance, the report listed its findings about aspects of the educational process that have contributed to the decline. First among the findings:

Secondary school curricula have been homogenized, diluted, and diffused to the point that they no longer have a central purpose. In effect, we have a cafeteria-style curriculum in which the appetizers and desserts can easily be mistaken for the main courses. . . . This curricular smorgasbord, combined with extensive student choice, explains a great deal about where we find ourselves today. . . . Twenty-five percent of the credits earned by general track high school students are in physical and health education, work experience outside the school, remedial English and mathematics, and personal service and development courses, such as training for adulthood and marriage. (pp. 18–19)

The Commission on Excellence recommended tougher standards for high school graduation, more time on academic subjects, and various measures to strengthen the teaching profession. This program, in various versions, quickly swept the country. By the fall of 1983, twenty-six states had adopted more stringent requirements for high school graduation, and the remaining twenty-four states were all actively considering proposals to do the same (Education Week , 7 December 1983).

The renewed insistence on high academic standards immediately prompted expressions of concern about abandoning the struggle against inequality. Harold Howe (1984, p. 24), formerly United States Commissioner of Education, urged policymakers "to give equity a chance in the excellence game. Unless we find a way to do this, our 'excellence' will be so flawed that it will haunt us." He charged that "a major component missing" from most of the recent commission reports "is any recognition of the importance of motivation" for students. He warned that the "recommendations for more homework, more demanding courses, longer

school days, and more tests are likely to be implemented in ways that further increase the number of dropouts." As someone else put it, "If a kid isn't clearing six feet, what's the point of raising the bar to six feet and two inches?"

The current stress on excellence is much like the response after Sputnik in 1957. Then, too, schools were charged with lack of academic rigor (see Conant 1959; Rickover 1963). Curricular reforms in science and mathematics were designed to produce more and better scientists and engineers, in the hope of restoring America's technical supremacy. Now, again, new resources are being earmarked for education in science and mathematics, to preserve the nation's technical edge against foreign competition. But these priorities and programs in the 1960s and 1970s came to be regarded as elitist. For instance, Gallup polls in 1971 and 1972 found large majorities of the general public agreeing that "the schools spend too much time in preparing students for college and not on occupations which don't require a college degree," and approving of "schools reducing the amount of classroom instruction to make greater use of educational opportunities outside the school" (Timpane et al. 1976). Work experience and other items now regarded as educational junk food were just what the public ordered ten or fifteen years ago.

Political tides may turn again. Today's educational reforms may once again be denounced as elitist and unfair to children whose home environment puts them at a disadvantage even before they enter school. American education has oscillated between these poles for at least a century (see Edson 1983). But oscillation is not progress, and changing the direction of policy entails considerable transaction costs as old programs are dismantled and new ones put in place. Are current reforms merely another swing of the pendulum, or will they produce some lasting improvement?

Incentives and Information

There do seem to be one or two new things under the sun. One is the systematic attention now being given to the incentive structure in education. Another is the development of better technology for managing information.

Neither revenues of school districts nor compensation of staff usually depends on actual performance. Possibly local taxpayers are willing to pay more for schools they think are doing a good job, but this is a clumsy mechanism for quality control. Furthermore, states now pay a larger share of the nation's school costs than localities pay themselves. Most

grants by states to school districts are allocated simply on the basis of enrollment or attendance, sometimes with a correction for a district's own ability to pay, but they do not depend on any measure of what students learn. Some state grants, along with federal grants, are categorical awards for specific purposes such as compensatory education, vocational classes, or special services for handicapped students. But categorical grants "throw money at" problems; they are paid whether results are achieved or not.

For teachers, it has long been recognized that public schools do not offer incentives like those provided for employees in a typical private enterprise. To begin with, most business firms can adjust levels of compensation to attract qualified applicants for particular jobs and can even tailor compensation packages for individuals. Once hired, individuals can be fired if they perform below standard, and if they perform above standard they may well be rewarded with additional pay (Henderson 1979). In contrast, compensation of public school teachers, almost everywhere in the United States, is governed by a salary schedule in which the only two factors that determine differences in pay are experience and training. The level of the salary schedule and the precise amount of increments for experience and training do vary from district to district and change from time to time. But local administrators do not have the power to make adjustments as they see fit. Availability of revenues from the federal and state governments, the politics of local school boards, and collective bargaining (which covers a much larger proportion of public school teachers than of employees in the private sector) all exert powerful influences on salary decisions. As a result, compensation does not serve as a flexible tool for recruitment, motivation, and retention of employees in public schools, as it does in private business.

Lack of incentives has been a problem in other parts of the public sector. The past fifteen years have seen considerable experimentation with management techniques, including new compensation schemes, to make government work better. Productivity has been improved in street repair, garbage collection, police and fire protection, hospitals, libraries, public utilities, and other public services (National Commission on Productivity and Work Quality 1975; National Center for Productivity and Quality of Working Life 1977; Greiner et al. 1977). However, among the hundreds of cases described in the literature on improving productivity in the public sector, examples of incentives for teachers have until very recently been extremely rare.

Why have public elementary and secondary schools been particularly

inhospitable to incentive plans? One obvious reason is that teachers' unions have consistently opposed any differentiated compensation except that based on experience and training. One of the unions' arguments against merit pay is that if administrators had the power to decide which teachers deserved extra pay, the decisions would sometimes be arbitrary and capricious. Underlying this objection to merit pay is the basic technical problem of deciding which teachers are more productive—a problem which would exist even if there were no teachers' unions. The problem was well summarized in a report by the Committee for Economic Development (1976) on improving productivity in government:

[P]roductivity in education is difficult to measure both because many of the goals of education are intangible and because complex outside factors come into play. A decline in reading scores does not necessarily indicate a drop in the productivity of public schools; it might also reflect a change in the composition of the school population. However, a sustained rise in reading scores does not necessarily indicate better teaching; it might result from increased enrollment of students from family backgrounds that stimulate learning. (p. 56)

It is possible—with sufficient good data—to control for students' characteristics by means of statistical regression analysis or other procedures. But the multiplicity, subtlety, and long-range nature of educational goals still create a problem. Explicitly or implicitly, performance incentives emphasize some goals more than others. This was evident in the federally sponsored demonstration of educational performance contracting in 1970–1971 (Gramlich and Koshel 1975).

Finally, even if the technical problem is somehow solved, a political problem will remain—and this, too, would arise even if teachers' unions did not exist. The problem is, if some teachers are better than others and are given higher pay in recognition of that fact, how to justify, and explain to parents, why the better teachers are assigned to some students rather than others.

Currently the problem of incentives for teachers is receiving an extraordinary amount of attention. Present shortages of qualified math and science teachers, and impending shortages of teachers in the elementary grades due to the echo of the baby boom, are one theme of current discussion. Comments by the National Commission on Excellence in Education (1983) on the desirability of rewarding good teachers, followed by President Reagan's strong statements in support of merit pay, have also generated much debate. In response to both concerns,

teachers' unions are currently reexamining their positions on differential compensation.

In 1983 Florida and California both established merit-pay programs for outstanding teachers. In both plans, a winning teacher receives a cash bonus of several thousand dollars a year for one to three years. The selection process is controlled by the state in Florida, while in California it is carried out locally. In California the "mentor teachers" also have extra responsibilities for curriculum development and supervision of new teachers.

Merit pay for individuals has the well-known drawback of encouraging individuals to try to look good at their coworkers' expense; for instance, teachers may try to have troublesome students assigned to other classrooms, or they may undermine other teachers in relations with parents or supervisors.

Conversely, merit pay for individuals does not encourage collaboration, and some collaboration among teachers can benefit students. Teachers can collaborate by coordinating curricular content across classrooms, enforcing consistent standards for student conduct in the common areas of the school (auditorium, hallways, cafeteria, school yard), communicating information to and from parents, supervising extracurricular activities, sharing ideas about individual students, and teaching each other new instructional techniques. This last kind of collaboration may be especially important for improving performance, because there is ample evidence in the research literature that certain instructional practices are relatively effective (see Bloom 1984; Levin et al. 1984) but diffusion of these practices is very slow. Incentives for teachers ought to encourage such diffusion, not impede it further.

For reasons such as these, Florida and California have moved quickly to supplement their individual merit-pay programs with "merit school" plans, which would provide incentives to teachers in groups. Florida passed the necessary legislation in 1984. In California, state superintendent Bill Honig has begun distribution of "report cards" on each public school in the state, but legislation providing money to merit schools had not passed as of December 1984.

These incentive plans hold considerable promise for motivating schools toward "excellence" while promoting equal educational opportunity. Awards to schools can be based not only on average achievement test scores but also on measures of performance by disadvantaged students: average scores in the bottom quartile, for instance, or low dropout

rates. Schools with students at lower socioeconomic levels can be given bigger bonuses for achieving the same growth in students' learning as schools with students from more prosperous families.

Such refined incentive formulas require a good deal of data on individual schools. Because computers have become relatively cheap and widespread, these data are now much more readily accessible than they were a decade ago. Most states now have, or can easily construct, machine-readable files containing current data on test scores, enrollments, and some socioeconomic characteristics of students in each school. Indeed, the recent development of statewide data bases creates a temptation for state reformers to impose incentive systems that are too centralized. Incentive formulas are likely to seem more legitimate and less artificial if they are negotiated locally (see Stern 1984). States can encourage or require local districts to establish incentive plans. Making use of an "educational management information system" (Davis et al. 1984), incentives can motivate simultaneous progress toward higher standards and less educational inequality.

In the very short run, schools may face a real tradeoff between raising standards and reducing inequality. For instance, with a given staff it may not be possible to offer both advanced and remedial instruction. Such a tradeoff may also exist along some theoretical long-run frontier where, even with the best technology, resource constraints force a choice between helping the high achievers or the disadvantaged. However, most schools are a long way from this frontier. The task of educational improvement is to move schools toward the frontier by speeding diffusion of more effective practices. Creating new incentives may do that.

In theory, some sort of voucher plan could also provide the right incentives. But a bare-fisted, free-market voucher system, with price competition and caveat emptor , is likely to produce market segmentation along geographic and socioeconomic lines (Stern and Timar 1982). If publicly supported schools are to reduce the likelihood that children of poor parents will grow up poor themselves, a voucher system would have to be of the regulated, compensatory sort designed by Jencks et al. (1970).

Finally, it is important to reiterate that education by itself cannot eliminate poverty. As argued here, the commitment to compensatory education for children of the poor does not have to be dropped in the current quest for educational excellence. Compensatory education can give these children a better shot at getting good jobs. But educational policy cannot guarantee that there will be enough good jobs to go around.

References

Abramowitz, Susan, and Ellen Tenenbaum. 1978. High School '77. Washington, D.C.: National Institute of Education.

Bloom, Benjamin. 1984. "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring." Educational Researcher 13 (6): 4-16.

Brown, Byron W., and Daniel H. Saks. 1981. "The Microeconomics of Schooling." Chapter 5 in David C. Berliner, ed., Review of Research in Education, vol. 9. Washington, D.C.: American Educational Research Association.

Carter, Launor F. 1984. "The Sustaining Effects Study of Compensatory and Elementary Education." Educational Researcher 13 (7): 4-13.

Chiswick, Barry R. 1974. Income Inequality. New York: National Bureau of Economic Research.

1974.

Committee for Economic Development. 1976. Improving Productivity in State and Local Government. New York: Committee for Economic Development.

Conant, James B. 1959. The American High School Today. New York: McGraw Hill.

Davis, Otto A., Harry R. Faulk, and Holly H. Johnston. 1984. "An Educational Management Information System" (draft). Pittsburgh: School of Urban and Public Affairs, Carnegie-Mellon University.

Edson, C. H. 1983. "Risking the Nation: Historical Dimensions on Survival and Educational Reform." Issues in Education 1 (2 & 3): 171-84.

Gramlich, Edward M., and Patricia P. Koshel. 1975. Educational Performance Contracting. Washington, D.C.: Brookings.

Grant W. Vance, and Leo J. Eiden. 1980. Digest of Education Statistics 1980. Washington, D.C.: National Center for Education Statistics.

Grant, W. Vance, and Thomas D. Snyder. 1983. Digest of Education Statistics 1983-84. Washington, D.C.: National Center for Education Statistics.

Greiner, John M., et al. 1977. Monetary Incentives and Work Standards in Five Cities. Washington, D.C.: Urban Institute.

Hanushek, Eric A. 1981. "Throwing Money at Schools." Journal of Policy Analysis and Management 1: 19-41.

Henderson, Richard I. 1979. Compensation Management. Reston, Va.: Reston Publishing.

Honig, Bill. 1984. "Setting the Course for School Reform." Education Week, 18 April.

Howe, Harold, II. 1984. "More-of-the-Sane Reform Will Not Achieve Both Excellence and Equity." Education Week, 23 May, p. 24.

Jencks, Christopher, et al. 1970. Educational Vouchers. Cambridge, Mass.: Center for the Study of Public Policy.

———. 1972. Inequlity. New York: Basic Books.

Keeley, Michael C., et al. 1978. "The Estimation of Labor Supply Models Using Experimental Data." American Economic Review 68: 873-87.

Levin, Henry M., et al. 1984. "Cost Effectiveness of Four Educational Interven-

tions." Project Report 84-All. Stanford, Calif.: Institute for Research on Educational Finance and Governance, Stanford University.

Mincer, Jacob. 1974. Schooling, Experience, and Earnings. New York: National Bureau of Economic Research.

National Center for Productivity and Quality of Working Life. 1977. Improving Government Productivity: Selected Case Studies. Washington, D.C.: NCPQWL.

National Commission on Excellence in Education. 1983. A Nation at Risk. Washington, D.C.: U.S. Government Printing Office.

National Commission on Productivity and Work Quality. 1975. Employee Incentives to Improve State and Local Government Productivity. Washington, D.C.: NCPWQ.

Ravitch, Diane. 1982. The Troubled Crusade. New York: Basic Books.

Ribich, Thomas I. 1968. Education and Poverty. Washington, D.C.: Brookings.

Rickover, Hyman G. American Education, a National Failure. New York: Dutton.

Simms, Margaret C. 1984. "The Impact of Changes in Federal Elementary and Secondary Education Policy" (draft). Washington: Urban Institute.

Stern, David. 1984. Toward a Statewide System for Public School Acountability: A Report from California. Berkeley, Calif.: Policy Analysis for California Education, University of California.

Stern, David, and Thomas Timar. 1982. "Conflict and Choice in Public Education." Education and Urban Society 14: 485-510.

Timpane, Michael. 1978. "Federal Aid to Education: Prologue and Prospects." Pp. 1-20 in Michael Timpane, ed., The Federal Interest in Financing Schooling. Cambridge, Mass.: Ballinger.

Timpane, Michael, et al. 1976. Youth Policy in Transition. Santa Monica, Calif.: RAND.

Commentary

Jennifer R. Wolch

The view of welfare and education problems and policies provided by these two papers is discouraging, because the problems are worsening and because substantial increases in public resources to solve them are not likely to be forthcoming. Sheldon Danziger and Daniel Feaster find that the extent of poverty has grown over the past half-decade, with women as a greater percentage of the poor. Largely because of social spending reductions, government cash transfers are now less effective in relieving households in poverty. Richard Murnane provides a similarly depressing picture of problems in public education and the retreat from the federal role in education that had been established during the 1960s. In my commentary I raise questions that focus on the nexus of urban structure, poverty, and educational attainment.

Education and the Cities:
Questions for Research and Policy

Murnane and others regard the shortage of adequate teachers as one of the most pressing problems for public education policy, suggesting that shortages are due to relatively low teacher salaries. In fact, the salary differentials are astounding. Clearly, governmental arrangements for education mediate this situation, as do teacher unions. Nevertheless, public education has historically been remarkably responsive to the labor requirements of the economy (Bowles and Gintis 1976; Katz 1981). This implies that if the problems in primary and secondary education were of

greater concern to the private sector, we might see more progress toward overcoming them. Is the failure to commit more public resources to education, particularly for teachers' pay, indicative of fundamental changes in the economy, and as a result, in national as well as urban and regional labor markets? And is the failure to solve the problems of scarce teacher resources in inner-city schools related to central-city labor markets that appear to be dominated by services and labor-intensive industries which require only low-skill, low-wage labor from their nonprofessional/nontechnical workforce?

As Braverman (1978), Bluestone and Harrison (1982), and others argue, recent advances in production technology have been accompanied by a reduction in the demand for highly trained labor and thus in the need for high-quality educational services. Moreover, the industrial and occupational mix in metropolitan areas has shifted in the past decade; skilled and unskilled blue-collar jobs have been lost, while service-sector employment has increased dramatically (Phillips and Vidal 1983). The slackening of demand for skilled labor is most acute in the inner city, where—apart from the downtown office sector, which employs highly skilled professional and technical workers—labor markets are characterized by growing demand for low-wage clerical workers, and for unskilled workers for service firms and small-scale, labor-intensive manufacturing companies (Scott 1982). Many of these unskilled jobs pay very poorly; clerical jobs often do not provide median family incomes for workers, and many service jobs do not guarantee sufficient income to keep families out of poverty (Phillips and Vidal 1983). Since the marginal return on educational inputs in the inner city is typically lower than in suburban and more affluent central-city neighborhoods, any cost-benefit calculus might discourage heavy investment of public resources in inner-city education.

These factors raise troubling questions about the potential of less expensive policies such as minimum high-school-graduation requirements. Unless higher scholastic requirements are justified by the promise of better jobs and/or higher incomes, students are unlikely to make the extra effort. Can today's urban economy supply the variety and numbers of employment opportunities for students who meet additional requirements, or can policy levers succeed in creating such jobs? If not, we may be making the same mistake that befell job training programs of earlier years, i.e., assuming that the demand for labor will rise to meet an added quantity and improved quality of labor. The changing character of the urban economy and the mix of job skills demanded do not encourage

these assumptions. High-school graduates may not find meaningful jobs awaiting them; instead, they could be relegated to low-wage, unskilled positions, sporadic employment, or unemployment.

We face deepening divisions of class and race and growing disparities in the quality of life offered by neighborhoods in United States cities. Would the low-cost policy options open to state and local policymakers, such as minimum high-school-performance requirements or tougher teacher-certification programs, help or hinder this situation? The former policy may serve to increase dropout rates, particularly among inner-city minority youth. The latter policy typically increases the costs of becoming a teacher, thereby decreasing the supply and increasing the interdistrict competition for qualified staff. The losers in such a competition are likely to be poor, low-paying districts, exacerbating already severe deficits of teacher resources. The implications of such policies for income inequality and the extent of urban poverty must be explored further. If we do not undertake this task, the preeminent urban service and amenity—education—is likely to suffer from ever greater geographical disparities in quantity and quality, hindering the ability of central cities to maintain middle-income population groups.

The Urban Poor:
Service-Dependent Subpopulations

Danziger and Feaster analyze the declining effectiveness of anti-poverty cash transfers. A major gap in this analysis, however, is the exclusion of in-kind transfers and services, and the change in their antipoverty effect for subgroups of the poverty population who may be differentially reliant on in-kind assistance. As a means of bridging this gap and refocusing the discussion on urban dimensions of the poverty problems, it is useful to document the changing situation of in-kind service programs for the mentally ill, physically handicapped, retarded, ex-offenders, substance abusers, and the needy elderly. Unlike those supported by cash transfers and primary in-kind programs such as Medicaid and food stamps (mostly the AFDC population), these service-dependent groups not only rely on cash transfers and the major in-kind programs but also use a host of human services: community health and mental health care, vocational rehabilitation, sheltered workshops, adult day care for the elderly or retarded, juvenile diversion programs, crisis management services, nutritional assistance, drug- and alcohol-abuse treatment, job training programs, and residential treatment services.

Estimates of the service-dependent population are substantial, whether dependency is defined by various rough measures of "need" or by indices of program utilization. A lower-bound estimate of the population is the fraction of those receiving Supplemental Security Income who also receive services through the Title XX Social Service Program. In 1979, the last year for which data are available, almost 40 percent of the SSI population received Title XX services (U.S. Department of Health and Human Services 1981). Applying this proportion to the 1982 SSI population (because of cutbacks, the proportion actually receiving services is no doubt somewhat lower) yields a service-dependent population of about 1,489,000 persons, 13.4 percent as large as the AFDC population (U.S. Department of Commerce 1984).

A large portion of the service-dependent population now living in cities was formerly institutionalized (Wolpert and Wolpert 1976). These service-dependent poor concentrated in old, deteriorated parts of inner cities typically close to downtown. On the one hand, they were attracted by cheap housing—often large dwellings converted to congregate use—and by availability of urban services first developed there under the auspices of Great Society programs. On the other hand, they were pushed away from better-off neighborhoods and suburban jurisdictions. These communities mounted opposition campaigns aimed at systematically excluding the dependent poor and those public and voluntary service agencies that support them.

Just as with the wider poverty population, the problems of these groups have always been troubling. Since deinstitutionalization got underway in earnest, community-based services have been inadequate, as have been cash aid programs (see, for example, U.S. General Accounting Office 1977). Reagan administration cutbacks of in-kind services and cash assistance have created even greater problems for service-dependent populations, especially in states such as California, where tax limitations have curbed state and local spending.

In some cities the perils of being poor have intensified not only because of funding reductions but because of the changing economic and physical environment of the city. Anecdotal evidence from several United States and Canadian cities, as well as empirical research in California, suggests that the inner-city districts where the community-based dependent populations are concentrated have become the targets for private gentrification and public urban-renewal programs, market dynamics themselves prompted by the changing nature of the urban economy—mainly the growth of downtown service activities and skilled white-collar employ-

ment. As a result, dependent populations are being displaced to even less supportive environments. Their service networks are also being dispersed, unable to maintain their vital activities due to shrinking public resources which by and large have not been replaced by private philanthropy.

San Jose:
A Case Study

A study of the service-dependent population in San Jose, California, illustrates the urban dynamics of the service-dependent poor (Wolch and Gabriel forthcoming). Following deinstitutionalization of the mentally ill, retarded, and other groups in California during the late 1960s, a ghetto of service-dependent persons developed in downtown San Jose. By the mid 1970s, an area housing 15 percent of the city's population contained over 1800 residents in 140 board-and-care or rehabilitation facilities (83 percent of the city's total); 13 alcoholic halfway houses; 11 boarding homes for the aged; 4 job corps facilities; and a home for felons. A third of the city's SSI population lived in the same district, along with social security and AFDC recipients in much higher proportions than the city average.

This pattern of ghetto development changed after the mid 1970s. Growth in high-technology and service employment in the southern Bay Area combined with limited residential development to create a serious job/housing imbalance. The result was rapid home-price inflation and revived interest on the part of prospective homebuyers in downtown San Jose. New businesses were also increasingly drawn to San Jose. Local officials sought to encourage the changing economic functions of their city's core area by pursuing an aggressive downtown redevelopment program entailing massive new urban public facilities, transportation links, and office and residential construction.

Local officials saw the presence of service-dependent populations and their facilities as a barrier to successful revitalization. This change in attitude placed added political pressure on service operators, forcing them either to relocate or to close their facilities. These actions coincided with state and federal budget reductions. Proposition 13 cost Santa Clara County 92 human service positions, $42 million in human service funding, and two key service units; the Reagan cuts led to a loss of 363 social services positions from the county, forced regulatory changes that reduced the caseload of in-home support services for the aged, blind, and disabled by 24 percent and reduced services to over half the remaining

cases, and reduced benefits to 12 percent of the county's AFDC recipients, many of whom were removed from the program. The 1982–1983 budget for social services dropped by 20 percent, general assistance funds fell by 34 percent, and alcoholism services were curtailed by 19 percent.

The results of service cutbacks, program closures, and gentrification and renewal are already visible in San Jose. Between 1979 and 1982, a dozen facilities were closed or relocated; the size of the service-dependent population housed in community care facilities downtown fell by 20 percent; the number of cash transfer recipients there fell between 1 and 10 percent; and the number of people receiving food stamps and Title XX Social Services fell 32 percent and 65 percent, respectively.

Where are these service-dependent poor going? While there is limited hard evidence, it appears that in San Jose and other urban areas, clients are increasingly misassigned to services. For instance, more community-based mental patients seem to be taken into the criminal justice system or into purely custodial medical care. And more of the dependent are homeless. According to recent estimates, for example, there are about 25,000 to 30,000 homeless people in Los Angeles, a third of whom are either physically or mentally disabled (U.S. Committee on Banking, Finance and Urban Affairs 1984). Finally, at least in California, a return of clients to large-scale institutions may be in the offing. This is indicated by initiatives to build more penitentiaries, and by Deukmejian administration proposals to reaccredit all state mental hospitals so that their service costs can be reimbursed by the federal government (via Medicaid and Medicare). This of course could dramatically reduce the fiscal incentives for state and local communities to provide community-based care, which in turn could mean sending more people to institutions.

The example of the service-dependent poor in San Jose indicates that evaluation efforts which disaggregate the poverty population may be necessary in order fully to understand the dynamics of program changes and spending reductions. Groups reliant on in-kind services as well as cash transfers appear to be distinct from other welfare populations. Furthermore, urban conditions interact with social program changes, and so both must be assessed as to their implications for the well-being of the poor.

Prospects for Social Policy

I am not optimistic about the prospects for changing expenditure priorities in this country, at least in the short run. Danziger and Feaster recom-

mend that AFDC grants be indexed and that a targeted employment program be established for welfare recipients, but both entail major federal expenditures. And while the authors may be right when they argue that further social spending reductions are unlikely because of political resistance, it is equally improbable that either cash or in-kind transfer programs for the poor will expand.

The problems of the service-dependent poor cannot be solved by indexing cash aid or by providing expanded employment programs. This group faces two sets of problems: income and service inadequacy, and residential displacement. Policy solutions to these problems will require increased spending for local and metropolitan service delivery coordination and planning, and also greater expenditures for more and better human services, neither of which is apt to be forthcoming either from states or localities or from the federal government.

In the education area, the direction of federal programming is more uncertain. As Murnane indicates, Reagan administration initiatives in education policy may be a sign of renewed federal dedication of resources for solving education problems. Alternatively, they could simply be symbolic, election-year responses to public concern about education that do not fundamentally alter the administration stance of reducing the federal role in education.

In summary, urban poverty and education are worsening. The absolute size of the poverty population (in particular, the number of women in poverty) is growing and public programs are insufficient to ameliorate this situation; dilemmas of educational reform and school resources persist with no ready solutions. The changing structure of urban economies exacerbates the situation, reducing the attractiveness of low-cost education policies such as minimum high-school-graduation requirements. It also darkens the outlook for the service-dependent poor, as employment dynamics lead to downtown gentrification and renewal and a displacement of clients and human services from the urban core. Dedication of additional public resources to overcome difficulties in education, income maintenance, and other spheres of social life is improbable given the political-economic environment of the 1980s. I suspect that these challenges for public policy will still be with us as we enter the next decade.

References

Bluestone, B., and B. Harrison. 1982. The Deindustrializating of America. New York: Basic Books.

Bowles, S., and H. Gintis. 1976. Schooling in Capitalist America: Educational Reforms and the Contradictions of Economic Life. New York: Basic Books.

Braverman, H. 1974. Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century. New York: Monthly Review Press.

Katz, M. B. 1981. "Education and Inequality: A Historical Perspective." Pp. 111-25 in D. Rothman and S. Wheeler, eds., Social History and Social Policy . New York: Academic Press.

Phillips, R. S., and A. Vidal. 1983. "The Growth and Restructuring of Metropolitan Economies." Journal of the American Planning Association 49: 291-306.

Scott, A. J. 1982. "Production System Dynamics and Metropolitan Development." Annals of the Association of American Geographers 72: 185-200.

U.S. Committee on Banking, Finance and Urban Affairs, Subcommitee on Housing and Urban Development. 1984. "Homeless in America--II." Hearing before the Subcommittee on Housing and Urban Development. Committee on Banking and urban Affairs, House of Representatives, 98th Congress, 2nd session, Jan. 25. Washington, D.C.: U.S. Government Printing Office.

U.S. Department of Commerce, Bureau of the Census. 1984. Statistical Abstract of the United States, 1984. Washington, D.C.: U.S. Government Printing Office.

U.S. Department of Health and Human Services, Office of the Secretary. 1981. Annual Report to the Congress on Title XX of the Social Security Act, Fiscal Year 1980. Washington, D.C.: U.S. Government Printing Office.

U.S. General Accounting Office, Comptroller of the United States. 1977. Returning the Mentally Disabled to the Community: Government Needs to Do More. Washington, D.C.: U.S. Government Printing Office.

Wolch, J. R., and S. A. Gabriel. Forthcoming. "Dismantling the Community-Based Human Service System." Journal of the American Planning Association.

Wolpert, J., and E. Wolpert. 1976. "The Relocation of Released Mental Hospital Patients into Residential Communities." Policy Sciences 7: 31-51.

PART THREE—
URBAN PROGRAMS: TRANSPORTATION AND HOUSING

Two of the federal domestic programs that have the greatest effect on urban areas are transportation and housing. Jose Gomez-Ibañez argues for a reduction of the federal role in transportation. Sherman Maisel sets an agenda for housing policy that includes a substantial federal role in providing aid to new home purchasers and to low-income renters. John Kain suggests that federal support for fair housing has helped to mitigate the effects of racial discrimination in housing and that a recent trend toward black suburbanization may provide a glimmer of hope for the future. In the Commentary, Michael Goldberg uses the Canadian experience to support his contention that the federal government's role in housing and urban affairs should be smaller than Maisel advocates. Theodore Keeler questions Gomez-Ibañez's conclusions on the grounds that the benefits of urban highways far outweigh the costs and that the costs of fixed rail systems are substantial. Richard Muth disagrees with both Kain and Maisel, maintaining, first, that fair housing policies are not responsible for reported declines in housing segregation and, second, that there is little need for a federal role in housing policy. Melvin Webber faults Gomez-Ibañez for concentrating on the auto-transit dualism, suggesting that the appropriate dichotomy in most modern metropolitan areas is that between large-vehicle and small-vehicle systems.

Chapter Seven—
The Federal Role in Urban Transportation

Jose A. Gomez-Ibañez

The federal role in urban transportation increased in size and scope after World War II. In the late 1950s and early 1960s, federal involvement was largely limited to grants for urban segments of the Interstate Highway System, but by the late 1960s and early 1970s it included aid for other types of urban highways and for mass transit as well. At first, federal assistance was confined to aid to capital expenditures, but it gradually broadened to include some operating and maintenance expenses. By 1980, federal aid accounted for as much as 40 percent of all government capital spending on urban highways, 80 percent of capital assistance for urban mass transit, and 30 percent of mass-transit operating subsidies.

The Reagan administration proposed to reduce federal aid for urban transportation significantly, by restricting highway aid largely to the Interstate Highway System, by phasing out operating aid for mass transit, by reducing transit capital aid, and by allocating transit capital grants by formula rather than project by project. Congress rejected most of these proposals in 1982, when it restored or substantially increased funding for many of the existing programs.

While some of the proposals made by the Reagan administration may be undesirable, the idea of forcing local governments to assume more responsibility for decisions about the level and mix of urban transportation spending has merit. The rationale for extensive federal involvement in urban transportation is weak, and several of the federal programs, particularly grants for Interstate Highway construction and mass transit capital projects, probably distort local transportation decisions. Reduced

federal aid for highways and mass transit might impose temporary financial hardships, but in the long run most urban areas would probably be better off.

Development of Federal Programs

Urban Highway Aid

Early Federal Aid

Federal highway aid began in 1913 but was largely confined to rural roads until after World War II.^[1] The first highway program apportioned federal funds among the states to cover up to one-half the construction costs of rural post roads, roads over which the mail was carried. In 1921 the federal program was restructured and federal aid was restricted to two types of highways: primary or interstate highways, and secondary or intercounty highways. Each state, in consultation with the federal government, could designate a small portion of their highway mileage as part of these federal aid systems. Highway aid was apportioned among states on the basis of a formula that considered population, land use, and rural-post-route mileage.

Until the Depression, urban highways were specifically excluded from federal aid mileage. Public works relief legislation in 1932 and 1933 provided some temporary funding for urban highway construction since many of the unemployed were city dwellers, and in 1934 the prohibition on including urban mileage in primary or secondary systems was eliminated. Not until the Federal Highway Aid Act of 1944 was specific authorization made for urban highways. This act established three distinct categories of federal-aid highways: primary and secondary rural highways (which were to receive 45 and 30 percent of all aid, respectively) and urban extensions of the primary and secondary routes (which were to receive only 25 percent of the aid).

The Postwar Construction Boom

Pressures to increase highway spending mounted after World War II as postwar prosperity fostered automobile ownership and use. During the 1940s and 1950s urban highway traffic grew at almost double the rate of the urban population, as higher incomes encouraged travelers both to increase the number of trips they made and to make more of them by automobile rather than mass transit (Table 7.1).

Federal aid for urban highways became available on a large scale in 1956, when the federal government responded to the rapid traffic growth by funding the Interstate and Defense Highway System. Under the Interstate program, federal grants paid 90 percent of the construction costs of a congressionally designated system of 42,500 miles of limited access highways. The system was initially conceived as serving intercity travel and bypassing major cities but, under pressures from urban interests, about 20 percent of the final system was located in urban areas. Urban mileage accounted for an even larger share of Interstate program expenditures, since the urban segments were often wider and more costly than their rural counterparts. Unlike the primary and secondary road programs, Interstate grants were to be distributed to the states on a project-by-project basis rather than by formula. In 1970 Congress also established a new federal-aid urban system, instructing the states to designate for the new system mileage that supplements (but does not overlap) the older urban extensions; aid for the urban system is allocated to states on the basis of urban population. Thanks largely to the Interstate program, the federal share of highway capital spending by all levels of government grew from only 10 percent in 1950 to around 40 percent in the 1960s. While in 1950 federal grants were almost exclusively for rural roads, by the 1960s and 1970s many federal highway dollars were spent in urban areas.

Disenchantment with Highways

During the 1960s many social and environmental costs of highway construction and automobile use became apparent. The first focus of concern was the social and aesthetic costs of expressway construction, particularly in urban areas. In response, Congress in 1962 required that metropolitan areas receiving federal highway aid prepare a comprehensive land use and transportation plan, which included consideration of mass transit alternatives. These regulations were substantially strengthened in 1969 by passage of the National Environmental Policy Act, which required the preparation, with public comments and hearings, of an environmental impact statement for any federally funded project. In 1973 Congress also allowed local governments to trade in the federal funds designated for locally contested segments of the Interstate System for use on mass transit improvements, as long as the United States Secretary of Transportation certified that the contested segment was not essential to the national Interstate System. By 1980, over 8000 miles of urban

Interstates had been opened to traffic, although in almost every metropolitan area controversial segments remained unfinished.

The 1960s and 1970s also saw the development of federal programs to regulate the safety, air pollution emissions, and fuel economy of new motor vehicles. In 1965 Congress established the National Highway Traffic Safety Administration, with authority to set safety standards for new cars and to order manufacturers to recall and correct defective vehicles. Air pollution was the next focus of concern, as the problems caused by automotive emissions of hydrocarbons, carbon monoxide, and nitrogen oxides became more noticeable in major urban areas. The Clean Air Act of 1970 required 90 percent rollback in the emission rates of these pollutants from new automobiles by the 1975 model year. High costs and technological uncertainties forced several extensions of the deadlines and a relaxation of the nitrogen oxide target to a 75 percent rollback, but by the 1982 model year the goals set by the Clean Air Act were achieved. Finally, the oil crisis of 1973–1974 stimulated Congress to pass a law requiring that average new-car fuel economy be improved in stages from 18 mpg in model year 1978 to 27.5 mpg in 1985. United States automakers were in compliance with the schedule as of model year 1984, although weakening gasoline prices were making it more difficult to induce customers to buy fuel-efficient models.^[2]

The level of highway construction and investment peaked during the late 1960s. Although nominal highway spending continued to increase during the 1960s and 1970s, capital outlays failed to keep pace with the rapid inflation in building costs (see Table 7.2). The decline in real spending was due in part to the heavy reliance on federal and state fuel taxes to finance highway improvements. Since fuel taxes are assessed in cents per gallon, real receipts fell with rapid inflation and the automobile fueleconomy improvements of the 1970s.

The scope of federal highway aid expanded in the 1970s, despite the spending decline. Highway maintenance became a major concern toward the end of the 1970s, as many of the highway pavements and structures built in the decades immediately following World War II began to reach the end of their useful lives. Until the 1970s, federal highway aid could only be used for new construction or for major reconstruction projects in which the road was widened or substantially rebuilt. In 1976 Congress responded to the maintenance crisis by expanding the definition of highway construction to include "resurfacing, restoration, and rehabilitation" (3R) and establishing a separate program of grants for Interstate 3R projects. In 1970 Congress also established a special aid program for

bridge repair and reconstruction on the federal aid highway systems, activities that had been exclusively state-financed.

Urban highway traffic has continued to grow despite the slowdown in highway construction, higher fuel prices, and increased regulation of new car pollution and fuel economy. Traffic growth was slightly lower in the 1970s than in earlier decades, probably largely because the growth in urban population also slowed during that decade. As in the previous decade, urban traffic increased about four times faster than the urban population during the 1970s (see Table 7.1).

Urban Mass Transit Aid

Almost as rapid as the growth in urban automobile use was the decline of urban mass transit ridership in the 1940s and 1950s, falling from a wartime high of 18 billion trips per year to 7 billion trips in 1960. In an attempt to stem the ridership loss, in the 1940s and 1950s many large metropolitan areas began to provide government aid to their mass transit systems, and several metropolitan areas planned new rail systems. The rapidly deteriorating financial condition of mass transit strained local resources, however, and by the early 1960s had led to strong pressure for federal assistance.

Federal aid began in 1964 with a small program of capital grants that would finance up to two-thirds of the cost of a capital project.^[3] Funds were allocated for specific projects at the discretion of the Secretary of Transportation. Funding for the capital grant program expanded rapidly during the 1970s, reaching levels of almost $3 billion per year, and in 1973 the maximum federal share was raised to 80 percent of costs. Federal grants were used to finance local government buyouts of ailing private bus companies, replacement or refurbishing of much of the existing transit rolling stock and equipment, construction of several new rail systems, and extensions to many of the existing rail systems.

In 1974 Congress established a new federal transit grant program to fund operating expenses, partly in response to pressure from metropolitan areas that felt they did not benefit greatly from the capital grant program, either because their rail systems predated federal aid or because they were too small to consider rail seriously. Funds from the new grant program could be used for capital as well as operating expenses, although in practice almost all was used for operating aid. The federal grants had to be matched by an equal amount of state or local aid if operating expenses were being subsidized, but the federal share could be as high as

80 percent for capital projects. In a significant departure from the original capital-grant program, the operating grants were allocated among urbanized areas according to a formula based on population and population density instead of at the discretion of the Secretary. Federal operating aid grew steadily from around $200 million in 1970 to nearly $1 billion in 1980.

Transit aid from all levels of government increased so rapidly during the 1970s that sometime in the middle of that decade government aid became a more important source of revenue to the industry than passenger fares (Table 7.3). Passenger fares had been enough to cover operating costs and make a small contribution to capital expenses through the 1950s, despite the fact that the industry was contracting. In 1964 passenger receipts fell below operating expenses for the industry as a whole and by the 1980s covered only about 40 percent of operating costs and made no contribution to capital expenses. Government aid grew from $518 million in 1970 to $7.8 billion in 1982, and the federal share of all aid grew from 26 to 53 percent, as shown in Table 7.4.

The influx of government aid was not enough to restore mass transit ridership to its former levels. The steady postwar decline in nationwide mass transit ridership was reversed between 1974 and 1980, when patronage grew by 15 percent, but in 1981 it began to fall once again. The most important urban travel market is commuting, since work trips dominate rush hour travel and are key contributors to traffic congestion and other urban transportation problems. Public transportation's share of commuting trips in all metropolitan areas fell steadily from 19 percent in 1960 to 12 percent in 1970 and 8 percent in 1980 (Table 7.1). As of 1980, mass transit captured more than 10 percent of the commuter market only in a dozen of the largest, densest, and oldest metropolitan areas, where high levels of traffic congestion and parking fees discourage auto ownership and use. Only in one metropolitan area (New York) does transit's share of the commuting market exceed 20 percent.^[4]

The performance of the first new rail systems—opened in San Francisco in 1972 and Washington in 1976—was particularly disappointing; construction costs soared far beyond original projections and the ridership gains were smaller than originally forecast. These results failed to discourage local interest in new rail systems, however, in part because federal aid was available to fund up to 80 percent of construction costs. To force local governments to consider smaller and more cost-effective alternatives, in 1976 the federal government required that cities applying for rail grants submit an analysis of non-rail alternatives, including low-

cost express bus systems. Federal regulations also restricted new rail starts in each metropolitan area to one major corridor at a time, rather than to an entire system of many lines. In 1978 these regulations were strengthened to limit federal liability in the case of construction cost overruns and to specify in more detail the types of low-cost, non-rail alternatives that had to be considered.

Policy under the Reagan Administration

By the time President Reagan took office in 1981, federal grants had increased to cover roughly 20 percent of all government highway aid and 50 percent of all transit aid (Table 7.4). The Reagan administration initially proposed dramatic changes in urban transportation policy as part of its "new federalism" initiative. The theme of the new federalism was to return responsibility for largely local programs, such as urban transportation, to state and local governments. The Reagan administra-

tion proposed both to reduce the level of federal highway aid and to restrict it to highways, such as the Interstate System, in which the federal interest was clearest. Federal gasoline taxes would be lowered (to the levels necessary to fund the smaller program) and state governments would be encouraged to increase their own fuel taxes as needed. For mass transit, the Reagan administration proposed to eliminate federal aid for operating expenses, reduce federal aid for capital projects, and allocate

capital grants on the basis of a formula rather than at the discretion of the Secretary of Transportation.^[5]

The administration achieved some of these goals in its first year in office. The FY1982 budget reduced federal transit aid below FY1981 levels and left highway aid unchanged despite inflationary pressures. The administration also imposed a moratorium on federal aid for starts of new rail mass transit systems or extensions, although funding would continue for projects already under construction.

These victories did not last long. Frustrated by the Reagan administration's moratorium on new rail starts, the House and Senate appropriating committees began to order the Urban Mass Transportation Administration (UMTA) to fund the planning, engineering, or construction of new rail systems in specific cities. The FY1982 appropriations conference report earmarked federal planning grants for six projects and construction grants for five others (two in Miami), while the FY1983 report earmarked funds for alternatives analysis of seventeen projects and engineering or construction of eleven. If all these new rail projects are built with federal aid, the cost to the federal treasury will be between $12 and $19 billion.^[6]

During 1982 pressure also grew within the administration and Congress to increase highway spending as a means to alleviate the deterioration of the nation's highways and bridges and to combat unemployment from the 1981–1982 recession. To gain support from urban areas for increased highway spending, Congress proposed to increase mass transit aid as well.

The Surface Transportation Assistance Act, passed in December 1982, greatly increased federal funding for highways and restored the cuts to mass transit. The law raised the federal gas tax from 4 to 9 cents per gallon, the first increase since 1959. Four cents of the increase is dedicated to the Federal Highway Trust Fund to finance a near-doubling of 1982 levels of highway aid (Table 7.5). The remaining cent of the new gas taxes is earmarked for a new Federal Mass Transportation Trust Fund, which will be used to finance the discretionary capital grant program. Additional transit funding is authorized out of general revenues for new operating and capital formula grant programs and total transit aid is restored to the 1981 funding levels (Table 7.6). The Reagan administration's only victories were in the composition of the spending authorized by the act. Only highway aid for Interstate and rural primary systems is increased, for example, while funding for rural secondary and urban extensions

was held constant. Transit operating assistance is also to be gradually lowered.

Evaluating Highway Aid

Benefits and Costs of Postwar Highway Expansion

Although some benefit-cost analyses have been made for isolated highway projects, there have been no recent and reliable attempts to measure the benefits and costs of the federal urban highway program as a whole.^[7] A comparison of highway user tax receipts with highway expenses suggests, however, that the benefits of the postwar boom in urban highway construction, partly financed by federal aid, have on the whole exceeded the costs.^[8]

User Payments as Benefit Measures

Motorists pay a variety of special taxes to use the highways, including gasoline taxes, tolls, and vehicle registration fees. Since motorists are willing to pay at least this much (and maybe more) for highway use, these user charges provide a minimum estimate of the benefits motorists receive from highway investments. In theory, a particular segment of the highway system might be deemed worthwhile if users pay enough in highway taxes to cover the costs that segment imposes on society.^[9]

Estimating the user component in highway taxes involves differentiating between special charges made for highway use and general taxes levied on consumption or personal property. The commonly used rule for making such separations is to regard a tax as a highway-user charge if collected from highway users or on motor vehicles but not on most other comparable goods or services. By this rule, highway-user charges clearly include federal gasoline, vehicle, spare parts, and excise taxes; state motor vehicle and driver registration fees; and special highway-user taxes on trucks. State gasoline and motor vehicle sales and property taxes are not necessarily included, since many states or localities levy sales or personal property taxes on a variety of other goods, although typically at lower rates. Estimates of United States highway-user-charge receipts derived by following this separation principle are shown in Table 7.7 in current dollars for the years 1956 through 1975.

Government Highway Expenditures

The costs of the urban highway system include both direct government highway expenditures, such as highway construction and maintenance outlays, and the externalities or social damages that highway users impose on others in the form, for example, of automobile air pollution and noise. Table 7.7 also shows estimates for 1956 to 1975 of direct highway expenditures paid by all levels of government, but excluding any external social costs of highway use. The estimates assume that all government highway capital expenses are amortized over the life of the facilities using a 5 percent real discount rate—thus, for example, the 1975 costs include charges for the depreciation of capital investments made in previous years. Capital outlays include all expenditures for the purchase of right-of-way (including relocation assistance), grading and drainage costs, construction of bridges and tunnels, and paving costs. Operating expenses include routine maintenance, snow removal, mowing, painting, and similar activities as well as highway administration, research, safety, and law enforcement. Outlays for all police assigned to traffic and highway safety duties, as well as correctional expenses attributable to enforcing traffic laws and curtailing automobile theft, are included as highway law enforcement costs.

Most highway expenses and user revenues can be attributed to particular parts of the highway system, such as urban or rural roads or Interstate and non-Interstate highways. Some costs and revenues, such as the administrative expenses of highway agencies or annual vehicle registration fees, are, however, not clearly assignable. These costs and revenues amount to about one-quarter of the total and are allocated, somewhat arbitrarily, to the different highway types according to the vehicle mileage occurring on those highways.

For the national highway system as a whole, highway user charges exceeded direct government costs by 50 to 80 percent in the 1950s and 1960s, but fell to rough parity with costs by 1975. If a more conservative 10 percent discount rate is used instead of 5 percent, revenues exceeded expenses by about 25 percent in the 1950s and 1960s but fell short of costs by about an equal amount in the 1970s.^[10] The ratio of revenues to expenses fell largely because total amortized highway outlays increased in real terms (although current capital investment has declined), while user revenues did not keep pace with inflation.

Most of the shortfalls between highway user revenues and direct gov-