Preferred Citation: Keith, Bruce E., David B. Magleby, Candice J. Nelson, Elizabeth A. Orr, Mark C. Westlye, and Raymond E. Wolfinger The Myth of the Independent Voter. Berkeley:  University of California Press,  c1992 1992. http://ark.cdlib.org/ark:/13030/ft1m3nb12x/

The Myth of the Independent Voter Bruce E. Keith, David B. Magleby, Candice J. Nelson, Elizabeth Orr, Mark C. Westlye, and Raymond E. Wolfinger UNIVERSITY OF CALIFORNIA PRESS Berkeley · Los Angeles · Oxford © 1992 The Regents of the University of California

Preferred Citation: Keith, Bruce E., David B. Magleby, Candice J. Nelson, Elizabeth A. Orr, Mark C. Westlye, and Raymond E. Wolfinger The Myth of the Independent Voter. Berkeley:  University of California Press,  c1992 1992. http://ark.cdlib.org/ark:/13030/ft1m3nb12x/

To the University of California
at Berkeley

Preface

Thirty-five years ago a cartoon by Bill Mauldin had a man explaining his outlook on politics. "Me, I vote the man, not the party. Hoover, Landon, Dewey, Eisenhower. . . ." Most Americans resemble that man: they claim to vote for the best candidate, but their choices have a certain uniformity. Such predictable voting patterns reflect durable individual attachments to a party. Although these inclinations have long been recognized, their measurement has been possible only since the advent of survey research. A Democrat or Republican is someone who picks that identity when given a chance to do so by an interviewer. Almost all others call themselves Independents.

Most Americans who tell an interviewer that they are Independents will, if asked, concede that they are closer to the Democratic or Republican party. What should be made of this additional piece of information? For decades most academic specialists evidently concluded that the correct answer was "Nothing." The concession of closeness to a party was widely ignored and everyone who claimed to be an Independent was left in that one category, assumed to consist of people who were without partisan feelings.

These apparently neutral citizens did not attract much attention when they amounted to little more than a fifth of the electorate. But it was quite a different matter when, beginning in the late 1960s, their share of the voting-age population almost doubled to the point where nearly two out of five Americans seemed uncommitted to either major party. This

was judged an event of the profoundest importance, perhaps the most significant domestic political development since the Vietnam War.

This book results from our discovery in the fall of 1974 that the key to understanding Independents is the hitherto ignored second question about closeness to a party. Independents who say they are closer to a party are very similar to outright identifiers with that party. Genuine neutrals amount to little more than a third of all those who first claim to be Independents. Our finding contradicted two groups of scholars—those architects of the first national surveys whose published writings treated all Independents as similar and the later revisionists whose thesis that the public was adrift without familiar reference points depended on the assumption that all Independents were alike in their neutrality.

As might be expected of a project that spans three decades, we have been helped by many people. Our earliest ventures into data analysis were aided by Margaret Baker and Ilona Einowski of the University of California, Berkeley. Steven J. Rosenstone and John R. Zaller gave us sage advice at the outset, as did Henry E. Brady many years later. We are grateful to Steven Rosenstone and Edward R. Tufte for comments on earlier drafts of parts of this book, and to Warren E. Miller for advice in the final stages. Our greatest debt is to Laura Stoker, whose meticulous, wise, and learned comments on our manuscript went well beyond the call of collegiality.

We gratefully acknowledge the contributions of our talented research assistants, including Martin I. Gilens, James M. Glaser, Michael G. Hagen, Benjamin Highton, and Jonathan S. Krasno of the University of California at Berkeley and Debbie Owen Bell, Jamie Cook, Mark R. Devey, Harlan Hatfield, Earl Marshall, Robert C. Miller, George I. Monsivais, Kristin Sandberg, Rodney J. Symes, Kathleen E. Tait, and Erin A. Wilson at Brigham Young University. Preparation of the various drafts of the manuscript was accomplished with speed and good cheer by Lyn Lake at the University of California and Lani Gurr and Lisa Miller at Brigham Young University.

We appreciate the patient professionalism displayed by Valeurie Friedman, Amy Klatzkin, and Naomi Schneider of the University of California Press.

Financial support was provided by the College of Family, Home, and Social Sciences, the Department of Political Science, and the Young Scholar Award program, all at Brigham Young University.

This book is based on data from surveys by the Institute of Social Research of the University of Michigan, conducted since 1978 by Warren E. Miller and the National Election Studies, supported by long-term grants from the National Science Foundation. The data were made available by the Inter-university Consortium for Political and Social Research through UC DATA (Data Archive and Technical Assistance) of the University of California and the Social Science Data Archive of Brigham Young University. Neither the collector of the data nor the Consortium is responsible for our analyses or interpretations.

1—
Partisanship and Independence

Few contemporary domestic political themes have attracted more attention than the growing number of Americans who choose to call themselves Independents rather than Democrats or Republicans. As one textbook observed, "None of the other trends . . . can match the decline in partisanship and party vitality with respect to the sheer number of words written."^[1] The decline of partisanship has been a rare point of consensus among commentators in all parts of the political arena. Scholars and journalists, liberals and conservatives, Democrats and Republicans—all agree on this point. The conservative publicist Kevin Phillips wrote in 1982 that "the most enduring phenomenon of the years since the war in Vietnam and Watergate has been the rise in the ranks of independents."^[2] The authors of The Changing American Voter , judged the most distinguished contribution to political science in 1976, concluded that the "most dramatic political change in the American public over the past two decades has been the

[1] David B. Hill and Norman R. Luttbeg, Trends in American Electoral Behavior , 2d ed. (Itasca, Ill.: F. E. Peacock, 1983), 32.

[2] Kevin P. Phillips, Post-Conservative America (New York: Random House, 1982), 227.

decline of partisanship."^[3] This view was shared a dozen years later by the New York Times reporter Hedrick Smith, who wrote that "the most important phenomenon of American politics in the past quarter century has been the rise of independent voters who have at times outnumbered Republicans."^[4]

As we will see, not everyone shares this view. However, it remains the preferred interpretation of at least some specialists in voting behavior and a great many students of political parties. The decline of the parties is a favorite theme in American government textbooks, the repository of our discipline's conventional wisdom and the main source of undergraduates' understanding of the political system. Most writers outside political science—historians, authors of big-picture interpretations of the United States, journalists—seem to take for granted the proposition that many fewer American voters are now affected by party ties.

The implications attributed to this putative growth in independence are numerous and momentous. Independents are unconstrained by partisanship from responding to election-year appeals. If more and more voters are becoming Independents, there could be wider and wider swings between the parties. Presumably this is what leads some observers to write that the increase in the number of Independents portends greater political instability. Other anticipated consequences include weaker presidential mandates, richer opportunities for third parties, partisan realignment, a more fragmented Congress, and the end of the current party system. We will examine these predictions in more detail shortly.

The implications for social science are scarcely less important. Party identification is the "foundation" of the "edifice" of conceptualizing and measuring voting associated with the National Election Studies (NES) conducted since 1952 by the

[3] Norman H. Nie, Sidney Verba, and John R. Petrocik, The Changing American Voter , enlarged ed. (Cambridge: Harvard University Press, 1979), 47.

[4] Hedrick Smith, The Power Game: How Washington Works (New York: Random House, 1988), 671.

University of Michigan's Center for Political Studies.^[5] For a generation, political scientists had believed that party identification powerfully affected voting decisions and perceptions of political events; it was "the central thread connecting the citizen and the political process."^[6] Assumed to be a point of stability in individual behavior, party identification was the baseline against which researchers measured the impact of short-term forces like issues, candidates, and assessments of governmental performance.^[7]

In its modern meaning, "party identification" is a product of survey research. Democrats, Republicans, and Independents are identified by asking respondents to put themselves in one of these three categories. The portentous trend from partisanship to independence reflects an increase in the proportion of respondents who tell interviewers that they consider themselves not Republicans or Democrats but Independents. In the Michigan NES, the share of Independents went from 19 to 23 percent in the 1952–1964 period to a high of 37

[5] Donald R. Kinder and David O. Sears, "Public Opinion and Political Action," in Handbook of Social Psychology , vol. 2, 3d ed., ed. Gardner Lindzey and Elliot Aronson (New York: Random House, 1985), 684.

The Michigan election research has had various labels since 1952. In the early years it was identified with the university's Survey Research Center and then took the name of the Center for Political Studies (CPS) when that unit was formed in the 1960s. Although still conducted by the CPS, the research has been called the National Election Studies (NES) since the creation in 1977 of a national board of overseers consequent to the first long-term multimillion-dollar grant to the CPS from the National Science Foundation.

The Michigan Survey Research Center conducted a national sample survey of sorts in 1948. For a variety of reasons, this study has had little visibility or influence, so our chronology of "the Michigan school" begins with the 1952 study.

[6] Nie, Verba, and Petrocik, The Changing American Voter , 73.

[7] One well-known example is the central place of party identification in calculating the "normal vote," a method used to identify groups that respond in different ways in a given election. See Philip E. Converse, "The Concept of a Normal Vote," in Angus Campbell et al., Elections and the Political Order (New York: John Wiley and Sons, 1966), 9–39.

percent in 1978. In 1990, 36 percent of the NES sample called themselves Independents.

Any NES respondent who claimed to be an Independent has always been asked if he or she was closer to one or the other party. Until we began our research, analysts of the NES data—the major source of information about individual political behavior—seldom paid attention to this follow-up question. Independents were all those people who claimed this status when answering the initial question about party identification. This book results from our discovery more than fifteen years ago that the second question is essential to an understanding of Independents because it enables analysts to distinguish genuine Independents from those who initially claim this status but then concede that they lean toward the Democrats or Republicans. Thus we can dispel the assumption, once almost universal and still widespread, that all Independents—"the second largest group in the electorate"—share some characteristics that differentiate them in important ways from Republicans and Democrats.^[8]This assumption is wrong. Independents, defined inclusively, have little in common. They are more diverse than either Republicans or Democrats. Most of them are not uncommitted, and they are not a bloc. They are largely closet Democrats and Republicans .

Therefore, none of the large generalizations about Independents is correct because as Independents are usually defined, the category includes not one but three kinds of people. Most of the increase in Independents has occurred among the hidden partisans; the high-level speculations apply only to genuine Independents, whose increase has been rather modest.

The gist of our findings about Independents began to circulate among some political scientists as early as the summer of 1975 and was reported at the end of that year in a textbook

[8] Gerald M. Pomper et al., The Election of 1976 (New York: David McKay, 1977), 73. This estimate is more modest than that of Nie, Verba, and Petrocik, who call Independents "the largest group in the society" (The Changing American Voter , 346).

coauthored by one of us.^[9] Our first full statement was in a paper delivered in the summer of 1977.^[10] Many of our early findings could have been found one place or another in the literature. Some were mentioned as an aside, an unremarked table, or a paragraph ignored in subsequent chapters. Despite some published clues and an unimaginable number of similar findings in computer printouts, our case against the monolithic conception of Independents was very much a minority view in 1977. Since then, our view has been accepted by some, considered and rejected by others, and (apparently) ignored by others. These developments are discussed at length in chapter 5.

Although we disagree with the revisionist scholars who think that party identification lost much of its importance in the 1970s, our findings are also incompatible with the earlier conventional wisdom associated with researchers at the University of Michigan. The latter difference is less conspicuous because Independents did not occupy such a prominent place in their work. It should be emphasized, however, that none of our findings describes a recent development. With one exception, everything that we say about different types of Independents was as true in the 1950s as it is today. The only difference is the lower level of interest and participation by the genuine Independents, those who deny that they are closer to one party or the other.

Parties in Disrepute

A jaundiced view of political parties is part of our most distinguished intellectual heritage. Parties were, after all, among the most unattractive of those "factions" that the Founding

[9] Raymond E. Wolfinger, Martin Shapiro, and Fred I. Greenstein, Dynamics of American Politics (Englewood Cliffs, N.J.: Prentice-Hall, 1976), 133, 151.

[10] Keith et al., "The Myth of the Independent Voter," paper delivered at the 1977 annual meeting of the American Political Science Association.

Fathers so roundly deplored.^[11] Long after parties had become an essential component of the American political system, they were still regarded with suspicion by acute and influential observers. For example, Lord Bryce's model citizen

will give close and constant attention to public affairs, recognizing that this is his interest as well as his duty. He will try to comprehend the main issues of policy, bringing to them an independent and impartial mind, which thinks first not of his own, but of the general interest. . . . If, owing to inevitable differences of opinion as to what are the measures needed for the general welfare, parties become inevitable, he will join one, and attend its meetings, but will repress the impulses of party spirit. Never failing to come to the polls, he will vote for his party candidate only if satisfied by his capacity and honesty.^[12]

Bryce was not alone in urging Americans to keep their distance from parties if they had the astuteness and strength of character to do so. This was the outlook in an 1891 high school civics text:

As on the playground, some do not care always to go with the crowd, or even prefer to be by themselves. Such as these, who think for themselves, and dare to stand alone, make the Independents in politics.

They are likely to prefer the good of their country to the success of their party. They will not act with their party, or will leave it, if it is wrong. If the other party changes, as parties sometimes change, and advocates measures that they believe in; if they change their own minds as sensible men sometimes must; or if the other party puts forward better candidates; or if a new party

[11] See Austin Ranney, Curing the Mischiefs of Faction: Party Reform in America (Berkeley: University of California Press, 1975), 22–57.

[12] James Bryce, Modern Democracies , vol. 1 (New York: Macmillan, 1929), 47–48.

arises, the independent voters are willing to act wherever they believe they can best secure the public welfare.^[13]

In short, partisanship had its place, but it was no substitute for the voter's independence of thought and action.

Other students of politics thought that it was up to Independents to rescue the parties:

If the politicians must look after the parties, there should be somebody to look after the politicians, somebody to ask disagreeable questions and to utter uncomfortable truths; somebody to make sure, if possible, before election, not only what but whom the candidate, if elected, is going to represent. . . . The old parties are not to be reformed from within. It is from without that the attempt must be made, and it is the Independents who must make it. If the attempt should fail, the failure of the experiment of democracy would inevitably follow.^[14]

More recently, serious students of politics came to prize parties as an essential part of the polity:

The political parties created democracy and . . . modern democracy is unthinkable save in terms of the parties. . . . The parties are not therefore merely appendages of modern government; they are in the center of it and play a determinative and creative role in it.^[15]

Few contemporary political scientists would challenge this view, but most Americans share the Founding Fathers' distaste. In the fall of 1974, in the immediate aftermath of Watergate—a scandal to which neither party contributed—72 percent of the American public said that the parties were

[13] Charles F. Dole, The New American Citizen (Boston: D. C. Heath, 1891), 127.

[14] James Russell Lowell, "The Place of Independents in Politics," in J. R. Lowell, Political Essays (Boston: Houghton, Mifflin, 1888), 318.

[15] E. E. Schattschneider, Party Government (New York: Farrar & Rinehart, 1942), 1.

the part of the government they "least often trust to do what's right." A mere 13 percent mentioned the president in this connection.^[16]

The fundamental explanation for the parties' unpopularity seems to be a belief that they stir up conflict. Jack Dennis, the leading student of the parties' public image, reported that nearly two-thirds of those interviewed agreed that "the political parties more often than not create conflicts where none really exists." Fifty-three percent thought that "our system of government would work a lot more efficiently if we could get rid of conflicts between the parties altogether."^[17] In 1980, 56 percent of all NES respondents agreed that "the parties do more to confuse the issues than to provide a clear choice on issues," and just 24 percent dissented from this verdict. Half the sample went so far as to agree that "it would be better if, in all elections, we put no party labels on the ballot."^[18]

In view of such sentiments, it is not surprising that the vast majority of Americans deny that a candidate's party is important and insist that they vote for the best candidate, irrespective of party. A month before the 1986 elections, less than 10 percent of registered voters said that the biggest factor in their vote decision would be the candidate's political party.^[19]

If party affiliation really influenced the votes of so few people, it would not be the keystone of most efforts in the past thirty-five years to understand voting behavior. We will see that despite their expressions of distaste for the parties, most

[16] Unattributed statements about voting and public opinion are based on our analysis of National Election Studies (NES) data. The multi-year tables throughout this book are based when possible on the NES 1952–1988 Cumulative Data File.

[17] Jack Dennis, "Support for the Party System by the Mass Public," American Political Science Review 60 (September 1966): 605. See also Dennis, "Public Support for the Party System, 1964–1984," paper delivered at the 1986 annual meeting of the American Political Science Association.

[18] This item has not been asked of NES respondents since 1980.

[19] CBS News–New York Times poll release dated October 6, 1986. Nearly three-quarters of the votes cast in the 1986 House elections were along party lines.

Americans are powerfully affected by their affiliations with the Democratic or Republican party.

The Michigan View of Party Identification

Affiliation with a party can be defined in various ways: legal voter registration, formal "card-carrying" membership, or personal identification. The first of these alternatives is a function of state laws, which vary enormously. In some states there is no registration by party; in others, membership is achieved simply by claiming to be a Democrat or Republican when voting in that party's primary. This diversity makes it impossible to use legal registration as the basis of a nationally uniform definition of party membership. The drawback of the second definition, formal membership in an organization, is its rarity; less than 4 percent of all Americans belong to a party in this sense. The third and most useful manifestation of partisanship is party identification, a subjective identification as a Republican or Democrat.

Reporting their analysis of nationwide surveys in 1952 and 1956, four social scientists at the University of Michigan described party identification as the most stable and important determinant of individual voting decisions and the point of departure for analysis of many aspects of public opinion:

Few factors are of greater importance for our national elections than the lasting attachment of tens of millions of Americans to one of the parties. These loyalties establish a basic division of electoral strength within which the competition of particular campaigns takes place. And they are an important factor in assuring the stability of the party system itself. . . . Most Americans have this sense of attachment with one party or the other. And for the individual who does, the strength and direction of party identification are facts of central importance in accounting for attitude and behavior.^[20]

[20] Angus Campbell et al., The American Voter (New York: John Wiley and Sons, 1960), 121.

The Michigan researchers were not the first to assert that many Americans have durable commitments to a political party. This idea had long been a staple of political commentary and was an important element in the work of such influential scholars as V.O. Key, Jr.^[21] But they were the first visible academic survey researchers to identify Republicans and Democrats in other than election-specific terms. Earlier researchers had assigned respondents to a party on the basis of their voting intentions in the impending presidential election.^[22] The Michigan contribution in this respect was an operational measure of partisan commitment that was consistent with the previous understanding of this concept, as represented in the work of scholars like Key.

If party identification were identical to vote choice, it would be an unnecessary term; if it were unrelated to vote choice, it would be without explanatory value. The crux of the concept is that it is antecedent to, distinct from, and influential of individual voting decisions. By conceiving of party identification as an affective attachment, the Michigan scholars made it possible to understand electoral decisions as the interplay of historical forces and such contemporary factors as candidate appeal, disputes about policies, and evaluations of governmental performance.

Few analysts dispute the brilliant simplification involved in the initial Michigan approach, that voting is affected both by long-term predispositions and by short-term cues. Predispositions continue to be much the same from election to election and can be invoked to account for the

[21] An especially interesting example is V.O. Key, Jr., and Frank Munger, "Social Determinism and Electoral Decision," in American Voting Behavior , ed. Eugene Burdick and Arthur J. Brodbeck (New York: Free Press, 1959).

[22] Bernard R. Berelson, Paul F. Lazarsfeld, and William N. McPhee, Voting (Chicago: University of Chicago Press, 1954), 25–26. The Gallup Poll began asking about party identification as early as 1937. See Everett Carll Ladd, "Party Identification: The Idea and Its Measure," The Public Perspective 2 (May/June 1991): 17.

substantial numbers of electors who always vote for the same party.^[23]

An "affective orientation to an important group-object,"^[24] party identification was an enduring aspect of the individual's political outlook. It imparted stability, a certain degree of predictability, to elections. Committed partisans—most of the public—were less receptive to candidates appealing to election-year passions. This was nicely illustrated by the pattern of support in 1968 for the third-party presidential candidacy of George C. Wallace, which was based largely on Wallace's hostility to civil-rights laws. These measures were most popular among the young, but so was Wallace, who "captured less than 3 percent of the vote among people over seventy outside the South, but 13 percent of those under thirty, with a regular gradient connecting these two extremes."^[25] The explanation for this apparent inconsistency was the strong relationship between age and strength of party attachment. Younger voters, less anchored by party identification, were more vulnerable to Wallace's appeal.

One can easily understand why party identification was not only "the key concept of U.S. electoral research"^[26] but also of great interest to a broader scholarly constituency:

In the course of a mere two decades, party identification has become as pervasive a concept as power, authority, legitimacy, stability, or any other element in the professional political scientist's vocabulary. . . . It offers a

[23] Ian Budge, Ivor Crewe, and Dennis Farlie, "Introduction: Party Identification and Beyond," in Party Identification and Beyond , ed. Budge, Crewe, and Farlie (New York: John Wiley and Sons, 1976), 18.

[24] Campbell et al., The American Voter , 121.

[25] Philip E. Converse et al., "Continuity and Change in American Politics: Parties and Issues in the 1968 Election," American Political Science Review 63 (December 1969): 1103.

[26] Michael B. MacKuen, Robert S. Erikson, and James A. Stimson, "Macropartisanship," American Political Science Review 83 (December 1989): 1125.

new explanation of democratic stability and legitimacy, which in turn confer authority and power. The more electors are attached by enduring psychological links to political parties, the argument runs, the more the polity is insured against . . . sudden demagogic incursions.^[27]

Measuring Party Identification

For more than a generation, social scientists have been identifying Democrats, Republicans, and Independents by asking respondents to apply one of these labels to themselves. The NES uses this question:^[28]

Generally speaking, do you usually think of yourself as a Republican, a Democrat, an Independent, or what?

If the respondent answers "Republican" or "Democrat," the interviewer follows up with

Would you call yourself a strong Republican [Democrat] or a not very strong Republican [Democrat]?

If the respondent answers "Independent," the interviewer probes:

[27] Budge, Crewe, and Farlie, "Introduction," in Party Identification , 3.

[28] The NES questions are often used in other academic surveys and by media and political pollsters. The wording ("Generally speaking, do you usually . . .") was designed to encourage respondents to look beyond the immediate present. This contrasts with the Gallup organization's question, which emphasizes the here and now: "In politics, as of today, do you consider yourself a Republican, a Democrat, or an Independent?" Compared to the Gallup measure, responses to the NES question vary much less over time and are considerably less responsive to economic conditions and assessments of presidential performance. See Paul R. Abramson and Charles W. Ostrom, Jr., "Macropartisanship: An Empirical Reassessment," American Political Science Review 85 (March 1991): 138–47.

Gallup seldom uses a probe that permits identification of partisan Independents. For a thoroughgoing review of responses to Gallup questions on party affiliation, see Everett Carll Ladd, Jr., "Declarations of Independents," Public Opinion , April–May 1984, 21–32.

Do you think of yourself as closer to the Republican party or to the Democratic party?

Nearly two-thirds of those who initially label themselves Independents concede that they are closer to one or the other party. These are the "leaners," "partisan Independents," or "Independent Republicans" and "Independent Democrats." "Pure Independents" are those who continue to deny any partisan inclinations when asked about closeness to a party. Together with the Strong and Weak ("not so strong") Republicans and Democrats, the NES questions provide seven categories of party identification:^[29]

Strong Democrat
Weak Democrat
Independent Democrat
Pure Independent
Independent Republican
Weak Republican
Strong Republican

The full time series from 1952 through 1990, with all seven party identification categories, is in table 1.1.

One obvious decision in any data analysis using the Michigan party identification questions is how to treat respondents who call themselves Independents. Should their initial label be honored? This would create an inclusive category comparable in size to either party's adherents. Or should attention be paid to their response to the probe about closeness to a party? This would produce a small Pure Independent category, less than one respondent in six.

This problem was not mentioned in The American Voter 's discussion of measuring party identification, except perhaps for the observation that "by treating Independents as a single

[29] Once in a while someone identifies with a minor party and a handful of people manifest no interest in parties or politics. These apolitical respondents are discussed in chapters 2 and 9 and will be excluded from data analyses after chapter 2.

group we may reduce seven categories to five."^[30] The emphasis was on party identification as "a continuum of partisanship extending from strongly Republican to strongly Democratic." ^[31] In his article "The Concept of a Normal Vote," Philip E. Converse said that the seven categories of party identification "are often collapsed, as in this article, to five or three classes, in response to needs for greater numbers per class, or under certain circumstances to assure monotonicity " (emphasis added).^[32]

The authors of The American Voter opted to aggregate all Independents for purposes of analysis, setting a precedent that was almost universally observed for nearly twenty years.^[33] Independents were the midpoint on a continuum running from Strong Democrats and Weak Democrats to Weak and

[30] Campbell et al., The American Voter , 123.

[31] Ibid., 122–23.

[32] Converse, "The Concept of a Normal Vote," 20n.

[33] The exceptions to this generalization in the three seminal books by the scholars who conducted the first Michigan studies are sparse enough to be enumerated in one note. The first book from Michigan contained several tables based on the seven-way classification but did not acknowledge in the text that Democratic leaners were more loyal to the Democratic presidential candidate than Weak Democrats or the same pattern on the Republican side. See Angus Campbell, Gerald Gurin, and Warren E. Miller, The Voter Decides (Evanston, Ill.: Row, Peterson, 1954), 101, 108–9.

This evident lack of interest in differentiating leaners and Pure Independents continued in The American Voter , which presented a great many data analyses that combined all varieties of Independents and just seven tables where they were not combined (pp. 124, 125, 126, 134, 148, 201, 390). Only the tables on pp. 125, 126, and 148 bear on the controversy about how to classify leaners. All three tables show leaners less willing than weak partisans to report a personal history of party regularity. (We discuss this substantive topic and the validity of such retrospective data in chapter 5.)

The single seven-way table in Elections and the Political Order (p. 218) concerns one of the few variables that always has a monotonic relationship to strength of party identification: respondents' reports about the partisan consistency of their past voting choices. The same finding appears on p. 125 of The American Voter . Another table in Elections and the Political Order (p. 197) included leaners with strong and weak partisans in a three-way distribution.

Strong Republicans. Once all three kinds of Independents were in one category, strength of partisanship was positively associated with turnout and other measures of civic virtue, including interest in the campaign and concern about its outcome. The relationship was U-shaped, with Independents at a low point between the partisans of each party. By the same token, the relationship of this five-point measure to vote choice was monotonic. Strong Democrats were most likely to vote for their party's candidate, Weak Democrats less so, then Independents, and so on to the Strong Republicans at the opposite end of the party identification scale.

In sharp contrast to the pre-survey-research view of Independents, The American Voter found them ignorant, apathetic, and inactive.^[34] Other than impugning their civic virtue, The American Voter had little to say about Independents. They were the midpoint on a continuum; their beliefs and behavior were interesting chiefly as a basis of comparison with people of varying degrees of partisan affiliation.

This picture of partisans and Independents became the unchallenged conventional wisdom until the late 1960s when large increases in the number of Independents raised new questions. Although this development led some scholars to characterize Independents in new ways, for some years no one challenged the five-category and three-category scales of party identification that treated Independents as a single group.

The Rise of Independents and the "Decline of the Parties"

In the next few pages we will examine some of the implications that observers of the American political scene have found in the growing number of Independents. Many of these fairly portentous conclusions are associated with explicit rejections of a view of individual political behavior that their authors attribute to The American Voter . But both these revi-

[34] Campbell et al., The American Voter , 143–44.

sionists and the objects of their criticism generally shared the analytic mistake that gave rise to this book: They combined all three kinds of Independents in analyzing data, or they assumed that this combination was appropriate.

We begin with the simple three-way distribution of Democrats, Republicans, and Independents that is commonly used to demonstrate the trend. As fig. 1.1 shows, Independents out-numbered Republicans as early as 1966, became more than a third of the sample in 1972, and first came within two percentage points of the Democrats in 1974. By the end of the 1980s Independents and Democrats had equal proportions of the sample.

The decline of the political parties has of course been a staple item in descriptions of the American political scene for most of this century. In the words of one believer, it is "a frequent refrain among experts in American political parties."^[35] The growth in Independents has become an important exhibit in discussion of this putative trend.^[36] "While claiming to be an Independent does not necessarily signify a positive commitment to 'Independence,' as a principle, it does reveal a lack of commitment to either of the major parties."^[37]

The implications of fewer committed adherents are explicated in this passage from H. G. Nicholas's recent interpretation of American politics:

Indeed one of the most conspicuous features of the contemporary electorate is the growing proportion—

[35] Howard L. Reiter, Parties and Elections in Corporate America (New York: St. Martin's Press, 1987), 36.

[36] See Leon D. Epstein, Political Parties in the American Mold (Madison: University of Wisconsin Press, 1986), 4, 240; Nicholas Henry, Governing at the Grassroots , 3d ed. (Englewood Cliffs, N.J.: Prentice-Hall, 1987), 111–12; James L. Gibson et al., "Whither the Local Parties? A Cross-Sectional and Longitudinal Analysis of the Strength of Party Organization," American Journal of Political Science 29 (February 1985): 139–40.

[37] Paul R. Abramson, John H. Aldrich, and David W. Rohde, Change and Continuity in the 1980 Elections , rev. ed. (Washington, D.C.: CQ Press, 1983), 240.

Figure 1.1
Democrats, Republicans, and Independents in the United States, 1952–1990

upwards of 30 percent—which has developed an indifference to party and describes itself as 'Independent.' As such it is unavailable for the day-to-day business of politics, the canvassing, the drumming up of interest in the minor offices, the attendance at meetings, the raising of funds—in short, the maintenance of the structure of participatory politics.^[38]

Some writers think that the trend toward more Independents portends not just weaker parties but the end of the party system.^[39] One respected scholar has suggested that "if the decline of partisan politics continues, . . . the loss will be that of democracy."^[40]

If the proportion of Americans who could be affected by party affiliation was shrinking, it seemed likely that another consequence of the trend would be greater electoral volatility:

Independents operate under no such constraints [as party loyalty]. By definition, their vote (and, for that matter, whether they choose to vote at all) is not anchored by party ties. . . . It is an unstable vote (and turnout) that is basically unpredictable over time, and introduces into elections an increased volatility that today's fluid politics do not need. . . . The independent vote is up for grabs. It has no allegiances. Its volatility and malleability does [sic ] little to ease the concern of those who value stability and order in American politics.^[41]

One casualty of this increased volatility might be effective government. In the opinion of Everett Carll Ladd, Jr., "The

[38] H. G. Nicholas, The Nature of American Politics (Oxford: Oxford University Press, 1986), 57.

[39] Walter Dean Burnham, Critical Elections and the Mainsprings of American Politics (New York: W. W. Norton, 1970), 131–34.

[40] Gerald M. Pomper, "The Decline of the Party in American Elections," Political Science Quarterly 92 (Spring 1977): 41.

[41] William J. Crotty, American Parties in Decline , 2d ed. (Boston: Little, Brown, 1984), 37. For the same diagnosis of "much wider swings from one election to the next" from a scholar whose political perspective differs from Crotty's, see John H. Kessel, Presidential Campaign Politics , 2d ed. (Homewood, Ill.: Dorsey Press, 1984), 279.

fact that a very large proportion of the electorate has been cut loose from relatively stable party ties and has been left free to float necessarily compounds the fragility of the mandate that a modern president enjoys."^[42] Hedrick Smith draws a similar conclusion: "This free-floating independence of voters has encouraged the individualism of a new breed of politicians. Since the party brand name lacks the old punch, some candidates run away from their party label when it suits them." Smith goes on to assert that this new breed of politicians with "highly independent campaign styles" is in turn less responsible to party leaders in Congress.^[43]

Some writers go beyond characterizing Independents as volatile and see them as something of a bloc.^[44] The authors of one textbook observe that "in a strong sense, independents are one of the two major parties."^[45] Well-known specialists in voting behavior say that "the large portion of the electorate, mostly young, who have no party ties remains available to

[42] Everett Carll Ladd, Jr., Where Have All the Voters Gone? 2d ed. (New York: W. W. Norton, 1982), 77.

[43] Smith, The Power Game , 685, 686.

[44] A list of other scholars who combine the three types of Independents for analysis would include Robert C. Luskin, John P. McIver, and Edward G. Carmines, "Issues and the Transmission of Partisanship," American Journal of Political Science 33 (May 1989): 444; Carole J. Uhlaner, "Rational Turnout: The Neglected Role of Groups," American Journal of Political Science 33 (May 1989): 417; Helmut Norpoth, "Under Way and Here to Stay: Party Realignment in the 1980s?" Public Opinion Quarterly 51 (Fall 1987): 385–86; Lee Sigelman et al., "Voting and Nonvoting: A Multi-Election Perspective," American Journal of Political Science 29 (November 1985): 757; Charles H. Franklin, "Issue Preferences, Socialization, and the Evolution of Party Identification," American Journal of Political Science 28 (August 1984): 469; and Stanley Feldman and Pamela Johnston Conover, "Candidates, Issues, and Voters: The Role of Inference in Political Perception," Journal of Politics 45 (November 1983): 824.

[45] Kenneth S. Sherrill and David J. Vogler, Power, Policy and Participation (New York: Harper and Row, 1982), 317. Sherrill and Vogler also think that "we may be witnessing the disarray and deterioration of both parties" (emphasis in original) (p. 282).

attach itself to a new party coalition"^[46] Others think the beneficiaries of this big new bloc of potential supporters might not be limited to the existing two parties. Bruce A. Campbell wrote that one result of the increase in Independents "is the advantage that large numbers of Independents present to third parties. . . . The feeling lingers that sooner or later those Independents will find a cause around which they can rally."^[47] A few years later, Abramson, Aldrich, and Rohde took the same line: "The great hope of third parties is the large percentage of the electorate (roughly one-third) who claim to have no party ties."^[48]

The trend toward independence is believed by some political scientists to be the first step toward partisan realignment, that is, one or more changes in the profile of identifiers with the two parties. At the least, more Independents make such a reshuffling of party affiliations more feasible: "Since the mid-1960s, the uncommitted share of the electorate has increased throughout the nation, thus raising the possibilities for realignment."^[49] By the mid-1980s, a historically significant realignment appeared imminent to William Schneider:

The Democrats are no longer the nation's normal majority party. That was the big political news of 1984. The big question for 1985 is whether the Republicans can take their place. One-third of all Americans call themselves Republicans, one-third call themselves Democrats and

[46] Kenneth Prewitt, Sidney Verba, and Robert H. Salisbury, Introduction to American Government , 5th ed. (New York: Harper and Row, 1987), 359.

[47] Bruce A. Campbell, The American Electorate (New York: Holt, Rinehart and Winston, 1979), 268.

[48] Abramson, Aldrich, and Rohde, Change and Continuity , 237.

[49] Paul Allen Beck, "Realignment Begins? The Republican Surge in Florida," paper delivered at the 1981 annual meeting of the American Political Science Association, 14; see also Martin P. Wattenberg, "The Hollow Realignment: Partisan Change in a Candidate-Centered Era," Public Opinion Quarterly 51 (Spring 1987): 58–74.

the remaining third are Independents. Is this the long-awaited realignment of American politics?^[50]

A more popular interpretation of the trend, however, or a more popular label for it, is "dealignment":

The present alignment is accompanied, indeed distinguished, by the continued weakening of voter loyalties to political parties in general. I erred four years ago when I argued that the question was whether a realignment or a dealignment was occurring. The larger change in voter alignments we are experiencing includes a greater reluctance on the part of many voters to express any underlying party preference, and an increased willingness to leave the candidate of their professed party with little hesitancy. In general party indentification is a more casual matter than it ever before has been.^[51]

The dealignment thesis, which is based on an assumed diminution of the importance of partisanship in voting, is often asserted without support.^[52] Advocates of the thesis are among the best-known students of public opinion and voting in America. William Schneider has called it "a major change in American politics."^[53] Paul Allen Beck sees the evidence of dealignment as "undeniable."^[54] Even Paul Kirk, former chairman of

[50] William Schneider, "Demos No Longer the Majority Party," Los Angeles Times , December 30, 1984, pt. IV, 1–2.

[51] Everett Carll Ladd, Jr., "As the Realignment Turns: A Drama in Many Acts," Public Opinion , December 1984–January 1985, 6.

[52] James E. Campbell and Joe A. Sumners, "Presidential Coattails in Senate Elections," American Political Science Review 84 (June 1990): 513, 520, 521; Edward G. Carmines, John P. McIver, and James A. Stimson, "Unrealized Partisanship: A Theory of Dealignment," Journal of Politics 49 (May 1987): 376–400.

[53] William Schneider, "Antipartisanship in America," in Parties and Democracy in Britain and America , ed. Vernon Bogdanor (New York: Praeger, 1984), 100.

[54] Paul Allen Beck, "Incomplete Realignment: The Reagan Legacy for Parties and Elections," in The Reagan Legacy: Promise and Performance , ed. Charles O. Jones (Chatham, N.J.: Chatham House, 1988), 165.

the Democratic National Committee, has claimed that "realignment is a myth but so-called dealignment is a reality."^[55]

An Overview of the Book

Political observers have not been bashful about discerning significance in the apparent diminished importance of party identification. In a few years, Independents were transformed from the forgotten respondents in The American Voter into the central characters in a series of melodramatic scenarios about contemporary American politics: electoral volatility, the decline of the parties, the loss of governmental authority, the end of the current party system, the birth of third parties, realignment, and dealignment. Any or all of these notions may be vindicated by events, but it is important to understand that as of 1991, trends in responses to the NES questions about party identification have not provided much supporting evidence. This is true because researchers who combine all three kinds of Independents when analyzing these data are making a crucial mistake. Most respondents who answer the first NES question by calling themselves Independents are no such thing; as the follow-up question lets us discover, they are covert partisans. Therefore, virtually no generalizations about broadly defined "Independents" are correct. Since the interpretations cited in the previous section are based on this mistaken operational definition of Independents, they are wrong. What is true of the revisionist work of the 1970s and 1980s applies equally to the earlier conventional wisdom based on the Michigan trilogy.^[56]

Some early work about the growth of Independents con-

[55] Paul Kirk at the conference "Picking the President: Is There a Better Way?" sponsored by the Public Policy Program and the Department of Government, College of William and Mary, Williamsburg, Virginia, November 10, 1989.

[56] Campbell, Gurin, and Miller, The Voter Decides; Campbell et al., The American Voter; Campbell et al., Elections and the Political Order .

cluded that it reflected in large measure the Southern shift away from the Democratic party that began in response to the passage of the Civil Rights Act of 1964 and Barry Goldwater's explicit appeal to the South in the election that fall. At the same time, blacks seemed to be moving in the opposite direction from other Americans, becoming more, not less, identified with a party. In chapter 2 we examine regional and racial trends in party identification to ascertain the extent to which blacks and white Southerners differed from the rest of the country.

In the era before survey research, independence was the mark of the ideal citizen. As we have seen, this picture was shattered by The American Voter . One part of the revisionist argument challenged this new conventional wisdom, arguing that there were now "two kinds of Independents," one of which was characterized by civic virtue. Chapter 3 shows that neither of these formulations is satisfactory. There are indeed two kinds of Independents: leaners and Pure Independents.

Chapter 4 describes the voting choices of Independents in presidential and congressional elections. It shows that leaners vote very much like the outright partisans of the parties toward which they incline. This finding persuaded many observers that leaners were indeed partisans. Other scholars, however, developed arguments that explained how leaners could vote as they did and still be considered Independents. Chapter 5 deals with this round of responses to our earlier work by examining leaners' participation in presidential primaries, the durability of their party identification, and their views of the two parties.

In chapter 6 we examine the distribution of party identification among people of different ages and educational attainment to test various propositions about the appeal of the parties to different generations and educational groups.

Other, related explanations for the growth of Independents concern the divisive political issues that confronted Americans from 1966 to 1972: Vietnam and race relations. These are the topics of chapter 7. One of the presumed consequences of

this pair of issues was widespread disaffection with existing political institutions. In chapter 8, therefore, we examine relations between this phenomenon and party identification.

In chapter 9 we consider two other attempts to explain Independents: that they dislike parties as institutions and that party identification is really a multidimensional phenomenon that was not adequately measured by the Michigan questions. Finally, in chapter 10 we summarize what we have found and examine the implications for diagnoses of the contemporary American political scene.

2—
Partisan Shifts among Blacks and Southerners

This chapter has two purposes. First, we show that the increase in Independents was confined to the white population. Overt partisanship spread among blacks while it declined in the white majority. For this reason, most of our analysis in subsequent chapters excludes blacks.

Second, we test the argument that changes in the Southern electorate account for a disproportionately large share of the swelling population of Independents. If this were the case, the South should be discussed separately because any analysis of dealignment using national samples would attribute to the entire country changes that characterized only one region, thus underestimating the regional trend and overestimating the effect everywhere else. We show that the trend toward independence, while evident in the South, is a national phenomenon.

Party Identification among Blacks

For most of American history, black partisanship was of little consequence because most blacks were disfranchised. In the North, where blacks were permitted to vote, their numbers

were modest. In the South, home to most blacks until recently, the end of Reconstruction in the 1870s was followed by a series of legal and extralegal measures that kept them from voting in all but a few places.^[1] Intimidation dampened other forms of political expression; many blacks doubtless concluded that it was prudent to treat politics as white people's business.

In 1940 a mere 3 percent of Southern blacks were registered to vote. A combination of legal pressure and political mobilization caused a gradual increase in registration, to 20 percent in 1952 and 29 percent in 1960.^[2] Over a quarter of Southern blacks interviewed by the National Election Studies in 1952, 1956, and 1960 were classified as apolitical; when asked about their party identification, they denied any affective involvement in politics, even as Independents. Less than 3 percent of Northern blacks and 2 percent of all whites were apolitical in these three elections. Table 2.1 presents the party identification of Northern and Southern blacks for the period 1952–1990^[3]

After the Civil War, most blacks identified with the party whose leader freed the slaves.^[4] This historical legacy was strongest in the South: "From Reconstruction until Franklin D. Roosevelt, most southern Negroes, insofar as they had parti-

[1] See, for example, Charles V. Hamilton, The Bench and the Ballot (New York: Oxford University Press, 1973).

[2] Gerald David Jaynes and Robin M. Williams, Jr., eds., A Common Destiny: Blacks and American Society , Committee on the Status of Black Americans, National Research Council (Washington, D.C.: National Academy Press, 1989), 233. White registration was far higher, particularly in the Deep South, the last stronghold of the old tradition that politics was reserved for white people.

[3] Rather than the clumsy locution nonblack , we will use the word white to describe all respondents who are not black, including Asian Americans, American Indians, persons of Hispanic heritage, and other racial minorities. These groups accounted for a substantially larger part of the sample in 1988 than in 1952, but none of them was numerous enough to permit valid analysis.

[4] James L. Sundquist, Dynamics of the Party System , rev. ed. (Washington, D.C.: Brookings Institution, 1983), 219, 347–48.

san inclinations, were habitually Republican in the tradition fixed when they had a taste of political power. Eventually Negroes became the only Republicans."^[5] The shift toward the Democratic party began with the New Deal. Government intervention on behalf of the deprived moved millions of blacks from the party of Lincoln to the party of Franklin Roosevelt^[6] As early as 1937, 44 percent of blacks identified with the Democratic party^[7] By 1952, one of every two blacks was a Strong or Weak Democrat; about one in eight was a Republican.

The climax of the civil rights movement in the mid-1960s accelerated these trends toward full electoral participation and affiliation with the Democratic party. The Voting Rights Act of 1965 provided equal access to the ballot box in the South.^[8] Wholesale repression of black political activity soon became little more than a memory.

The events of 1964 had already modified the two parties' diffuse and ambiguous images on racial issues.^[9] The momentous Civil Rights Act enacted in June of that year was proposed by a Democratic president, passed by a Democratic Congress, and signed into law by another Democratic president. Republican congressmen had been crucial to its enactment, but their hopes of sharing in the credit were lost when their presidential candidate, Senator Barry Goldwater, cast a well-publicized negative vote and then campaigned against the law that fall. Not until 1964 did the Democratic party have the unambiguous image of support for civil rights that has characterized it since then. Nineteen percent of all blacks

[5] V. O. Key, Jr., Southern Politics in State and Nation (New York: Alfred Knopf, 1950), 286.

[6] Nancy J. Weiss, Farewell to the Party of Lincoln: Black Politics in the Age of FDR (Princeton, N.J.: Princeton University Press, 1983).

[7] Everett Carll Ladd, Jr., and Charles D. Hadley, Transformations of the American Party System (New York: W. W. Norton, 1975), 60, 159.

[8] Abigail M. Thernstrom, Whose Votes Count ? (Cambridge: Harvard University Press, 1987), chap. 1.

[9] Angus Campbell, "The Meaning of the Election," in The National Elections of 1964 , ed. Milton C. Cummings, Jr. (Washington: Brookings Institution, 1966), 267–68, and "Civil Rights and the Vote for President," Psychology Today , February 1968, 29–30.

had been Strong and Weak Republicans in 1956, as had 16 percent in 1960. This dropped to 7 percent in 1964 and has been below that level since then.

The salience of civil rights in the 1960s and the polarization of the parties' images on those issues raised and clarified for blacks the stakes of electoral conflict. So did their full enfranchisement in the South and the cessation of restraints on their political activity there, which held out the promise of real political power. An obvious consequence of these developments was a much higher level of black political mobilization. One feature of this mobilization was a sharp decline in the number of apolitical Southern blacks—from 25 percent in 1960 to 8 percent in 1964 and then to the same negligible level of Northern blacks and whites everywhere.

A second aspect of black mobilization was an increase in overt identification with one of the two parties; that is, in the proportion of black respondents who said "Democrat" or "Republican" in response to the first question in the NES sequence. Such acknowledged partisans accounted for anywhere from 60 to 69 percent of all blacks from 1952 to 1960. This proportion jumped from 60 percent in 1960 to 81 percent in 1964 and then to a high of 87 percent in 1968. Since then, it has averaged over 70 percent of black respondents.

In 1960, when 27 percent of blacks were Independents, another 14 percent professed no political interest at all. Apolitical black respondents became scarce after 1960. The number of black Independents also declined. Indeed, as fig. 2.1 shows, the 1968 election witnessed the first sharp increase in Independents among whites and another drop among blacks. The rate of black independence more than doubled in 1972, from 10 to 23 percent, but even in that year there were half again as many white Independents.

The clearest conclusion from the trends depicted in fig. 2.1 is that from 1964 onward, blacks have been less inclined than whites to call themselves Independents, usually by a substantial margin. Together with the findings about their increasing willingness to identify with a party, this leads us to conclude

Figure 2.1
Independents by Race, 1952–1988

that for the most part they have not shared in the dealignment that is supposed to have begun in the late 1960s. Because blacks have not contributed in full measure to the trend we are examining, we will exclude them from further consideration. Unless we specifically indicate otherwise, no data reported henceforth include black respondents.^[10]

[10] Because blacks are the most disaffected of any major population group, omitting them also avoids complications if one examines relationships between alienation and independence, as we do in chapter 8. For a similar decision, see Paul R. Abramson, "Generational Change and the Decline of Party Identification in America: 1952–1974," American Political Science Review 70 (June 1975): 469–78.

The South

The direct and immediate consequences of the civil-rights revolution were felt principally in the South: enfranchisement of blacks; the end of Jim Crow laws; legal prohibition of segregated lodging, eating, and entertainment facilities; and great strides toward racial equality. Many Southerners had thought that their familiar way of life required second-class status for blacks. Another part of that way of life had been loyalty to the Democratic party, but the fateful legislation had been signed by a Democratic president. Although Barry Gold-water had forthrightly opposed those measures, the next Republican presidential candidate, Richard M. Nixon, had always been a racial moderate.^[11] Hubert H. Humphrey, the Democratic candidate in 1968, was more conspicuously identified with civil rights than any other party leader. This left an opening for a third-party contender: Governor George C. Wallace of Alabama, the country's most famous segregationist. The candidate of the American Independent party, he won 34 percent of all Southern votes and was particularly successful with whites who disliked both Nixon's and Humphrey's stands on civil rights.^[12]

In 1971, E. M. Schreiber attributed the nationwide increase in Independents largely to changes in the South.^[13] This interpretation had some momentary validity, as can be seen in fig. 2.2, which displays the incidence of independence among white Northerners and Southerners. From 1952 through 1964, the number of Northern Independents was essentially level,

[11] Stephen E. Ambrose, Nixon: The Education of a Politician, 1913–1962 (New York: Simon and Schuster, 1987), 412–14, 433–36, 538, 609.

[12] Steven J. Rosenstone, Roy L. Behr, and Edward H. Lazarus, Third Parties in America (Princeton, N.J.: Princeton University Press, 1984), 112–14, 163–64.

[13] E. M. Schreiber, "'Where the Ducks Are': Southern Strategy versus Fourth Party," Public Opinion Quarterly 35 (Summer 1971): 157–67.

Figure 2.2
White Independents in the North and South, 1952–1988

around a quarter of the sample, and then rose five percentage points from 1964 to 1968. Among white Southerners, on the other hand, Independents increased 17 percentage points from 1964 to 1968. In this last four-year period, white Southerners accounted for more than half the total national increase in Independents.^[14]

Between 1968 and 1980, Independents declined in the South and gained elsewhere. Perhaps the Southern high-water mark in 1968 reflected the special circumstances of Wallace's appeal as the candidate of a party that was labeled

[14] We define the South as the eleven former Confederate states: Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia.

Independent.^[15] In any event, from 1972 on, Independents accounted for just over one in three white Southerners and for roughly the same proportion of white Northerners. In 1988, 39 percent of Southern whites were Independents; Northern whites were 37 percent Independent. Because the rise in Independents was a national phenomenon, in no way limited to the South, we include that region in our analysis.

The Democratic party's loss of its monopoly in the South is one of the most significant developments in recent American political history. By the 1980s, Republican presidential candidates fared better in the South than in any other major region. The trend in congressional elections was similar if less dramatic. In the House of Representatives, for example, Southern Republicans held six seats in 1960 and 38 in 1991. Some political scientists attributed these gains not to a broad shift toward Republican affiliation—realignment—but merely to weakening Democratic identification—dealignment. This conclusion was based on the undeniable fact that outright Republican identifiers (Weak and Strong Republicans) were only slightly more numerous among white Southerners in 1972–1988 than in 1960: 21 percent in the earlier year, 24 percent in 1988. Including the leaners as Independents led to such conclusions as these by Paul Allen Beck, author of perhaps the most widely cited study of Southern dealignment:

Gains in identification with the Republican party in the last two decades have been meager indeed, and the few which appeared through 1960 have been reversed since. . . . GOP prospects for the future seem dim. . . . In short, the future seems bright for neither party. Too large a share of the southern electorate has repudiated party ties. The politics of dealignment are upon the South and should remain dominant for quite some time. . . . Parti-

[15] Like other "parties" that are created as vehicles for the presidential campaign of a politician disappointed by his prospects in one of the major parties, the American Independent party faded into obscurity after 1968.

sanship will play a restricted role in the future, just as it has in the past few years.^[16]

On the other hand, researchers who classified Independent Republicans as partisans reported a steady increase in Republican identification that led them to conclude that the region was indeed undergoing a realignment.^[17] If leaners are combined with outright identifiers with each party, we find that by 1988, 45 percent of all white Southern voters were Republicans, compared to 50 percent who were Democrats.

At this point we anticipate what is to come in this book by reporting that from 1952 through 1984, 92 percent of Southern white Independent Republicans voted for the Republican presidential candidate. In 1988, Southern Independent Republicans gave George Bush 97 percent of their votes. What is more,

[16] Paul Allen Beck, "Partisan Dealignment in the Postwar South," American Political Science Review 71 (June 1977): 495. Two other authors who also treated leaners as Independents concluded that the "Republican success in attracting votes [in the South] has been accompanied by failure to secure large numbers of conversions to the party." See Edward G. Carmines and James A. Stimson, "Issue Evolution, Population Replacement, and Normal Partisan Change," American Political Science Review 75 (March 1981): 114.

[17] See, for example, Raymond E. Wolfinger and Robert B. Arseneau, "Partisan Change in the South, 1952–1976," in Political Parties: Development and Decay , ed. Louis Maisel and Joseph Cooper (Beverly Hills, Calif.: Sage, 1978). Although reluctant to accept that leaners are truly partisan and therefore attracted to the "dealignment" label, Earl and Merle Black present data showing the importance of Independent Republicans to partisan changes in the region; see their Politics and Society in the South (Cambridge: Harvard University Press, 1987), 239–40.

An article asking how researchers using the same data base could come to such conflicting diagnoses concluded that the answer lay in different choices about "data management" and then had nothing to say about the alternative treatments of Independents that fully explain the conflicting interpretations. See Charles D. Hadley, "Survey Research and Southern Politics: The Implications of Data Management," Public Opinion Quarterly 45 (Fall 1981): 393–401. For a thorough survey of the literature on partisan change in the South, see Harold W. Stanley, "Southern Partisan Changes: Dealignment, Realignment or Both?" Journal of Politics 50 (February 1988): 64–88.

a growing proportion of votes in the South are cast along party lines. Using our definition of Independents and partisans, Harold Stanley concluded that the evidence pointed to "realignment rather than dealignment insofar as realignment covers an altered balance of partisan identifiers, identifiers remaining loyal in voting, and an increased perception of important, general party differences."^[18]

Summary

We have seen that party identification strengthened among blacks at the same time that more whites were becoming Independents. For this reason, our analyses of Independents will exclude blacks. The growth in Independents among whites was a nationwide trend, not a phenomenon whose understanding requires special attention to white Southerners. Finally, our analysis suggests that Republican gains in the South reflect a broad shift toward identification with that party, not merely a weakening of Democratic ties. This conclusion rests on our major theme: The leaners are partisans, not neutrals in the contest between the parties. Chapters 4 and 5 present our case for this conclusion.

[18] Stanley, "Southern Partisan Changes," 85.

3—
The Civic Virtue of Partisans and Independents

As we have seen, since the beginning of the Republic, few Americans have been strong admirers of political parties. The rise of corrupt and unresponsive machines after the Civil War provided more reasons for preferring political action that was "prosecuted without the necessarily onerous alliances with political parties."^[1] In response to the parties' evident moral squalor, attempts were made to turn them into more satisfactory instruments for political participation. Some thought that this was a futile strategy because the parties were irredeemable. This outlook led to the nonpartisan movement, which generated attempts, with varying degrees of success, to circumvent, subvert, and minimize the parties.

People devoted to nonpartisanship found it easy to believe that independent voters were superior citizens who did not need the crutch of party to make their voting decisions. Howard Penniman nicely summed up this view of party identification: "Faith in the independent voter was thus closely linked to an opposition of intellectuals to political machines—and, indeed, to parties as such. . . . Since the thoughtless ones were

[1] M. I. Ostrogorski, Democracy and the Organization of Political Parties (London: Macmillan, 1902), 440.

the supporters of the corrupt party machines, then almost by definition the thinking members of society had to become independents."^[2]

This belief in the superior civic virtue of Independents was one of the first casualties of survey research, which yielded evidence that Independents were the least likely of any citizens to be interested, informed, and active. The widely accepted epitaph to this ideal was written in The American Voter:

Far from being more attentive, interested, and informed, Independents tend as a group to be somewhat less involved in politics. They have somewhat poorer knowledge of the issues, their image of the candidates is fainter, their interest in the campaign is less, their concern over the outcome is relatively slight. . . . What is more, a further division of the Independent group would show in each case that those who refused to say they were closer to one party or the other are even less involved than other Independents.^[3]

The addendum about "a further division of the Independent group" was not elaborated, and few readers seemed to heed it. The harsh verdict on Independents became the new orthodoxy, an application of the generalization that "the stronger the individual's sense of attachment to one of the parties, the greater his psychological involvement in political affairs."^[4]

At the beginning of the 1970s, Walter Dean Burnham challenged this conventional wisdom. Burnham noted that some socioeconomic groups with relatively high proportions of Independents were thought to be among the more active and informed parts of the population:

A new breed of independent seems to be emerging as well—a person with a better-than-average education,

[2] Howard Penniman, The American Political Process (Princeton, N.J.: Van Nostrand, 1962), 38.

[3] Angus Campbell et al., The American Voter (New York: John Wiley and Sons, 1960), 143–44.

[4] Ibid., 143.

making a better-than-average income in a better-than-average occupation, and, very possibly, a person whose political cognitions and awareness keep him from making identifications with either old party.^[5]

It may be entirely likely that there are at least two sets of independents: "old independents" who correspond to the rather bleak classical survey-research picture, and "new independents" who may have declined to identify with either major party not because they are relatively politically unconscious, but because the structure of electoral politics at the present time turns upon parties, issues, and symbolisms which do not have much meaning in terms of their political values or cognitions.^[6]

Burnham conceded that this portrait did not include young people and Southerners, who were particularly likely to be Independents.^[7] Nor did he actually contrast the civic virtue of Independents and partisans. Nevertheless, his revision of the Michigan position gained considerable acceptance during the 1970s.

Morris Janowitz came to a conclusion very much like Burnham's: "The 'independents' have changed their character and include a high concentration of persons with strong political involvements."^[8]

A few years later, Gerald M. Pomper presented another, more empirically grounded defense of Independents. One should not expect them to be particularly concerned or informed about elections and campaigns, he argued, because they have refused to choose sides about these aspects of politics. A more reasonable citizenship test for Independents

[5] Walter Dean Burnham, Critical Elections and the Mainsprings of American Politics (New York: W. W. Norton, 1970), 130.

[6] Ibid., 127.

[7] Ibid., 123.

[8] Morris Janowitz, "Mass Media: Institutional Trends and Their Consequences," in Reader in Public Opinion and Mass Communication , ed. Morris Janowitz and Paul Hirsch (New York: Free Press, 1981), 318.

would focus on topics other than elections fought between Democrats and Republicans. Pomper found that while broadly defined Independents were less informed about the contemporary election than Weak Democrats and Republicans, they surpassed them on questions about public affairs outside the electoral arena.^[9]

The conclusions of Burnham, Pomper, and the Michigan scholars all suffer from the same mistake: combining Pure and partisan Independents. Whether the civic virtue of Independents is high, low, or average is a matter of definition. Partisan Independents always have been relatively interested, informed, and active, and Pure Independents have been notably uninterested, ignorant, and inactive. As we will see, however, Pomper's distinction between partisan and other kinds of politics is a useful one.

Interest in Politics

We begin with involvement in activities other than campaigns. Since the 1960s, the Michigan researchers have asked about interest in "what's going on in government and public affairs most of the time, whether there's an election going on or not." Examining the responses across all seven party-identification categories introduces the finding that will recur in this chapter: Pure Independents are at the bottom and partisan Independents at or near the top on measures of political involvement. Table 3.1 shows these data for presidential election years from 1964 to 1988. Pure Independents usually expressed the least interest in politics. Independent Democrats were invariably more interested than Weak Democrats and equaled or surpassed Strong Democrats in five of the seven years. The same is true for Republicans. Independent Republicans were more interested than Weak Republicans in all

[9] Gerald M. Pomper, Voter's Choice (New York: Harper and Row, 1975), 32–33.

seven years and more interested than Strong Democrats in four of the seven years.

The same pattern appears with respect to two other noncampaign activities—writing letters to editors and to public officials. Pure Independents were the least likely correspondents, followed by weak partisans. Strong partisans and partisan Independents wrote most frequently.

Thus the only Independents out of touch with politics are those who deny any inclination toward a party; that is, about a third of those usually classified as Independents. The partisan Independents are among the most politically involved of all respondents.

What about the argument that partisan Independents, al-

though interested in public affairs, do not care too much about narrow partisan matters like campaigns and elections? They are less interested in campaigns than strong partisans but considerably more so than weak partisans. Table 3.2 shows the proportions of each of the seven categories who said they were "very much interested" in the current campaign in every presidential election year from 1952 through 1988. Over this period, Pure Independents usually have been the least interested group. In 1988, for instance, only 14 percent were very interested in the campaign, compared to 24 percent of Independent Republicans and 33 percent of Independent Democrats. Independent Democrats have been at least as interested as Weak Democrats; in seven of the ten election years, they were more interested. The same relationship is found on the Republican side; the mean proportion of Independent Republicans highly interested in the campaign is seven percentage points greater than that of Weak Republicans.

Concern about the election outcome is distributed the same way. Pure Independents were most likely to say they did not "care very much which party wins." Independents who leaned toward either party were generally as concerned as weak partisans and a bit less than strong partisans. At each level of identification, Republicans usually expressed more concern than Democrats.

All these findings suggest that the distinction between campaign and noncampaign politics has no validity for Pure Independents but does reflect variations in the involvement of partisan Independents compared to outright partisans. This might suggest that for partisan Independents, general political interest is weightier than concern about campaigns. Among strong partisans, a larger proportion cares about the election results than displays an interest in politics generally. On the other hand, a relatively higher percentage of partisan Independents are interested in politics than care about the outcome of the party struggle. For outright identifiers, therefore, the campaign may contribute a larger share of the stimuli to their political interest. For partisan Independents, it may be the

other way around: general political involvement is what pricks their interest in party politics.

Information

Our findings about levels of political information resemble those about interest in the campaign. Pure Independents are consistently the most ignorant of all Americans. Strong partisans are a bit better informed than leaning Independents, and weak partisans are in third place. These relationships are illustrated by the data in table 3.3, on the extent to which each of the seven groups knew which party had a majority in the House of Representatives before the election in eight presiden-

tial election years. In 1988, for example, only 47 percent of Pure Independents knew that the Democrats controlled Congress, compared to 63 percent of Independent Democrats and 64 percent of Independent Republicans. In 1984 and 1988, only Strong Republicans clearly surpassed the two groups of leaners on this question, and Pure Independents were the least knowledgeable group. The same pattern of differences between leaners and Pure Independents is found with respect to respondents' ability to recall the names of candidates for the House. And in 1976 partisan Independents were more likely to know if there had been a presidential primary in their state.

One might think that partisan Independents resemble Burnham's informed, interested "new independents" and that "old independents" is another label for the sluggish Pure Independents. In fact, the civic virtue of partisan Independents did not blossom in the fevered political climate of the late 1960s. They were just as virtuous during the 1950s and did not subsequently become relatively more informed or interested. The relative levels of concern and information for all party-identification categories have been fairly consistent since the 1950s. The American Voter's characterization of all Independents as uninformed about politics and uninterested in the campaign was true for Pure Independents then and now, but it was never an accurate portrait of partisan Independents.

Our findings about sharp distinctions among the different kinds of Independents also explain data that seemingly puzzled other analysts of voting behavior. Examining relationships between "political sophistication" and party identification, Philip Converse thought it "most intriguing" that

among the most highly sophisticated, those who consider themselves "independents" outnumber those who consider themselves "strong" partisans, despite the fact that the most vigorous political activity, much of it partisan, is carried on by people falling in this cell. If one moves diagonally toward the center of the matrix, this

balance is immediately redressed and redressed very sharply, with strong partisans far outnumbering independents. In general, there is a slight tendency (the most sophisticated cell excepted) for strength of party loyalty to decline as one moves diagonally across the table, and the most "independent" cell is that in the lower right-hand corner.^[10]

Converse thought this finding an anomaly. In light of our discussion, it is not surprising to find Independents who are sophisticated about politics—partisan Independents.

Turnout

The most important test of political involvement is voting. The conventional wisdom on this subject was summarized by John F. Bibby:

As the strength of commitment to a party increases, so does the likelihood that a person will turn out and vote. Strong partisans, therefore, have higher rates of turnout than weak partisans, who in turn are more likely to vote than independents.^[11]

The evidence for this generalization is a cross-tabulation of turnout and party identification in which all three varieties of Independents are combined. If one compares turnout among all seven party-identification categories, however, it is obvious that turnout is not monotonically related to strength of party identification.

Table 3.4 displays the turnout of the seven categories of party identifiers in presidential elections from 1952 through

[10] Philip E. Converse, "The Nature of Belief Systems in Mass Publics," in Ideology and Discontent , ed. David Apter (New York: Free Press, 1964), 227.

[11] John F. Bibby, Politics, Parties and Elections in America (Chicago: Nelson-Hall, 1987), 259. Bibby used the broad definition of Independents.

1988. At each level of strength of identification, Republicans on average vote more than Democrats; the average gap between the two parties is about seven percentage points. Within each party, strong partisans have the highest turnout, about 10 percent more than partisan Independents and weak identifiers. The voting participation of leaners and weak identifiers in each party is similar. The turnout of Independent Democrats has remained constant since 1956, and while voting by Independent Republicans has slackened a bit since their high-water mark in 1960, their turnout is substantially exceeded only by Strong Republicans. As we might expect by now, Pure Independents are the lightest voters. Since 1964, Pure Independents have averaged 11 percent lower turnout than the next-lowest group.^[12]

The story is very much the same in midterm contests, as can be seen from table 3.5, which shows turnout in eight recent House elections. The major difference is the tendency of Independent Republicans to vote somewhat less than their fellow Republicans.

These two tables provide no support for the "new breed of Independent" argument. The more interested and knowledgeable segments of the broad Independent grouping, partisan Independents, are no more likely to go to the polls now than they were in the supposedly somnolent Eisenhower years. On the other hand, they are not notably lighter voters now than before. Thus these data are also inconsistent with the notion that the growth of partisan Independents signifies a withdrawal from partisan politics.

[12] The data summarized in table 3.4 are for self-reported turnout. Because some respondents falsely claim they voted, the NES now verifies reports about registration and voting by checking county records. This "vote validation" was not done in the early period and has been a consistent practice only in the 1980s; hence our use of self-reported turnout for purposes of continuity. Use of the validated turnout measure for the years in which it is available produces essentially the same pattern as in table 3.4. For instance, the validated turnout for 1988 was Strong Democrat, 73; Weak Democrat, 60; Independent Democrat, 62; Pure Independent, 43; Independent Republican, 60; Weak Republican, 68; Strong Republican, 79.

The same cannot be said about Pure Independents, however. Until the midsixties, they voted nearly as much as anyone else. In the first three presidential elections in table 3.4, their mean turnout equaled that of Weak and Independent Democrats. Since then, their turnout has fallen off steeply. From the high point of 78 percent in 1956, their voting rate in presidential contests declined almost steadily to 56 percent in 1980, fully 10 percentage points below any other class of identifier. Although Pure Independents' turnout rose to 61 percent in 1984, it remained well below that of either group of leaners. In 1988 only half the Pure Independents voted, 14 percent below the next-lowest category.

The decline in Pure Independents' turnout vitiates the electoral impact of their growing numbers. In 1956, 9 percent of all presidential-election voters (including blacks) were Pure Independents. As table 3.6 shows, this proportion remained remarkably constant in the ensuing thirty-four years. In 1988, 7 percent of all voters in the presidential election were Pure Independents. For the period 1956–1988, the average proportion of all presidential voters who were Pure Independents was 9 percent.

The same is true in both presidential-year and midterm elections for the House. This is why we have avoided talking about increasing numbers of independent voters . This finding should put a damper on speculations about a larger and larger pool of uncommitted voters who are "up for grabs" in election campaigns. As we shall see, Pure Independents are the only Independents who fit this description, and they contribute the same share of the total vote now that they did in the 1950s.^[13]

[13] The stable proportion of Pure Independents in the voting population also disposes of James L. Sundquist's attempt to shrug off our earlier work by pointing out that "all three categories of independents have grown as a proportion of the electorate, so limiting the term to the nonleaning independents would not alter the interpretation in this chapter." See his Dynamics of the Party System , rev. ed. (Washington, D.C.: Brookings Institution, 1983), 397n.

Participation in Campaigns

Campaign activity other than voting calls for greater commitment and more energy. In his attempt to rehabilitate Independents' reputation for civic virtue, Pomper conceded that "in regard to electoral politics, Independents do show less activity and do conform to the accepted description of limited involvement."^[14] The most widespread type of campaign activ-

[14] Pomper, Voter's Choice , 32.

ity is trying to talk someone into voting for a candidate. Predictably, Pure Independents have been least likely to do this. It comes as something of a surprise, however, to learn that Weak Democrats are frequently as inactive. In the middle are Independent Democrats, Independent Republicans, and Weak Republicans, surpassed by Strong Democrats, Independent Republicans, and—most active of all—Strong Republicans. Pomper's conclusion (which accurately described the "accepted position") is true only for Pure Independents; leaners are much more like partisans than Pure Independents in their level of campaign involvement.

The same is true for the less common forms of electioneering: financial contributions, attendance at meetings, and other activities helpful to candidates and parties. Because a small proportion of the population does any of these things, we created a participation index by combining all three types of behavior. The results for ten presidential elections are presented in table 3.7. Perhaps because of the relatively small number of cases in any cell, the numbers jump around a bit; nevertheless, the outlines are clear. The familiar pattern appears again: Republicans are more active than Democrats, strong partisans are most active, and Pure Independents are least active. In either party, leaners and weak partisans are about equally likely to participate in the campaign.

These campaign activities are seldom performed in an organizational or emotional vacuum. They are the sorts of things that party activists do and in fact are usually the identifying signs of party organization membership.^[15] The relatively high level of such behavior by partisan Independents is yet another reason to question the customary assumption that they are without party ties. So is the finding that fully 30 percent of the delegates to the 1972 Democratic National Convention identified themselves as Independent Democrats. The counter-

[15] See, for example, David Nexon, "Asymmetry in the Political System: Occasional Activists in the Republican and Democratic Parties, 1956–1964," American Political Science Review 65 (September 1971): 716–30.

part identity was claimed by 12 percent of the Republican delegates in 1972.^[16]

Previous Findings

The relationships reported in this chapter have not gone wholly unnoticed in the literature. Indeed, the first book by the Michigan group contained tables showing that partisan Independents voted and participated more than weak partisans.^[17] This finding was not acknowledged in the text, however. As we said earlier, the data analyses presented in The American Voter did not generally separate Pure and partisan Independents. Robert Cantor remarked on the nonmonotonic relationship between party identification and general political interest. But he seemed interested primarily in attacking the importance and validity of the whole notion of party identification, and the thrust of his argument went in that direction.^[18] Theodore J. Macaluso reported, "Somewhat surprisingly in light of the conventional wisdom . . . leaning Independents appear to be the most informed segment of the electorate."^[19] Warren E. Miller and Teresa E. Levitin noted that leaners vote more than weak partisans and "seem to be among the most well-informed, involved, issue-oriented of citizens. . . . The Independent Inde-

[16] Joseph H. Boyett, "Background Characteristics of Delegates to the 1972 Conventions: A Summary Report of Findings from a National Sample," Western Political Quarterly 27 (September 1974): 477.

[17] Angus Campbell, Gerald Gurin, and Warren E. Miller, The Voter Decides (Evanston, Ill.: Row, Peterson, 1954), 101, 108–9. For another early report of the same finding, see Robert Agger, "Independents and Party Identifiers: Characteristics and Behavior in 1952," in American Voting Behavior , ed. Eugene Burdick and Arthur J. Brodbeck (New York: Free Press, 1959), 319–20.

[18] Robert D. Cantor, Voting Behavior and Presidential Elections (Itasca, Ill.: Peacock, 1975), 37.

[19] Theodore J. Macaluso, "Political Information, Party Identification, and Voting Defection," Public Opinion Quarterly 41 (Summer 1977): 257–58.

pendents are, in contrast, predominantly disinterested nonparticipants in political affairs.^[20]

John Petrocik made the most elaborate analysis of the relationship between partisanship and civic virtue that appeared before our 1977 paper.^[21] He began by noting that contrary to expectation, many measures of interest and activity were not monotonically related to party identification. Pointing out that these behaviors were also related to differences in education, income, and direction of partisan choice, Petrocik argued that if one wanted to isolate the effect of partisan affiliation on civic virtue, it was necessary to control for the other independent variables. Controlling for education, income, and Republican affiliation reduced or reversed the apparent advantage of partisan Independents over weak partisans. Petrocik concluded that the familiar index of party identification was problematic.

Our interest, however, is different from Petrocik's. We are concerned not with the relative importance of different independent variables as explanations of political involvement but with the election-level and systemic issues raised in chapter 1. None of these issues is affected by the undisputed finding that independent variables other than party identification help explain civic virtue. In short, we seek to characterize Independents, not to explain political activity.

Petrocik did not discuss the most interesting aspects of party identification—its relationship to various measures of partisan choice—except to observe that "the measurement of party identification given to the profession by the [Michigan] Survey Research Center is useful for explaining the phenomenon for which it was designed; it adequately accounts for party-specific attitudes and behaviors."^[22] The evident ratio-

[20] Warren E. Miller and Teresa E. Levitin, Leadership and Change: Presidential Elections from 1952 to 1976 (Cambridge, Mass.: Winthrop, 1976), 99.

[21] John R. Petrocik, "An Analysis of Intransitivities in the Index of Party Identification," Political Methodology 1 (Summer 1974): 31–47.

[22] Ibid., 41.

nale for this verdict was Petrocik's belief that "while weak identifiers are partisans, leaners are basically independents."^[23] We will consider the evidence for this assertion in the next two chapters.

John C. Pierce and Paul R. Hagner, writing after our findings were first presented, found the same difference between Pure Independents and leaners. Compared to leaners,

The pure independents exhibit a much higher proportion of no-issue orientations and a lower proportion of ideologues. This pattern is repeated in each of the six elections. . . . It is only this pure classification of independents which fits the description of having low interest and awareness of electoral politics.^[24]

Some scholars have attempted to explain variations in turnout by means of multivariate analysis in which leaners and Pure Independents are combined.^[25] Others, in our opinion more wisely, have combined leaners and weak partisans.^[26] Dealignment, "whose most obvious manifestation is the growing proportion of Independents,"^[27] has been nominated as a prime cause of falling turnout:

The behavioural consequences of dealignment include increasing voter apathy, growing volatility, and greater

[23] Ibid., 35.

[24] John C. Pierce and Paul R. Hagner, "Conceptualization and Party Identification: 1956–76," American Journal of Political Science 26 (May 1982): 384. See also Herbert B. Asher, Presidential Elections and American Politics , 4th ed. (Chicago: Dorsey Press, 1988), 86–87, 115–17.

[25] Lee Sigelman et al., "Voting and Nonvoting: A Multi-Election Perspective," American Journal of Political Science 29 (November 1985): 749–65; Carole J. Uhlaner, "Rational Turnout: The Neglected Role of Groups," American Journal of Political Science 33 (May 1989): 390–422.

[26] Paul R. Abramson and John H. Aldrich, "The Decline of Electoral Participation in America," American Political Science Review 76 (September 1982): 502–21.

[27] William Schneider, "Antipartisanship in America," in Parties and Democracy in Britain and America , ed. Vernon Bogdanor (New York: Praeger, 1984), 100.

interest in unconventional political alternatives. It is well known that Independents exhibit these characteristics and that the proportion of Independents has been growing steadily since 1964.^[28]

We have condensed some of the data from table 3.4 in table 3.8 to enable readers to assess more easily this proposition from William Schneider, based on the broad-spectrum definition of Independents. It is accurate regarding Pure Independents. They have become more numerous and have been voting at a much lower rate than in the sixties. As for the leaners, however, their aggregate turnout decline is matched by that of weak partisans. The two categories—partisan Independents and weak partisans—appear to be equal contributors to falling turnout.

[28] Ibid., 114.

Summary

There is a wide disparity in the political interest and involvement of Pure and partisan Independents. Pure Independents are consistently the least interested, informed, and active of any partisan classification. Partisan Independents, in contrast, are not only more interested and involved than Pure Independents but as, or more, interested and involved than weak partisans. This was as true for the 1950s as the 1980s. The only notable change over this period has been a decline among Pure Independents, who display even less civic virtue now than they did a generation earlier.

4—
How Independents Vote

The Michigan measure of party identification reflects the extent to which individuals consider themselves Democrats, Republicans, or Independents. The most useful and powerful application of this identity is in voting decisions. The importance of party identification to the political system comes from the combination of its relationship to vote choice and its persistence from one election to the next. As Nie, Verba, and Petrocik put it, "A voter with no party identification cannot vote on the basis of party. Thus the growth in the number of Independents automatically reduces the number who can give a party vote.^[1]

Because they have nothing to be loyal to, it has been argued that Independents respond more freely than partisans to the short-term forces of the campaign, the candidates and issues that provide the dynamic element in elections. If nearly 40 percent of all Americans are "simply and tautologically unconstrained by partisanship,"^[2] it is easy to see why the decline of

[1] Norman H. Nie, Sidney Verba, and John R. Petrocik, The Changing American Voter , enlarged ed. (Cambridge: Harvard University Press, 1979), 157.

[2] Gerald M. Pomper, Voter's Choice (New York: Harper and Row, 1975), 40.

partisan affiliation has led many observers to predict fundamental political change (see chapter 1). On the other hand, if this description applies only to the much smaller number of Pure Independents, then the threat to political stability posed by the increasing numbers of Independents has been vastly exaggerated.

Do leaners vote like outright partisans or like Pure Independents? This simple question was not directly answered in The American Voter , which introduced the concept of party identification and established the analytic categories that guided political scientists' use of that concept. The questions that elicited each respondent's "partisan self-image" made it possible for "us to place each person in these [1952 and 1956] samples on a continuum of partisanship extending from strongly Republican to strongly Democratic."^[3] The impression of a monotonic relationship between party identification and vote choice suggested by this quotation was reinforced by passages like this:

The effect of party [on vote choice] is seen at once in the changing location of the distributions [of the probability of voting for the Republican presidential candidate in 1956] along this probability dimension as we consider successively Strong Democrats, Weak Democrats, Independents, Weak Republicans, and Strong Republicans.^[4]

This refers to a multivariate analysis of the relationship between a cluster of attitudes and the presidential vote. On the next page is a simple cross-tabulation of the presidential vote in 1952 and 1956 with the five-point classification of party identification usually employed in The American Voter , combining all three varieties of Independents.^[5] Fig. 4.1 reproduces this cross-tabulation for 1956. It shows a smoothly monotonic decline from Strong Democratic to Strong Repub-

[3] Angus Campbell et al., The American Voter (New York: John Wiley and Sons, 1960), 122–23, emphasis added. For the questions, see pp. 12–13 above.

[4] Campbell et al., The American Voter , 138.

[5] Ibid., 139. See note 33, chapter 1.

Figure 4.1
Five-Point Party Identification and Vote for Democratic
Presidential Candidate, 1956, 1976, and 1988

lican in the share of the vote received by Adlai Stevenson in his rematch with Dwight D. Eisenhower. We also plotted in fig. 4.1 the same relationship in 1976, a Democratic year when Jimmy Carter edged out Gerald R. Ford and for 1988, when George Bush defeated Michael Dukakis. These lines are even straighter than the one representing Stevenson's vote from each of the five party-identifier categories.

When they introduced their measure of party identification, the authors of The American Voter discussed the possibility that many self-designated Independents might really be covert partisans:

Sometimes it is said that a good number of those who call themselves Independents have simply adopted a label that conceals a genuine psychological commitment to one party or the other. . . . We have incorporated in our measure of party identification a means of distinguishing Independents who say they lean toward one of the parties from Independents who say they do not. We do not think that the problem of measurement presented by the concealed partisan is large. Rather it seems to us much less troublesome than the problems that follow if psychological ties to party are measured in terms of the vote.^[6]

Dismissal of fears about covert partisans masquerading as Independents was justified after analysis of responses to this question: "Have you always voted for the same party or have you voted for different parties for President?"^[7] The proportion of respondents who said they always or mostly voted for the same party went from 82 percent of strong party identifiers and 60 percent of weak identifiers to 36 percent of leaners and a mere 16 percent of Pure Independents. Even more striking results were achieved by dividing the sample at age 35 to control for younger voters' fewer opportunities to opt for different parties. After presenting these data, Campbell and his associates concluded that "we will describe relatively few genuine partisans as Independents by using self-classification to measure party attachments."^[8]

As we will show in this and the following chapter, this is a defensible conclusion only if leaners are excluded from the Independent category. Before starting to demonstrate this, the central proposition in this book, let us examine briefly the candidate preferences of Pure Independents. One would predict that lacking any party to which they could be loyal, they would be more susceptible to influences specific to each elec-

[6] Campbell et al., The American Voter , 123, 125.

[7] Ibid., 125.

[8] Ibid., 125–26, quoted passage at p. 126. We discuss the inadequacy of such retrospective data in chapter 5.

tion: current issues, the appeal of that year's candidates, and judgments on the administration's performance. This is why the putative increase in Independents moved observers to anticipate wider and wider swings from one election to the next—in a word, volatility.

Volatility is the right word for Pure Independents' voting patterns. Fig. 4.2 depicts the percentages of white Pure Independents and of all white respondents who voted for the Democratic presidential candidate in every election from 1952 through 1988. As a general proposition, the Pure Independent vote reflects in an exaggerated way the election outcome. Candidates who win by a landslide take an even more lopsided share of the Pure Independent vote. When Lyndon B. Johnson crushed Barry Goldwater with 65 percent of all votes cast by whites in 1964, he received 75 percent of the Pure Independent vote. Both numbers were about the same in Richard M. Nixon's one-sided victory over George McGovern in 1972. Eisenhower's appeal to Pure Independents was even more disproportionate when he defeated Stevenson with 55 percent of the total vote in 1952 and 58 percent four years later. In 1980, when Ronald Reagan had a 5:4 advantage over Jimmy Carter in the popular vote, he won among Pure Independents by a 3:1 ratio, 66 to 22 percent (John Anderson took the other 12 percent). In closer elections, Pure Independents usually vote for the winner (1976 was an exception), but not by very much. George Wallace's candidacy muddied the waters in 1968 when he won 18 percent of Pure Independent votes, compared to 13 percent of the total popular vote.

The sharp peaks and deep troughs in fig. 4.2 leave no room for doubts about Pure Independents' volatility. We have already seen, however, that they amount to less than 10 percent of voters in presidential elections, a proportion that has remained essentially unchanged since the 1950s (see table 3.6, p. 52). If there is any justification for the fears of so many political scientists that we are in an era of heightened electoral instability, it must lie in the behavior of partisan Independents, who are responsible for most of the growth in the

Figure 4.2
Vote for Democratic Presidential Candidate of Pure Independents
and All White Respondents, 1952–1988

inclusive Independent category and account for two-thirds of all Independents. The conventional wisdom, and the root cause of all the worry about partisan instability, is that Independents "are not guided by a party affiliation."^[9] As we have just seen, this seems a fair description of Pure Independents' voting practices.

Leaners are a very different story. They display an impressive tendency to vote for the candidate of the party they feel closer to; indeed, in presidential elections they are generally more loyal to their party than weak partisans . In seven of the ten

[9] Nie, Verba, and Petrocik, The Changing American Voter , 50.

presidential elections since 1952, Independent Democrats gave a higher proportion of their vote to the Democratic presidential candidate than did Weak Democrats. Sometimes the difference was substantial, as in 1960 when 71 percent of Weak Democrats voted for John F. Kennedy, compared to 86 percent of Independent Democrats. Four years later, 89 percent of Democratic leaners voted for Johnson, as against 81 percent of Weak Democrats. In 1972, McGovern actually lost the Weak Democrat vote but won 58 percent of the ballots cast by Democratic leaners. In the 1984 Republican landslide, Walter Mondale captured fully 76 percent of Democratic leaners' votes and just 63 percent of the ballots cast by Weak Democrats. In 1988, Michael Dukakis won 88 percent of Democratic leaners and 67 percent of Weak Democrats. Only in 1980 did Weak Democrats' loyalty exceed that of the Independent Democrats by more than three percentage points; 53 percent of the former and just 41 percent of Democratic leaners voted for Jimmy Carter. This gap reflects the third-party candidacy of John Anderson, who won fully 20 percent of the Independent Democratic vote. Ronald Reagan ran equally well among these two groups of less-than-strong Democrats.

The story for Independent Republicans is very much the same. They exceeded Weak Republicans in party loyalty by a considerable margin in two years, 1964 and 1976. On average since 1952, 89 percent of Independent Republicans have voted for the Republican presidential candidate, compared to 87 percent of Weak Republicans. Only in 1980 when 11 percent of them voted for Jimmy Carter, compared to just 5 percent of Weak Republicans, were they significantly less loyal. On average, both groups of leaners were more likely than their counterpart weak identifiers to vote for their party's candidate.

Patterns of party-line voting among Southern whites are somewhat different. All three varieties of Southern Republicans are very loyal. In the past nine presidential elections, fully 98 percent of Strong Republicans voted for their party's candidate, as did 92 percent of Independent Republicans and 91 percent of Weak Republicans. In 1984, every white Southern Independent Republican in the NES sample voted for

Reagan. They were just as likely as Weak Republicans to vote for Republican House candidates and more likely to choose the Republican Senate candidate. Among white Southern Democrats, on the other hand, only strong identifiers consistently back their party's presidential candidates, and even they vote 8 percent less Democratic on average than their Northern counterparts. The other two groups of Southern Democrats are more likely than not to defect. Weak identifiers are more loyal than leaners, who nevertheless vote for Democrats much more than do the Pure Independents.^[10]

These findings about the voting behavior of independent leaners are a spectacular refutation of the orthodox belief that party regularity is proportionate to strength of party identification. The extent of the damage is illustrated by fig. 4.3 where we have plotted the same relationship between party identification and vote for the Democratic presidential candidate shown in fig. 4.1, but with one alteration: Party identification is now measured by a seven-point rather than five-point scale. The result is dramatic; the nearly straight parallel trend lines in fig. 4.1 are transformed, in fig. 4.3, into jagged strokes that move in an irregular stair-step pattern across the page.

Table 4.1 also sheds light on another commonplace assertion about contemporary political life: the alleged diminishing relevance of party identification to the decision Americans make on election day. For example, Marjorie Randon Hershey had this to say:

Although most voting-age Americans still express a psychological attachment to one of the major parties, that attachment has weakened and means less and less in elections. . . . Party has less influence on people's voting choices than it once did; . . . the parties, quite simply, are becoming irrelevant to large numbers of Americans.^[11]

[10] Raymond E. Wolfinger and Michael G. Hagen, "Republican Prospects: Southern Comfort," Public Opinion 8 (October–November 1985): 8–13.

[11] Marjorie Randon Hershey, Running for Office (Chatham, N.J.: Chatham House, 1984), 24.

Figure 4.3
Seven-Point Party Identification and Vote for Democratic
Presidential Candidate, 1956, 1976, and 1988

William Crotty's view of the importance of party identification is even more negative about

an increasing disparity between [party] identification and the vote. There is an increasing willingness on the part of many party members to break with their party and its candidates and to vote for the opposition . . . the odds are not much over fifty-fifty today that a party identifier will vote for his party's presidential nominee. Republicans are little different from Democrats in this regard.^[12]

[12] William Crotty, American Parties in Decline , 2d ed. (Boston: Little, Brown, 1984), 31, 34.

Against these sweeping generalizations we contrast table 4.1, which shows that in the ten-election period from 1952 through 1988 there was no significant increase in defection by any of the six party-identifier categories. Other than simple inspection, this also can be seen by computing each category's average Democratic vote for the three elections of the 1980s and comparing that number to the mean for the previous seven elections. The difference between the two numbers is two or fewer percentage points in three of the six cases. Defection increased by three percentage points among Weak Democrats and declined by about the same amount among Weak Republicans and Strong Democrats.

The data in table 4.1 are no kinder to Crotty's proposition about the similarity in the party loyalty of Democrats and Republicans in presidential elections. Simply put, in nine of the past ten elections, Democrats have been substantially more prone to defect. (This has not been the case in House elections, as we will see shortly.)

The finding that leaners display more party loyalty than weak partisans strikingly illustrates our basic proposition that leaners are partisan, not neutral. Although a nonmonotonic relationship of the kind shown in table 4.1 and fig. 4.3 is a particularly arresting demonstration of our thesis, it is not the appropriate criterion. We will see that leaners do not always match or exceed weak identifiers on all measures of partisan affect, but they are never neutral, and the extent of their affect almost invariably resembles that of weak partisans .

These remarks introduce our findings on voting in House elections, which show little nonmonotonicity but consistent party-line voting by leaners. Table 4.2 displays how the seven categories of white identifiers voted in the fifteen House elections from 1962 through 1990. Comparison with table 4.1 reveals that Independent Democrats were more loyal in House than in presidential voting; 74 percent were party-line voters in House elections, compared to 67 percent when voting for president. But Weak Democrats were much more loyal when voting for Congress, so much so that over the twenty-eight-

year period, they were slightly stronger party voters in House races than the leaners were. A closer look at table 4.2 shows that this difference was found mostly in the 1960s. Beginning in 1972, Independent Democrats equaled or surpassed Weak Democrats in party-line voting in every House election year but two. In 1978, 79 percent of Weak Democrats and only 60 percent of leaners voted for Democratic congressional candidates. Even with this deviant year included, Democratic leaners have been faintly more loyal in these ten elections: 77 percent, compared to 75 percent for Weak Democrats.

Compared to Democrats, Republicans defect less in presidential elections and more when voting for the House. This reflects the pull of incumbency and the much greater number of Democratic incumbents. Even so, Republican leaners choose their party's House candidates by about a 2:1 ratio; Weak Republicans are five percentage points more loyal. Better-informed voters are more likely to defect in House elections, which may explain why leaners desert their parties more than the less-informed weak identifiers.^[13]

The discovery that leaners are slightly less loyal than weak identifiers in House elections gave heart for a time to defenders of the old view that leaners are not partisans. At the beginning of the 1980s, John R. Petrocik drew this conclusion:

The "biased" Independents [that is, leaners] appear rather partisan only in their presidential choices. In other contests Independents are much less likely to support the candidate of the party toward which they lean than are the weak identifiers.^[14]

[13] Thomas E. Mann and Raymond E. Wolfinger, "Candidates and Parties in Congressional Elections," American Political Science Review 74 (September 1980): 621–22.

[14] John R. Petrocik, Party Coalitions (Chicago: University of Chicago Press, 1981), 159. See also his "Contextual Sources of Voting Behavior: The Changeable American Voter," in The Electorate Reconsidered , ed. John C. Pierce and John L. Sullivan (Beverly Hills, Calif.: Sage, 1980), 265–66.

At the end of the decade, Petrocik reported that data from 151 surveys conducted for congressional, state, and local candidatesshowed this pattern of voting for Republican candidates: 33 percent by Weak Democrats, 36 percent by Independent Democrats, 71 percent by Independent Republicans, and 75 percent by Weak Republicans. The multivariate analysis on which the study was based "separates independents into three groups: two groups of leaners and those who insist they have no preference for either party." See his "An Expected Party Vote: New Data for an Old Concept," American Journal of Political Science 33 (February 1989): 44–66, quoted passage at 53.

Two years earlier, he had written that "all data referring to Democrats and Republicans are based on a definition of partisanship which includes strong, weak, and leaning identifiers as partisans." See his "Realignment: The South, New Party Coalitions and the Elections of 1984 and 1986," in Warren E. Miller and John R. Petrocik, Where's the Party? An Assessment of Changes in Party Loyalty and Party Coalitions in the 1980s (Washington, D.C.: Center for National Policy, 1987), 54n.

It is difficult to reconcile this conclusion with leaners' degree of party-line voting; 74 percent of Independent Democrats and 33 percent of Independent Republicans vote for Democratic House candidates. We see no reason to ignore this gap, deny that leaners are partisan, and combine them in an omnibus Independent category that mingles three distinct types of voters.

The partisanship of leaners was reported in the literature before we presented our findings in 1977. Some scholars presented tables or charts depicting the nonmonotonic relationship but then failed to mention it in the text.^[15] William Flanigan and Nancy Zingale, however, did draw the appropriate conclusion:

On crucial attitudes and in important forms of political behavior, the "leaning" independents appear quite partisan. Independents who lean toward the Democratic party behave rather like weak Democratic partisans, and the same is true on the Republican side. . . . independent

[15] Robert D. Cantor, Voting Behavior and Presidential Elections (Itasca, Ill.: Peacock, 1975), 34; Peter B. Natchez and Irvin C. Bupp, "Candidates, Issues, and Voters," in Political Opinion and Behavior , 2d ed., ed. Edward C. Dreyer and Walter A. Rosenbaum (Belmont, Calif.: Wadsworth, 1970), 447.

leaners are more interested in campaigns than are weak partisans, and they are as likely to vote and even as likely to be loyal to "their" party as are weak partisans.^[16]

Having said this, they proceeded in the rest of this edition of their book to analyze data combining all three types of Independents.

Warren Miller and Teresa Levitin clearly described the leaners' partisan voting habits:

In fact, between 1952 and 1976 the two groups of Independent leaners voted more solidly in support of their party preferences than did weak identifiers in almost all of the several electoral opportunities to do so. This persistent partisan behavior on the part of self–designated Independents raises questions about the meaning of the self–classification.^[17]

However, there is too little research on the political perceptions, evaluations, and behaviors of Independent leaners to answer these questions.^[18]

Miller and Levitin concluded that the leaners' voting patterns suggested that "the simple fact of an increase in the incidence of Independents cannot be taken as evidence of a decline in the importance of partisanship."^[19] Few scholars or other observers heeded this prudent warning. Another passage from Miller and Levitin seemed more consistent with most interpretations of the last two decades: "Since they are not tied to a party and respond primarily to the short–term issue and candidate forces of each particular election, the

[16] William H. Flanigan and Nancy H. Zingale, Political Behavior of the American Electorate , 6th ed. (Boston: Allyn and Bacon, 1987), 46.

[17] Warren E. Miller and Teresa E. Levitin, Leadership and Change: Presidential Elections from 1952 to 1976 (Cambridge, Mass.: Winthrop, 1976), 210.

[18] Ibid., 252.

[19] Ibid., 210.

Independent leaners are also a relatively volatile segment of the electorate."^[20]

Since the late 1970s, many specialists in voting behavior have acknowledged the partisan character of leaners. This shift is so far from universal, however, that even scholars as well informed as Robert L. Lineberry and George C. Edwards II could write in 1989:

For most white Americans, though, party desertion—the abandonment of either party for a nonparty attachment—is well advanced. This abandonment occurred at all age levels in the electorate, though it was more pronounced for those with the weakest party ties, that is, younger voters.^[21]

In the next chapter we consider various reactions to the finding that leaners vote like outright partisans and then examine a variety of other evidence that sheds light on the central question: Are they really neutral, or are they committed to the party toward which they lean?

[20] Ibid., 99.

[21] Robert L. Lineberry and George C. Edwards II, Government in America , 4th ed. (Glenview, Ill.: Scott, Foresman, 1989), 252. Lineberry and Edwards relied on Gallup data that showed 28 to 33 percent of Americans were Independents in the 1970s and 1980s.

5—
Partisans or Independents?

The discovery that most leaners are loyal party voters called into question the prevailing view of party identification. By the beginning of the 1980s, study of this topic, once a model of consensus, was marked by controversy and confusion. Some scholars accepted our interpretation of leaners' voting behavior: They vote as they do because they are more partisan than independent—not neutral, but nearly as committed to a party as those who readily acknowledge their affiliations. Analyses using all seven party-identification categories, once scarce, became commonplace.^[1]

[1] Media and political pollsters dealt with the topic in various ways. As early as 1976, the CBS News–New York Times polls divided leaners and Pure Independents. Newsweek and the NBC–Associated Press surveys did this during the 1980 election season. One advantage of this practice was realized in the 1980 preconvention period: classifying leaners as partisans made it easier for pollsters to accumulate enough cases for analysis of intraparty candidate preferences.

Richard Wirthlin, President Ronald Reagan's pollster, reported, "We find that when people say they are independent, and then we ask them which party they lean toward, they are stronger partisans than many of those who declare a party." Quoted in Elizabeth Drew,"A Political Journal," The New Yorker , February 20, 1984, 116. See also Wirthlin, "Partisan Change in the 1980s: A Rolling Realignment," The Public Perspective 1 (November–December 1989): 14.

For some years after 1980, however, the Gallup organization still did not ask about leaning—or did not report the results if it did ask—and so continued to report that over 30 percent of all Americans were Independents. A Gallup release in the spring of 1983 appeared under this headline: "Republican Affiliation Falls to 1 in 4," San Francisco Chronicle , March 10, 1983, 13.

Not everyone accepted these distinctions, however. A surprising number of scholars did not seem aware that it was important to differentiate between leaners and Pure Independents. Experienced students of elections continued to produce variations on the "partisanship is dying" theme:

Independents have increased markedly in number to the point where they now rank in some polls as equal to or slightly behind the Democrats and well ahead of the Republicans. The day is not far off when independent will clearly be the identification preferred by most Americans. In fact, there is nothing to indicate a substantial reversal or leveling off of the trend shown. Thus, independents will most likely continue to increase in significance and number with each passing election year. From an electorate in which 75 percent of its members loyally associated themselves with the major parties to one (in future elections) in which the majority may well claim to be independents is a major transition that, in itself, has changed, and will continue to change, the practice of American politics.^[2]

[2] The short and long quotations are both from William Crotty, American Politics in Decline , 2d ed. (Boston: Little, Brown, 1984), 42, 28–30. Like some other scholars with his point of view, Crotty presented tables showing differences between leaners and Pure Independents that he ignored in the text.

For other examples of the same theme, see Paul R. Abramson, John H. Aldrich, and David W. Rohde, Change and Continuity in the 1980 Elections , rev. ed. (Washington, D.C.: CQ Press, 1983), 237; William J. Keefe et al., American Democracy: Institutions, Politics, and Policies (New York: Harper and Row, 1990), 237; Everett Carll Ladd, Jr., Where Have All the Voters Gone? , 2d ed. (New York: W. W. Norton, 1982), 77–78, 124; Kevin P. Phillips, Post-Conservative America (New York: Random House, 1982), 227–28; Kenneth Prewitt and Sidney Verba, An Introduction to American Government , 4th ed. (New York:Harper and Row, 1983), 392–93, 400, 402; and James L. Sundquist, Dynamics of the Party System , rev. ed. (Washington: Brookings Institution, 1983), 394–408.

In a lead editorial during the 1980 campaign, the New York Times said that the one-third of the population who initially called themselves Independents could be indicative of

a large and growing corps of voters located between the major parties [who might provide support for] an independent movement, a Center Party, a Third Force, made up of people unsatisfied with both sides of a Carter-Reagan race, and fastening mostly on John Anderson.^[3]

In addition to the converted and the apparently oblivious, there was a third category: the explicit skeptics, scholars who knew that there are two ways of looking at Independents but nevertheless carried on with the old way. The skeptics often relied on various interpretations of leaners' voting choices that preserved the conventional wisdom, but some did not seem to think it necessary to justify their orthodoxy. Charles H. Franklin, for example, after noting that "some recent research has questioned" the old assumption of monotonicity, announced (without benefit of argument or citation), "I prefer to retain it."^[4] Other skeptics, however, did

[3] The New York Times , March 20, 1980, A26. Readers will notice that the New York Times and Abramson and his collaborators have been on both sides of the dispute about Independents.

In the middle of 1983, John Anderson was reported to believe that the "one-third" of the population who "claim no ties" to a party provided an opportunity for a third-party movement; George F. Will, "Is a Third Party Necessary?" Newsweek , May 23, 1983, 88. A year later, the Washington Post's director of polling wrote that the "between 35 and 40 percent" of American adults who "considered themselves independents . . . appear to present fertile ground for the kind of effort Anderson says he is willing to undertake"; Barry Sussman, "Why Anderson's Third-Party Bid May Hit Responsive Chords," Washington Post (National weekly ed.), May 21, 1984, 37.

[4] Charles H. Franklin, "Issue Preferences, Socialization, and the Evolution of Party Identification," American Journal of Political Science 28 (August 1984): 469.

explain their adherence to the traditional position that leaners should not be considered party identifiers. Perhaps the most comprehensive, if compressed, statement is Warren E. Miller's:

Although it is true that the so-called Independent Democrats and Independent Republicans may behave very much as partisans in the short-term, even outdoing weak identifiers in the homogeneity of their voting partisanship—they are more volatile in their partisanship over the long-term; and, when tested by measures of their psychological ties to the parties, they are demonstrably the independents they claim to be, more moved by candidates or issues and less identified with their preferred party of the moment.^[5]

In this chapter we test explanations of leaners' voting choices that preserve the conventional wisdom about their essential neutrality.^[6] We employ a variety of measures other than voting choice that shed light on the basic question: How partisan are the leaners? We will answer this question by describing their participation in presidential primaries, the stability of their voting choices and party identification, and their assessment of the two parties.

[5] Warren E. Miller, "Party Identification Re-examined: The Reagan Era," in Warren E. Miller and John R. Petrocik, Where's the Party? An Assessment of Changes in Party Loyalty and Party Coalitions in the 1980s (Washington, D.C.: Center for National Policy, 1987), 24n. Neither data nor citations were presented to support this conclusion. For a more theoretically grounded defense of the traditional position, see Miller, "Party Identification, Realignment, and Party Voting: Back to the Basics," American Political Science Review 85 (June 1991). We will consider this argument near the end of this chapter.

[6] As we observed in the previous chapter, some writers found support for their belief in the leaners' nonpartisanship in the finding that they were faintly less loyal than weak partisans in voting for House candidates. See, for example, William Schneider, "Antipartisanship in America," in Parties and Democracy in Britain and America , ed. Vernon Bogdanor (New York: Praeger, 1984), 116–17.

Participation in Presidential Primaries

American political theory and practice have provided several definitions of party membership. All are plausible, and none is conclusive.^[7] The major practical implication of these endless legal, political, and scholarly controversies is eligibility to vote in party elections. Primaries have usually been considered reforms that weakened the parties as institutions because voters rather than party leaders choose nominees. But Leon D. Epstein argues that at the individual level, primaries necessarily strengthen partisanship because they "must encourage voters to think of themselves as Republicans or Democrats."^[8] This point is particularly important for independent partisans. If their designation of a party to which they are closer is nothing more than a signal of their intentions in the general election, one would expect them to be under no impulsion to participate in one or the other party's primary. What is more, because most states have registration by party and closed primaries, choosing a primary to vote in cannot be a spur-of-the-moment decision.

For both of these reasons, we can learn something about the stable partisan identifications of leaners by looking at their participation in presidential primaries when competition for the nomination was stiff in both parties. One such year was 1980; 38 percent of the NES sample who lived in a state with a presidential primary in 1980 reported voting in that primary.^[9] Leaners were as likely as weak partisans to

[7] See Austin Ranney, Curing the Mischiefs of Faction (Berkeley: University of California Press, 1975), 147–69.

[8] Leon D. Epstein, Political Parties in the American Mold (Madison: University of Wisconsin Press, 1986), 246. Epstein cites a study by Steven E. Finkel and Howard A. Scarrow that shows a relationship between registration by party and more partisan responses to the party-identification question; see their "Party Identification and Party Enrollment: The Difference and the Consequence," Journal of Politics 47 (May 1985): 620–42.

[9] For similar findings about the 1976 presidential primaries, see Keith et al., "The Partisan Affinities of Independent 'Leaners,'" British Journal of Political Science 16 (April 1986): 165–66.

participate; just over a third of both groups voted, with no difference between Democrats and Republicans. This was double the turnout rate of Pure Independents. Asked which party's primary they had participated in, 76 percent of Independent Democrats said they had voted in the Democratic contest, compared to over 90 percent of the other two types of Democrats. The corresponding figures on the Republican side were 70 percent for leaners, 87 percent for weak partisans, and 97 percent for the strong partisans. These data are in table 5.1, which includes the counterpart findings for 1988.

In 1988, 36 percent of the sample reported voting in a presi-

dential primary or caucus. In both parties, leaners' turnout rate was five percentage points below that of weak partisans. As before, Pure Independents were much the least active, and strong partisans the most active. Participation in the Democratic primary ranged from 98 percent by Strong Democrats to an almost identical rate of 95 percent by Independent Democrats. Seventy-six percent of Republican leaners chose to participate in a Republican primary; Weak Republicans were a mere five percentage points more likely to do so.

We can summarize our findings as follows: Leaners are not notably more reluctant than weak partisans to participate in presidential primaries. When it comes to choosing one or the other party's primary, the four possible comparisons in 1980 and 1988 yield three different findings. In one case, that of 1988 Democrats, leaners were almost unanimous participants in their party's primary. In two other cases, the odds were better than three to one that a Democratic leaner in 1980 and a Republican leaner in 1988 would pick his or her party's primary. The weakest case for our general theme is the 1980 Independent Republicans, of whom "only" 70 percent voted in the Republican primary.

More evidence in support of our argument comes from Gary D. Wekkin's study of voting in the 1980 presidential primaries in Wisconsin, where citizens may choose to participate in either party's primary. He found that weak and leaning Democrats and Republicans were equally likely to vote in their respective party's primary. "What Keith et al. found to be true of independent-leaning voting behavior in general elections may also hold true in presidential primaries."^[10] We interpret Wekkin's finding and our own as evidence that when respondents say they are closer to a party they are revealing an enduring identity more than a response to passing issues and personalities.

[10] Gary D. Wekkin, "The Conceptualization and Measurement of Crossover Voting," Western Political Quarterly 41 (March 1988): 112.

The Stability of Party Identification

The essence of party identification is "some engagement of partisan feelings with self-identity."^[11] Few people redefine themselves frequently; hence self-definition implies stability. Something closely resembling enduring party identification is also implicit in the presurvey-era studies based on analysis of election returns that characterized particular regions or groups as partisan for reasons attributed to distant historical experience: eastern Tennessee, blacks before the New Deal, Catholics. The acquisition of party affiliation by children in advance of any political understanding and the transmission of party identification from parents to children also suggest stable partisan orientations.^[12] The role of party identification in the "edifice" of thinking about political perceptions, beliefs, and choices associated with The American Voter assumed that it was fairly stable:

In the period of our studies [1952 and 1956] the influence of party identification on attitudes toward the perceived elements of politics has been far more important than the influence of these attitudes on party identification itself.^[13]

Defining party identification as a long-term predisposition made it possible to ascertain the importance in a given electoral setting of more ephemeral factors such as the candidates' appeal, issues, and evaluations of governmental perfor-

[11] Philip E. Converse and Roy Pierce, "Measuring Partisanship," Political Methodology 11 (1985): 144. The other element they consider "absolutely central to the whole notion of party identification: an extended time horizon."

[12] Fred I. Greenstein, Children and Politics (New Haven: Yale University Press, 1965), chap. 4; M. Kent Jennings and Richard G. Niemi, The Political Character of Adolescence (Princeton, N.J.: Princeton University Press, 1974), chap. 2.

[13] Angus Campbell et al., The American Voter (New York: John Wiley and Sons, 1960), 135.

mance. The elections of the 1950s seemed well suited to this analytic goal. The Republicans gained the White House because millions of Democrats voted for Dwight Eisenhower without changing party or (particularly after 1952) voting in equal measure for Republican congressional candidates.

While all these findings were consistent with the proposition that party identification was a relatively stable variable, they did not directly measure stability. The same was true of the similar distributions of party identification that were observed in the nationwide survey data that the Michigan researchers began to accumulate at two-year intervals after 1956 (see table 1.1). It could be argued, however, that the impressive aggregate stability in those repeated cross sections could conceal considerable shifting among the various categories of party identification. There were two ways to explore stability directly. One was to ask respondents if there was "ever a time when you thought of yourself as a Democrat (Republican) rather than a Republican (Democrat)?" As we will see shortly, this is not a valid way to measure change in party identifications.

The second method is through a "panel study," repeated interviews with the same sample over a period of years. Panel studies are very expensive, subject to considerable attrition, and likely to produce a preternaturally informed and interested sample of survivors. The Michigan researchers have incorporated a panel design in their nationwide electoral research only twice—in the 1956–58–60 period and in 1972–74–76. This summary of the results from the first panel applies equally to the second one: "the stability of party identifications in the 1956–60 period vastly outstripped the stability of individual positions on even the most stable of the major political issues of the period."^[14]

[14] Philip E. Converse and Gregory B. Markus, "Plus ça change . . . : The New CPS Election Study Panel," American Political Science Review 73 (March 1979): 33. The standard study of the comparative stability of party identification and political attitudes in the 1952–54–56 panel is Philip E. Converse, "The Nature of Belief Systems inMass Publics," in Ideology and Discontent , ed. David Apter (New York: Free Press, 1964), 206–61.

Such comparisons did not directly address the stability of party identification, a topic that was not illuminated in print until the 1970s, when one could learn that nearly one in six respondents in the earlier panel had moved from one to another of the three major groups—Democrat, Pure Independent, or Republican—and that less than 60 percent of the sample had given the identical response to the party-identification questions in any two-year period. Instability was slightly higher in 1972–74–76.^[15]

If party identification had changed so much, what was causing it to do so? Studies began to appear showing that partisanship was influenced by various contemporary phenomena, including evaluations of the presidential candidates,^[16] assessments of presidential performance,^[17] respondents' comparisons of their own issue preferences and those they attributed to the parties,^[18] vote intentions in the current election,^[19] and appraisals of current economic conditions.^[20] These had all been considered responsive to party identification, endogenous variables in models of vote choice. Party identification

[15] Morris P. Fiorina, Retrospective Voting in American National Elections (New Haven: Yale University Press, 1981), 87–88. For a summary of the earlier studies of change in party identification, see Fiorina, pp. 86–89.

[16] Benjamin I. Page and Calvin C. Jones, "Reciprocal Effects of Policy Preferences, Party Loyalties, and the Vote," American Political Science Review 73 (December 1979): 1071–89.

[17] Fiorina, Retrospective Voting; Michael B. MacKuen, Robert S. Erikson, and James A. Stimson, "Macropartisanship," American Political Science Review 83 (December 1989): 1125–42.

[18] Franklin, "Issue Preferences"; John E. Jackson, "Issues, Party Choices, and Presidential Votes," American Journal of Political Science 19 (April 1975): 161–85; Charles H. Franklin and John E. Jackson, "The Dynamics of Party Identification," American Political Science Review 77 (December 1983): 957–73.

[19] Gregory B. Markus and Philip E. Converse, "A Dynamic Simultaneous Equation Model of Electoral Choice," American Political Science Review 73 (December 1979): 1055–70.

[20] MacKuen, Erikson, and Stimson, "Macropartisanship."

had been treated as an "unmoved mover," an exogenous variable. Now it appeared that the relationship might well run as powerfully in the opposite direction, indeed that party identification was little more than an amalgam of the individual's reaction to contemporary events and conditions. This is how some defenders of the conventional wisdom had explained leaners' partisanship; now the suggestion was extended to all categories of party identifier.^[21]

This interpretation probably went beyond the intentions of the scholars cited in the previous paragraphs, although one can find support for it here and there, particularly in Fiorina's book, which attacked the conventional wisdom with exceptional clarity, force, and color. His reduction of party identification to "a running tally of retrospective evaluations" seemed more memorable than his "disclaimer" that some people "may never forget Hoover's depression or Sherman's march to the sea," which carried retrospective evaluation to a previously unimagined historical dimension.^[22]

The entire literature alleging the instability of party identification has been challenged on methodological grounds by Donald Philip Green and Bradley Palmquist, who argue that the imprecision of survey instruments calls into question both apparent changes in party identification and responsiveness of partisanship to short-term forces.

The problem of response error has been largely ignored in recent models of partisan change. As a result, estimates that purport to show the lability of party identification with respect to candidate evaluations, voting behavior, issue proximities, and performance evaluations turn out to be statistical artifacts. In retrospect, it would appear that in their enthusiasm for the use of sophisticated non-

[21] Some of these studies classified leaners as partisans when defining the variables in their models; others adhered to the traditional wide-spectrum view of Independents. None of them, however, shed any light on whether leaners were particularly susceptible to short-term forces.

[22] Fiorina, Retrospective Voting , 94, 91.

recursive estimation techniques, researchers in this area have tended to lose sight of the inherent deficiencies of survey data.^[23]

One source of the measurement error discussed by Green and Palmquist is suggested by Converse and Roy Pierce:

Citizens marginal to the electoral process, for example, such as chronic non-voters, give responses to the party identification item which are very unstable and which seem to move dynamically in tune with whatever party the respondent would vote for at the moment (assuming interest enough to get to the polls, which is usually absent).^[24]

The controversy about the stability of party identification in general sets the stage for our consideration of leaners' stability. Some political scientists do not think that leaners' indisputably partisan voting habits are inconsistent with the conventional wisdom. They consider leaners not hidden Democrats and Republicans but neutrals who decide how to vote and then use that decision when answering the question about which party they are closer to. "Since I'm going to vote for Dukakis, I guess I'm closer to the Democrats."

We continue our appraisal of this interpretation by comparing the stability of different categories of identifiers. The first data source is the repeated interviews of the 1972–1976 panel in the Michigan NES. We plotted the full array of 1972 respondents according to their identification in 1976, thus producing a seven-by-seven matrix of partisan affiliation. This revealed considerable movement from category to category.^[25]

[23] Donald Philip Green and Bradley Palmquist, "Of Artifacts and Partisan Instability," American Journal of Political Science 34 (August 1990): 899. The MacKuen, Erikson, and Stimson study was the only one cited above not criticized in detail by Green and Palmquist.

[24] Converse and Pierce, "Measuring Partisanship," 150.

[25] Including respondents' party identification in 1974 would increase the sense of movement since some respondents who were in the same partisan category in 1972 and 1976 were somewhere else in 1974.

Relatively few of these changes were from one party to the other, however; most were shifts in the intensity of partisan identity, not in its direction. The overwhelming majority of people who professed some affiliation with a party in 1972 were still on that side four years later. We will call these respondents directionally stable . Many others shifted only to the Pure Independent category, leaving a handful who crossed the divide from one party to the other. This latter group is the directionally unstable . The proportions of different categories of identifiers in the directionally stable and unstable groups are shown in table 5.2.

As we might expect, strong partisans were the most stable; 95 percent of 1972 Strong Democrats and 92 percent of Strong Republicans were on the same side of the fence in 1976. They were not far ahead of weak partisans, however; 85 percent of Weak Democrats and 88 percent of Weak Republicans were directionally stable. Independent Democrats were scarcely less fixed in their partisan identity than Weak Democrats; 79 percent were directionally stable. The same could not be said for Republican leaners; "only" 63 percent of them were directionally stable. The remainder did not become Democrats, however, as a glance at the "directionally unstable" column in table 5.2 reveals. Just 15 percent of Republican leaners crossed to the other party. The gap on the Republican side between leaners and outright partisans is nine percentage points, which is scarcely overwhelming. To put it another way, the chances that an Independent Republican would be driven into Democratic ranks by Watergate and other shocks of the years from 1972 to 1976 are less than one in six.^[26] The directional instability of Democratic leaners is a bit lower, and scarcely higher than that of Weak Democrats: 12 percent as opposed to 8 percent.

Another source of data on change in party identification is a national sample of high school seniors and their parents who were interviewed in 1965, in 1973, and a third time in 1982.^[27]

[26] Richard A. Brody has shown that the events of 1973 and 1974 were associated with shifts away from Republican identification; see his "Stability and Change in Party Identification: Presidential to Off-Years," in Reasoning and Choice: Explorations in Political Psychology , ed. Paul M. Sniderman, Richard A. Brody, and Philip E. Tetlock (New York: Cambridge University Press, 1991).

[27] The first wave of interviews was reported in Jennings and Niemi, The Political Character of Adolescence , the first two waves in Jennings and Niemi, Generations and Politics (Princeton, N.J.: Princeton University Press, 1981). Some results from the third wave may be found in Jennings and Gregory B. Markus, "Partisan Orientations over the Long Haul: Results from the Three-Wave Political Socialization Panel Study," American Political Science Review 78 (December 1984): 1000–1018. Of the 1,562 parents in the original sample in 1965, just 898 were interviewed in 1982 (p. 1001).

This provides much more time for change than the 1972–1976 panel. The disadvantage is the nature of the two samples. The students were 17 or 18 in 1965. They could not vote and were much less developed politically than the adult population. Their parents are a better bet for our purposes, but they are still a long way from being a representative sample. For one thing, their age range is fairly restricted; three-quarters of them were between 40 and 54 in 1965. And since their children had stayed at school, they were themselves somewhat above the national average in education.^[28] These disparities do not seem to be a significant source of bias.^[29]

Data on the parental generation from the 1965–1973 and 1973–1982 panels are also summarized in table 5.2. The most striking thing about these figures is their resemblance to those from the 1972–1976 NES panel. Aggregate party switching was not notably greater in eight or nine years than in four. Twice as much time did not produce twice as much net change.^[30] Indeed, the biggest difference between the data sets is the lower directional stability of Independent Republicans in the NES panel, from 1972 to 1976. Although not so one-sided in their partisan impact, the years from 1965 to 1973 were a time of extraordinary turmoil and governmental fail-

[28] Jennings and Niemi, "The Persistence of Political Orientations: An Over-Time Analysis of Two Generations," British Journal of Political Science 8 (July 1978): 333–63.

[29] Converse and Markus, "Plus ça change," 39.

[30] The same finding emerges from comparisons of stability in two-year periods in the 1972–74–76 panel and from analysis of changes in the four-wave panel conducted as part of the 1980 National Election Study (the latter data can be found in Green and Palmquist, "Of Artifacts and Partisan Instability," 876–77). The fact that the aggregate change is not proportionate to the time available for change should inhibit anyone who thinks that a simple extrapolation of a two-year or four-year change will yield a useful measure of long-term party switching. Limiting the analysis to respondents who voted in both 1972 and 1976 produces slightly higher figures for directional stability, which is consistent with Green and Palmquist's measurement-error thesis. Looking only at voters does not alter our conclusions about the relative stability of leaners and outright partisans.

ure, full of challenges to established institutions, not least to the two parties. The modest differences in stability between leaners and outright partisans in the face of these pressures are further evidence that the former were far from neutral observers of the partisan struggle.

The younger generation, as one would expect, were considerably less constant in their party preferences, particularly during their first years after high school. Among those who had been Democrats in 1965, directional instability in 1973 ranged from 12 percent of strong identifiers to 18 percent of weak partisans and 24 percent of leaners. On the Republican side, all three identifier types were 26 or 27 percent directionally unstable. These data provide scant justification for judging younger leaners to be any less partisan than other members of their age group. Change in party identification in the next nine years was a good deal more modest, although still well above that of the parents; 7 percent of Strong Democrats, 14 percent of Weak Democrats, and 22 percent of Independent Democrats were directionally unstable. The eight-percentage-point gap between the latter two groups is just two points greater than in the 1973 reinterview. On the other hand, 16 percent of Weak Republicans were directionally unstable, compared to 13 percent of leaning Republicans and 8 percent of Strong Republicans. Once again, we find little in these comparisons to justify characterizing leaners as neutrals rather than partisans.

The earlier (1952–54–56) Michigan panel data on partisan stability have been used by W. Phillips Shively for an ingenious and widely cited defense of the conventional wisdom:

If leaners are in fact partisans who call themselves independent, then they should act like other partisans when they decide to vote for the opposite party—they should generally retain the same party loyalty, while deviating from their expected vote. On the other hand, if their reported "leaning" is simply another statement of how they intend to vote, then a change in the party for which

they intend to vote should bring a corresponding change in the party to which they lean.^[31]

Shively tested this idea by examining changes in partisan voting and party identification from 1956 to 1958 and from 1958 to 1960 among those respondents in the Michigan panel who voted in both elections in either pair of years. For example, did people who in 1956 called themselves Democrats and voted for a Democratic candidate and in 1958 voted Republican change their 1958 party identification to match their vote in that year? Since we assume party identification to be fairly stable, occasional defections in voting should not lead to a change in party identification. Thus those people who "follow" a change in vote with a change in party identification may be presumed to have a weak attachment to the party they claim to be closer to or with which they identify.

Comparing each of the three kinds of Democrats or Republicans, Shively found that strong and weak identifiers who defected to the other party tended not to change party, a finding consistent with the conventional understanding of party identification. About half the floating leaners, however, changed their identification in each pair of years, leading Shively to conclude that "based on still scant evidence, it appears that most 'leaners' should be considered independents who have used the follow-up question to indicate their intended vote."^[32]

Several problems with Shively's analysis cast considerable doubt on his conclusion, however. One difficulty, which he concedes, is the very small number of cases: just twelve vote-changing leaners in 1956–1958 and nine in 1958–1960. With cell sizes this small, a change of two respondents yields a percentage change of over twenty points—hardly an adequate basis for conclusions as fundamental as whether leaners are really independent.

There is a reason for the small cell sizes that points up an-

[31] W. Phillips Shively, "The Nature of Party Identification: A Review of Recent Developments," in The Electorate Reconsidered , ed. John C. Pierce and John L. Sullivan (Beverly Hills, Calif.: Sage, 1980), 234.

[32] Ibid., 235.

other problem with this analysis. The cell sizes are so small because most leaners did not change their vote from one party to the other in either two-year period. Over 80 percent of partisan Independents voted for the same party in both years in each of the comparisons Shively examined. In fact, weak partisans were more likely than leaners to vote for different parties from one election to the next, a finding ignored by Shively that speaks directly to the point of whether leaners are really so different from outright partisans. Since the vast majority of leaners were constant party-line voters, one should not single out the handful of exceptions and use their behavior to characterize the entire group.

We have exploited the 1972–1976 panel to repeat Shively's analysis with two successive presidential elections rather than contrasting a presidential vote with one cast in an off-year election. There are slightly different ways of doing this, and Shively's article did not make clear which alternative he used. We chose to take all 1972 strong identifiers who voted in both 1972 and 1976 and divide them into two groups—those whose votes in the two years were cast for the same party's presidential candidate and those who voted for different parties. Then we compared the proportion of each group who changed parties from 1972 to 1976. The same procedure for weak partisans and leaners produces for each category of party identification a probability that people in that group would switch parties to "follow" their vote intention. These calculations are presented in table 5.3.

This quasi replication yields findings similar to Shively's about the tendency of the different groups of identifiers to "follow" their votes. The leaners who voted for different parties in the two years are indeed more likely to switch parties in accordance with their changed votes. The difference in changed identification between the constant and floating voters is greater for leaners (25 percent) than for outright partisans (10 and 12 percent). The number of cases is considerably larger than in Shively's analysis, but the same considerations nevertheless destroy one's faith that these calculations show leaners to be truly independent. For one thing, 70 percent of

all leaners voted for the same party's presidential candidate in both elections. This level of constancy, just a few percentage points below that of the outright partisans, is inconsistent with the notion that the leaners' partisanship only reflects short-term forces.^[33]

Less than a third of the floating leaners switched parties, while Shively's theory predicts that all of them would. In other words, less than 10 percent of the leaners changed their vote between 1972 and 1976 and switched parties to reflect that changed vote choice. The behavior of the other 90 percent was consistent with conventional expectations of partisans: either voting the same way in consecutive elections or, if defecting, nevertheless retaining their party identification.

Much more comprehensive evidence on the leaners' relative susceptibility to short-term and interelection forces comes from John R. Petrocik's analysis of 151 surveys conducted for congressional, state, and local candidates from 1980 through 1983:

Weakly identified Democrats appear slightly less responsive to short-term forces than leaning Democrats. . . . The

[33] These calculations count as constant voters those people who voted for the same party's presidential candidate in the two elections, even if both votes were for the other party's candidate. That is, Strong Democrats who voted for Nixon in 1972 and Ford in 1976 are not distinguished from Strong Democrats who voted for McGovern and Carter. By the same token, Democrats in 1972 who voted for Nixon and then Carter are mixed in with Democrats who voted for McGovern and Ford. We repeated our analysis after removing those respondents whose 1972 vote did not match their 1972 party identification, defining "constant voters" as those who voted in both years for the candidate of the party with which they identified in 1972 and "floaters" as those who cast a party-line vote in 1972 and switched to the other party's candidate in 1976. This increased the proportion of constant voters by eight to 10 percentage points in each party-identification category. The difference in change in party-identification between floaters and constant voters was about the same for leaners and weak partisans but rose to 32 percent for the strong partisans.

We also categorized respondents who voted in both elections according to their 1976 party identification and then repeated the original calculation. The results were very similar to those in table 5.3, except that leaners and strong partisans had about the same proportion of constant voters.

difference is statistically insignificant. With Republicans, the difference between weak and leaning identifiers is the reverse of expectations and (fortunately) also statistically insignificant. . . . Moreover, independents are not consistently more responsive to interelection tides than other partisans.^[34]

Attitudes toward the Two Parties

A dozen years ago, Philip Converse and Gregory B. Markus referred to "a rising proportion of the electorate refusing to report identification with either party, and insisting on being classed as independents, thereby swelling the neutral middle of the party identification continuum."^[35] Although skeptical of the verb "insist," we think this passage does identify the heart of the controversy about Independents: Is "neutral" an appropriate label only for the Pure Independents, or does it accurately describe the attitudes toward the parties of all three varieties of Independents? As we have seen (p. 79) Warren Miller wrote that leaners should be considered true Independents partly on the basis of the weakness of "their psychological ties to the parties." He did not mention any actual measures of such affect. The Michigan questionnaires have numerous items about general and specific attitudes toward the two parties that we will analyze in this section to see if leaners are less favorable to their own party, friendlier or more hostile to the other party, or merely more indifferent than outright partisans.

One such measure is the "feeling thermometer" on which respondents are asked to indicate the warmth of their affect for various persons and groups: "Ratings between 50 degrees and 100 degrees mean that you feel favorable and warm toward the person. Ratings between 0 and 50 degrees mean that

[34] John R. Petrocik, "An Expected Party Vote: New Data for an Old Concept," American Journal of Political Science 33 (February 1989): 55, emphasis in original.

[35] Converse and Markus, "Plus ça change," 33.

you don't feel too favorable." "Democrats" and "Republicans" have been rated often on the thermometer, but "Democratic Party" and "Republican Party," which are probably more pertinent objects, have been rated only since 1978.^[36]

We are not the first researchers to use the feeling thermometer as an aid to classification of leaners. Charles H. Franklin and John E. Jackson reported their findings as follows:

There is a very strong, monotonic relationship between the seven-point [party-identification] scale and an alternative measure of partisanship, defined by the difference between Democratic and Republican feeling thermometer scores. . . . Thus the seven category traditional variable is consistent with the relative evaluation of each party, or with the unidimensional concept of partisanship.^[37]

Unfortunately, Franklin and Jackson said no more about the relationships they observed, hence we have no way of reconciling their report with our very different findings, some of which may be found in table 5.4.

As expected, strong partisans on both sides are by far the most favorably inclined toward their own party, the coolest toward the other party, and generally least likely to give an indifferent 50 rating to either party. In 1988, Weak and Independent Democrats evaluated their party almost identically. They were not far apart in their views of the Republican party; more Weak Democrats gave it both negative and positive ratings. These two groups' mean thermometer readings for the Republican party were both 50. On the Republican side, differences between weak partisans and leaners were also modest (and short of statistical significance) when rating their own party. Thirty-four percent of the Independent Re-

[36] Differences between leaners and weak partisans on thermometer ratings of "Democrats" and "Republicans" were tiny to nonexistent from 1972 through 1980.

[37] Franklin and Jackson, "The Dynamics of Party Identification," 963.

publicans rated the Democratic party favorably, as did 27 percent of Weak Republicans. On the other hand, just 31 percent of Republican leaners were neutral about the Democratic party, compared to 40 percent of Weak Republicans. The data for 1980 and 1984, not shown here, are equally cold comfort to scholars who believe that leaners and weak partisans differ in their evaluations of either party.^[38]

Attitudes toward the two parties can also be measured by responses to the four "master code" questions: "Is there anything in particular that you like (don't like) about the Democratic (Republican) party?" We classified respondents by the combinations of positive and negative answers they gave. Most respondents fell into one of these six categories:

1. +Democrats, –Republicans
2. +Democrats, ±Republicans^[39
] 3. ±Democrats, –Republicans
4. –Democrats, ±Republicans
5. ±Democrats, +Republicans
6. –Democrats, +Republicans

A seventh category is respondents who gave the same type of answer (positive, negative, or balanced) about both parties. An eighth category accommodates respondents with nothing to say about either party. We define the first three categories as pro-Democratic, the next three as pro-Republican, and the seventh as balanced. Table 5.5 shows what percentage of each type of party identifier made pro-Democratic, balanced, and pro-Republican comments in 1988, as well as the percentage who had nothing to say about either party.

[38] The data for 1980 may be found in our "Partisan Affinities," 171.

[39] Most respondents whose comments were coded ± had both favorable and unfavorable things to say about the indicated party. This category also includes respondents who had nothing to say about one party and answered one or both questions about the other party. Readers should note, therefore, that a small number of respondents in the "favorable" categories in table 5.5 said nothing about the party we say they favor; they made only negative responses about the other party.

Sixty-one percent of Pure Independents had neither favorable nor unfavorable opinions of either party, more than twice as many no-comment responses as any other category. This is another illustration of the difference between the largely apathetic Pure Independents and the more opinionated leaners and outright partisans. Because so many of these genuine Independents have nothing to say, even the number with balanced comments is low: 7 percent in 1988, 13 percent in 1980, and 22 percent in 1976. The leaners, on the other hand, are no more reticient than weak identifiers, no more inclined to offer balanced observations about the two parties. Just as many of them have only favorable comments about their own party and as few have anything good to say about the opposition. The findings for other years are not always quite as gratifyingly neat as those for 1988, but they all show that the leaners are similar to

weak partisans in their positive affect for their own party, their lack of enthusiasm for the other party, and their disinclination to make balanced comments about either party.^[40]

These master-code items provide the weakest evidence for our thesis. Other measures of partisan affect show the leaners to be as favorably disposed to their respective parties as the weak partisans; in some cases, more so.

This conclusion emerges from analysis of responses to the question about which party "would be most likely to get the government to do a better job in dealing with [the most important problem the country faces, in the respondent's opinion]" (see table 5.6). From year to year, the public's assessments of the two parties change dramatically, as do the issues on which they base those assessments. In 1980, the Democratic party was held in much lower esteem than in previous years by all types of identifiers. In 1982, the Republican party suffered this fate.

Over the years, leaners have been at least as likely as their weak partisan counterparts to choose their own party. More Independent Republicans than Weak Republicans selected their own party in four of the last five presidential elections. By the same token, they were less inclined to prefer the Democratic party in every year except 1972. In the same period differences between Independent and Weak Democrats in preferences for the Democratic party were never greater than three percentage points and since 1980 Weak Democrats have been a bit more inclined to choose the Republican party to solve the most important problem.

Except in 1972, Independent Democrats were somewhat more likely than Weak Democrats—the gap ranges from seven to eleven percentage points—to say there was no difference between the parties. This scrap of contrary evidence is balanced by the opposite finding on the Republican side where weak partisans tended to exceed leaners in denying that either party would do a better job.

[40] The counterpart of table 5.5 for 1976 and 1980 may be found in our "Partisan Affinities," 173.

The pattern displayed in table 5.6 was also found in responses to questions asking which party handles inflation better and is better at keeping the country out of war. By the same token, leaners in each party gave appropriately partisan answers when expressing their feelings about whether "Democrats" and "Republicans" have "too much influence in American life and politics," "just about the right amount" or "too little influence."

Behavior versus Self-Image

Up to this point, we have measured leaners' partisanship by looking at their voting and attitudes toward the parties. In all these respects they resemble outright party identifiers.^[41] These similarities fade when we look at another dimension—leaners' beliefs about their own behavior. A good starting point is the pattern of responses to this question, asked by the Michigan researchers between 1952 and 1980: "Have you always voted for the same party or have you voted for different parties for President?" Responses for eight presidential elections are tabulated in table 5.7.

Unlike the findings reported so far, these data support the conventional wisdom about Independents. Table 5.7 depicts the sort of smoothly monotonic relationship between party identification and voting loyalty that the traditional literature would lead us to expect. The great majority of strong partisans, about half the weak partisans, less than a third of partisan Independents, and about one-sixth of Pure Independents say they always vote for the same party's presidential candidate.

Viewed in isolation, these findings seem perfectly reasonable. If being an Independent implies lack of partisan ties, we would expect little year-to-year consistency in the way self-professed Independents vote. At every election they would

[41] Their attitudes on issues also resemble those of outright partisans in each party, as we will show in chapter 7.

make their decisions after pondering the issues, judging the candidates, and reflecting on the nature of the times. Table 5.7 suggests some partisan regularity on the part of the leaners, but at a much lower level than for weak partisans. Relationships similar to those shown in table 5.7 appear to be the principal evidence adduced in The American Voter in support of the fundamental proposition that the party identification scale measures a unitary dimension that proceeds in linear fashion from one pole to the other.^[42] The same item was used by Nie

[42] The only tables in The American Voter separating leaners and weak partisans on any measure of partisan affect or behavior are those about respondents' reported consistency in past presidential voting or past changes in party identification (pp. 125, 126, and 148). By the same token, the only such table in Elections and the Political Order (p. 218) is about claims to have voted for the same or different parties' presidential candidates.

and his associates to illustrate "the change in the strength of party ties from the 1950s to the beginning of the 1970s."^[43] Some scholars believe that these retrospective reports on party loyalty are convincing evidence in favor of the conventional wisdom, even in the face of our findings about the actual voting choices of leaners.^[44]

Table 5.7 becomes problematic when compared to other data on party loyalty. Although Strong and Weak Republicans are much more likely than their Democratic counter-parts to vote for their party's presidential candidate in any given year, they are somewhat less likely to claim consistency in all past elections. Judging by their reported votes just after the election, Strong Republicans averaged 97 percent loyalty in the eight elections from 1952 through 1980. Yet during the same period only 73 percent of Strong Republicans said they always voted for the same party. These findings are difficult to reconcile, even if one assumes complete turnover in the ranks of disloyal Strong Republicans from one election to the next.^[45] In order to take the data in table 5.7 as a measure of actual voting behavior, one would have to believe that at every level of party identification, Democrats are more loyal presidential voters than Republicans and that, in either party, the weak identifiers are more loyal than the leaners. This is the opposite of findings from data based on reports of voting in a single election.

The problem might be one of question validity: Which of the two questions is the better measure of party regularity? The item about voting in the immediately past election asks about a recent act; it does not severely test the respondent's

[43] Norman H. Nie, Sidney Verba, and John R. Petrocik, The Changing American Voter , enlarged ed. (Cambridge: Harvard University Press, 1979), 54.

[44] See, for example, Herbert M. Kritzer and Robert B. Eubank, "Presidential Coattails Revisited: Partisanship and Incumbency Effects," American Journal of Political Science 23 (August 1979): 617n–18n.

[45] Another possibility is substantial turnover from one election to the next among people claiming to be Strong Republicans. The data in table 5.2 lead us to reject this alternative.

memory. This is not the case with the other question, which calls on respondents to remember behavior that may not be important to them and occurred at various times in the past. Doubts in general about such retrospective data are sustained by Richard A. Brody's analysis of panels in which the same respondents were reinterviewed at two-year intervals. Brody found that 71 percent of those people who identified with one party in 1956 and the other party in 1958 denied that they had ever switched parties. In the 1972–1974 panel, 58 percent of switchers reported that they had always identified with the same party.^[46]

The 1972–1976 panel provides an opportunity to verify respondents' claims about their voting histories by comparing responses to these three items: (1) 1972 reports about past consistency, (2) the reported vote in 1972, and (3) 1976 reports about past consistency. Those respondents who said in 1972 that hitherto they had always voted for the same party and then voted for the other party's candidate in that year should without exception report four years later that they had voted for different parties. By the same token, people who claimed consistency and voted consistently in 1972 should all repeat their claim that they always voted for the same party. In fact, 30 percent of the consistent voters said in 1976 that they voted for different parties, and 27 percent of the actual inconsistents claimed that they had always been faithful to one party.

If such misreporting reflected nothing more than memory lapses, we would expect it to be evenly distributed across different categories of partisans, modified only by variations in political involvement. But a question about always voting for the same party is probably not value-free. Some people prize a self-image as loyal partisans, and others might value the appearance of open-mindedness. The severity ("always")

[46] Brody, "Stability and Change." Because of the obvious potential for motivated memory lapses when responding to the question about past changes in party identification, we dismiss The American Voter table (p. 148) that shows a monotonic relationship between this variable and the strength of party identification.

of the question compounds the problem, as does its location in the questionnaire. The item on past party regularity was asked shortly after the series of questions on party identification. Independents, having just declared this status, might want to describe their behavior consistently with their self-identification. Respondents unsure about their voting history might take the self-description as a cue: "I just called myself an Independent; therefore, I must vote for different parties."^[47]

In short, there are reasons to suspect that some people who had been wholly faithful to one party might claim otherwise and that some inconsistent voters might insist on their loyalty. If this were true, we would expect that avowed partisans whom we discovered to be inconsistent voters in 1972 and 1976 would be more likely to make the false claim that they had always voted for the same party. This is what we found. Indeed, almost every one of these "concealed defectors" was a strong or weak partisan. Similarly, we found that both leaners and Pure Independents who voted consistently were much more likely than consistent partisans to deny that they always voted for one party.

The same pattern of systematic misreporting emerged when we examined the claims about past voting of only those respondents who in 1976 were old enough to have voted in just one previous presidential election. Along with the rest of the sample, they were asked in the preelection interview if they had voted in any past presidential elections for which they had been eligible to vote. Those who answered affirmatively were

[47] In 1980, the question did not follow the standard set of items about party identification. Instead, it followed a new battery, tapping attitudes toward partisan commitment, the last item of which asked: "Do you ever think of yourself as a political independent, or not?" Leaners were by far the most likely to respond affirmatively, followed by Pure Independents, as the following figures for each type of identifier demonstrate: SD, 22; WD, 31; ID, 78; I, 64; IR, 79; WR, 33; SR, 27. It is no more surprising than for earlier years, therefore, that when asked to report on their partisan consistency as a sequel to affirming their independence, all three types of Independents in 1980 overwhelmingly reported not having voted for the same party over the years. We will discuss these 1980 questions in chapter 9.

asked whether they had always voted for the same party for President or for different parties. Their replies, tabulated against the seven-point party-identification scale, are presented in table 5.8.

The responses of these young voters differ in one respect from those of the entire sample: Republicans are more likely to profess party regularity. Otherwise, the distribution of responses resembles that for the whole sample. Strong identifiers are most likely to claim fidelity to one party, followed by weak identifiers, leaners, and Pure Independents. The catch, of course, is that none of these respondents could have voted in more than one prior presidential election. Thus it is difficult to see how any of them could have managed to avoid always voting for the same party. Yet only two-thirds of these young voters would acknowledge the irrefutable.

Given the findings in table 5.8, we believe that replies to the question about past voting measure self-image as well as

previous behavior. Considering this possibility, the authors of The American Voter wrote:

In this discussion we assume that the concealed partisan is less likely to distort his voting record than his description of his party attachment; that is, we assume that what the undercover partisan values is chiefly the designation "Independent." To the extent this is untrue, the analysis of voting consistency by strength of party identification fails to enhance our understanding.^[48]

With the benefit of many years of hindsight since these words were written, a better conclusion might be that the value the "undercover partisan" places on the designation Independent is sufficient to distort his memory.

The halo effect of being an Independent was anticipated by the Michigan researchers when they designed the questions about party identification. Warren Miller told one of us that the probe about leaning toward a party was based at least partly on the assumption that "being an Independent was a matter of civic virtue; for many people it was more socially acceptable" than being a Democrat or Republican. The probe, in other words, was designed to smoke out the Democrats and Republicans who were reluctant to come clean about their partisanship. The message of this book is that the goal was important and the method effective. If these simple propositions had been acted on consistently by the Michigan scholars and their successors and critics, this book would have been unnecessary.

The patterned misperceptions about past voting are useful data in several respects. On the negative side, they show why the voting-history item does not provide valid evidence for the conventional position on Independents. More generally, these findings are another nail in the coffin of the notion that

[48] Campbell et al., The American Voter , 125n. The quoted passage refers to Pure Independents' answers to the question about past partisan consistency.

long-term retrospective data are a measure of reality.^[49] Positively speaking, the findings help us think about the leaners' initial reluctance to acknowledge an affiliation consonant with their beliefs and behavior. Leaners resemble avowed identifiers in their voting choices and opinions about the two parties, but evidently are uncomfortable about being partisan. We interpret the data here and in chapter 4 as evidence that the leaners are party identifiers who hesitate to own up to their party ties despite an affinity scarcely less stable than that of Weak Republicans or Democrats. We suspect that their reluctance reflects at least in part the parties' low popular image. We view the second question in the NES party identification series as a probe to overcome that reluctance.

A completely different interpretation has been developed by Warren Miller, who argues "the utility of distinguishing between the overlapping concepts of partisanship and party identification."^[50] The leaners "are demonstrably and overtly partisan," but they are not party identifiers.^[51] Miller's distinction is rooted in the different wording of the successive NES questions. Party identification is a durable self-identity that is appropriately measured by the first question's emphasis on the long view: "Generally speaking, do you usually think of yourself . . . ?" In contrast,

the follow-on question, "Do you think of yourself as closer to the Republican or Democratic party?" does not attempt to elicit a qualified or limited sense of an "enduring engagement of partisan feelings with self-identity;" the question is asked only in the present tense, and it calls only for a cognitive assessment of current circum-

[49] For more evidence on this point, see Richard G. Niemi, Richard S. Katz and David Newman, "Reconstructing Past Partisanship: The Failure of the Party Identification Recall Questions," American Journal of Political Science 24 (November 1980): 633–51.

The flaw in the question about past voting history that is explained in this chapter led to its deletion from the NES questionnaire after 1980.

[50] Miller, "Party Identification," 557.

[51] Ibid., 558.

stance. The answers may indicate partisanship but they do not reflect a sense of party identification.^[52]

Miller acknowledged that "the reasons for this strict interpretation of the meaning of party identification are more firmly grounded in theory than in data."^[53] His distinction is not germane to our central purpose, which is to evaluate the assumptions about Independents that underlay the widespread discussions of dealignment described in chapter 1. Leaners exhibit none of the uncommitted traits attributed to Independents, hence trends in party identification since the late 1960s do not pose much of a threat to the American political system. From this perspective, it may not matter whether leaners are more properly considered party identifiers or only long-term fellow travelers.

In this chapter we have examined measures of partisan affinity other than vote choice to explore leaners' degree of neutrality and commitment. We conclude that the simplest explanation of their voting patterns is correct: They vote as they do because they are partisan, not neutral, about the two parties. Their partisanship is evident not only in their voting but in their choice of presidential primary elections, the stability of their party identifications, and their attitudes toward the two parties. In all these respects they resemble outright partisans and differ from Pure Independents, who profess no attachment to either party.

[52] Ibid. The quoted passage is from Converse and Pierce, "Measuring Partisanship," 143.

[53] Ibid.

6—
Age, Education, and Dealignment

Our analysis up to this point has been largely static. What we found about 1988 or 1976 was equally true of the 1950s. Nothing changed very much, except that Pure Independents have even less civic virtue than they did a generation ago. Having made our case that leaners are essentially partisans who have nothing in common with the neutral Pure Independents, we turn now to describing the people most responsible for the growing numbers of Independents, a trend that we will call dealignment without prejudice to our argument about the leaners' partisanship.

We begin with the familiar and plausible proposition that independence, always more common among the young, was particularly appealing to those who reached voting age at the time of the Vietnam War. Was dealignment confined to the young, or did it also reflect abandonment of party identifications by some older Americans? A second question concerns the persistence of these independent identifications. Are the baby boomers becoming more partisan as they reach middle age? In other words, how general was dealignment, and how durable?

A third line of inquiry concerns the next generation, to whom Vietnam is history and Eugene McCarthy as remote as Joseph McCarthy. Are they as disinclined to identify with a

party as their immediate predecessors, or do they have the more modest inclination to independence of those who first could vote before the death of President Kennedy?

This is not unexplored territory, but as we will see, differentiating leaners from Pure Independents makes possible a clearer description of trends in party identification in the past twenty-five years. The distinction is central to our analysis of education, the concluding topic of this chapter and the only other demographic variable helpful in understanding dealignment.

Age and Partisanship

Most scholars agree that the increase in Independents was a reaction to "shocks" that seemed sharpest in the decade that began in the mid-1960s: the failure of political institutions, including the two parties, to achieve satisfactory solutions to the country's most salient problems, the Vietnam War and racial conflict.^[1] We will examine these interpretations directly in the next chapter. Here we concentrate on those citizens who seemed most susceptible to such shocks. Young people are widely believed to be more politically labile. Whether because they are more impressionable than their elders or because their political "anchors" are weaker, their reaction to dramatic political events is sharper and more durable than that of older citizens. It seems reasonable, then, to expect that young people would be found "at the cutting edge" of dealignment, that is, more inclined to be Independents.^[2]

[1] The most meticulous tracing of declining party identification is in Philip E. Converse, The Dynamics of Party Support (Beverly Hills, Calif.: Sage, 1976). He found the first marked drop in "mean party identification strength" occurred between 1964 and 1966; a second from 1971 through 1974. Converse did not measure an increase in Independents as such, however. His measure of party identification strength assigned a score of 3 to strong partisans, 2 to weak partisans, 1 to leaners, and 0 to Pure Independents.

[2] M. Kent Jennings and Gregory B. Markus, "Partisan Orientations over the Long Haul: Results from the Three-Wave Political Socialization Panel Study," American Political Science Review 78 (December 1984): 1015.

The American Voter reported that even in the 1950s "young people, just entering the electorate, are more likely than any of the older age groups to call themselves Independents."^[3] If Independents were progressively scarcer in older age groups, this might be a life-cycle effect : "party identifications strengthen with age."^[4] The mechanism is thought not to be age as such; rather, age is a surrogate for the number of years an individual has been affiliated with a party. "Perhaps older citizens are more staunchly partisan because the repeated use of party labels to interpret and understand the political world through the years reinforces and strengthens one's partisan ties."^[5] This interpretation "fits very well a more general thesis that group identification is a function of the proportion of a person's life he has been associated with the group."^[6]

But any analysis of age and partisanship must account for conflicting interpretations of a relationship observed at a single point in time or over a brief period. If older respondents in the 1950s were likelier to be strong partisans, this might reflect not so much their place in the life cycle as the fact that their political perspectives were formed in an era when partisanship was more enthusiastic, widespread, and unquestioned. This is a generation effect , defined by Philip Converse as "a kind of early imprinting which . . . will always leave its characteristic mark" distinguishing the affected age cohort from other generations not so affected by the imprinting

[3] Angus Campbell et al., The American Voter (New York: John Wiley and Sons, 1960), 161. This seemed to be the accepted position, but there were dissenters. See William H. Flanigan and Nancy H. Zingale, Political Behavior of the American Electorate , 4th ed. (Boston: Allyn and Bacon, 1979), 68.

[4] Philip E. Converse, "The Concept of a Normal Vote," in Angus Campbell et al., Elections and the Political Order (New York: John Wiley and Sons, 1966), 18n.

[5] Gregory B. Markus, "Dynamic Modeling of Cohort Change: The Case of Political Partisanship," American Journal of Political Science 27 (November 1983): 721.

[6] Campbell et al., The American Voter , 163.

event or condition.^[7] The defining characteristics of a generation effect are its limitation to an age-defined group and its durability. A period effect , on the other hand, makes its mark irrespective of age.

These formulations are somewhat problematic. No matter how much one may believe in the susceptibility of the young, it is difficult to imagine an event that would imprint one generation for life without leaving a discernible mark on at least some older people. Moreover, the effect of a gripping event may linger for only a few years before the familiar life cycle of strengthening partisanship resumes. If so, one must be chary about judging any relationship to be evidence of a generation effect. One might also discern an apparent period effect limited to people beyond the first flush of youth, but still short of middle age. As we will see, these caveats are illustrated by evidence on the party identification of different age groups from the 1960s through 1990.

We begin by examining the proportions of various age groups in each of the three categories of independent identifi-

[7] Converse, The Dynamics of Party Support , 74. For evidence of a strong life–cycle effect in the "steady–state period" from 1952 to 1964, see ibid ., chap. 3. A number of studies have shown that for at least a decade, beginning in the mid–1960s, party identification did not strengthen as people grew older. See ibid.; Paul R. Abramson, "Generational Change and the Decline of Party Identification in America: 1952–1974," American Political Science Review 70 (June 1976): 469–78, "Generational Replacement and Partisan Dealignment in Britain and the United States," British Journal of Political Science 8 (July 1978), and "Developing Party Identification: A Further Examination of Life-Cycle, Generational, and Period Effects," American Journal of Political Science 23 (February 1979): 78–96. See also M. Kent Jennings and Richard G. Niemi, "The Persistence of Political Orientations: An Over-Time Analysis of Two Generations," British Journal of Political Science 8 (July 1978): 333–63; Norman H. Nie, Sidney Verba, and John R. Petrocik, The Changing American Voter , enlarged ed. (Cambridge: Harvard University Press, 1979), 59–73.

These studies seldom distinguished among types of Independents. The exceptions are Abramson and Converse, who used the coding scheme described in note 1.

cation. These data for white respondents in each presidential election year from 1952 through 1988 are arrayed in table 6.1. The only one of these groups that is truly neutral between the parties—Pure Independents—had no special youth appeal in the 1950s. Not until 1964 were people under the age of 29 significantly more likely than the entire population to be Pure Independent. In 1968 a monotonic relationship between youth and pure independence emerged. Since then, the facts have matched the conventional wisdom that young people are especially disposed to independence and older people are not.

Much the same is true of Independent Democrats, although here reality matched the stereotype more readily. In 1952 everyone up to the age of 65 was equally likely to be an Independent Democrat. This category did not have any special youth appeal until 1956, when a consistent pattern began. Consistency is the last term one would apply to the relationship between youth and Independent Republicans. They are the only group to display a satisfying monotonic relationship in the 1950s, but this disappeared in 1960 and 1964. For a dozen years thereafter, the Independent Republican identity also seemed to appeal more to young people. In the 1980s, the only age group notably less attracted to this label was people 65 and over.

By 1968, the three varieties of independent identification were consistently most popular among people in their twenties. What is more, increasing proportions of youngsters reaching voting age were eschewing outright partisanship in favor of one or another sort of independent status.

Did this trend continue unabated through the 1970s and 1980s? Was the popularity of independence limited to the youngest set, or did older citizens come to share it? And what happened to the children of the sixties as they entered the second decade of their political maturity? Did they continue the disavowal of partisanship with which they entered political life? To answer these questions, we have made use of not only the simple cross-tabulations in table 6.1, but also analysis of eight age cohorts, which let us track party identification

in each age group as it moved through the life cycle. In order to reduce sampling error by accumulating larger numbers of cases, we used eight-year cohorts, the youngest of which in 1960 were those respondents who had not been old enough to vote in 1952. We began with 1960 because it preceded the "shocks" to party identification described by Converse; 1968, of course, was the year of peak political tumult. Because of the ratification of the 26th Amendment in 1971, the group too young in 1976 to have voted eight years earlier had an age span of eleven years and our youngest cohort in 1984 had an upper age limit of 25 rather than 28. The availability of 1990 NES data shortly before this book went to press let us take one more look at the earlier cohorts and also at a new cohort of those aged 18 to 23. Our oldest cohort takes in everyone over 74 in 1990. The cohort data are presented in tables 6.2, 6.3, and 6.4.

All three varieties of independence increased, in a different pattern for each category. The proportion of Pure Independents grew most among young people, but we did find a limited amount of change among some older cohorts. There was no change among people who were at least 45 in 1960 (Cohort VII) and a four-percentage-point increase in Pure Independents from 1960 to 1968 in the next two cohorts. Thereafter, the only shift in these three cohorts was a gradual decline in their proportion of Pure Independents. The oldest group in which party identification declined sharply was Cohort IV, who came of age in the Eisenhower era and weathered the Johnson administration without gaining Pure Independents. But in 1976, when Cohort IV was 37 to 44, approaching middle age, its proportion of Pure Independents nearly doubled, to 19 percent. Just 8 percent of them were Pure Independents in 1984 and 11 percent in 1990, which is about where they were in 1960 and, for that matter, where everyone 45 and older was in 1960.

Among the prime candidates for disillusionment are those who reached political maturity in the 1960s, our Cohort III. Eighteen percent were Pure Independents in 1968, at least

half again as many as any other cohort in 1968. Cohort III were at the same place in 1976, when 19 percent of them eschewed any partisan affiliation. By 1990 this dropped to 10 percent, the norm for everyone but the youngest cohort thirty years earlier.

In 1976 the next generation, Cohort II, had spent their entire brief political lives in a time of disorder, discontent, scandal, and apparent bipartisan failure. Twenty-two percent of them were Pure Independents in 1976, the highest proportion for any cohort in any year. By 1990 this figure had dropped by half, to 11 percent, the same as the two preceding cohorts in 1990.

Next is Cohort I, those aged 18 to 25 in 1984, whose political perspectives were largely developed in the era after Vietnam and Watergate. Fourteen percent of them were Pure Independents in 1984. This is substantially lower than the reading for the first appearance of the two previous cohorts and is a statistically insignificant two percentage points higher than the proportion of Pure Independents in Cohort IV at its debut in 1960. Six years later Cohort I's share of Pure Independents rose from 14 to 16 percent. This may be nothing more than a blip, which we consider more plausible than taking it as a portent of another round of dealignment.

Finally there is the truncated Cohort O, aged 18 to 23, who made their debut in 1990. Their relative youth may explain the 18 percent of this group who were Pure Independents.

In 1976 the three youngest cohorts had about the same one-fifth proportion of true partisan neutrals. In view of their age range, from 18 to 44, we consider this evidence of a limited period effect that stilled whatever relationship between age and increasing partisanship would usually be associated with movement through the life cycle. The three cohorts were also fairly similar eight years later, but at a sharply diminished level of independence. And by 1990 Cohorts II–IV had virtually the same proportions of Pure Independents as their counterpart age groups in 1960. In other words, after its suspension in the time of troubles, the familiar life cycle of strengthening partisan identification seems to have resumed in the (temporarily at least) more reassuring atmosphere of the Reagan administration. With the advantage of more time for hindsight, we can see that Paul Abramson was premature in concluding that the Vietnam-Watergate cohorts were fated to remain unusually withdrawn from partisan commitment:

The overall weight of evidence for nearly a quarter of a century casts doubts on the life-cycle formulation. Instead, the overall results show considerable stability in partisan strength within cohorts as they age, suggesting that once formative socialization occurs persons tend to

retain their level of partisan strength as they move through the life cycle.^[8]

Table 6.2 also displays the sharp differences betwen Cohorts II–IV and older citizens, those who were at least 29 in 1960 and therefore were well into their thirties when the time of troubles began. This generation gap can be seen most clearly in the 1976 column. In that year, one in five respondents under 45 were Pure Independents, double the proportion in the rest of the population. The proportion of Pure Independents among Cohorts V, VI, and VII was scarcely greater in 1976 than in 1960; differences for single cohorts observed in table 6.2 are below the level of statistical significance.

This conclusion is important enough to warrant repetition: There was no increase in partisan neutrality among Americans who had reached their forties before the nation was convulsed by Vietnam and racial conflict. The increase in Pure Independents was concentrated entirely among those who attained voting age in the 1960s and 1970s. In this conclusion we differ from Converse, who reported that strength of partisanship declined among the elderly and middle-aged as well as the young; and Warren Miller, who saw the disruptions and frustrations of the era creating "a period effect felt throughout the electorate."^[9]

[8] Abramson, "Developing Party Identification," 91. Later studies showing a revival of partisanship include Warren E. Miller and J. Merrill Shanks, "Policy Directions and Presidential Leadership: Alternative Interpretations of the 1980 Presidential Election," British Journal of Political Science 12 (July 1982): 308–9; and J. Merrill Shanks and Warren E. Miller, "Policy Direction and Performance Evaluation: Complementary Explanations of the Reagan Elections," British Journal of Political Science 20 (April 1990): 160.

[9] Converse, The Dynamics of Party Support , 95–97. In 1968 Converse's three cohorts were aged 25–40, 41–56, and 57 and older. Miller, "The Electorate's View of the Parties," in The Parties Respond , ed. L. Sandy Maisel (Boulder, Colo.: Westview Press, 1990), 103. Miller found particularly significant contributions to dealignment by the young. His findings were based on the ratio of strong partisans to Pure Independents among four-year cohorts. Using a differ-ent measure, Miller and Shanks reported a decline in partisanship among people who came of age as early as the 1940s; see "Policy Directions and Presidential Leadership," 308.

In contrast to Pure Independents, who continued to gain among young people in the 1970s, Democratic leaners reached their peak with this constituency in 1968. They lost ground by 1976 and continued to do so in the 1980s. In 1968, 20 percent of all whites under 29 (Cohort III) were Independent Democrats. Their proportion of this cohort fell to 14 percent in 1976. It fell again to 9 percent in 1984 and then rose to 14 percent in 1990. Cohort II, who arrived on the scene in 1976, were 16 percent Independent Democratic in that year, 13 percent in 1984, and 14 percent in 1990. The proportions were similar for Cohort I in 1984 and 1990 and Cohort O in 1990.

Democratic leaners became more numerous in the 1970s not because young people were more attracted to this identity, nor as a result of a broadly-based shift in the same direction. Instead, the increased proportion of Democratic leaners reflected the baby boom, which generated many more young citizens, more likely than older people to be Independent Democrats, but not more so in the 1970s than previously.

In all older cohorts (IV–VII) the increase in Independent Democrats from 1960 to 1976 was barely perceptible (and short of statistical significance); thereafter, there was a trace of movement in the other direction. Even more than with Pure Independents, increases in Independent Democratic identification were confined to the younger crowd.

Republican leaners resemble the other two types of Independents in one respect: their increase was greatest among younger cohorts. (In saying this we acknowledge the mild swings of Cohort VI as well as the sharp increase among Cohort V. After registering no change from 1960 to 1984, this latter group showed 17 percent Independent Republicans in 1990, when most of its members were in their sixties.) Cohort IV, who came of age in the Eisenhower administration, were just 5 percent Independent Republicans in 1960. This proportion doubled by

1968. The next cohort was 16 percent Republican leaners in 1968, and the three succeeding cohorts displayed about the same proportion when they made their debuts in 1976, 1984, and 1990.

In contrast to the other Independents, however, Republican leaners did not become scarcer as the 1970s passed into history; a decline from 1984 to 1990 was notable only in Cohorts VI and III. And as we just observed, when the next three cohorts arrived, they included as many Republican leaners as had their counterparts in past years.

How can we summarize this welter of findings? First, it is

clear that most of the increase in all three types of Independents came from those who entered the electorate from the middle 1960s through the 1970s. Fully 54 percent of Cohort III, aged 21 through 28 in 1968, were one or another type of Independent in that year. The same was true of 52 percent of Cohort II in 1976. In addition, Cohort IV became somewhat more independent. The older cohorts seemed scarcely affected. There was no increase in Independents among people who were at least 30 in 1960, and only a modest amount among those eligible to vote in that year. This finding is consis-

tent with the proposition that longtime party identification builds resistance to short-term influences.

Second, it is equally clear that the cohorts largely responsible for the increase in Independents are becoming more explicitly partisan as they age. The familiar life-cycle effect of the steady-state period is reemerging. This trend is most complete with respect to Pure Independents, who in 1990 were a much smaller proportion of Cohorts II–IV than in 1968 or 1976.

Third, youngsters reaching voting age after 1976, Cohorts I and O, were less averse to identifying with a party than their counterparts in 1968 and 1976, but still were a bit more inclined to be Independents than were Cohort IV, those arriving on the scene in 1960. It is too early to tell how much of this difference reflects their lower age. These new additions to the electorate, socialized primarily in the era of Carter and Reagan, contribute importantly to maintaining the proportion of Independents at a higher level than in the steady-state period. The other dynamic factor here, of course, is the passing of the oldest, quite partisan, generation.^[10]

What do these findings suggest about the future stability of the political system as the dealignment generation enters the age of maximum participation? The leaners are partisans. Thus what results from the distribution of partisan Independents as the population ages is not likely to have much to do with volatility, party responsibility, or similar issues. Wanting to feel independent, as leaners do, may well have psychological implications, but our concern, and the concern of most observers who discuss dealignment, is its political implications. In this regard, leaners do not differ substantially from outright partisan identifiers.

The shift toward Independents coincided with the entry into the electorate of the baby-boom generation and the enact-

[10] For similar findings see Shanks and Miller, "Policy Direction and Performance Evaluation," 160.

ment of the 26th Amendment. Those people most inclined to be Independents became a much larger share of the voting-age population. The extent of this phenomenon can be seen in table 6.5, which shows the proportion of people under the age of 29 in each type of Independent and the white population as a whole. In 1960, young people accounted for just 12 percent of all whites and anywhere from 7 to 18 percent of the three kinds of Independents. In 1972, young people were 25 percent of the entire white sample and 32 to 42 percent of Independents. By 1990 these proportions had fallen to 21 percent of the sample, 25 percent of the leaners, and 32 percent of the Pure Independents.

These data provide further reason for thinking dealignment is fading. Young people are a smaller proportion of the electorate, relatively fewer of them are without partisan affinities, and more people in the older dealignment cohorts are moving from independence to partisan identification.

Education and Independence

As we noted in chapter 3, the earliest conception of the Independent was an ideal citizen, one who could "give close and constant attention to public affairs."^[11] In large part, the rise of the Independent was "an opposition of intellectuals to political machines—and indeed, to parties as such."^[12] The view that Independents were educated citizens continued through the first half of the twentieth century. This image, developed largely from observations by scholars and pundits, was affirmed by early survey research. In 1948 George Gallup reported that "the higher the voter is in educational scale, the

[11] James Bryce, Modern Democracies , vol. 1 (New York: MacMillan, 1929), 47–48.

[12] Howard Penniman, The American Political Process (Princeton, N.J.: Van Nostrand, 1962), 39.

more likely he is to be independent."^[13] This conclusion was based on these figures:

The American Voter 's oft-cited low assessment of Independents' civic virtue included no reference to their educational attainment, except this remark: "There is no significant relationship between strength of party identification and formal education."^[14] A few years later Walter Dean Burnham's suggestion of a "new breed" of Independent was based on 1964 NES data showing that the "peak share of independents falls among high school graduates who have had some college education."^[15] Burnham concluded that the "political parties are progressively losing their hold upon the electorate" and that the losses "have largely been concentrated among precisely those strata in the population most likely to act through and in the political system out of proportion to their numbers."^[16]

Burnham was right to wonder about the apparent inconsistency between the 1964 data on Independents' education and the low levels of interest, knowledge, and participation attributed to them in The American Voter . But as we will see, his explanation of the anomaly and the implications he derived from it miss the mark. Nevertheless, well into the 1980s many journalists and scholars noted Independents' relatively high education, and some cited Burnham's new-breed hypothesis. Others thought that rising education contributed directly to a

[13] Cited in Dayton David McKean, Party and Pressure Politics (Boston: Houghton Mifflin, 1949), 183.

[14] Campbell et al., The American Voter , 479.

[15] Walter Dean Burnham, Critical Elections and the Mainsprings of American Politics (New York: W. W. Norton, 1970), 123.

[16] Ibid., 129, 130. For a more recent exposition of the new-versus-old thesis, see Frank Sorauf, Party Politics in America , 6th ed. (Boston: Little, Brown, 1984), 173–74.

putative loosening of party ties. These two passages exemplify many others:

The growth of Independents has come particularly in those persons with sufficient education to permit freedom from party cues.^[17]

Better educated Americans are less apt to affiliate with parties, which they perceive as lacking substance, and are more apt to identify themselves as free-thinking independents.^[18]

Implicit in these statements is a belief that education reduces the individual's need for the cognitive and perceptual road map provided by party identification.^[19] This notion is not timebound; it should be as applicable to the 1950s as to the contemporary scene. One simple way to explore this proposition is to examine relationships between respondents' education and their identification as Independents. Table 6.6 displays the percentages of people at various levels of schooling who were Independent Democrats, Pure Independents, and Independent Republicans in five of the ten presidential-election years from 1952 through 1988. These five years exemplify all ten.

Table 6.6 shows that the more educated people are, the lower the probability they will be Pure Independents and the higher the probability they will be leaners. In 1952, 7 percent of the least educated were Pure Independents as opposed to only 2 percent of college graduates. In 1988, the figures were 15 percent as opposed to 6 percent. The only exceptions were in 1960 and 1964, when there was no relationship between education and identification as a Pure Independent. The story is just the opposite for leaners. In 1952, 20 and 13 percent of

[17] Gerald M. Pomper, Voter's Choice (New York: Harper and Row, 1975), 34.

[18] Richard T. Saeger, American Government and Politics (New York: Scott, Foresman, 1982), 228.

[19] For a formal statement of this proposition, see W. Phillips Shively, "The Development of Party Identification among Adults: Exploration of a Functional Model," American Political Science Review 73 (December 1979): 1039–54.

college graduates were Independent Democrats and Republicans, respectively, as opposed to 10 and 6 percent of the least educated. This greater tendency for the better educated to be leaners is found throughout the period, with occasional modest exceptions that do not disturb the generalization.

Following trends from 1952 to the 1970s and 1980s reveals a considerable increase in the proportion of uneducated people who were Pure Independents, no change in their inclination to be Independent Democrats, and a modest rise in the percentage who were Independent Republicans. The picture for the college educated is very different. In the 1970s there was a modest increase in their inclination to be Pure Independents that reversed in the next decade. By 1988 the proportion of college-educated people who were Pure Independents had reverted to the level of the early 1960s. On the other hand, educated people in 1988 were much more likely to be leaners than their counterparts in the late 1950s and early 1960s.

Because middle-aged and older Americans did not contribute to dealignment, we examine, in table 6.7, the relationships between education and independent identification of respondents under 29 and those aged 29 to 44. These more focused cross-tabulations elaborate our earlier findings and