Self-Esteem and Social Problems:
An Introduction

Neil J. Smelser

The well-being of society depends on the well-being of its citizenry. This is the central proposition on which the chapters in this volume—as well as the work of the California Task Force to Promote Self-Esteem and Personal and Social Responsibility—are based. This proposition is somewhat unorthodox, because a great deal of Western social and political thought would have it the other way around. Classical economic theorists, for example, regarded the productive and market arrangements of capitalism as an apparatus by which the greatest good for the greatest number could be realized; classical democratic theorists regarded the perfect democratic polity as the set of arrangements that would bring forth the best and most rational choices from an informed electorate.

The more particular proposition that informs our enterprise here is that many, if not most, of the major problems plaguing society have roots in the low self-esteem of many of the people who make up society. It is supposed that those citizens who appreciate themselves and have a sense of personal empowerment will cultivate their own personal responsibility and will attend to the tasks that are necessary for the welfare of the community and the society. It is further supposed that those in society who are burdened with the conviction that they are not worthy will take refuge in behaviors that are unproductive, costly, deviant, and dangerous to society and will, by that measure, contribute disproportionately to serious social problems. Bearing these two propositions in mind, it becomes essential for the leaders of society, first, to

establish social conditions that will maximize the development of selfesteem among the population and, second, to establish social arrangements that will rescue and rehabilitate those who have emerged from families and communities with a sense of diminished self-worth. That is the agenda of the California Task Force to Promote Self-Esteem, and that is the agenda that we in this special volume on self-esteem and social problems are putting to the best critical test in light of the best social scientific literature available to us.

My task in this introductory essay is to develop a statement that synthesizes the issues raised and discussed, the knowledge compiled, and the conclusions assessed by the individual contributors to this volume. This will mean covering some of the same ground they do, but from a more general point of view. To that end, I will consider several major questions:

• What constitutes a social problem?

• How do we define the nature of a social problem, and how do we define self-esteem?

• What are the linkages between individual self-esteem and the generation of a social problem?

• How do we measure both of these?

• How do we go about establishing scientifically that connections exist between diminished self-esteem (cause) and the kind of behavior that constitutes a social problem (effect)?

• What are the scientific findings relevant to those connections?

• How do we intervene in a way that will attack the causes of the identified social problems and prevent the loss of self-esteem of those who are part of them?

What Constitutes a Social Problem?

In our accepted ways of thinking, a social problem is a kind of carbuncle on the social body, a tear in the social fabric that signals some kind of malfunctioning in society and sets up demands for its own amelioration. A social problem is some kind of tangible, identifiable, unwanted thing in society.

This conceptualization of a social problem, which treats the existence of a problem as a matter of verifiable fact, is, I argue, unrealistic. Before

an empirical state of affairs can qualify as a social problem, it must be linked with a whole range of cultural and ideological phenomena. Let me illustrate:

First, behavior that we identify as constituting a social problem must be relevant to some institution that we endow with cultural value . Pregnancy out of wedlock, for example, is a problem in large part because it stands in violation of the value we place on the family as the legitimate locus for childbearing and child-rearing. Dropping out of the educational system is a problem because of the value we place on learning, both in itself and as preparation for entering the occupational structure. Chronic welfare dependency is a problem in part because it involves not participating in an established job or career; it is an interruption of involvement in that valued social role.

Second, behavior that becomes defined as a problem is also regarded as deviant in relation to some role expectation. This is clearest with respect to violence and crime, which are deviant because they are against the law; the same can be said for the use of illegal drugs. Dropping out of school also may be illegal because it deviates from the established code calling for compulsory schooling up to, say, age sixteen. Other behaviors, such as excessive alcohol consumption, teenage pregnancy, and being out of work, are not illegal, but they violate social norms relating to substance dependency, premarital sexual and childbearing behavior, and holding a job. A social problem thus involves some kind of social deviance.

The necessity of these value and normative references means that even though a certain kind of behavior may be prevalent in society, it is not considered a social problem unless and until these linkages are established. Child labor, for example, has always existed, but it did not come to be viewed as a social problem until moral crusaders holding humanitarian values deemed it one, and until child labor legislation supplied norms against which it could be considered deviant. The same is true of child abuse: only with the rise of humane values relating to parental discipline, and only with the rise of new expectations about punishment and degradation, did this widespread practice become a social problem. It follows that social problems may become such by virtue of value and normative drifts in society as much as by virtue of the appearance of new kinds of behavior. It is the linkage of a behavior to relevant value and normative considerations that gives it its character as a social problem.

An additional ingredient in the definition of a social problem—an

empirical ingredient—is that the behavior in question must have a great enough incidence or prevalence to be regarded as significant; there must be some numbers of people perpetrating such behavior. The isolated occurrence of mass murders by snipers, for example, is not customarily defined as a social problem; rather, it is seen as a matter of individual psychopathology. The fact that almost all authors in this volume begin their expositions by referring to "the scope of the problem" also emphasizes the importance of incidence. A further ingredient, though not always an essential one, is evidence that the kind of behavior in question—drug dependency, mental illness, dropping out of school, child abuse—has been on the increase in the recent past. And a final part of this "numbers" aspect is that in order for a behavioral phenomenon to be considered a social problem, a sizable number of people have to be able to successfully define it as such and to make the required symbolic linkages to the relevant cultural and normative references; otherwise, it will not be perceived as a social problem, but only as the private and perhaps idiosyncratic preoccupation of a few.

Another element that goes into identifying a social problem is that it must be regarded as involving some economic or social cost. Crime is a good example: paying for the long-term incarceration of convicted criminals is an expensive proposition, to say nothing of the cost in human lives and private property that crime involves. The social cost of premarital pregnancy—defined in terms of welfare costs for mothers and the psychological costs to the children—is another ready example. It is the cost component of chronic welfare dependency that really endows that phenomenon with its characterization as a problem. Absenteeism, inefficiency, and low worker morale are among the high economic costs of alcoholism and drug dependency.

Finally, another necessary part of what defines a social problem is that we believe we can do something about it. It has to be something at which we can successfully throw resources; something we can ease by getting people to shape up; something that can be cured through social policy legislation and decisions and the application of knowledge; something that can be ameliorated. Otherwise, it is seen as one of those ineradicable scars on the social body that we have to live with, a necessary evil, one of those inevitable frailties of human nature. Better put, a social problem is something that we believe we have a way of dealing with; and if a significant number of people succeed in defining a social phenomenon as something we can do nothing about—it is "in the genes," it is in the nature of social life—it loses its status as a social problem.

This faith that we can do something about a social problem rests on two frequently unspoken assumptions about causality: first, that we understand what the main causes of the social problem are and therefore can target them in order to ameliorate the problem; and second, that we know that the recommended legal, policy, reformative, and therapeutic interventions will have an effect on those causes and thus will achieve that amelioration. It goes without saying that these assumptions are frequently not verified and that they are often not much more than matters of faith on the part of social diagnosticians and social reformers. Nevertheless, they are always present as part of a concern with social problems.

This discussion demonstrates that, in the last analysis, the existence of a social problem is a matter of persuasion and a matter of politics. To get a social problem on the agenda, sufficiently visible and powerful people have to persuade those who officially name social problems that a kind of behavior exists that is costly to society and that it has a high incidence in society—an incidence that is perhaps on the increase. Furthermore, they have to be persuasive in arguing that the problem constitutes an erosion of some institution we consider valuable or sacred, that it involves behavior that is deviant in light of some established law or norm, and that it is a problem we can do something about with the right social policy and the right investment of resources. All kinds of groups in society are forever jockeying with one another in attempts to have their favorite societal ill officially defined as a social problem and thereby placed on the table before the legislature or other concerned body; but of course only some of them succeed. For this reason, it is highly problematic—and it is the outcome of a complicated political process—that a social problem ever gets to be defined as such, much less gets the public attention it may deserve.

One final observation emerges from this discussion of social problems. The California Task Force to Promote Self-Esteem—and the scholars who have contributed to this volume—have focused on certain major concerns in American and California society that seem to qualify as social problems, that is, they include all the ingredients identified above. These social problems are child abuse, crime and violence, teenage pregnancy, academic failure, alcohol and drug abuse, and chronic welfare dependency. Obviously, these are only a few of the problems that could have been selected; an equally good case could have been made for including ecological and environmental threats, risks from nuclear power, threats to international security, the complex of sexismracism-ageism, divorce and separation, and, above all, physical and

mental health. We could easily have expanded the list to several hundred; in fact, there is no end to the catalogue. The problems considered here must be regarded as illustrative only, constituting but a small fraction of society's ills.

A Plausible Case for the Link Between Self-Esteem and Social Problems

As an intuitive matter—based on our own personal experiences and our observations of others—we know what it is to experience high selfesteem. It means, fundamentally, that we appreciate ourselves and our inherent worth. It also means that we have a positive attitude toward our own qualities; that we evaluate them highly; that we are imbued with a sense of our own ability, competence, and power to do what we want; that we compare ourselves favorably with others; and that we can organize our daily round of activities and performances in keeping with these feelings of self-worth. We also know what it means to experience diminished self-esteem; it means the opposite of all those positive elements just described, and it results in self-deprecation, helplessness, powerlessness, and depression.

Also as an intuitive matter, we know how we are likely to behave, depending on whether we think well or poorly of ourselves. If the former is the case, we can take command and control of our lives; can consistently behave responsibly and well toward our duties and to others; and can do this on our own, without relying on any kind of psychological or social crutches. If we are down in the dumps, however, the tendency is to withdraw from performance into passivity; to seek some ready way to pick ourselves up or to have someone else do so; to seek out some activity that will make us feel better about ourselves, at least in the short run. As often as not, that kind of behavior is likely to be antisocial and deviant from some point of view: we take several drinks too many; we find a way to get high; we quit work; we drop out of school; we take our depression or hostility out on someone else; or we thrash around for some momentary impulse gratification, whatever that might be. This withdrawing, self-defeating, and deviant behavior is the stuff of which social problems are made, because it involves getting out of responsible, expected behavior patterns and getting into just the opposite. When this kind of behavior is aggregated, it becomes a social problem, and it gains the attention of those who care about how society is faring.

We are intuitively aware of what it takes to turn high self-esteem into

low, and what it takes to turn low self-esteem into high. In the former instance, constant failures and a constant bombardment with the message that one does not count as a person or with others gradually add up to the feeling that one is a cipher in this world. In the latter case, the first step is acknowledging that one regards oneself as worthless or helpless or in the grips of self-abasement, combined with a sense of personal suffering and a desire to drag oneself out of the hole. When this is linked with a personally meaningful experience—perhaps finding a mate, perhaps getting a new job, perhaps entering psychotherapy, perhaps finding a new social or religious identity—the individual finds a new sense of capability, power, and self-control; a liking for oneself and who one is; and a capacity to get out of the realm of behavior that is damaging to oneself and to others and to reach toward the realm of responsible and pro-social behavior. Moreover, it is this significant turnaround from low to high self-esteem—and its behavioral consequences—that allows the ebbing away of those kinds of behavior that constitute social problems.

Intuition also tells us that both benign and vicious cycles characterize various levels of self-esteem and their supposed behavioral consequences. This is particularly true when we consider self-esteem in its interpersonal reference. As for the benign cycle, consider the appearance of a respected and admired role model in the life of a young and disadvantaged schoolchild, a model who, by giving encouragement and support and being a good mentor, fills that pupil with a sense of value, competence, and power. These feelings are reflected in the superior academic and social performance of the child; this feeds positively into the child's self-perception and turns into an upward spiral that may continue throughout life. Consider also the confirmed and self-abased alcoholic who finally acknowledges to friends and supporters in Alcoholics Anonymous that he or she is completely helpless in the face of alcohol dependence, who through membership and assistance from alcoholic colleagues is able to find strength through abstinence, and who builds from this a greater involvement and achievement in work, marriage, and friendship. As for the vicious cycle, we need consider only that dynamic described in this volume by Scheff, Retzinger, and Ryan: one partner in an interpersonal relationship is subtly but tangibly shamed by the other and does not acknowledge this debasement but instead lashes out in a violent rage toward the other person. The violence sets off a frenzy of greater shame and guilt, which can find expression only in another bout of destructiveness.

In this case, the causal link is clear: low self-esteem is the causally

prior factor in individuals seeking out kinds of behavior that become social problems. Thus, to work on social problems, we have to work directly on that which deals with the self-esteem of the individuals involved. Or, as we say in the trade, diminished self-esteem stands as a powerful independent variable (condition, cause, factor) in the genesis of major social problems. We all know this to be true, and it is really not necessary to create a special California task force on the subject to convince us. The real problem we must address—and which the contributors to this volume address—is how we can determine that it is scientifically true.

Further intuitive reflection, however, should convince us that citing diminished self-esteem as an independent variable in the generation of social problems is only a partial diagnosis. We know equally well that diminished self-esteem is often the product of something outside the individual, something in one's personal and social environment. If a child is singled out as the family dummy, is the one voted least likely to succeed, or is abused by parents and siblings, that child is a poor candidate for having high self-esteem throughout any part of life. If a child in school is continually identified by teachers as stupid, is made to sit in the corner with the dunce cap on, or is relegated to the inferior academic track, he or she is a good candidate for developing a poor self-concept. If a child is continually scapegoated by peers and is always the follower and never the leader, he or she may don that role throughout life. Or if individuals are members of a group in society—usually a minority group—that is routinely abased, thought to be inferior, and denied access to chances for advancement and a share of the good things in life, those individuals may pick up and wear the image that they do not count for much or deserve much. In short, self-esteem is as much an effect of psychological and social causes as it is a cause of social problems. Or, as we say in the trade, it is an intervening variable in the genesis of social problems. Those who would ease social problems, moreover, would have to have social reform on their agenda—that is, the effort to reshape the practices and institutions in society that make for human suffering, degradation, and injustice. In this conception of social problems, such problems are ultimately the product of culture and society, and efforts at amelioration involve social transformation, not only individual therapy and rehabilitation.

In practice, of course, the diagnosis of and assault on social problems involve working on both theories at once—that social problems are the products of individuals with psychological dispositions who act out or act up in society and thus create the problems, and that they are the

products of imperfect social institutions that generate the dispositions in individuals to act this way. Many social problems seem almost intractable; they do not go away easily; and they call for a multifaceted attack, an attack that involves both many kinds of therapy and the rehabilitation of individuals, on the one hand, and the reform of practices and institutions, on the other.

The big issues facing those social scientists and policymakers who wonder about social problems and their causes and cures are to formalize this intuitive knowledge: to develop the proper definitions of what behaviors constitute social problems and what factors constitute their causes; to develop proper measures of those behaviors and those factors; to find associations or correlations between the behaviors and the factors; to assign causal significance to those associations; and to devise interventions that will modify the factors and thus change the behaviors. To these difficult issues we now turn.

Defining Self-Esteem and Its Consequences

Two theoretically minded colleagues of mine in West Germany under-took a major research project about two years ago. It concerned the topic of "the major social crises in postmodern Western society." After two years of concerted work, they submitted their conclusion: "We have been able to determine that we can neither define nor measure either 'major social crises' or 'postmodern Western society.' That concludes our research report." Things are not quite so bad in our case, but it has been found—and it is reported by the contributors to this volume—that we encounter a number of conceptual problems in getting at the essence of self-esteem and its various behavioral consequences.

As suggested before, we have a fairly firm grasp of what is meant by self-esteem, as revealed by our own introspection and observation of the behavior of others. But it is hard to put that understanding into precise words. Often, when we try to describe something we are uncertain about, we flounder around with a number of different words, not one of which captures its essence. In addition to self-esteem, our authors have identified the ideas of self, self-concept, self-respect, awareness, identity, image, congruence, and consciousness. None of these hits the nail on the head. Neither do the negative opposites, such as lack of self-concept, powerlessness, helplessness, ineffectiveness, inefficacy, self-derogation, anxiety effects, depression, or lack of self-esteem. Is there anything we can salvage from this definitional maze?

One starting point is to identify the almost universally accepted com-

ponents of the concept. There is first a cognitive element; self-esteem means characterizing some parts of the self in descriptive terms: power, confidence, agency. It means asking what kind of person one is. Second, there is an affective element, a valence or degree of positiveness or negativeness attached to those facets identified; we call this high or low selfesteem. Third, and related to the second, there is an evaluative element, an attribution of some level of worthiness according to some ideally held standard.

What that standard is is another feature of self-esteem. Sometimes the standard is an absolute sense of self-regard, measured against an ego ideal that one holds out for oneself. Or it may be a relative standard, measuring one's sense of self-worth in relation to an internal aspiration or desired level of attainment. Sometimes the standard or point of reference is mainly internal or psychological; sometimes it involves measuring one's self-worth in relation to another person or group. The latter case is particularly important for members of minority groups, who compare their own group's fortunes to the situations enjoyed by other groups in society.

Greater definitional difficulties arise when we consider the stability of the concept of self-esteem. Should it be regarded as some kind of global attribute, one that enjoys a constant strength and level of organization for the individual? Or should it be treated as largely situarional? All of us know that we feel good about ourselves at one moment, bad about ourselves at another, depending on mood and social situation. Add to this complication the fact that high self-esteem is sometimes experienced as an inherent set of feelings, but at other times as a defensive reaction against feelings of inferiority or inadequacy. In the latter case, high selfesteem is better described as vanity, arrogance, or egotism. All these definitional difficulties will come into play when we consider problems of measurement or try to arrive at operational definitions of the concept.

With respect to the supposed behavioral consequences of high or low self-esteem, we are somewhat better off conceptually, but not too much so. Definitions are most easily formulated in the areas that are clearly marked off behaviorally from their opposites. For example, we know what it is not to be enrolled in school, because of the institutional character of that attachment; it is more difficult to determine when one is not learning, however. It is relatively easy to define when one is out of work, though the phenomena of part-time work, seeking or not seeking employment, and unpaid labor complicate that definition. It is relatively easy to define teenage sexual activity, and quite easy to define teenage

pregnancy. It is easy to define what we mean by the ingestion of some drug substance, much more difficult to define the ingestion of "too much." We know fairly well what we mean by "crime," but we are often hard put to come up with a clear definition of what constitutes violence. In the area of child abuse, we find the greatest difficulties. Bhatti, Derezotes, Kim, and Specht cite legal, medical, sexual, and psychological definitions of abuse, each of which yields a different range of meanings.

It is particularly important to raise these questions of definition, because we want to be certain that definitions of supposed causes (e.g., low self-esteem) are kept separate from definitions of their supposed behavioral consequences (e.g., psychological stimulation by drugs or alcohol). Otherwise, we will be caught naming the same things twice, thereby making it difficult to argue that they are independent from one another and difficult to avoid circular reasoning.

Measuring Self-Esteem and Its Consequences

In measuring what we have conceptually defined, it is customary to ask two kinds of questions: Is the measure we adopt reliable (that is, is it a consistent measure of some definite phenomenon)? Is the measure we adopt valid (that is, does it constitute a measure of what we think we are measuring)?

In assessing self-esteem, we have been supplied primarily with measures designed by professional psychologists, and we are short on other kinds of measures. The typical psychological measure is arrived at by posing a question or set of questions to a group of people, asking them how they describe themselves, how effective they believe themselves to be, or how much in control of things they are. Or the questions may be framed to get at how people feel about themselves, for example, how positively or how negatively they regard themselves. Or the questions may be devised to discern how people compare their own well-being with that of some comparison group. Several true-false questions from the Tennessee Self-Concept Scale may serve as an example:

1. I have a healthy body;

2. I am satisfied with my moral behavior;

3. I have a lot of self-control.

This set of questions touches the internal or psychological aspect of selfesteem and the individual's sense of self-competence. From this list, we

already can see one measurement problem, namely, that one set of questions (self-descriptive) may be getting at something entirely different from what another set of questions (self-evaluative) addresses, thus tapping different facets of the mental set we define as self-esteem. Be that as it may, the "measure" of self-esteem is usually taken to be the kind of response evoked by administering the questions in an interview or experimental setting. The responses are usually recorded, averaged, and given some quantitative representation. The object of seeking these sons of measurements is to identify some index that we can compare and correlate with the supposed behavioral consequences of the measured attribute.

Two measurement problems are associated with this kind of selfreport procedure. First, no measure of self-esteem taps more than one or two of the facets of that phenomenon (cognitive, emotive, evaluational, or interpersonal), and for that reason any measure will yield an incomplete representation. Second, most of the measures assume that selfesteem is global, not situational, and does not vary over time. This assumption is questionable. It is true that with many measures of reliability (split-test, inter-judge, repeated administration over time) the phenomenon of self-esteem seems to have some integrity; it seems to be something that can be tapped on a fairly consistent basis. Yet investigators have discovered that measurements of children's self-esteem fluctuate over time and from situation to situation and that measurements of children's self-esteem yield results that are different from those found by measurements of adults' self-esteem, even when the instruments designed to measure the one kind are adapted to measure the other. When it comes to assessing the validity of these measures, the concept of selfesteem often becomes slippery and elusive. Validity of such a measurement is usually assessed by holding it up against some other kind of measure of self-estimation or of performance (e.g., on intelligence tests) or some other kind of self-assessment (e.g., depression, feelings of empowerment or self-control). These measures of validity are very erratic, suggesting either that the phenomenon of self-esteem is not very stable or that its measures are imperfect representations of something else.

Measuring the behavioral consequences of self-esteem also poses some difficulties, which parallel the definitional difficulties outlined above. We can measure welfare dependency by checking with various welfare agencies or by asking people whether and how long they have been out of work, but these measurements of "spells" often do not re-

cord erratic or short-term work. We can also track rates of school attrition or dropout rates by consulting school enrollment records, but these too may be inaccurate; it is often not known whether dropouts have left school altogether or are in school somewhere else, because they are not usually tracked after they drop out. We can measure whether a teenager is sexually active (although the line between sexually active and inactive is difficult to draw, because level of activity may differ over time), and we can determine medically if a teenager is pregnant. We have consistent measures of crime (although police arrest records, self-reports of victims, and self-reports of perpetrators yield a wide range of figures, especially for major crimes). Measures of violent behavior, however, are difficult to come by, largely because violence appears in so many psychological and social forms. With respect to one form of violence in particular—child abuse—the measurement problems mount, because of the vagueness of the line between nonviolent and violent behavior and because one kind of violence is often indistinguishable from another. We have good medical and behavioral measures on the intake of drugs and alcohol, but we do not have any consensus on what it takes to declare someone "dependent" on these substances. Thus, when attempting to determine the "scope" or incidence of the problems posed by these behaviors, all our authors found themselves referring to various measures; but all of them also took refuge in discussing the difficulties that stem from "unrecorded" or "invisible" occurrence and thus confound any consistent effort to measure prevalence, much less changes in prevalence over time.

There are good sociological reasons why measures of the supposed behavioral consequences of self-esteem are so elusive. Many of these behaviors are either illegal or deviant from some point of view. Both the perpetrators and the recorders of such behaviors may have a vested interest in keeping a low profile. This secrecy is well understood by the perpetrators, who keep their behavior "in the closet." As for the recorders, police, attorneys, physicians, and insurance companies often keep the lid on various kinds of behavior (alcoholism, for example) to avoid stigmatizing individuals or blemishing a person's official record.

As we will indicate later, the main purpose of social science investigations of the phenomena we have been discussing is to establish meaningful correlations between one set of reliable and valid indicators and another set and to work out the best ways to establish causal relations between them. But if the measures are not independent from one an-

other, and if they are not reliable and valid, we have to assume that the measures of association are bedeviled with an unknown amount of error.

Establishing Associations and Causes

We now move from the intuitive models of high self-esteem and low self-esteem and their consequences—representations of which make great sense—to the adoption of scientific procedures that will establish relations among these variables and permit the inference of causal priority. In this operation, we must first establish what our expectations are with respect to causal priority. For most of the social problems listed in the legislative charge to the California Task Force to Promote Self-Esteem, the expected causal direction is fairly clear: low self-esteem (and perhaps the tendency to abuse) on the part of the abused; depression leading to alcohol and drug abuse; low self-esteem leading to quitting school or work; and so on. In some cases, however, the logic is only half-plausible. For example, because so much unemployment is involuntary and so many of these unemployed people are chronically dependent on welfare, we can hardly expect that the causally prior condition for chronic dependence is low self-esteem. It might be the case, however, that such dependency causes low self-esteem, because chronic dependency is stigmatized in our society and because it contributes to feelings of helplessness and powerlessness.

The next step is to identify a population that scores low on selfesteem and to see if these low scores can be related to one or more suspected behavioral consequences, particularly as this relationship is compared with that of another population that scores "normal" or "high" on self-esteem measures (often referred to as the "control group"). Or, to take another and more indirect tack, the investigator may take a group that is assumed to be low in self-esteem because of its disadvantaged economic or social position (for example, blacks) and ask whether members of this group engage in the behaviors predicted, again in relationship to members of other groups (for example, Asian Americans or Anglos).

The procedures by which these plausible relations are tested for correlation are highly standardized. We need first a quantitative measure of self-esteem, gathered by administering one of the many measures designed to tap that variable, and a quantitative measure of the problem behavior in question. These measures are converted into scales (how

much of the time do I feel dissatisfied with myself? how often do I take drugs?), and a standardized measure of association (correlation coefficient) is applied. Often, this is combined with an estimate of how much of the variance in the dependent variable (e.g., drug-taking) is accounted for by scores on the independent variable (e.g., low self-estimation). A further test of the strength of association is made by estimating, by standardized statistical procedure, how often this association might have been expected to occur by chance alone (say, 5 percent of the time).

One of the disappointing aspects of every chapter in this volume (at least to those of us who adhere to the intuitively correct models sketched above) is how low the associations between self-esteem and its consequences are in research to date. In some cases, consistent relationships are found. Substantial numbers of studies show a low level of selfesteem among socially isolated parents who abuse their children; high self-esteem is associated with the use of contraceptives by teenage girls; measures of high self-esteem correlate positively with achievement in the classroom, and, as self-esteem decreases, so does academic achievement; within different sociocultural groups, drug and alcohol abuse are positively correlated with low self-esteem.

With respect to the sources of low self-esteem, the results also appear to run in the expected directions. Children with alcoholic parents have lower self-esteem than other children; children who have been abused by their parents show low scores on self-esteem measures; adults who have been on welfare for long periods show low self-esteem; and students who fail in competitively based school situations suffer from blows to their self-esteem.

Sometimes, however, the associations run in unexpected directions. For example, the use of psychoactive drugs seems to have a positive effect on self-esteem (though the association is questionable, because other changes in consciousness appear to take place as well). In a curious report on intervention, Bhatti, Derezotes, Kim, and Specht describe a TV ad developed by the National Committee on Child Abuse that seems designed to convince abusive parents that their behavior is reprehensible (and thus to lower their self-esteem), which suggests—strangely—a positive association between high self-esteem and child abuse!

The news most consistently reported, however, is that the associations between self-esteem and its expected consequences are mixed, insignificant, or absent. This nonrelationship holds between self-esteem and teenage pregnancy, self-esteem and child abuse, self-esteem and most cases of alcohol and drug abuse. The question naturally arises:

why are the correlations so low? The following possibilities come to mind:

• As often as not, an association between self-esteem and its behavioral consequences is not found because the variable of self-esteem is never even measured or brought to bear. That is the case in most of the studies of drug and alcohol abuse among different minority groups. And self-esteem or other psychological factors as independent variables are almost never represented in studies of chronic welfare dependency.

• The anticipated relationships may be valid but are counteracted by a counterassociation that arises from actually engaging in the problem behavior. For example, a teenage girl, plagued by feelings of low selfesteem and loneliness, may engage in sexual behavior (and even become pregnant) because she needs love and affection at any cost. But through the mechanisms of reinforcement and anticipatory association, she may experience feelings of increased self-esteem from the sexual relationship or the pregnancy. Lonely, dispirited, and depressed individuals may abuse their bodies by ingesting too much drink or too many drugs, but they may receive temporary boosts to their self-esteem through those actions. These feelings may be reinforced by rewards from supportive peers. Through the vicious cycle of shame, guilt, and self-disgust, an individual may strike out in rage at an intimate, but the temporary relief experienced may generate stronger feelings of self-worth and selfrealization. In all these cases, the causal priority is reversed, and the positive association is washed out by the negative one.

• Another reason why posited or predicted associations may be low has to do with the character of the variables being related. Self-esteem is represented as a global, dispositional variable with a great many possible behavioral outcomes, including substance abuse, crime and violence, compulsive seeking for social support, and a variety of withdrawal behaviors, including dropping out of school and quitting work. Because there are multiple consequences, it stands to reason that selfesteem will be correlated with many outcomes, but that correlation will be weak in any given case. From a state of self-esteem alone, in other words, it is not plausible to expect a high correlation with any of these consequences.

• The main reasons for low correlations are probably methodological; that is, the measures of self-esteem do not tap that variable reliably or validly, and the measures of the behavior in question are similarly flawed. The error built into associations (or nonassociations) between

variables is always unknown, because its magnitude is always a mystery. The logic that intuitively links self-esteem with some kind of outcome does not correspond with the logic that leads us to construct quantitative indices of dispositional qualities such as self-esteem and quantitative indices of actual behavior; thus the link that we all know exists is shrouded in error.

• Potentially positive associations are not found because additional plausible factors are not held constant. If level of intelligence is controlled, it, rather than self-esteem, may turn out to be the major determining factor in academic performance. The availability of contraceptives may be an opportunity factor that confounds the link between low self-esteem and both teenage sexual behavior and teenage pregnancy. Crockenberg and Soby report that because race and ethnicity may affect teenage sexual behavior and pregnancy, any association between those behaviors and psychological dispositions may be clouded. Also, in a given cohort of teenagers (for example, in the 1970s and 1980s), all of the young people may have been sexually active at an earlier age than in the past (because of more lenient sexual mores); certainly it is difficult to assume or demonstrate that increased incidence was caused by a jump in the incidence of low self-esteem.

If the association between self-esteem and behavior is so often reported to be weak, even less can be said for the causal relationship between the two. Why, we ask next, must this causal relationship be regarded as so weak?

The first answer to this question is based on the common scientific knowledge that correlation does not establish cause. This has been acknowledged ever since John Stuart Mill demonstrated it in his discussion of methods of scientific proof and ever since Émile Durkheim was forced to concede, much against his methodological preferences, that concomitant variation (association or correlation) cannot be taken as evidence for positing a causal connection.

The second answer is found in the specific limitations of the studies reported in the various chapters. As each author indicates, most of the research on which the findings and interpretations are based has been done with small samples, with relatively simple psychological measures of self-esteem held against some outcome, with simple correlational methods, with no control groups or control variables taken into account, and with temporal priority of variables not taken into account. Thus, each author uses well-advised caution in assigning causal signifi-

cance to whatever associations were observed. In several chapters, longitudinal designs—which are one of the best means of gaining insight into causal priority of one variable over another—are discussed. In one case, there appeared to be a longitudinally based, causally significant relationship between length of time on welfare and the variable of learned helplessness (low self-esteem). In another case, Crockenberg and Soby report the results of three longitudinal studies relating self-esteem and subsequent pregnancy; two studies found a relationship between the two variables, and the third reported no association between self-esteem and pregnancy. The strength of these studies lies in the fact that measures of self-esteem were taken before the reported or observed pregnancies, but even here there was little effort to control for the intervention of other variables.

New Research

The main conclusion to be drawn from this extended discussion of definition, measurement, and association is that the social-psychological variable of self-esteem is simultaneously one of the most central and one of the most elusive factors in understanding and explaining the behaviors that constitute major social problems. It is central because it is the omnipresent variable that intervenes between personal and institutional histories of individuals with productive, responsible, and self-realizing behavior, on the one hand, and deviant, self-defeating, socially costly behavior, on the other. Furthermore, it is a variable that is not reducible to those personal and institutional histories, because we can produce both instances of well-endowed, well-loved, and well-supported individuals who move through life with the gloomiest assessment of themselves and instances of children who are disadvantaged and abused from almost every standpoint but who manage to tough it out and preserve a high estimation of themselves and their capacity for personally and socially responsible behavior. The variable of self-esteem is elusive, however, because its precise role in the drama of self-realization is difficult to pinpoint scientifically; by using the conventional kinds of scientific methods we possess, it is difficult to arrive at strong associations between self-esteem and its supposed causes, on the one hand, and selfesteem and its supposed outcomes, on the other. Or, to put the matter more simply, the scientific efforts to establish those connections that we are able to acknowledge and generate from an intuitive point of view do not reproduce those relations.

When we are confronted with the situation I have just described—an evidently important variable or set of variables whose importance is nevertheless not readily discernible in the available research—a multifaceted attack is usually called for. Some of the necessary work must be mainly theoretical, directed toward recasting the definition, meaning, and behavioral significance of the concept. Perhaps the global concept of self-esteem should also be disaggregated conceptually into isolable subvariables. More reliable and valid measures should be sought. And above all, our scientific procedures need to be improved; we must have more ambitious and better-devised scientific investigations if the state of our knowledge in the several areas in which self-esteem plays a role is to be advanced. The new kinds of research I have in mind should have some particular characteristics.

First, the samples studied should be large. The research reviewed in the following chapters has been carried out primarily with small, ad hoc samples generated by researchers who have pulled together the sample from groups that were available to them through some personal or institutional contact. Small samples yield relations that cannot be regarded as statistically significant; when uncovered, these relations cannot permit causal inferences; and, above all, small samples do not permit the holding constant of other variables suspected of affecting the relationships between self-esteem and some outcomes.

Second, major control variables should be taken into account in the design of the studies and samples. Researchers should draw separate samples from populations of low-income, middle-income, and high-income groups; from populations of children with low intelligence and children with high intelligence; from children of intact families and from children of broken or one-parent families; from children in minority groups and children in the majority group; and so on. All these group factors are of obvious significance in the development of selfesteem and its consequences; but if representatives of these groups are mixed indistinguishably in the same samples, the finer play of the individual and the institutional histories of the persons studied cannot be traced.

Third, the research must involve more than simple correlational studies. The imposition of controls described above is one way of going beyond simple correlations and controlling for variables. In addition, research designs should take into account that both the factors affecting the genesis of self-esteem and the behavioral consequences of selfesteem are multiple. This calls for imposing methods of multiple regres-

sion, partial correlation, path analysis, and the like, so that the relative strength of each of those factors is singled out, identified, and assigned an appropriate designation.

Fourth, the studies must be longitudinal. Samples can be drawn either prospectively—that is, from groups of those whose "career" outcomes with respect to some kind of behavior (crime, substance abuse, educational persistence) are not yet known—or retrospectively—that is, from groups of individuals who have "arrived" at one or another of these outcomes and groups of individuals who have not. In the samples drawn prospectively, the task is to track the subjects over time, recording the distinctive features of their family circumstances (e.g., birth order, divorce history of parents, kinds of abusive behavior); the parade of successes and failures they experience in school, and the treatment they receive from teachers; their history of experiences with peers in the classrooms and on the sports fields; their encounters with advisers, scoutmasters, camp leaders, and other potential mentors; the brushes they have with police and juvenile authorities. In samples drawn retrospectively, histories should be rewritten "backwards," as it were—reconstructed to take into account the same sorts of variables that impinge on individuals' lives. When possible, experimental or quasiexperimental designs should be devised, so that proper controls can be attained over third factors; and the administration of survey and other instruments should be supplemented by in-depth interviews and direct observations of the subjects, so that the meaning-context of the events to which they are exposed through their developmental years can be ascertained. Above all, these studies should document the vicissitudes of the subjects' self-esteem in its various measurable facets.

A model for these kinds of studies is found in the beginning efforts to study criminal careers, in which there is systematic examination of large samples of those who seem destined for a life of crime or those who have become involved in such a life, as contrasted with those who have never committed crimes or those who became involved as juveniles but later "dropped out" in favor of more law-abiding lives. Important events in the lives of these subjects are family experience (including discipline and nurturance), school experience, work history, marital history, and experiences in prison. These are the same variables, moreover, that affect those who engage in various kinds of self-abuse, those who end their schooling early, those who are unable to hold a job, and so on.

Such longitudinal, controlled studies should involve every kind of sample and group available. It is apparent, however, that from the

standpoints of urgency and policy relevance the most careful and well documented study should be devoted to those groups who tend to contribute most to social problems and whose self-esteem appears to be at greatest risk (disadvantaged minorities, low-income groups, children from poverty-stricken and single-parent families, for example).

Fifth, the research must be adequately and substantially funded. Launching large-scale, longitudinal study involves both the expenditure of much greater sums of research money than have been available heretofore and the mobilization of the finest research talent available. Supporting such research means changing our basic assumptions about empirical investigation, because it involves funding research projects that will be based on an investment of money over many years; moreover, this investment will have to be renewed periodically, so that the longterm tracking of the samples can continue. Only ample funding of this sort can produce an understanding of the development of the social problem behaviors that concern us. It seems fair to conclude that failure to provide such support for sound, policy-related research may prove far more costly in the long run, both in terms of dollars and in terms of human suffering.

The Problem of Intervention Strategy

It follows from this brief overview of scientific research that the guidelines for intervention and policies aimed at rehabilitation, prevention, and institutional reform will be as murky as the state of our knowledge of the phenomena concerned. And so they are. The authors who have assessed the state-of-the-art knowledge of factors important in the genesis of many social problems have been unable to uncover many causally valid findings relating to that genesis—and they have therefore been correspondingly unable to come up with systematic statements relating to cure or prevention. Most settled for reviewing that portion of the literature that discussed possible interventions and evaluated these studies as best they could.

In their review, Bhatti, Derezotes, Kim, and Specht identify several possible theoretical explanations of the increase in child abuse in recent generations. These include the weakening of taboos against the exploitation of children, the continued existence of poverty and economic stress, men who are intimidated by "liberated" women (men who subsequently strike out at their children to regain a sense of their own manhood), and the growing prevalence of those types of families in which

child abuse may result from insufficient bonding (stepfamilies, adoptive families, single-parent families, and families in which both parents are working). Policy interventions based on this range of possible causes should be correspondingly varied, including psychotherapy, police action to prevent abuse, and punishment of abusers, as well as institutional changes that would ameliorate the conditions that contribute to abuse, such as poverty and family frailty. Yet the authors are able to identify only two kinds of intervention that have been tried: psychotherapy of various sorts that aims to restore self-esteem to those who have been abused, and limited sorts of prevention aimed at protecting children from being abused. Neither intervention has produced more than inconclusive results, and no efforts whatsoever seem to have been directed at any of the possible underlying institutional causes of child abuse. Crockenberg and Soby argue that the family and the school are the most logical focus of policy interventions to affect the incidence of teenage pregnancy, but the basis of their argument appears to be little more than that these institutions are the most important sources of control over impulsiveness in the teen years.

Other possibilities for intervention are mentioned in the various chapters, but in many instances the form of intervention appears to involve certain other social costs that we may be unwilling to pay. One line of research—actually, some of the best research—has shown that black children who experienced their early years of schooling in integrated school settings appear to exhibit lower self-esteem than do black children who have been taught in schools that remained effectively segregated. Does this mean that we wish to turn back the clock and return to a system of segregated schools? Few would argue that that policy is worth the social cost. Covington suggests that primary schools should maximize what he calls equity systems, in which all students are provided with equally attractive incentives for pursuing academic goals and are not subjected to the zero-sum competitive school settings in which few students are winners and most students compete mainly to avoid failure. But such a system appears to run the risk of minimizing the competitive, individual-achievement basis for incentives, which runs through most of the institutions of our society. Do we want to scrap the competitive principle in schools, only to revive it in later occupational settings and thus make schooling irrelevant to later life? Many would agree that the negative effects on self-esteem that schooling almost inevitably involves should be conditioned by humane practices, but few

would argue that the deeply institutionalized competitive principle can be realistically minimized in school settings in the short run.

Wisdom appears to suggest caution, then, in the area of policy. Certainly we should encourage interventions and public expenditures that could strengthen rehabilitation and construct institutional arrangements to rescue or bolster self-esteem. But these interventions should be cautious, with experimentation, model programs, and continuous evaluation the order of the day. As scientific findings become more reliable, however, and as the involvement of self-esteem in the causes of social problems becomes better understood, policy interventions can profitably become correspondingly bolder.