Ambition and attainment: a study of four samples of American boys 9780912764092

Book by Kerckhoff, Alan C

106 19 22MB

English Pages 106 [112] Year 1974

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Ambition and attainment: a study of four samples of American boys
 9780912764092

Table of contents :
Frontmatter
PREFACE (page iii)
LIST OF TABLES AND FIGURES (page v)
1. THE STATUS ATTAINMENT PROCESS (page 1)
2. DESIGN OF THE STUDY (page 10)
3. PRELIMINARY FINDINGS (page 19)
4. SCHOOL EXPERIENCE (page 37)
5. BELIEFS ABOUT OPPORTUNITY (page 54)
6. PARENTAL INFLUENCE (page 63)
7. PEER INFLUENCE (page 73)
8. SUMMARY, SYNTHESIS, AND INTERPRETATION (page 85)
REFERENCES (page 105)

Citation preview

AMBITION AND

ATTAINMENT A Study of Four Samples of American Boys

by

Alan C. Kerckhoff Duke University

The Arnold and Caroline Rose Monograph Series of the American Sociological Association

i

THE ARNOLD AND CAROLINE ROSE MONOGRAPH SERIES IN SOCIOLOGY

A gift by Arnold and Caroline Rose to the American Sociological Association in 1968 provided for the establishment of the Arnold and Caroline Rose Monograph Series in Sociology. The conveyance provided for the publication of manuscripts in any subject matter field of sociology. The donors intended the series for rather short monographs, contributions that normally are beyond the scope of publication in regular academic journals.

The Series is under the general direction of an editorial board appointed by the Council of the American Sociological Association and responsible to the Publications Committee of the Association. Competition for: publication in the Series has been limited by the Association to Members and Student Members.

Arnold Rose was my teacher and my friend. I was fully aware, before his untimely death, of his sense that sociology needed a publication outlet of the sort provided by this Series; and I was dimly aware of his hope that his and Caroline’s gift would meet that need. I am grateful to the American Sociological Association for providing me the opportunity to help fulfill Arnold’s hope. Sheldon Stryker Editor

January 1974 Library of Congress Number 74-75001 International Standard Book Number 0-912764-09-0

©American Sociological Association

1722 N St. N.W.

Washington, D.C. 20036 : ii

PREFACE A study of this kind can be conducted only if the necessary elements are available. The first is a population of subjects from whom the data can be collected.

This was made available through the generous cooperation of the Fort Wayne | | Community Schools, whose Superintendent, Lester L. Gile, was a strong supporter

of the project from the outset. The facilitation of the day-to-day work of the project in Fort Wayne was skillfully provided by Douglas Baugh, Assistant Director of Personnel. To those two, as well as the many others there who contributed to the project, I am very grateful. Access alone, of course, is not enough. The boys’ serious participation in the research was essential and greatly appreciated.

A second critical element is the funds to support the research, which were provided by the U.S. Office of Education, Project No. 8-0053, Grant No. OEG-3-808-0053-0057 (085). They are gratefully acknowledged.

A third necessary element is a set of ideas to guide both the plan of the data collection and their analysis, once collected. Although the final form of the data and my treatment of them are solely my own responsibility, the work has profited greatly from the consultation, assistance and advice of Otis Dudley Duncan, Robert M. Hauser, William M. Mason, James N. Porter, Sharon Sandomirsky Poss, and William H. Sewell.

A fourth important element is technical assistance in the conduct of the study. Among the numerous contributors, James Porter and Sharon Poss were particularly helpful. Porter administered the in-school questionnaires and participated fully in the construction of the several instruments used. Sharon Poss helped in so many

ways that it would be impossible to list them here. Suffice it to say that without her the work would have been impossible, certainly within the time limits. Such an able and motivated assistant is a rare find: I have been exceedingly fortunate. Finally, the earlier drafts and the final manuscript were typed with care and skill by Jean Brenner, Madge Lee and Marcia Spray. Their efforts are much appreciated Alan C. Kerckhoff iii

TABLE OF CONTENTS Page LIST OF TABLES AND FIGURES ............ 0... cece cece ere ceeeee V Chapter

1. THE STATUS ATTAINMENT PROCESS ..............cceeeeee. 1 The Social Context Six Factors Affecting Status Attainment The Process From Factors to Variables and Back

2. DESIGN OF THE STUDY ............ cee eee eee cece ee eee 10 Method Selection of the Research Site Selection of Samples Design of Instruments Summary and Outline of the Report

3. PRELIMINARY FINDINGS ........... 0... ccc ccc ee ee ee eee ee ee LY The Basic Attainment Model The Basic Ambition Model Summary and Interpretation

4. SCHOOL EXPERIENCE ........,..... ccc cece cece cece eee ST Academic Performance IQ, School Experience and Expectation Implications for the Basic Models Two Elaborated Models

5. BELIEFS ABOUT OPPORTUNITY ............ 2... cece eee eee D4 An Ambition Model Incorporating Fatalism An Attainment Model Incorporating Fatalism

6. PARENTAL INFLUENCE ............... 0.02 cece cece etc ccc ee 6 03 Social Characteristics of Mother Parental Encouragement The Quality of the Parent-Son Relationship

7. PEERINFLUENCE ................ cee cece ce eee cece ets eeeeee dS Peer Similarity Evidence of Peer Influence The Effect of Peer Similarity

8. SUMMARY, SYNTHESIS, AND INTERPRETATION ............. 85

Explanation and Explication

A Synthetic Cohert Model Discussion iv

Table Page LIST OF TABLES AND FIGURES

2.1 Fort Wayne Compared With “The Average City” ............... 14 2.2 Fort Wayne Sample Loss, by Cohort and Race ............... 16 3.1 Descriptive Summary, In-School Cohorts ................... 20 3.2 Descriptive Summary, Class of 1963 ..............02020006- = 2] 3.3. Correlation Matrix, Basic Model of Duncan ................. 25 3.4 Correlation Matrix, Basic Model, Fort Wayne Graduate Cohort ..... 25 3.5 Correlation Matrices, Basic Ambition Model, In-School Cohorts ..... 28 3.6 Path Coefficients, Basic Ambition Model, In-School Cohorts ....... 3] 4.1 Characteristics, Over-, Under- and Normal Academic Performers,

In-School Cohorts ....... 0.0... cece eee ew eee eee eee) AQ 4.2 Correlation Matrix, Grade and Grade-Partic Models, Graduate Cohort . 45 4.3 Correlation Matrix, Grade and Grade-Partic Models, In-School Cohorts 46

4.4 Path Coefficients, Grade Model, In-School Cohorts ............. 47 4.5 Path Coefficients, Grade-Partic Model, In-School Cohorts ......... 5] 5.1 Correlation Matrix, Grade-Fatalism Model, In-School Cohorts ..... 58 5.2 Path Coefficients, Grade-Fatalism Model, In-School Cohorts ....... 59 5.3. Correlation Coefficients, Grade-Fatalism Model Using Ultimate

Educational and Occupational Expectations, Graduate Cohort ..... 61 5.4 Path Coefficients, Grade-Fatalism Model Using Ultimate Educational

and Occupational Expectations, Graduate Cohort ............. 62 6.1 Correlation Coefficients, Basic Model, by Mother’s Employment

Status, In-School Cohorts ........... 0... cee eee eee eee es 66 6.2 Path Coefficients, Basic Model, by Mother’s Employment Status,

In-School Cohorts ..... 2... 2... cee ee eee eee ee eee 8)

6.3 Correlation Coefficients, Parental Encouragement Model,

In-School Cohorts 2... 0.0... ccc eee ee eee ee eee eee eee 69

6.4 Path Coefficients, Parental Encouragement Model, In-School Cohorts . 70 7.1. Correlations between Educational Expectations of Friends,

In-School Cohorts ..... 2... 0. cee eee eee eee ee eee IS

7.2 Agreement of Long-Term and Short-Term Friends on Educational _ Expectation at Two Points in Time, Ninth- and Twelfth-Grade

Cohorts 6. ee ee ete ee eee ee ee eee eee TT

7.3. Agreement of Previous and Current Friends on Educational Expectation in Twelfth Grade and Six Years Later, Graduate Cohort . 78

7.4 Correlation Matrix, Grade-Friend Model, In-School Cohorts ....... 80 7.5 Path Coefficients, Grade-Friend Model, In-School Cohorts ....... 81 7.6 Correlation Matrix, Grade-Friend Model, Graduate Cohort ....... 82 7.7 Path Coefficients, Grade-Friend Model, Graduate Cohort ......... 83 8.1 Summary of Ninth- and Twelfth-Grade Models, EdExp as

Dependent Variable... 0... 0... 0... eee ee eee ee te eee ee eee = 87 8.2 Summary of Graduate Models, EdAtt as Dependent Variable....... 88 Vv

8.3. Correlation Matrix, Synthetic Cohort Model ................. 92 8.4 Path Coefficients, Full Synthetic Cohort Model ............... 95 8.5 Path Coefficients, Delimited Synthetic Cohort Models ........... 98

Figure Page 8.6 Implied Correlations and Correlations of Residuals, Delimited Models . 99

3.1. Basic Path Model, Fort Wayne Graduate Cohort ............... 24 3.2 Basic Ambition Model, Twelfth-Grade Cohort ................. 30

4.1 Grade Model, Graduate Cohort .................. 00 ee eeeee 4 4.2 Grade-Partic Model, Graduate Cohort ...............00020- 49 5.1 Pattern of Relationships Among Beliefs about Self and

Opportunity, In-School Cohorts ............ 0.0.22 eee eee ee 96

8.1 Structure of Synthetic Cohort Model ....................... 93 8.2 Delimited Synthetic Cohort Models ....................... 97

vi

CHAPTER I

THE STATUS

ATTAINMENT PROCESS

In essence, this volume reports the findings of a study of the process of intergenerational mobility. I have used the term “status attainment process” instead of the more traditional term “intergenerational social mobility” not simply because it is less cumbersome, but to emphasize that the research is based on a social psychological conception of social mobility rather than on the more traditional demographic view.

A social psychological approach to intergenerational social mobility focusses on the process by which individuals move from points of origin in stratification systems to later destinations in them.' Mobility is viewed as a flow of individuals occurring over time, with different individuals experiencing different outcomes. The primary concern is to explain the differential outcomes by reference to general factors which are thought to influence the process. The character of the stratification system and the basic societal mechanisms provided for movement within it can be viewed as given, but such contextual factors play an extremely important part in

the process. They both establish the limits of possible attainment and provide potential sources of influence which vary according to the social position of the developing individual. It will therefore be necessary to look at the varied implications of the social context for different kinds of individuals, if the process by which they move from origin to destination is to be charted adequately. The Social Context

For present purposes, the most important characteristics of the social context may be discussed in terms of the bases of stratification and the means available for intergenerational movement. It is possible to differentiate several bases of stratifica-

AMBITION AND ATTAINMENT ]

tion in American society (Lenski, 1966), but perhaps the occupational structure reflects more clearly than any other its major dimensions (Blau and Duncan, 1967, _ 6-7). The prestige level of an occupational position is strongly, though not perfectly, associated with the power and privilege enjoyed by its incumbents. Thus one’s position in the stratification system can be indexed by the social level of his occupation. The combination of this summary quality of occupational prestige and the relative ease with which it can be measured has led to its widespread use in _ research on stratification and mobility. In the present report, level of occupational prestige will be used as a measure of both the social origin and the early adult placement of the young males studied, origin being indexed by the boy’s father’s occupation, and destination by the boy’s occupation as a young adult. The major vehicle of mobility provided by American society is the public school system. Although occupational opportunities are not wholly contingent on performance in school, there is a strong association between educational and occupational attainment. Also significant in the process of status attainment is the norma-

' tive prescription of open access to educational and occupational opportunities. Within the school system, a salient norm supports the right of all children to an opportunity to perform the required tasks and to be evaluated on the basis of universalistic standards. Similarly, the criteria of recruitment into occupations are supposed to be limited to characteristics directly relevant to job performance and not to such qualities as race, ethnic identity, social origin, and so on. The prescription of equal access is at least partially due to the importance of educational attainment in determining occupational access, academic performance being viewed as mastery of job-relevant tasks in an “open” system with universalistic standards of evaluation. Neither the structural nor the normative features of the system guarantee that every individual will have an equal chance of reaching a given level of occupation, of course, althouth they increase the likelihood of equal opportunity. The actual flow of individuals from origin to destination exhibits both mobility and continuity from generation to generation. In effect, the openness of the stratification system may be indexed by our inability to predict an individual’s destination from knowledge of his origin; since less than one-fifth of the variation in destinations in our society can be explained by origins (Blau and Duncan, 1967), it may be argued that ours is, indeed, a relatively open system. Destination is far from fully determined by origin, but neither are the two wholly independent. Research on the process of movement from origin to destination must thus attempt to clarify the bases of both mobility and continuity. It can neither ignore origins nor assume that they are the most important factor in the status attainment process. Six Factors Affecting Status Attainment The structural and normative contexts of mobility help us raise basic theoretical

questions about the factors which influence the flow of persons from origin to destination in a society. In the terms used above, the problem is to specify the factors which influence the attainment of occupational positions enjoying varying levels of social prestige, given varying levels of social origin. Six factors are: ability, opportunity, performance, sanctions, understanding, and ambition.

2 THE STATUS ATTAINMENT PROCESS

Ability. Presumably, to the extent that individuals are endowed with different amounts of intelligence, they have varying chances of attainment in an open system. One would expect, therefore, that level of destination might be explained at least in part by native ability. The significance of native ability in the actual distribution of persons in our society is a highly charged topic, given the racist arguments

sometimes based on its presumed significance. Such arguments, of course, run backwards, taking as true both the openness of the opportunity structure and the significance of native ability, and arguing that those who have achieved less are therefore less able. In research which is designed to test such propositions rather than assume them, one simply investigates the degree to which attainment can be explained through information about ability. Even here, however, there are almost insurmountable problems since none of our measures of ability can be viewed as measuring native ability. They are, rather, measures which quite obviously reflect experience as well as innate qualities. To that extent, such measures are influenced by the other five factors, and an individual’s score is thereby subject to change during his lifetime.

Opportunity. If one looks at the process of social mobility rather than the outcome, it is obvious that many different kinds of opportunity at different points need to be taken into account. There are differences in opportunity to learn various skills at home, to attend schools with highly qualified teachers, to associate with

various kinds of peers, to perform in ways that will be evaluated for future refer- ) ence, to have another chance after failure, to learn how society really works, and so on. Thus opportunity is not a simple matter of being given equal consideration once job openings occur, but a very complex set of circumstances which influence the attainment process throughout life. The presence of interdependent contingencies makes it particularly difficult to consider all the facets of opportunity. That is, if

one has missed or been denied an earlier opportunity, later ones are likely to be fewer. Later opportunities are thus at least partially dependent on earlier ones. Later opportunities are also dependent on what one has (or has not) done with those earlier opportunities—which leads to the third of the six factors. Performance. In most areas of performance the individual, as he matures, moves

from the relatively simple to the more difficult and complex, and his ability to perform at some later date will depend heavily on his mastery of the earlier tasks. Early performance thus influences later performance in two ways: by making it possible for one to learn the more complex tasks and, as noted above, by influ-

encing his opportunity even to try to learn them. The school system is heavily committed to this view of growth in performance potential. Children are moved through the school grades in keeping with their growing ability to master more and more difficult tasks. They are constantly evaluated, and this evaluation is fed back to them and to their parents so that they may adjust their efforts accordingly. The age-graded structure of the school leads to a strong pressure on the individual child

to keep up with his age cohort and to move through the system according to a pre-conceived and externally imposed schedule. The child who fails to perform satisfactorily at a given level either is left behind (to try again), or is moved along with his cohort, with the knowledge that he will not be able to perform later tasks as difficult as those presented to his age-mates. In fact, as one moves into adoles-

AMBITION AND ATTAINMENT 3

cence, the school system provides for whole segments of the age cohort who have performed less adequately earlier—through “streams” or “programs’’ pitched to

different levels of acquired information and skill. |

Sanctions. Feedback about earlier performance in school serves as a set of sanctions as well as a source of information. In effect, the child is punished with low grades and rewarded with high ones. But there are many other kinds and sources of sanctions. Parental and peer approval and disapproval iof not only academic but other kinds of performance are constantly being experienced. Such sanctions may or may not be consistent with those received at school. Not only may others punish the very behavior the school rewards (and vice versa), but they may reward behavior ignored by the school and thus present to the young person alternative means of

gaining rewards. In the latter case the goals of the school are being subverted through distraction rather than opposition, but the result may. be equally detrimental to the young person’s academic performance.

, Understanding. I refer here to an understanding of “how the system works,” to the ability to answer such questions as: What happens to students who fail in their school work? What are the advantages of going to college? Why should anyone want to learn algebra? What are the conditions of work in a factory as compared with an office? What are the long-range prospects for income from different kinds of jobs? What does one have to do to get into different kinds of jobs? Probably no one ever

has a wholly adequate understanding of such matters, but it is likely that most adults know a great deal more about them than most children. A critical feature of such practical knowledge of the world is when one gains it. If the significance of early steps in the attainment process is not appreciated, those who, for whatever reason, perform inadequately early may not “catch on” until it is too late. Evenif — such ignorance is rather general among young children, some will actually perform better than others according to the criteria of the school system, and their early performance will increase their opportunities as well as their growing skills. It seems likely, however, that there will be some association between understanding the system and adequacy of early performance, either because both depend on native intelligence or because of differences in opportunities to learn out-of-school. In any event, some understanding is ultimately necessary if the individual is to guide his own efforts effectively to achieve specific goals. Ambition. The above discussion assumed that if one understands the system he will gear his efforts to reach a relatively high position in it. That is, it assumed that “success’’ is defined similarly in all segments of the society. There is ground for that assumption—otherwise there would not be such a stable ranking of occupations— but at the same time there are differences in values among various segments of the population. The term “ambition” will be used here to refer to one’s willingness to work to achieve goals. Thus, although some goals may be highly desirable in the eyes of everyone, the specific referents of ambition may vary considerably from one part of society to the next. In terms of the dynamics of attainment, this means both that occupations may vary in their desirability, depending on one’s position, and that people may have different occupational goals and still be equally ambitious. If nothing else, a middle-range occupation may look much more desirable to

4 THE STATUS ATTAINMENT PROCESS

a boy who comes from a lower-class family than to one who comes from an upper-middle-class family: what would constitute upward mobility for the first boy

would be downward mobility for the second. Such an occupational goal would imply very different things about the ambitions of the two boys. It is also important to see how all of the other factors discussed above affect ambition. The individual’s working knowledge of the stratification system and his own place in it will tell him how desirable different occupations are; his earlier performance and the evaluation it has received will provide him with a view of his own abilities and his prospects of future performance; and his conception of the details of the opportunity structure will influence his willingness to commit himself to goals, however desirable they may be and however competent he may think he is. Thus the present view of ambition involves more than a wish for a desirable goal; it requires that the individual believe he has a reasonable prospect of attaining it. The Process

None of the six factors can be viewed as wholly fixed at the birth of a child. Although presumably there are differences in ability which are innate, measured ability certainly is not a direct result of innate qualities, and changes in measured ability during a lifetime strongly suggest continuing experiential effects. Similarly, though much of the opportunity structure faced by a child is given and unchanging, at least some of it is contingent on his own early performance and/or on decisions made by his parents (e.g., to move to another school district) or by others (e.g., to

change district lines). Also, even if sanctions are forthcoming on the basis of a predetermined contingency pattern, which rewards he receives from whom will depend on what he does. It is even more apparent that performance, understanding, and ambition are problematic at the outset. Thus all six of these factors are likely to vary during the lifetime of the individ-

ual. Such variation will depend on the kinds of experience he has, and, for the social psychologist, the most significant are those he has with other people. These others act as models, as teachers, as sources of sanctions: their characteristics are part of his opportunity structure, though he can usually choose to stress one social relationship rather than another. It will be noted that two of the six factors (opportunities and sanctions) are located in the social milieu within which he moves while

the other four refer to certain of his personal characteristics. The changes that occur during his lifetime, therefore, are in both the individual and in his social milieu; and these are interdependent. The social milieu affects his personal characteristics, but he also learns to differentiate and to choose among the available social relationships. But the order of events is important. To the extent that ability is innate, and to the extent that the provision of opportunities within the family and neighborhood is constant during the early years, it will take exceptional inputs from other sources to alter the direction of his development. Some teachers may be able to counteract such effects by the skillful manipulation of opportunities and sanctions. Sometimes a special talent exhibited by the child will gain him rewards in school which will sustain ambition even in the face of relative failure. It may be that friends will serve as models and even give assistance so that he can fare better in the future than in

AMBITION AND ATTAINMENT 5

the past. It seems even more likely, however, that the process will tend to be self-reinforcing: that he who performs well early will be rewarded, will be provided with further opportunities, will develop a view of the world as a place where effort pays, and thus will remain relatively ambitious. In turn, ambition and a sanguine view of possibilities of achievement are likely to promote the quality of performance, thereby increasing his opportunities and his measurable abilities. In order of significance in the attainment process, therefore, native ability and family-based opportunities may be seen as most basic, early school performance — and the sanctions associated with it being contingent on them, and understanding and ambition as later developments. At the same time, the persistent significance of

all of these factors must be acknowledged. Those with great native ability are undoubtedly better able to alter their course once they gain sufficient practical knowledge of the attainment process to recognize the relationship between academic performance and occupational attainment. Family-based opportunities continue to be important as the child matures, families with greater resources being better able to help him surmount or rectify earlier mistakes. Children who have performed well early have both the records and the skills to make possible a high quality of performance later on, even if they are not very ambitious. And so on. Viewed thus, the attainment process is seen to be exceedingly complex, countless contingencies making each individual case unique. No single research project

can hope to consider all this complexity. It is necessary to focus on a limited set of the more salient patterns rather than on the infinite variety of individual experience, and to confine research operations to those which are feasible within the social and economic limits faced by all research. We do not know how to measure everything that seems conceptually relevant (e.g., the differential effects of tuition

and modeling vis-a-vis a particular significant other). And we may be unable, because of economic or normative limitations, to measure some things we know how to measure (e.g., the quality of parent-child interaction on a day-to-day basis). The research design is thus always a simplified version of its conceptualization. From Factors to Variables and Back

Few of the six factors discussed above can be measured in any direct way. Perhaps one can measure ambition rather directly by asking the individual what goals he has set for himself and whether he expects to reach them. It may be also that IQ is. the most reasonable measure of ability we can expect, although it must be seen as a measure of functional rather than innate ability. Even in these two cases, however, it may be argued that it is unlikely that fully adequate measures can be obtained in large-scale research which is dependent on paper and pencil instruments. The problems of measurement are even greater when one considers the other four elements.

Opportunity as such is probably impossible to measure directly; certainly any total assessment of any individual’s opportunities is impossible. But it is possible to obtain a crude index of aspects of his opportunity structure. Any measure of the parents’ social characteristics can be interpreted in this way. Parents who have reached a high level of education can at least provide the child with interpersonal

| experiences which a poorly educated parent cannot supply. Similarly, wealthy

6 THE STATUS ATTAINMENT PROCESS

parents can buy toys, books, travel, music lessons and so on which the poor parent cannot afford. The personal characteristics of parents (attitudes, values, interpersonal styles, etc.) can also be seen as an index of kinds of influence which might be exerted upon the child at home. The number and characteristics of siblings also

form part of his family structure of opportunity, because of their effect on the child’s interpersonal experiences for one thing, and their effect on family finances for another. Whether or not the mother works also influences the family economy and her interaction with the child. At school, the socio-economic composition of

the student body, the characteristics of the teachers, and the nature of the academic and extracurricular program all affect his opportunities to learn. Few if any of these, however, can be viewed as direct measures of opportunity, as such. What this means, above all, is that associations between such measures and attainment need to be interpreted; their meaning is not always obvious. If boys of

high socio-economic status (SES) reach higher levels of attainment, it may be because they are smarter or because they have wealthier fathers who can assist them (Blau and Duncan, 1967), or because high-status parents relate to their sons differ-

ently and encourage them to strive more purposefully (Rosen, 1956). In some cases, the relative merit of such interpretations can be tested with further information; but not always. There will usually remain a gap between factor and variable, between concept and datum. The conceptualization leads to the adoption of measures which are meaningful and feasible within the limitations of the research design. The outcome provides a test of the conceptualization, but it is not complete without the interpretive link between factor and variable.

As to performance and sanctions, it is difficult to consider one without the other. In general, sanctions are contingent on performance, and performance is responsive to prior sanctions. Within the family and in school, the child’s behavior

leads to responses from others which are rewarding or punishing, and his later behavior reflects the earlier sanctions. To some extent, academic performance can be directly measured through achievement tests, but it is doubtful if that is the most useful measure. More significant in the on-going process of achievement in the school are the grades received on particular exercises and periodically through the official grading system. The latter are known to all relevant parties—student, parent, teacher (and often even peers)—and they are likely to have a significant effect on his future. But such grades are not simply objective measures of performance; they

are also an expression of the teacher’s evaluation of that performance. Thus they measure both performance and reward for performance. Other measures of performance and reward may be equally important. Parents’ responses to the child’s school work, especially to his grades, add further to the sequence: performance-evaluation-performance. Performance in the non-academic activities at school provides possible alternative sources of sanctions, both there and at home. Participation in extra-curricular activities, being a “behavior problem,” being popular with one’s peers, all index both performance and sanctions, and are thus probably significant for the individual’s attainment pattern. In all such cases it is difficult if not impossible to separate performance and sanction, but that does not reduce the relevance of the measures to the conceptualization. The interpreta-

AMBITION AND ATTAINMENT 7

tion of such measures, however, must usually involve some important assumptions which go beyond the measures themselves, such as that high grades stand for either superior performance or teachers’ preferences, or both, and further that they exert

a favorable effect on future performance through their influence on ambition or opportunity, or both. Similarly, being defined as a behavior problem in the school is assumed to index both the child’s past performance and his probable subsequent involvement in official school activities. In the same way, participation in extracurricular activities and popularity with one’s peers are assumed to provide rewards at school and thus have the effect of keeping the child involved enough to increase his chances of performing well academically.” In its most general form, understanding refers here to an accurate expectation of the probable outcomes of various kinds of behavior. Clearly, no one has full understanding in this sense, and it would be meaningless to try to measure the extent of it in a given individual. Yet there are some kinds of understanding of “how society

works” that are important in planning one’s future and that a young person can obtain. Such matters as the association between education and occupational opportunity or the income characteristic of various kinds of jobs are both knowable (at least in a gross sense) and significant. We say the child who wants to be a doctor but who hates school is “unrealistic’’ because he lacks an adequate understanding of the link between educational performance and admission to the professions. The child

with a low IQ and high occupational aspirations is also unrealistic, though in a different way.

Yet such knowledge of the world is more than a purely cognitive grasp of the manifest characteristics of the social system: it also involves an image of oneself in the system. The child with low IQ who wants to be a doctor may be unrealistic because his image of himself is false rather than because he is ignorant about society. It is equally true that an individual’s understanding of society and himself is very difficult to separate from his attitudes toward them both. For instance. people from different segments of the population differ in their confidence in their ability to control their own destinies (Coleman, Campbell, et al., 1966) and in their belief in ‘“‘fate’(Kahl, 1965). Whether one argues that these findings indicate levels of understanding of how the system “‘really” works, may depend as much on one’s ideology as on the data themselves. It is manifestly true that people of lower status actually have less chance to control their own destinies. But whether one accepts that as true, it seems very unlikely that the individual who believes that it does not pay to strive for a goal will do so. Thus, whether such measures are called “understanding”’ or “knowledge” or “‘belief’? may not be so important as that their relevance in the attainment process is recognized. In the case of the great majority of variables used in research on the attainment process, therefore, the connection between measure and conceptualization is complex; the findings produced with the use of such variables require interpretation in

light of the broader conceptualization. The variables are often proxies for desired but impossible “pure” measures (e.g., grades, instead of academic performance as such). In other cases, they are based on a limited sample of relevant observations (e.g., proportion of high-status classmates as an index of extra-familial opportunities to learn middle-class values). In still other cases, the measures are summary

8 THE STATUS ATTAINMENT PROCESS

indices of many possibly relevant factors (e.g., father’s occupation as an index of opportunities to learn in and through the family). Such variables differ, of course, in the adequacy with which they represent the given factors, and interpretations of research findings using them are thus subject to criticism. Interpretation is almost always necessary , however, because a wholly adequate measure is indeed rare. Much of the interpretation of findings using such measures postulates the impor-

tance of interpersonal influence, a factor which can seldom be measured directly. Similarity between the attitudes, goals, or values of two closely associated persons (e.g., two friends or a parent or child) may often be taken as evidence of interpersonal influence, but demonstrating it to be so is another matter (see Kerckhoff and Huff, 1973). In the later chapters of this report some of the difficulties of such interpretations are examined more closely. Here it is only necessary to stress the fact that conceptualization and interpretation will inevitably go beyond the limits of the data: the data seldom “speak for themselves.” This is not intended as an apology. It is simply to emphasize the necessity of theoretical activity before, during, and after data are collected and analyzed. It is necessary beforehand to determine which variables ought to be measured, a decision which must balance feasibility against theoretical significance. It is necessary during analysis to increase the ability to choose among ways of analyzing the data. And it is necessary after the study is completed to highlight the points of greatest and least clarity of outcome. It is one thing to find that SES of origin is related to SES of destination to a statistically significant degree; it is a much more important and difficult thing to provide a convincing interpretation of how this association comes about. The latter requires both a great deal more information about the process of status attainment and a considerable amount of theoretically cogent interpretation of whatever data are available.

The general conceptualization of the attainment process which guided the present research has been presented in terms of six factors as they operate within the context of several kinds of interpersonal relations. This study was necessarily concerned with a much less ambitious task than charting the whole process in these conceptual terms. It is limited especially in that it has used easily measured, familiar variables requiring interpretation in light of that conceptualization. Throughout the

design, data collection, and analysis, however, the conceptualization was kept in mind, and as far as was possible the research took it into account.

FOOTNOTES

lit is true, of course, that during the period covering the move from origin to destination, the nature of the stratification system itself may change. For present purposes, however, that

added source of complexity will be ignored. |

2 This, of course, also depends on the link between these sources of rewards and the academic program. If extracurricular groups are anti-intellectual (as some athletic teams are) or if the peer group’s value system is anti-academic, the effect may be negative rather than positive (Coleman, 1961).

AMBITION AND ATTAINMENT 9

CHAPTER II

DESIGN OF THE STUDY

The study reported here is an attempt to examine the relevance to the attainment process of the six factors discussed in Chapter I. Since attainment extends through the entire period from birth to early adulthood, and since the interrelationships among the many variables are admittedly complex, a design had to be found that permitted an examination of some of the time period and part of the complexity, but in which the overall pattern of mobility was still kept in mind. In effect, the design used here started out with the large-scale view of the flow of individuals within a system of stratification and mobility and attempted to add more detail. Within the basic framework of our knowledge of the pattern of intergenerational mobility, as outlined by Blau and Duncan (1967) and others, an attempt was made to look at various relevant factors as they influence the pathways from social origin to destination followed by young males during the pre-adult years. Using ambition, or the boys’ views of what they can accomplish in the future, as an index of the direction in which they are moving at any point in time,

the aim was to assess the role played by various influences on ambition, and ultimately on attainment. To do so, it was necessary to adopt methods which permit the simultaneous analysis of a number of different factors at different points during the pre-adult time, and to use measures which are effective and feasible in relatively large-scale research. Clearly, it was not possible to measure everything one might consider important, nor was it possible to examine every item of information in full detail. Method

Briefly, the method was chosen to fulfill three important requirements: (1) to

10 DESIGN OF THE STUDY

permit analysis of a large set of variables relevant to the conceptualization outlined in Chapter I and subject to empirical investigation; (2) to supply a set of data which

includes the critical measures taken for the same set of cases, rather than one relationship measured on one set of cases, and another relationship measured on another set; (3) to provide measurement at a number of points during the most critical period of the life cycle as a means of illuminating the process involved. Each of these issues is discussed below. (1) Multivariate techniques such as multiple correlation are of only limited value in the situation being studied because they are designed to examine the effects of a

number of independent variables on a single dependent variable, rather than to explicate the structure of a set of links among variables, some of which may be best viewed as intervening. Moreover, after three or four independent variables have been used, additional variables do little, in most cases, to explain the variance in the dependent variable. Thus, if all the variables involved in this problem area were used

to explain the variation in, say, the level of occupational placement of the son, most of them would contribute little to the analysis. A more effective strategy in the study of a process is to view the relationships as links in a chain of influence rather than as coequal sources of simultaneous influence. The most promising technique for the purposes at hand seems to be path analysis, introduced to sociology by Boudon (1965) and Duncan (1966).' This method is appropriate because it requires a view of process as a flow of influence. Although behavioral scientists often shy away from explicit acknowledgment of it, most of our theory contains a causal logic. In the present case, the logic is not always fully explicit, but with respect to many of the links there would be general agreement about the direction of the flow of influence. Father’s occupational position is seen as preceding parental childrearing practices, and thus if there is any dominant flow

of influence between the two, it must be from the first to the second. Certainly son’s academic performance influences his educational attainment rather than vice versa. And, although there may well be interaction, the general theoretical position taken here calls for parental childrearing practices to influence the son’s characteris-

tics more than the reverse. Although not all of the links can without debate be placed in a flow diagram representing the direction of influence, the merit of attempting to construct such a diagram is obvious. It would simplify and make explicit a theory that is currently implicit and fuzzy at best. (2) The previous studies of relevance here are almost all confined to the examination of a limited set of the crucial variables. One study yields a measure of the relationship between father’s and son’s occupational position; another gives a meas-

ure of the relationship between family SES and childrearing practices; a third provides data on the link between SES and the son’s academic performance; and so

on. Even more critical is the fact that these studies present data on cases from various settings (rural-urban, regional, etc.) and relevant to boys at various points in

the life cycle. It is thus difficult to know if they provide pieces of the same or different theoretical puzzles. In spite of such difficulties, however, it is possible to make some progress toward the development of a coherent model of the attainment process. The work of Duncan et al. (1968) was devoted to the construction of a

multiple-factor model, using the technique of path analysis, based on bits and

AMBITION AND ATTAINMENT 11

pieces from several demographic and social psychological studies. Duncan’s innovative work in this area encouraged me to believe that, with more explicitly relevant data, considerable progress can be made. More recently, Sewell and his associates have published analyses similar to the type used here (Sewell, Haller, and Ohlen-

dorf, 1970). (3) If it seemed likely that the pattern of interrelationships among the many factors just discussed were constant through the period of the son’s development,

the diversity of the sources of information about the process would not be so troublesome. Since a shift would be expected in the pattern of relationships as the boy moves through adolescence and into adulthood, the variation leaves the development of a summary model of the pattern open to serious criticism. The ideal solution would be a continuing longitudinal study. The state of development of work in this area, however, is not sufficiently advanced at this point to warrant the

necessary investment of time and money. As a result, it was considered more efficient for the present to conduct what has been called a synthetic cohort analysis: data are collected at one point in time from a series of age cohorts drawn from the same larger population, and comparisons are made across cohorts as if they represented successive measures on the same cohort. Analysis across cohorts can be

by interpolation from the structure of the model at one age to the structure at another age; it may also be by use of data from any cohort on its characteristics at

an earlier point in time. The latter kind of analysis requires either recorded or retrospective data. Although retrospective data are not the strongest basis for an analysis, the “real” data available from younger cohorts to compare with the retrospective data from older cohorts provide the basis of some reasonable assumptions about the adequacy of the retrospective data. Through such means, links between the models for successive cohorts may be at least tentatively inserted in the conceptualization of the overall process. Although such a method does not solve all of the problems involved (Schaie, 1965), considerable clarification may be gained. The research which the present volume reports involved, therefore, the collection of data from several age cohorts of young males and the development of a coherent multiple-factor model at each age level and of a tentative processual model made by linking the several cohort models together.

Three major planning decisions had to be made at the outset: selection of a research site, selection of samples, and design of instruments. 7 Selection of the Research Site At first, it seemed desirable to obtain some kind of national sample as a basis for the study, but the difficulties were imposing. The expense and logistical problems, given the need for several age cohorts, were serious deterrents. Also, information was wanted from and about the close peers of the subjects, and this required either

that all of the boys of a particular age in some population be studied or that a two-stage sampling design be used, calling for, first, data from a sample of boys, and

then data from those they named as friends. Because of these problems, it was decided to use a single city as the research site. In choosing a city, a careful review of many characteristics of all U.S. middlesized cities was made, a middle-sized city being one with a population of between

12 DESIGN OF THE STUDY

90,000 and 495,000. All cities whose population characteristics deviated sharply from those of the total U.S. urban population were eliminated.” From among those remaining, selection was made according to accessibility from Duke University. Permission to conduct the research within the school system of the chosen city was then requested. The school officials in Fort Wayne, Indiana, were interested and highly cooperative in the arrangements for the study, so it was conducted there.

Table 2.1 shows how the Fort Wayne population in 1960 compared with the average of American cities, using two definitions of the “average” city. Selection of Samples

Although the whole of a boy’s pre-adult life may be seen as relevant to the | attainment process, this study concentrates on the latter portion of it. The choice of a lower limit of the age span was dictated by two considerations: first, subjects needed to be old enough to understand questions about the future, dealing with educational and occupational goals; second, they needed to be literate and mature enough to cope with a questionnaire. Although the findings are somewhat mixed, several studies have shown that children in the upper elementary school grades have some understanding of the occupational and social class characteristics of American society (Stewart, 1959; Simmons and Rosenberg, 1971; Tudor, 1971). Moreover, the experience of others, together with pre-testing of the questionnaires used in this

study, demonstrated that it was possible for children as young as ten or twelve years of age to use pre-coded questionnaires. It was therefore decided to include in _ the study an elementary school sample, and to insure the maximum effectiveness of data collection the oldest group, the sixth grade, was chosen. Two older school cohorts were also included, the ninth and twelfth grades. The ninth grade contains a wide range of students, including most of those who will ultimately drop out before graduation. The twelfth grade is the most critical point of reference for further educational potential; most previous studies have dealt with

twelfth graders. Finally, there is a cohort of boys who had graduated from high school six years earlier. Once the decision was made to study these four cohorts, the next problem was to choose the specific samples within the cohorts. In view of the interest in the boys’ relations with their peers, the school was made the sampling unit, and all of the boys in the designated grade in any chosen school were included.

A central issue was to determine the high schools to be used; the other schools chosen were feeders to them. Fort Wayne has five public high schools, one of which at the time of the data collection (Spring, 1969) was only three years old. Thus for

purposes of selecting a graduate cohort the choice was among only four high schools. Furthermore, since the building of the new school the district lines had been shifted; consequently, to make the graduate and twelfth-grade cohorts comparable in the geographic areas represented would have required a highly selective

inclusion and deletion of cases. It thus was reasonable to include all five high schools in the twelfth-grade sample and all graduates of the four existing high schools in the graduate sample. The class of 1963 was chosen as the graduate cohort. Since placement in the labor force was a focus of interest, it was necessary to choose a class which had been out of high school long enough for most of the

AMBITION AND ATTAINMENT 13

S ve nN oO \O WwW) tas = + + —Samm

SS

RB

ss) ®

2>

= —) WY) —“~oY2 E Via) =

— SS

.2| 33 >

> 2 121 | S| 2¥#38 $5

.a ; > IS =©

:N

1S) o ' = > ‘< on)on on)Se on

Sang ’ = (=)

= ~ eo Interviews were also conducted with the parents of samples of the boys, but the inclusion of those data in the present analysis would add greatly to the bulk and complexity of this report. That feature of the study will be reported elsewhere.

18 DESIGN OF THE STUDY

CHAPTER III

PRELIMINARY FINDINGS

Since the great majority of the boys studied were white, and since there were not enough blacks to warrant a comparison by race in the graduate cohort, it was not possible to conduct a wholly comparable analysis of the two races, although interest in such comparison partly instigated the research. Parallel analyses were carried out wherever possible, but the outcomes were strikingly different. The space limitations of the present report make it impossible to present a racial comparison, but a more detailed report will be published elsewhere.’ This report is thus limited to analysis of the data from the white subjects.” Table 3.1 reports descriptive data on the three in-school cohorts. With the exception of IQ, days absent, and behavior problems, all these data come from the boys themselves, although some refer to their parents. There are few large differences among the three cohorts. It may be significant that the twelfth-grade boys have fewer siblings than the younger boys; those from larger families may be more likely to drop out of school before reaching the senior year. The average IQ of twelfth graders is also somewhat higher than that of the two younger cohorts, presumably because boys with low IQs are more likely to drop out. Fewer older boys expect to go to college, and the sixth graders are the most likely to say so. Furthermore, the difference between “wanting” and “expecting” a college education is greater for the older boys. In spite of this, there are no differences by age in the level of expected first job, although the younger boys more often find low prestige jobs at least “satisfactory.” These findings probably reflect the younger boys’ less adequate knowledge of the world of work and of the relationship between educational and occupational attainment—a point which will be referred to later.

AMBITION AND ATTAINMENT 19

Table 3.1

Descriptive Summary, In-School Cohorts

Characteristic Grade Twelve Grade Nine Grade Six

Now 84 84 88 3 years ago 88 88 90 When in first grade 95 95 93

% Living with Both Parents:

Avg. Number of Siblings 2.94 3.39 — 3.59

Avg. Father’s Occupational Level* 47.26 47.95 46.81

% Fathers H. S. Graduates 80 83 86 % Fathers College Graduates 20 24 33 % Mothers H. S. Graduates 86 81 86 % Mothers College Graduates 12 13 19

Average IQ 109.8 107.0 104.6 Avg. Days Absent Last Year 6.5 8.8 7.4

Problems 7 8 7

% Who are Severe Behavior |

% Expecting H. S. Graduation 99 98 98 % Expecting Further Education 86 75 83 % Expecting to go to College 50 51 67 % Strongly Wanting H. S. Graduation 80 77 72 % Wanting Further Education 83 75 83 % Wanting to go to College 57 51 66 % Mothers Wanting College 63 62 . 77

% Fathers Wanting College 61 59 76 % In College Prep. Program 57 — —

Avg. Expected First-Job Level* 57.7 57.9 57.7 Avg. Wanted First-Job Level* 59.5 57.0 56.0

Avg. Satisfactory Job Level* 56.4 50.4 47.2

Level* 60.1 56.4 52.6 Level* 57.7 51.6 49.2

Avg. Mothers’ Satisfactory Job Avg. Fathers’ Satisfactory Job

* All occupation measures are in the form of Duncan Scores (See Duncan,

1961).

20 } | PRELIMINARY FINDINGS

While most of the data from the in-school boys discussed above dealt with , images of the future, most of the data from the graduates dealt with accomplishments rather than goals. A useful comparison, therefore, may be made between the goals of twelfth graders and the accomplishments of the men of the class of 1963. For this purpose, the data in Table 3.2 may be compared with those in the twelfth-

| grade column of Table 3.1.

Table 3.2 Descriptive Summary, Class of 1963

% Living with Mother in Twelfth Grade 95 % Living with Father in Twelfth Grade 87

Average Father’s Occupational Level* 48.2 % Fathers High-School Graduates 70 % Fathers College Graduates . 15 % Mothers High-School Graduates 72

Average IQ 108.1

% Mothers College Graduates 8

% Educated beyond High School 68

% College-educated % College Graduates 55 31 % Who Want Eventually Education beyond High School 84 % Who Either Have Gone or Want Eventually to go to College 64

Average Level of Present Job* 50.7 Average Level of First Full-Time Job* 46.4

Average Level of Satisfactory Jobs* at Age 30 62.3 Average Level of Satisfactory Jobs* in Mother’s View 61.3 Average Level of Satisfactory Jobs* in Father’s View 58.6

% Who Expected Further Education When Seniors 72

% Who Expected College When Seniors 56 % Who Wanted Further Education When Seniors 77 % Who Wanted to go to College When Seniors 59

1961).

* All occupation measures are in the form of Duncan Scores (See Duncan,

AMBITION AND ATTAINMENT 21

To the extent that comparisons can be made, the backgrounds of the seniors and the graduates are rather similar. There is some tendency for the graduates’ parents to have somewhat less schooling, while their fathers hold jobs slightly higher in

status. The average IQs of the two groups are almost the same; but when the graduates’ educational attainments are compared with the seniors’ expectations, more notable differences appear. Although 86% of the seniors expect to go further with their education, only 68% of the graduates have actually done so. As far as college education is concerned, so large a difference does not appear—50% of the seniors expect to go to college and 55% of the graduates have gone. (However, only 31% of graduates have graduated from college.) The main difference thus lies in the lower proportion of graduates who get other kinds of further education—in business, technical or vocational school or community or junior college. It is also noteworthy that a substantial number of graduates still hope eventually to go to college, the total proportion of graduates who have either gone or want to go to college being larger than the proportion of seniors wanting to go. The same general difference is found when occupational expectations of the seniors are compared with the occupational attainments of the graduates. The average level of the first job actually attained by the graduates is over ten points lower than that which the seniors say they expect to get (46.4 versus 57.7). Even if we consider the jobs the graduates held at the time of the survey, when they were about 24 years old, there is still a sizable gap (50.7 versus 57.7).°

In spite of their relatively low level of occupational attainment, however, the graduates’ level of occupational desire (specified as jobs with which they would be “satisfied” at age 30) is even higher than that of the seniors. One might speculate that their experience in the world of work had given them grounds of comparison

not available to the seniors, and they realize the desirability of jobs with higher status. Since they are approaching the age referred to in the question, however, and both their first and present jobs were considerably below the level they defined as satisfactory, their stated desires were probably not realistic.

The general picture is one of a group of graduates who come from families similar to those of the seniors, whose occupational accomplishments have fallen short of the goals defined by the latter, but whose hopes for the future remain undimmed. Whether the graduates actually had similar expectations and hopes - when they were seniors cannot, of course, be positively determined. However, they

, were asked about their educational expectations and wishes when in twelfth grade. The validity of such retrospective data may be questioned; nevertheless, the data suggest that their educational attainments to date have not been as great as either their expectations or wishes in their senior year. In spite of this, their wishes for the future are for even more education than they say they wanted when in high school.

The graduate data also provide a better basis for interpreting the differences among the three in-school cohorts. All three in-school cohorts have approximately the same average of expected first job, and that level is considerably higher than the average level actually attained by the graduates. To this extent, therefore, all three in-school cohorts seem to have unrealistic occupational expectations. With respect to educational expectation, however, the twelfth graders seem much more realistic than the younger boys, especially the sixth graders. There is nothing about the sixth

22 - PRELIMINARY FINDINGS

graders which could justify their high expectations, except that they are further removed in time from actual educational attainment and thus their aspirations are | not limited by hard realities. The fact that their expectations are not much different from their wishes also suggests that they have not yet had to consider the difficulties of goal attainment.* These findings are particularly impressive when it is

remembered that included in the younger cohorts are many boys who almost — certainly will not even complete high school. Thus the seniors have already reached a level of education beyond what some of the younger boys will ever attain. The Basic Attainment Model

The overall purpose of the analysis in this report is to illuminate the flow of influence which leads a young man from a point of origin to an adult destination in the stratification system. A first step in the explication of the pattern was offered by Blau and Duncan (1967) and later added to by Duncan (1968): father’s occupation (FaOcc), father’s education (FaEd), son’s IQ and number of siblings (Sib) were co-equal independent variables, son’s educational attainment (EdAtt) was an intervening variable, and son’s first job (OccAtt) was the dependent variable.* For a comparison with the national data used by Duncan the most comparable data from the present study are those for the graduates of the class of 1963. They were approximately 24 years old; Duncan’s men ranged from 25 to 34 years old.

All of the variables used by Duncan were available in the case of the graduates except number of siblings. In Table 3.3 are presented the intercorrelations of the variables analyzed by Duncan, and Table 3.4 presents the similar correlation matrix of the graduate data. Figure 3.1 presents the path model constructed with the five

available variables and showing the coefficients derived from both sets of data, those based on Duncan’s data being in parentheses. In Figure 3.1, the variables to the left in the diagram, referred to as exogenous,

are seen as “given;” that is, the analysis does not attempt to explain their values. The curved, two-headed arrows linking them simply indicate that they are interrelated, and the coefficient associated with each curved arrow is the zero-order correlation between the two variables the arrow links together. The straight, singleheaded arrows indicate an ordered relationship, the variable at the head of each arrow being caused by or dependent on the variable at the other end. The depend- | ent variables are viewed as ordered (one follows and is thus dependent on the other) and as being influenced by multiple variables (all those to the left of the dependent variable in the diagram). The coefficient associated with each arrow, called a path coefficient, is simply a standardized regression coefficient. Since the coefficients are standardized, their size corresponds to the relative importance of the several inde-

pendent sources of influence. Finally, each arrow which originates outside the system represents the influence on the dependent variables of other unmeasured variables and the coefficient associated with the arrow is the implied correlation between the dependent variable and all such unmeasured sources of influence. The

coefficients reported in the diagram thus account for all of the variance in the dependent variables; however, only those linking variables in the diagram measure known sources of influence.

AMBITION AND ATTAINMENT 23

— Oo + ov) On S RS — wee? bd ee’ CO = | OJ@| ¢ s+

o = § > —— be ° cO oO als [ta 2 S

© ‘= N a ow | SS rs) in oo — Li — © @ ome ©~~: a NJ |an8 @ mM Li. CO are= on 2 aD 3 Noe 7” (an)

Oo.

&

® =|

2

ee? ij = ©

'Z,

24 PRELIMINARY FINDINGS

Table 3.3 Correlation Matrix, Basic Model of Duncan*

FaEd FaOcc EdAtt OccAtt

IQ 28 29 59 45 FaEd 49 41 34 FaOcc 43 39

EdAtt 64 * Adapted from Duncan (1968) .

Table 3.4 Correlation Matrix, Basic Model, Fort Wayne Graduate Cohort

FaEd FaOcc EdAtt OccAtt Mean St.Dev.

IQ 303 266 506 404 108.5 11.93 FaEd .620 359 314 3.96 2.33

FaOcc 405 408 47.97 23.52 EdAtt 728 3.29 1.86

OccAtt 46.39 24.87

When Tables 3.3 and 3.4 are compared, three of the ten comparable coefficients

are found to differ by more than .05. One of these is the IQ-EdAtt correlation, Duncan’s coefficient being the larger. It is difficult to assess this difference with confidence. Duncan used a “correction” for his originally obtained coefficients

which increased their size. Furthermore, the IQ measure available in the Fort , Wayne records followed an alphabetical categorical form (A, B, etc.) rather than numerical form. These categories can be translated into numerical form, but they provide only six scale points, rather than the usual more refined measure.° These two disparities make direct comparisons suspect, though it is worth noting that the original (uncorrected) IQ-EdAtt coefficient which Duncan discusses was within .05 of mine. The other two differences both involve correlations of educational and occupational levels, the FaEd-FaOcc and the EdAtt-OccAtt coefficients. In both cases, the Fort Wayne coefficients are higher than Duncan’s. It is impossible to demonstrate it

with the data available, yet it may well be that these differences reflect temporal and chronological differences in the two sets of data. Duncan’s FaEd-FaOcc coefficient is based on measures as of the time his subjects were sixteen years old, that is,

| AMBITION AND ATTAINMENT 25

between 1943 and 1952 (for the cohort that was from 25 to 34 years old in 1962), while the Fort Wayne correlation is based on measures as of 1963. In the 1962 data of Blau and Duncan (1967, p. 178) relevant to men of approximately the age of the Fort Wayne fathers (i.e., 45 to 54 years old), the correlation between education and occupation is .593 (compared with .620 in Fort Wayne). In fact, the correlation between these two variables in the case of sons in the Blau and Duncan survey (all based on 1962 measures) are consistently higher than in the case of their fathers (all based on earlier measures), irrespective of the age of the sons. The coefficients for

fathers range from .481 to .535 and for sons from .576 to .657, of which the FaEd-FaOcc and EdAtt-OccAtt coefficients in Table 3.3 are one example. Thus, whatever the reason, correlations based on data from the early 1960s are higher than those based on data from an earlier time. The Blau and Duncan data also show that the correlation between education and occupation decreases with age. Although none of the coefficients they report is as high as the Fort Wayne coefficient (.728), it is also true that none is based on data from such young men, their youngest cohort ranging from 25 to 34 years in age. If it is generally true that the link between education and occupation weakens with age, one would expect the Fort Wayne coefficient to be higher than Duncan’s. ’ While the similarity between the path diagrams based on the two sets of data is very striking, there are some differences (e.g., in the EdAtt-FaEd paths), but only | the OccAtt-EdAtt path coefficients are appreciably different. Clearly, this is due to the higher correlation between these two variables in Fort Wayne. In both data sets,

however, the most striking thing about the models is that, although there are sizeable correlations between the exogenous variables and OccAtt, the direct paths from these variables to OccAtt are very small. Thus the independent effects of IQ and FaEd on OccAtt are almost wholly accounted for by their effect on EdAtt, and there is only a weak direct OccAtt-FaOcc path.®

The mediating role of educational attainment in transmitting the effects of background on occupational placement is apparent in both sets of data. It is thus clearly important to differentiate between a model based on the simple multiple correlation between OccAtt and all of the preceding variables (including EdAtt) and the flow of influence model presented in Figure 3.1. Although the two are identical in explaining the variance of OccAtt, Figure 3.1 tells much more about the process of attainment of occupational position, for it indicates that, once educational attainment is taken into account, intelligence and social origin make little difference, although father’s occupation (perhaps through income) seems to have some continuing significance.”

The kind of analysis just presented has made it possible to compare these more limited Fort Wayne data with the national data reported by Duncan. However, it may be argued that “‘first job” is not a very meaningful reference point for an age cohort. After all, the first job of the graduates was obtained at ages ranging from 18 to 24 (and even later by those still in school). Also, though the men were specifically asked to describe their first job after completing their education, some may have reported jobs held before finishing school. Blau and Duncan had evidence of

this error in reporting, and it undoubtedly occurred here also. In any event, an

% PRELIMINARY FINDINGS

analysis that describes the cohort at a single point in time (i.e., at age 24) has the merit of being more easily understood.

If the job held at age 24 is used as the point of reference, the data presented in Table 3.4 and Figure 3.1 remain the same except for the following: (1) The correlations in the OccAtt column of Table 3.4 become (top to bottom) .414, .338, .394, and .715. (2) The paths to OccAtt (from EdAtt through FaOcc) become .63, .06, .03, and .11. Probably the only noteworthy difference between this result and that for first job is the somewhat stronger OccAtt-FaOcc path using current job. Even

here, though, the difference is not large. In all of the further analysis of the graduate data the current job (at age 24) will be used, to provide a consistent point of reference. It is clear from the above, however, that the outcome is not appreciably different from an analysis using the first-job data. The Basic Ambition Model

Since most of the subjects of this research were still in school at the time the data were collected, it was not possible in their case to deal with educational and occupational attainment. As for the graduates, the future is defined in terms of education and occupation, but here the interest is in what they expected rather than what they actually attained.'° The three panels of Table 3.5 report the intercorrelations among the six variables (the four exogenous variables, including Sib, and the two expectation variables) among the boys in grades 6, 9, and 12. Several variations in the patterns of correla-

tion are worthy of comment: first, the correlation between FaOcc and IQ is stronger in grades 6 and 9 than in grade 12 (.44 and .44 versus .25). It seems likely

that this is at least in part a function of dropping out by the time the group had reached the twelfth grade, resulting in a higher mean IQ in grade 12 and a smaller variance in the IQ scores. Second, the correlation between IQ and Sib is lower among the twelfth graders than among either of the younger cohorts (—.10, —.27, and —.33 for twelfth, ninth

and sixth grades, respectively). The twelfth-grade correlation is also much lower than that found by Duncan (—.25) in his national sample of young men aged 25 to 34. Here again the drop-out pattern may be significant; those with lower IQs and those from larger families are evidently more likely to leave school before the twelfth grade. Third, large differences between cohorts are found in the correlations between

the two expectation measures: they are much more highly correlated in the two older cohorts (.35, .61, and .67). Probably this indicates, among other things, that the younger boys do not have adequate knowledge of the link between educational and occupational attainment. At the same time, it is important that the average level of their expectations is as high as that of the older boys as to occupation and higher as to education (Table 3.1). Clearly the sixth graders have very high educational expectations—a subject to which I shall return.

It is highly problematic how one should use the two expectation variables in a path model. If one orders them, it may imply that the boy decides on one goal (educational or occupational) before the other and that the first decision affects the second. There are those who argue that the boy decides on an occupation or kind

AMBITION AND ATTAINMENT 27

Table 3.5

Correlation Matrices, Basic Ambition Model, In-School Cohorts

Sib FaEd FaOcc EdExp OccExp Mean _ St.Dev. 12th Grade (N=778)

IQ 3312.95 110.2 11.52 . Sib-.104 -.158.274 116 .250 -221 487 8 -.161 1.98

FaEd 612 449 301 4.34 2.17 FaOcc 412 331 47.05 23.23 EdExp 671 3.16 1.34

OccExp 58.58 26.12 9th Grade (N=354)

IQ -.275 307 439 493 381 108.45 12.35

Sib -.103 -196 -.175 = -.156 3.22 1.97

FaEd 637 449 404 45424.19 2.22 FaOcc 444 374 48.53 EdExp 612 3.11 1.42

OccExp 59.06 28.01 6th Grade (N=280)

IQ -.327 .278 441 342 309 106.27 13.81 | Sib -218 = -.225 221 -.219 3.45 2.32

FaEd 522 340 1540 3.03 2.24 . FaOcc 333 339 47.39 24.82 EdExp 345 3.64 1.23

OccExp 59.52 26.01

of occupation first, and then seeks the education he needs to attain it. It may be equally well argued that a boy orients himself first to continuing or not continuing his education and then chooses an occupation to which the desired or attained education is suitable. Some of the research in this area has avoided the issue completely through one or other of two procedures: One is to combine the two kinds of expectation into a summary measure of “‘ambition”’; and the other is to build the

two measures into a model at the same point and to permit them to be freely

28 PRELIMINARY FINDINGS

correlated (Sewell, Haller and Ohlendorf, 1970). If one views education as a means of attaining various levels of occupation (or at least access to them), as I do, it is

undesirable to combine the two measures into a single index of ambition. It is especially undesirable in the present analysis in light of the very different levels of association between the two expectation measures in the three cohorts. More would be obscured than illuminated by comparisons across cohorts. For the present analysis, therefore, the two are considered separately. Also, they are built into the models in the same order as the attainment variables (education before occupation) in recognition of the prime significance of education as a means of attainment. Without being committed to the argument that a boy actually sets himself educational goals before occupational goals, this structure at least reflects the order in which he needs to cope with the specifics of attaining them: educational goals are more proximate than occupational goals. It is also true that most boys have more experience on which to base their educational goals. It is thus not surprising that in their educational goals the younger boys seem less realistic than the older. It is even more significant, however, that the occupational goals of all these boys are unrealistic, even of those who are very close to the point of attainment. The twelfth graders’ expectations exceed the graduates’ attainments by as much as do those of the sixth graders. The ordering here, therefore, follows this concern for realism, as well as the central significance of education in the process of achievement.

Figure 3.2 presents the path model for the twelfth-grade cohort,'! and standardized and (in parentheses) metric regression coefficients are presented for all three cohorts in Table 3.6; both are needed for the kind of analysis proposed here. The standardized coefficients, which are usually presented in path models, provide a basis for determining the relative importance of the several independent variables in

explaining a dependent variable in a particular model. They tell that for a unit | change in a given independent variable the dependent variable changes so much of a unit. Since such coefficients are standardized according to the standard deviation of each variable within a particular model, however, comparisons across models are of doubtful value if the samples being compared contain standard deviations of different sizes.

The metric coefficients, on the other hand, make it possible to compare the contribution of a particular independent variable across models, but they make it more difficult to compare the relative contributions of different independent variables within a model since each variable has its unique metric. The basic analysis is concerned with the relative contributions of independent variables in explaining a dependent variable. Thus the standardized coefficients are of primary interest. However, since such an analysis leads to statements about differences between cohorts, it is necessary to examine the metric coefficients also, to guard against basing comparative statements on differences in coefficients which simply reflect differences in the distributions and their standard deviations.!? What follows will be organized around the standardized (path) coefficients, and the metric coefficients will be mentioned only if they show a different pattern of results. Figure 3.2 may be used as a reference point in considering the process of influence in the three in-school cohorts. The path and metric coefficients of the other

_ AMBITION’ AND ATTAINMENT 29

Oo a ud ad RY = o/ + S :r: in S/2] %, w S I py fs oe) cO abd = & . O _ 8 rf 3Z Kk] y me @ = © CO ° 2 a E N = \ c no 6S o o—=@&ew < San * = Gump i ae D *« q—— ~ >< ro) © ——— LJ daw oO

Qa Mm

,

23 : = © 3 c & = < 2 ad e[aaoO oa wo oO

Qu»

nN

> "oO

=> Aei Sei en —* re Sa Bie + CO aw |

‘® q©"o) nw

st hal el —_— N7 © To!

>.

30 AMBITION AND ATTAINMENT

an 3

Oo 8 & : o x -Z~ a ca = aa \o.Ne © aoO NO < fs a | & mn Pome Cet

C—E Ss 00 \o+or© NN aow BD ~A wron) ON ©fom wa (ae) TS —~ Ya) N

A Pad = O © aon oo) Dy L & sm25 [= os 2 S S iS - Oo >2 S —5tom qm @

© ©

s

-© ‘-—¢ ” @ vs 52

= OS ~o tOonwwktnr & \O Cre) ™ tiled

&Zz, =MANA @ganaounan N~On \O o oe |Ss&© S&Osn | OSe0ae] —~ = STS] No} = a 0 0 a ee oe KS st Hm AO wH .o N Ye) a}s 5 z 0onrvrere eanan 0° oe WV ST} os &

3 rs so & _ Qu Pa Oo s o Z 2 S & 5 Sa mao po) S > —= — p> = © Oo mw OO} \S) co 4 OFB>0as ~ a,OS 0a % NO FFBes BOs Ss;OYa32 = 809 A imao 6 ~TO] 2oOo~co ° MO >S *E& €6 "3S o4 %714

Bz sa+es0OA SY aw 8Son ay 2| a Oo Fa Oo Ff he =rr oo nmS| Oo oO oS wy S$ 6 g AO Hino] oO Ss | ~ FEU Ss ~ BH OO FH{ eR oO F&O m~ YI Ss & 4 ooes5 OS a G6nm HSB 2S @ BSE SS=SV

cy oO ow Sr Fer SE GZ wvaiwae . ew, = eg o V nm ON ~~ Ve +_ amend od 060 F&F we F Ow»&HRB HAD w~AE v2» SITY Fe ©2396 PRE ZL 22 ZS ez e pars ee «> ee ee

5 :

MN) oe) -— oO NAY Vf | © 5

oe — s/ ;

YK! S / © cD) “> 4. | © — — & ~ ° —f © © ~— — fon) — re : CO os} = _ Of oO / NI 2 o>] a a @ NY « —e © > CO © aeneatlll * 65

=f 0 [*= 3 g 6 s0 + ae So > 28 oS— 2) a io) “, 2s i 38 3 ‘s S ‘O 2

i&) = = av) CO 0) +3=)—_—P %bene oooO av .bsS

ro) Sus . oF oat "D oO ara

] 3 we© S 25

. e: 52B 'O S9

=

Ss os

.=

o = 8 To) | NI cawi “ vw A 4 4 = 5 s& fae}

© Zz.

44 SCHOOL EXPERIENCE

make possible comparisons within and across cohorts. The relevant path and metric

coefficients of the in-school models are reported in Table 4.4.

Table 4.2 Correlation Matrix, Grade and Grade-Partic Models, Graduate Cohort

N=381 FaEd IQ Grade Partic EdAtt OccAtt Mean St.Dev. FaOcc .616 260 ~—.250 152 ~=.401 400 48.14 23.45

FaEd 297 261 184 348 316 3.97 2.32

IQ 526 .286 194 503 Grade 581418 461108.45 81.5811.89 5.99

Partic 335 2.671.86 2.37 EdAtt444 140 3.30

OccAtt 51.80 22.51

Figure 4.1 documents deviations from the basic model that does not include Grade. First, as expected, the inclusion of Grade increases appreciably the explanation of the variance in EdAtt (from the 33.9% reported in Chapter III to 44.6%). Second, its inclusion adds much less to the explanation of the variance in OccAtt

(from 52.7% to 56.2%). Third, the direct effects of the exogenous variables on EdAtt are decreased by including Grade in the model, but this is primarily true of the effect of IQ: previously the EdAtt-IQ path coefficient was .42, but it is now only .23. Finally, there is no notable alteration in the paths to OccAtt, and the OccAtt-Grade path is negligible.

Thus the inclusion of Grade in the graduate model has done two things: It has introduced an additional source of explanation of EdAtt not found in the basic © model, thereby increasing the explanation of EdAtt beyond that provided by the exogenous variables. At the same time, Grade does help explain the influence of the

exogenous variables on EdAtt. It is not just that boys with high IQs get more education; they earn good grades which, in turn, influence their educational attainment. Yet it is apparent that not just academic performance is involved; there are

still strong residual paths to EdAtt from IQ and FaOcc, even with Grade in the model. Smart boys tend to remain in school, even when they do not -get good grades; and boys whose fathers enjoy high prestige (and are rich?) tend to continue, even if they are not smart and their grades are not good. The model data of the twelfth graders (reported in Table 4.4) have some of the same features. First, the amount of variance explained in EdExp is increased from

37.8% (in Table 3.6) to 48.4%. Second, the variance explained in OccExp is not increased appreciably (from 45.6% to 46.6%). Third, the direct effects of the exogenous variables on EdExp are reduced, but the only sizeable change is in the EdExpIQ path (from .37 to .16). Finally, none of the paths from the exogenous variables

AMBITION AND ATTAINMENT , 45

$t

, >3. ot toondGaatag AseANNBDAK )Awn ren . e os° dodaaicn maadgenae Dn ats Hy N SONbADaAwH ~NN aarnge oO

Otnan 5 SARAH eVe) DATARS MOMS ~— Oo OM = —~ w+be0 ON o +MmNNo 0 Va)SE ©+ oOeos omMAHA wotAAN se on ma HE (x) ANON =~ MODAN S ~ HAMM 3 AIARERS Ss AAARASH

fas ' i ed RK Os | mn + -SSaynO =eroNn +S Vay -_ yf * mo mm OoTU 0 =~ AYO oO MN 3a a(3AEATERS TIEAS aaaaas a & ecgygc atsNM tA TM— 0 048a ON FTMm~ OM 37 fos2Ww Aaa Hae Hy

2 ~ 3 No +~OnF 0 T&S3° = .CSages oornmAN Mom 225/|é CRA atMY at 28 s6O yf %* ANAS x @~)

>

sg 3 on — A 0 ta -Mm i] r) (x) co cS (i, | . |4ANN

5 os

om

— SS

Fe

2olS= — op . ~“~~ on

Sf oO r~ o &S a/ 38s a | 38 Cf seg e8s| 5 sg 288/53 88285 Soi HORE S i SSzEAe Ci Begs Ka AZIR Ses 28 Sr,oeBSEeSS | SzoeRSsass 3

on

= SHO|RAECLRZEHLOA RO] SELB LSAERO

ase © a a=© “2 s S c= o 3 = 35 °®£oO ON + \Owv N +N Noe) i) cy ON © at st oO \O 0) =) N ~ al“ > se) ~T ~S a) a OA

Get it (2)

c=

ws

5om es x A es! iSs mm NN \O : moNee % AQ] & = ey —_ 2) é =

Nn

4 o 2 foot

aver ’ ‘ CO 00 ' t Oo a ‘ NN &

=) on ~~ =_~ o—~ “— oa & Ee c= mC} * ws * © LY a) iN oN S ef + $8 $8] . SE 88]. gE BE] = 8 58 & 5—_ —_aS —_68—SS vwS85] Now 5 n $

§ 82 .a6 z

o> Y a ad

n |2 8= Q om on#* am ~~t \o om on~ =~ — 3) Oo *on~ ODO + ons x ©$5) N Oo *£ Bele BO = ahs Ame mA ON = o4 aN Ae ANS 8S wv en Ww 48S AN cS o vo pas

\—e/ — mos — — — \nwe — — 6 2 .=3

acl EH os | sASs& © Ae | r=Na Zz

—&

2 &f a Ff o gg |38 ~ oy * faa © wy x aa oO wy * 1 S C3 8 jaa ) CO se ea) S) CT 3s eal g 2 & Ss oS oO Sm io) Oo S ae) oO =

g&

= 2 one sal eo) one ea) ‘e) cone i (oe) i

Oo

to OccExp is changed greatly, but there is a significant OccExp-Grade path, and the OccExp-EdExp path is reduced slightly (from .65 to .59). Thus, again, Grade provides both an additional source of explanation of variance in a dependent variable (EdExp) as well as a means of interpretation of the influence of the exogenous variables on the dependent variables. However, in contrast to EdAtt, all exogenous variables continue to have a significant effect on EdExp, and Grade has a significant independent effect on OccExp, but not on OccAtt. The path coefficients of the two younger cohorts (Table 4.4), clearly differ from

each other and from the twelfth-grade coefficients. In the ninth grade, all four exogenous variables have significant effects on Grade, IQ being strongest. In the sixth grade only IQ has a significant effect. In the ninth grade, Grade has a strong effect on EdExp and there are significant direct influences on EdExp from IQ and FaEd. In the sixth grade, only FaEd is significant although IQ and FaOcc approach significance. Finally, the OccExp-EdExp path is strong and there is a significant direct OccExp-FaEd path in the ninth grade. In the sixth grade the OccExp-EdExp

_ path is much weaker (though significant); it is about equal to the direct OccExpFaOcc path.

These models thus reflect the patterns discussed earlier in this chapter and in Chapter III. There is a general weakness in the sixth-grade model, no path except Grade-IQ being very strong, nothing going very far toward explaining EdExp, and OccExp evidently reflecting FaOcc as much as anything else. It is particularly striking that Grade does not add to the explanation of EdExp. Like the weak link between EdExp and OccExp, this seems to be an additional example of the limited understanding of means-ends relations in the sixth grade. It is equally impressive, however, that among these younger boys Grade clearly reflects IQ but not much else, while in the ninth grade family background contributes a great deal to the explanation of Grade. Although the EdExp-Grade and OccExp-EdExp paths are much stronger in the ninth than in the sixth grade, they are still weaker than in the twelfth grade. There is thus evidence, first, that previous performance and the boys’ image of their own ability is increasingly a basis for predicting their future, and second, that they achieve a growing understanding of the association in the “real world” between education and occupation. Yet it also seems to be true that as they

, grow older they see the significance for their futures of their own backgrounds. Especially with respect to EdExp, the older boys tend to respond to the effects of their family situation as well as to their own ability and performance. If Partic is added to the analysis as a second intervening variable, the resulting path diagram is as shown in Figure 4.2, where the path coefficients of the graduate cohort are reported. The relevant correlation coefficients (Tables 4.2 and 4.3) and the path coefficients of the in-school cohorts (Table 4.5) are identical to those just

reported for the Grade models as far as the correlations between the exogenous variables and the paths to Grade are concerned. Thus the discussion will be focussed on the other parts of the models. The data on the graduate cohort make it is clear that the exogenous variables are

much more effective in explaining Grade than Partic (29.3% versus 5.6% of the variance, respectively). The major difference is the size of the path from IQ, which is the only significant source of explanation of either Grade or Partic. Despite this

48 SCHOOL EXPERIENCE

A

& - ©_ an ow

\2 > 2 | gd? O\ cO ° ge . + Dpam -_ \ =| © CO WwW ae x us / Sy Q « =x °° a bo) Px g

5 g re a2 § 28 25 »=:2 o\|a\3S|S|2% = 3x 2rroe y — N oO Tt oO g En | “ oY WY > = aN ~] & = SP % sh 0) WAS ef S == ang “5aLy Tf+=mt aa “ry e /~ve)

oRD ida) : _— re< ©i -co3sa SOE oO To WwW ~ ead x = Os

as —

= RN

2s 3



33 Son

© Zz

AMBITION AND ATTAINMENT 49

weak link between the exogenous variables and Partic, the latter is almost as effective as Grade in explaining the variation in EdAtt, and the addition of Partic to the

analysis increases the explanation of EdAtt by more than six percent over the Grade model, 51.3% of the variance of EdAtt now being explained. With the exception of FaEd, all of the explanatory variables have significant direct effects on

EdAtt. The effects of the other variables on EdAtt are the same as those in the Grade model, except for the reduction of the EdAtt-Grade path (from .40 to .33). None of the paths to OccAtt is altered appreciably by the inclusion of Partic, and the OccAtt-Partic path is of negligible size. Thus the inclusion of Partic in the graduate model has the effect of increasing the overall explanation of EdAtt and reducing somewhat the direct effect of Grade on EdAtt. The reduction of the EdAtt-Grade path is a function of the common dependence of both Grade and Partic on IQ. However, the effect of Partic on EdAtt is much greater than its dependence on the exogenous variables suggests. It is thus, in effect, an additional exogenous variable, contributing to the explanation of EdAtt but not greatly dependent on the exogenous variables.

The Grade-Partic analysis of the twelfth-grade cohort (Table 4.5) has some of the same characteristics. Partic is much less fully explained by the exogenous variables than is Grade (9.5% versus 34.9%), the major difference being the effect of IQ. Here, however, a significant path runs to Partic from FaEd and Sib as well as

from IQ, whereas only IQ had a significant effect among the graduates. Undoubtedly this is a function of the more varied kinds of information included in Partic about the twelfth graders (number of hours spent on a job, attendance, and behavior problems as well as athletic and organizational activities). But again, as with its effect on EdAtt among the graduates, Partic has a strong independent effect on EdExp, its only other influence being a reduction in the EdExp-Grade path (from .41 to .32). It does not increase the coefficient of determination of EdExp as much as it did in the case of EdAtt, although there is some increase. Finally, as in the analysis of the graduates’ OccAtt, the inclusion of Partic alters

none of the paths to OccExp, and the OccExp-Partic path is very weak. | The effect of Partic on the ninth-grade model is somewhat different from what it

is in either of the older cohorts. Over twenty percent of the variance of Partic is explained by the exogenous variables, and all of the exogenous variables except Sib

contribute significantly to that explanation, the effects of IQ and FaOcc being equally strong. The addition of Partic to the model does not increase the explanation of EdExp, but it does reduce the EdExp-Grade path (from .31 to .23) in the same way as in the twelfth grade. Also as in the twelfth grade, the inclusion of Partic does not alter the paths of the other variables to OccExp, and Partic itself has no effect on OccExp. Thus, although the overall effects of including Partic are the same in all three older cohorts, so far as the pattern of explanation of EdExp and OccExp (or EdAtt and OccAtt) is concerned, Partic is most clearly an intervening variable in the ninth grade. It both helps explain the dependent variables and is itself explained by the exogenous variables.

The sixth-grade pattern is much the same as in previous models. Very little of the variance of Partic is explained by the exogenous variables (though there is a significant path from FaOcc), its inclusion does not increase the variance explained

50 SCHOOL EXPERIENCE

@ o! se}}=

. 4 5 , ; o Reta . Py ~ ey 3 a (@) wom

~a og &= =

ta 7tsabasse] a om

& a | e .S3a)S=5S cB) a @ | 3 9 = 3 oO joe AeOQ = a=a= © ae) Oo oS Oro 7 ror 2 © yam —_

ov oO

= oe) Ya) Nn oo=o §€ nr a = Noy ren) ro) =) ) N No) 00 ~ 5 Oo > ‘

i 9 a °+ve~ +Nn ~ “al+ys) + reQ° L c eR = e (1) ' ‘ ‘ | ote) ’ ‘ ’ oh) t ' ' Qs ] oS ‘Y— &. oo 7) 2 £2 ofS EF we we oe | = Ss , , CO ,Rab 00 en 8S=a8 >cB~ 8s = eo e| 8/5 5 28]$8oO £8sf}gk ak eS)st] BA za 22 S838 8] 8A sg =5 gg Ss] 8&8 oS]28&S BS) OGgy oeOY Oo fe] S232 =§ 28 “8&2 8%) 28SRSSHESS| a8oF88=F£8HF23)

Ctend

' w ' we , we me a) ‘os —— a) a moe ra, Nao, 0 v 0 os cD) oo co mn a °aa a” i a CO on rasa «oS = aD =) i c ~q3, 0S O65 “1S ow Bo 00 ON An ono i 00 ae on en z = co as ~~ Oo ~ A fon on oto) aN mo On r=) aban ag g =f “& 48 48 $8) YS FS aS S8| SR SS AS SN] §& cP om «© =m

2@ Se o. a = oO or Q, Pad Q, eed & O ©. r=] 3u 3 | 3 3% 2 ea Q 2 2 a4 ea y oO a7) 3 2 OD os ro © bo Ay eal © ss 5 ne 3) a es "O “Vw _—7 ree — ~— —, Nee” nd New’ — — + end 2 v

a= SsSaml =osneSO Sy ass ™So Lh — VO 2? Oo

) cD) EF

es | oo A. eal © GEO) a eal O oS oO |. EY BS N fom Sang fam MH Og ole) cv °oS ‘= Sim a0— ~~

s5 6s

) _ of either EdExp or OccExp, nor does it alter any of the other paths to those variables. The strongest path to EdExp continues to be FaEd and the strongest path to OccExp is still FaOcc. Although the measure of Partic used for the sixth grade is

different from that used for the older cohorts (it is based on attendance and behavior problems only), this outcome is clearly consistent with the pattern previ-

ously noted: sixth graders do not seem to use their own experience to form an image of their future but turn rather to their knowledge of their fathers’ status. By the ninth grade this has changed considerably, and the two older in-school cohorts apparently respond meaningfully to their own school experience, both academic and nonacademic. Consideration of the school experience of the boys and young men from Fort Wayne has thus altered the models of expectation and attainment very sharply in all cohorts except the sixth grade. Although it has had no appreciable effect on the

pattern of explanation of OccExp or OccAtt, its effect on the explanation ‘of educational expectation and attainment has been very striking. The inclusion of school grades as a basis of explanation has sharply increased the proportion of explained variance of both EdExp and EdAtt, especially in the two oldest cohorts. The further inclusion of Partic as a second measure of school experience again increases the explanation of EdExp and EdAtt, although not as much. Similar, but

smaller, increases are found in the ninth grade. But the school experience variables do more than add to the variance explained: they also help explicate the flow of influence of the exogenous variables on educa-

tional expectation and attainment. This is especially true of Grade, whose main effect is to lower the direct path from IQ to EdExp or EdAtt very substantially in all three of the older cohorts. The paths to EdExp and EdAtt from Grade are reduced considerably when Partic is added and the explanation of the dependent . variable is increased in all three older cohorts. This suggests, first, that Grade serves as a partial proxy for school experience more broadly conceived and second, that the broadened conception of school experience increases understanding of the process by which educational goals are set. In the ninth grade the exogenous variables most fully explain Partic as well as much of the variation in Grade. However, the strongest independent contribution of Partic to an explanation of the variance in the dependent educational variable is found in the graduate cohort. The ninth-grade data imply that the establishment of educational goals depends heavily at this point on the boys’ past nonacademic and academic experience. The graduate data show that, whatever goals they may have had, boys who participated actively in the nonacademic activities of the high school are more likely to pursue further education than those who did not. We may conclude, therefore, that school experience, both academic and nonacademic, is conditioned by the intellectual characteristics and social background of schoolboys, and that school experience in turn influences their educational goals and attainments. School experience serves both as a source of explanation of differences in goal and attainment and as a link between those two, on the one hand, and the boys’ intellectual and social characteristics, on the other.

, 52 SCHOOL EXPERIENCE

FOOTNOTES

The fact that there is a progressive drop with age in all performance groups in the proportions who say they enjoy their classes points to a general process of growing disaffection and felt restriction, as discussed by Stinchcomb (1964). Because the Involv measure was not available for the graduates, the discussion that follows deals only with the in-school whites. An analysis of graduate Partic scores will be presented later.

AMBITION AND ATTAINMENT 53

CHAPTER V , BELIEFS ABOUT OPPORTUNITY

Throughout the analysis thus far, the boy’s IQ has been used as an index of his ability to perform in ways that are rewarded in school and which provide access to higher occupational status. Ability is not the only personal quality which can make a difference in this process, however: the degree to which the boy strives toward future goals should also reflect his view of how long-range goals are to be achieved.

If he believes that opportunities for goal-achievement exist and that success depends only on skill and effort, he is likely to set goals and work toward them; but he is much less likely to do so if he sees the future as already fixed and uncontrollable, or as due wholly to chance. If his view of the world is the first, he will believe that the opportunity structure is “open.” At the same time, however sanguine his idea of the world, his reaction will also depend on his conception of himself and his

ability to take advantage of opportunity. The prerequisite of striving toward a difficult, distant goal is the belief that the goal is attainable and that one is capable of attaining it.

Many items in the student questionnaire dealt with these phenomena. They generally indexed the dimensions of self-esteem, a sense of potency, the usefulness of planning, and a fatalistic view of the world. In most cases, the items were taken from previous investigations. Scales were formed by grouping together sets of items presumed to be measuring the same quality and checking their inter-item consist-

ency by means of factor analysis. In the process, a few items were deleted, and some sets of items were broken up into sub-sets, the most important being a set of fifteen achievement-orientation items discussed by Kahl (1965). In the measures used, the factor loadings of the items were sufficiently consistent for item weights

54 BELIEFS ABOUT OPPORTUNITY

, to be judged unwarranted. In all cases, therefore, simple summation scores were used. The four most relevant measures were the following:

Control of Environment (ConEnv). This is the three-item measure used by Coleman, Campbell, et al. (1966); it is concerned with the boy’s belief in his ability to control his own destiny. In abbreviated form, the items are: ‘““Good luck is more important than hard work”; “Every time I try to get ahead, something stops me”; ‘People like me don’t have much of a chance to be successful.” On this measure, a high score indicates a sense of control. Fatalism (Fate). This is a six-item scale derived from the set of achievement-

Orientation items, which states that it does not pay to strive because one cannot , control the future. It includes such items as: “The wise person lives for today and lets tomorrow take care of itself’; ““When a man is born, the success he is going to have is in the cards, so he might as well accept it’’; ““The secret of happiness is not to expect too much out of life.’ Here a high score indicates a fatalistic view. Planning (Plan). This scale was derived from the achievement-orientation items. Its four items all deal with the advisability of making plans: “Planning only makes a person unhappy, since your plans hardly ever work out, anyway”; “It is important to make plans for one’s life and not just accept what comes.” A high score indicates a favorable view of planning.

] Self-Esteem (SelfEst). The nine items in this scale come directly from Rosenberg (1965). He used a more complex combination of the items in his research, whereas here the method was simple summation, a high score indicating high self-esteem. On the basis of their manifest content and the conceptual development which lies behind their construction, these measures might be expected to be interrelated. Their intercorrelations confirm the expectation, although none of the correlation coefficients goes much beyond .40. Perhaps more surprising than the general pat-

tern of intercorrelation is the fact that the strength of the relationships is practically the same in all three cohorts. I had expected the sixth graders to respond somewhat less meaningfully to these questions and their interscale correlations to be consequently weaker; but this was not the case.

_ It is thus possible to present a summary of the interrelations among the measures. Figure 5.1 reports coefficients which are approximate averages of those found _ for the three cohorts. Clearly boys who believe they have some control over their

favorable self-image.

environment like to plan ahead, reject a fatalistic view of life, and entertain a

The more relevant question for our present purposes, however, is to what extent such characteristics are associated with the key variables. Are the beliefs of boys with various backgrounds or various levels of ability dissimilar? Do boys with unlike beliefs set themselves different goals? To answer these questions the correlations

were examined between the four scales and the variables included in the basic model.

The most striking finding was that none of the exogenous variables except IQ was consistently associated with any of the measures; in fact, only in the case of Fate was even IQ correlated at the .25 level or better in all three cohorts. It is

noteworthy, however, that in most cases the link between IQ and belief was

AMBITION AND ATTAINMENT 55

Figure 5.1

Pattern of Relationships Among Beliefs about Self

and Opportunity , In-School Cohorts

0 ws ConEnv SelfEst K a0 aK . . Plan Fate

~

+.35

: stronger among the younger boys. All four scales were more clearly associated with the expectations measures than with the exogenous variables. Here, however, the associations were stronger in the older cohorts than in the sixth grade. This and the fact that the exogenous variables were more strongly associated in the younger cohorts means that the measures of belief about self and opportunity serve as links between the exogenous and expectations measures most clearly in the ninth grade. In view of the fact that Fate was the most consistent link in all three cohorts and that a similar measure was available for the graduates, it was decided to concentrate on Fate in the further analysis. An Ambition Model Incorporating Fatalism

To view fatalism as a link between the exogenous and expectations measures implies that fatalism at least in part explains the effects of the exogenous variables on expectation. It seems reasonable that expectation, in a boy with a low IQ, might be low partly because he thinks it is not possible to influence the future. Presumably his past experience would have shown him that “the secret of happiness is not to expect too much out of life.”” Similarly, there seems to be a general association between a low position in the stratification system and fatalism (see Kahl, 1965), and thus low-status boys might be expected to have low expectations precisely because they believe it impossible to control their own careers.

To view fatalism in this way leads to the idea of using it as an intervening variable in our model, in the same way as academic performance was used. Since a model was to be designed which would also include Grade, a further decision had to be made about the order of Grade and Fate. They could be included in the model without being ordered, as was done with Grade and Partic (Chapter IV), but here there seemed to be a firmer basis for choosing an order. Since Grade is based on

56 BELIEFS ABOUT OPPORTUNITY

performance in the past and Fate is based on questions asked at a later point in time, if there is any influence of one on the other, it is more probable that Grade influences Fate than the reverse. During one’s life a constant flow of influence in both directions would be expected between evaluated performance and fatalism. If a boy is fatalistic and thus does not strive, his performance is likely to be poor; if his performance is poor, he is likely to become fatalistic. Given the time sequence of the particular measures, however, it is reasonable to view past academic performance as influencing current fatalism. Table 5.1 presents the correlation matrix and Table 5.2 reports the path coefficients for the Grade-Fatalism model for the in-school cohorts. In Table 5.2 there is clear evidence of an effect of Grade on Fate in both older cohorts; the Fate-Grade

path is not significant in the sixth grade.’ In all three cohorts, there is a flow of influence from IQ to Grade to Fate. The association between IQ and Fate, indexed by the correlation coefficient in Table 5.1, is thus partially explained by Grade. Smart boys tend to perform better in school, and better performers are less likely to be fatalistic.’

There are actually few other differences between the results in Table 5.2 and those for the Grade model discussed in Chapter IV (Table 4.4) and those few are in the two older cohorts. The most obvious effect of including Fate is the reduction of the EdExp-IQ path in the twelfth and ninth grades, especially in the latter (from .21 to .13). This means that the combined effect of Grade and Fate is to reduce the

EdExp-IQ path very sharply from its original strength in the basic model discussed , in Chapter III. In the ninth grade the reduction is from .35 (in Table 3.6) to .13 (in Table 5.2), and in the twelfth grade from .37 to .13. Thus Grade and Fate contribute a great deal to the explanation of the effect of IQ on educational expectation, and they each make independent contributions to it in the two older cohorts.’ The inclusion of Fate, however, exerts practically no effect in the sixth grade. This analysis, therefore, reveals the same effect of an intervening variable as was observed of Partic (Chapter IV). The intervening variable serves to explicate the effect of a prior variable on the dependent variable, but it also makes an independent contribution to the model of the flow of influence. This is a pattern which will be found with respect to other intervening variables. An Attainment Model Incorporating Fatalism One of the reasons for choosing fatalism as an intervening variable in the analysis in the previous section was that a similar measure was available for the graduates. Four of the six items in the Fate scale used on the in-school cohorts were included in the graduate questionnaire.* A simple summation score of these four items was used in the analysis as an approximation of a Fate score. I have argued that prior evaluated performance in school should be viewed as a possible influence on fatalism. The graduates’ experience since they left school should be seen as another possible source of influence on fatalism. Thus, rather

than regarding educational and occupational attainment as partly the result of fatalism, fatalism will be considered as a partial result of attainment. Again, this is because of the time sequence: educational and occupational attainment were in the

past at the time the boy filled out the questionnaire, while his responses to the

AMBITION AND ATTAINMENT 57

>

e™ D]mmAIMN Mm Htm AOAO OMNOOK— aH St ONTHMAHNMO a~oWW HE SCOAMOMNNDD ONANANDWO =seaeNm atwonrnyn ANQABNIMNHAMmMAM MANMH HINA LO

nN = NA NQ ~ N N 1 N N

4'

= =a —

DaAnownrnerst = OMONROM WO woonnttno ANAMAA TH ANTAMOOM NOC9HTTYA OoFOOWNNON =ByONTE MONA ~MNOMOOM —

= | oS ' : co 1 s ’

— + WY ore) oO + oO ore) ra) —- oO 00

3 ~OMNMNNER 4 +t+ANOOMAA NaMaOthh 0 ONRAWONOOD NOM OKA tO O MONNOTN AAT ET. ONDOOFAA 1M MmAMNMENANAM

a =iS—™M~OOMNnM AtstAOr onunonawan =o = ON ONL ON ONnNKEM a=) MONanan Tawarn 6 8 i} ' ; re | [} $ 1 e 48QaaNnan ' OOaNr r) iy two a

S Annee IN DAM CO ANNOD ssws)eal = &© ONn~NHO =—=tarnon mS 5MA MAA Mm | © QYNAYN

= w BO

220% a eS am ss

t+ = ~SITS HAN AAI AM MS SaAN 2 SoS 2Sami ; 3 5 Oo +tO ~+ee "D orm Sent ma ON © TAM foes—= a00 e\O eDnwn .stm e eMOM e 06h

oo) § 1 ' ri moO moNN mr Ns NO t) (] ij 7” On NomNa ™~ =|=N =|

©)

ir m~ Ww in Oo oO + ae)

a} _~ feP) oO -—™~ oS CO mM a~~a, BD a, 32 a,

Or Qo Gx23] ol]O21 = | -=N © Ps gull UOQo KM a gux| OoVoOoRXxM sc = GVOxXM moms HOnDLE ] OnQo23 N ot asaawp oaNS & S=nhamawonwd ~ FZ oft sav oak =i ost Susxs “OSZS-~nhaawWowo oS SBnmMawmony

58 BELIEFS ABOUT OPPORTUNITY

Table 5.2

Dependent DetermiPath Coefficients, Grade-Fatalism Model, In-School Cohorts

Independent Variables | Coeff. of Variables IQ Sib FaEd FaOcc Grade Fate EdExp nation 12th Grade . (.2646) (-.0709) (.3114) (.0080)

Grade 525* -.024 .116* .032 - - - 340 Fate(-.0450) -.198* O17, = -.044 -.129* - - .129 (.0224) (-.1200) (-.0049) (-.05 28) EdExp 129% 113+ 157* 152* 405* -.128* , 510 (.0151) (-.0766) (.0971) (.0088) (.0939) (-.0660) OccExp -.059 -.018 -.049 .076* 147* -.036 591* A474 (-.1349) (-.2353) (-.5910) (.0856) (.6646) (-.3654) (11.51)

9th Grade

Grade 449* -.096* 182* AS3* - - - 430 (.2113) (-.3121) (.5010) (.0383) Fate (-.0812) -.344* 010 -.044 -.030 -.180* . . 242 (.0161) (-.0056) (-.0036) (-.0860)

EdExp .132* .006 .190* 056 .269* -.252* - .430 (.0150) (.0041) (.1212) (.0032) (.0622) (-.1217) OccExp .061 -.017 .128* .034 -.000 -.046 .486* 394 (.1382) (-.2501) (1.628) (.0397) (-.0019) (-.4429) (9.734)

6th Grade 664* -.037 080 059 541 (.3301) (-.1079) (.2433) (.0159)

Fate (.0436) -.227* .035 .064 -.082 -.178 - - .170 (.0410) (.0748) (-.0086) (-.0689)

EdExp 178 -.065 .211* 081 .038 -.037 - 188 (.0157) (-.0337) (.1132) (.0068) (-.0168) (-.0168)

OccExp .143 -.128 -.144 .228* .007 -.018 .199* 189 (.2702) (-1.419) (-1.652) (.2345) (-.0264) (-.1736) (4.263) NOTE: Main entries are the standardized path coefficients: those in parentheses are in metric form. Asterisks mark coefficients

that are more than twice their standard error. .

questions measuring fatalism were current. Unless it is assumed that fatalism is constant over six or more years so that the measure taken in 1969 would have been the same if taken in 1963, attainment must be seen as possibly affecting fatalism, rather than the reverse. Thus the graduates’ post-high school experience (EdAtt and OccAtt) is introduced in the model before Fate. If this were the only change made in the basic model, Fate would thus be the final dependent variable. However, two other variables from the graduate questionnaire are relevant to the role of a measure such as Fate. The graduates were asked how much more education they “really expect to get,” and, further, which occupation from each of two lists they thought was “‘the best you think you can have by

the time you are 30 years old.” The first of these measures is an almost exact counterpart of EdExp in the in-school cohorts; the second is a rough approximation of OccExp in the in-school cohorts, if we use the average Duncan score of the two occupations chosen from the lists. Since the measures are not really the same as for the in-school cohorts, however, the new measures are referred to here as UltEd and UltOcc.

AMBITION AND ATTAINMENT 59

If these two measures are the final variables in the model, it is possible to construct a model that is highly comparable to the ambition model of the in-school cohorts. In addition, by using EdAtt and OccAtt as intervening variables between

the exogenous variables and Fate, the effects of these early attainments on the graduates’ views of the future can be demonstrated. The structure of the resulting model is like the graduate Grade model in Chapter IV with Fate, UltEd, and UltOcc following OccAtt, in that order. Table 5.3 reports the intercorrelations of all variables in the analysis, and Table 5.4 presents the path coefficients. The first three rows of Table 5.4 simply reproduce the basic structure of the Grade model from Chapter IV; it is the last three rows that are of greatest interest. It is clear that the model does not explain much » of the variation in Fate among the graduates, not nearly as much as it does among the in-school cohorts. Only IQ has a significant effect on Fate, although EdAtt and OccAtt both produce sizeable, though not significant paths. More noteworthy is the fact that Fate is the only variable beside EdAtt and OccAtt that has a significant effect on UItEd and UltOcc; its effect on the last-named is stronger even than the effect of OccAtt. Thus, although attainment to date clearly has the strongest effect on expectation, expectation is also influenced by the young men’s view of how the world works. This is particularly impressive when one remembers that they were

asked about a future only six years away. Moreover, their responses were not significantly associated with their social origins implying that they saw themselves as on their own to a greater extent than did the high-school seniors.

Yet it is disquieting that so little of the variation in Fate is explained by the model. Among the younger boys a fatalistic view of the world seems to be affected ‘not only by ability (as it is among the graduates) but also by academic and proba-

bly other performance. Although both EdAtt and OccAtt have sizeable paths to Fate, neither is significant. In fact, none of the zero-order correlations with Fate reported in Table 5.3 is very large. It may be that with their wider experience these young men have been exposed to many influences not included in this analysis.

But, whatever the reason, Fate acts more like an exogenous than an intervening variable: it helps explain UltEd and UltOcc, but it is not explained by the antecedent variables. FOOTNOTES

'This difference in level of significance by cohort seems clearly to result from differences in size of sample in the three grades. Actually, the path and metric coefficients are both larger in the sixth than in the twelfth grade. 2 This explicative function of Grade with respect to the IQ-Fate relationship is more easily seen when the present model is compared with one in which Grade is not included. The Fate-IQ . path coefficients in the simpler model are .266, .425, and .336 in the twelfth, ninth, and sixth

Frees ay Pectvely, compared with the comparable paths in Table 5.2, which are .198, .344, 3It is not necessary, of course, to accept the ordering of Fate and Grade as used here for this

outcome to be meaningful. The EdExp-IQ path would be the same in the present analysis whatever order was used, so long as both intervened between IQ and EdExp. *It was not clear, of course, at the time the questionnaires were constructed, which of the achievement orientation items would form a scale for the in-school boys. A selection was made on substantive grounds for the graduate questionnaire and, fortunately, four of those selected were the same as those in the Fate scale.

BELIEFS ABOUT OPPORTUNITY 60

>

“=~ +t~ MM tT O NH DO YT NY Alam SGA BHA DI t= aA Nn se NAN BHGS ANA

“Ai ~™ N AN =

ce]ODM Ca tn OM Mm DBD OBO MH BDAOM we so ol Oo M Om HH Sse NN &

=i |© s+ °c Va) \O otf At oe Fe KR ODM = 2~]y Ne AOD Hh HN M 5 Oi nNnwunnuwoee iste nan ow

; ss a | a Ore ee ee

co mya) m$m NNN MH OE ON ee

—_—

yo Y

G Ox oOo WW co Or wo —

~™ Vv

asfa) wy Yo eon NN

E Gee fe oOo mM + pk ih ar a

oo | =i - gree © Oo OC6DOD NH ot ow $85 |31 28928 8 28meCys £OL2 16 " sis Summ &

6orziena ot Am4 & 52Ss o § [r] Cs

oY JEL Zi 6 .o + Se 8 8 ors & a » NN o= O = ba

Oo ~

; oe | ee

ail , Sl ong SiR ™~ CN

we

N

my ©}

SO 2) O zi Sgmgwmogm og i5 5 , roa a3 oO _—- +»

i} 5, 83 SBS RB 2EE

AMBITION AND ATTAINMENT 61

O . 3=: a to] é tS| t ] 8& =

CET \o tm =©wn oysa) t+ qe ct 92) aoOo &ON qAQ aLag) ve) faa) w e e e e ese Gay

5a

= Oo ~ 20 E 2% — em oO

~~ M&S * & =| 2 , KR Fe] eo ~ 5 SS at!) & es CL, : A = ran o , New . 3 O =) Naw A S par o~ om aN ~ SO = < tc + * 00 a 4 ai) ¢) ’ ' ' NAN © AN t om o O NA = 0o © + © a ® = o =S 66 a5] &§ ae, oe — aes .i3~a~=aiSSf5) A oO ~~ “ a oO & , OM AE ANH Of 7 OR sae tH MO 6 jaa e e : ny, "A = 35 —_ a pan}

oO

t+ | > = ~ Qa} 3st wpe ls Oo § a om co om a on aa om ot} of Ss = co WW No [~ — + D yw 2

“ C5 MA ey £6 a-_i

S a — os

2ss€& =Za© *om \O st (o.e) \Oxan) SoOo o oD ' NN t~ \O AS oo ee,°) aie 3) 8 sO OR OBA ue e) Feissoo © ma ©S—_ S69 SS ie. a, a, oeSCF] wwS8E& cs &3rs~ —~ os * OR ©Fon #A t~a a=&vom eo] as iK aren +O AoasoN SD No oF oom

SS ] >O.Om tonnes M ma oO CO OMAN ~~ —4 WO won 3 faa] = WW CO mm ANAH CO rs as}Mm YK TT Nf™ i Oo ss= ©aYeysear NQYIYNM

S

Mo nm SS Lo) © OD

=io) Nae naratoiAoo EoiaM one OO MAAS Am TNM NH 22 =|va88/8 abt oS i elOMWA AANA Se os 2, Pg 5aAeas = o=)”n ON °HO 1OtffonHO 8NN 8_OM e=iNANM 8CO8osCO 48 mSmwFS DY

2=

a} MmMw oO +eo m~99 A 60 A mM Oo v Co & ~™AO %

\®)

a o vt ~ oOon00rere) On No) © AQ

00eON mM © m~ cS Wm oO WwW oO MWY [Ly 8@ .- «@ _ ¢a

© wv wT + YQ

[L,

Ze 2,3pan OD Q, 4 o.

Vay jorOM aNVES oO Ou oS =F aOa he jori C™ a8 4oO Se ON a Ce OR ss! Qog Hose 2! Bog Foss 2c!OTCV8 Box Hos 38 cP)

f—

AZSESORSEES PwEBBBOL SDE SZSESQOR S=aOSsSmnoaengawavoond REGS aa O rnawnoo SE3 HnSESS Hne DOO

Table 6.4

Dependent DetermiPath Coefficients, Parental Encouragement Model, In-School Cohorts

Independent Variables Coeff. of

Variables MoEd FaOcc FaEd IQ Sib Grade ParEnc EdExp nation

Grade .093* 018 .090* 513* -.020 - - - 349 (3134) (0047) (2433) (.2602) ~—-(-.0602) 12th Grade

ParEnc .069* .182* .120* .184* -.104* .303* - - 413 (.0960) (.0187) (1324) (.0384) = (-.1265) (1241) EdExp .037 012 .083* 045 -.044* .207* .624* - .719 (.0289) (.0007) (.0518) (.0052) (-.0301) (0479) _(.3508)

OccExp -.090* .092* -.017 -.053 -.027 157* O51 562* 476 (-1.368) (.1030) (-.2055) (-.1193) (3630) (.7021) (5614) (10.90)

Grade .219* O95 .096 .430* -.109* - - - .466 (.6738) (0235) (2598) (.2074)_~—s (3482) 9th Grade

ParEnc _(.2345) .180* .068 .140* .251* -.030 118 : - 333 (.0072) (.1602) (0512) ~—_ (-.0412) —(.0502

EdExp 018 042 049 059 056 .221* .616* - 675 (.0132) (0025) (.0315) (.0067) (0424) (0525) — (.3463) OccExp -.086 .033 .148* .042 -.038 051 .197* 345* 402 (-1.242) (.0386) (1.888) (.0965) —_(-.5682) _(.2394) ~~ (2.200) _~—s (6.851)

Grade -.020 040 .090 .646* -.054 - - - 502 6th Grade

(-.0695) (.0109) (.2708) © (.3166) —(-.1639)

ParEnc (1148) .104 .093 .069 132 -.076 013 - - 115 (.0081) —(.0669)_ (.0209) (0738) __(.0043) EdExp -.001 107 112 .070 -.012 068 495* - .406 (-.0008) (.0052) (.0611) (0062) (-.0066) (.0123) (2787) OccExp -.008 .232* -.102 .112 -.133* 012 .017 .233* .222 (-.1103) (.2446) (-1.192) (.2141) (-1.562) —_(.0453) ~—- (.2031) += (5.007) NOTE: Main entries are the standardized path coefficients; those in parentheses are in metric form. Asterisks mark coefficients that are more than twice their standard error.

that parental encouragement reflected both the socioeconomic characteristics of the family and the boy’s ability, but the difference between the two studies lies in the explanation of variance in educational expectation: whereas ours shows a very strong path from parental encouragement to EdExp with rather weak direct paths from Grade and SES, theirs produces a path of moderate size from parental encouragement and paths that are nearly as strong from SES and IQ; consequently, parental encouragement is presented as less powerful than in our study. However, both studies show parental encouragement as responsive to the boy’s characteristics and a function of the family’s social status. And both indicate that, even given this kind of measure of parental encouragement, the boy’s view of his educational future is

...‘

independently affected by his academic qualities and his family’s social status.

The Quality of the Parent-Son Relationship

Parental influence on educational and occupational goals should depend to a considerable extent on the nature of the relationship between the boy and his

70 PARENTAL INFLUENCE

parents. If the relationship is badly strained, the boy may ipso facto resist what the parents want. In any general population this negative influence is presumably not important, but varying amounts of parental influence would be expected in any population. To explore this issue, several measures of the parent-child relationship, all based on the boy’s report, were used. These were measures of the respect the parents show for their son’s ideas, his sense of integration with his parents, and how far he accepts their rules and regulations as his own. None of these measures is consistently significantly related to the model variables, the measure of parental respect being the most consistent,” and clearly, none can contribute much to the explication of the association between the exogenous variables and the measures of expectation. They cannot even serve very well as independent sources of explanation of expectation.

Yet it might be argued that these measures of the parent-child relationship cannot be expected to provide so direct an explanation. Instead, perhaps they may possibly serve to explain how fully a boy accepts the parent’s goals, whatever they may be. Thus two other kinds of analysis might be more appropriate: first, one might attempt to use these measures to explain the differences between EdExp and ParEnc, anticipating that the closer the parent-child relationship, the smaller the difference between the two. Second, the cohorts might be divided into those with high and low levels of integration with their parents to see if, as one might expect, the model reported in Table 6.4 is stronger where the relationship is close. Both these procedures were carried out, but without the expected results. In fact, the measures of the general quality of the parent-child relationship failed completely to contribute anything of significance. Another variable, which deals with the parents’ response to their son’s academic performance, is somewhat more significant. The boys were asked if they agreed or disagreed with the following items: “My mother (father) doesn’t seem to care when I bring home a report card with high grades,” and “My mother (father) doesn’t seem to care when I bring home a report card with low grades.” A simple summa-

tion of the four responses to these questions was viewed as a measure of the parents’ concern. Although not very highly correlated with the other variables in

the model, this measure did contribute significantly to an explanation of EdExp in , both the twelfth- and the ninth-grade cohorts, even when Grade was also included. However, although the parents’ responses to the boy’s academic performance seem

to affect his level of educational expectation (at least among the older boys), neither the social status of the family nor the boy’s ability or academic performance helps much to explain those responses. That is, this measure of the parents’

variable. .

concern with school work functions more as an exogenous than as an intervening It will be recalled that the analysis of fatalism in Chapter V produced a similar

result. Although fatalism helped to explain ambition, it was not itself explained appreciably by the exogenous variables or by Grade. This similarity suggested that the present finding might be related to the earlier finding on fatalism. It might well be, for instance, that the way the parents respond to the boy’s activities in school and elsewhere influences him in deciding whether it pays to strive. If parents do not

AMBITION AND ATTAINMENT 71

respond to the boy’s performance, if they do not provide a context within which his efforts and their results are appreciated, he might well decide that it is not worth while to make the effort.®

In order to explore this line of reasoning, the several measures of the parentchild relationship were tested as possible sources of explanation of the boy’s fatalism, and it was found that the combination of the parents’ concern about his school

work and their respect for his judgment was the most effective.’ When these measures were used as exogenous variables along with those in the Grade-Fate model (Chapter V), the coefficient of determination of Fate was increased appreciably. In the twelfth grade it increased to .181, compared with .129 in the Grade-Fate model, while the increase in the ninth grade was even greater, from .242 to .331. The inclusion of these variables did not alter the Grade-Fate models except for their effect on the explanation of Fate in these two older cohorts. They had no effect in the sixth grade.

It may well be, therefore, that the quality of the parent-child relationship, though relatively independent of the family’s social status, makes a contribution to a boy’s ambition by affecting his view of the possibility of goal attainment. He is more likely to believe that it pays to work for goals not only if he.is smart and has been academically successful, but also if his parents respond to his performance and respect his efforts. Such parents are evidently not concentrated in any one part of

the hierarchy of social status, and thus social origin contributes very little to an explanation of the relationship. But if the quality of the parent-child relationship is considered itself as an exogenous variable, fatalism is better, fatalism and Fate is thus made a more integral part of the model.

FOOTNOTES

1The correlations between MoEd and FaEd were .52, .60 and .62 for the twelfth-, ninthand sixth-grade cohorts, respectively. *In a further analysis of these data, in which parent interviews that were part of the larger project were used as additional sources of information, it was shown that there is a considerable

and Huff, 1973).

difference between what the parent reports as his goal for the boy and what the boy reports that goal to be. The difference is especially striking among the younger boys (See Kerckhoff 3 This tendency to project one’s own goals upon the parent is strongly suggested by the analysis of some of the same data presented by Kerckhoff and Huff (1973).

4A strictly parallel analysis to theirs, using the twelfth-grade data from Fort Wayne, is reported in Kerckhoff and Poss, 1970. >There were 72 intercorrelations between the model variables and these measures of the parent-child relationship in the three in-school cohorts. Only eight were as large as .20, and none was over .30. Five of the eight over .20 involved the measure of parental respect. ©There is some parallel between the logic used here and the theory Rosen and D’Andrade

(1959) propose to explain the development of boys’ motivation to achieve. Their theory, however, is much more complex and detailed.

7The measure of respect consisted of five items, including: “Do your parents give you a chance to share responsibilities?’’, “In family discussions, do your parents encourage you to say

what you think?”’, and “My parents respect my judgment.” .

72 PARENTAL INFLUENCE

CHAPTER VII

PEER INFLUENCE A second potentially potent source of influence on the boy’s expectations about his future is the peer group in which he spends much of his time. The influence of peers is presumably more important in the United States than in some other socie-

ties. This seems to be true for a number of reasons. First, the continuation of formal education well into adolescence and even early adulthood places the individ-

ual in an age-graded social context which is certain to foster a strong sense of . collective identity. Second, the American ideal is that of independence in adulthood, which makes it clear to the young person that he must eventually disengage , himself from intimate ties with his family. The peer group constitutes a kind of half-way house in this process, providing him with social support and a competitive,

achievement-oriented setting in which he can learn to fend for himself. Finally, rapid social change and the value placed on improving the system and one’s place in it make adults less capable than they are in other societies of providing guidance; the young must find their own way. There are two ways in which peer influence may be viewed in such an investigation as this. The first is to focus on the few friends who are most significant to the individual and seek evidence of their influence on him. The second is to consider the whole peer group as the source of influence. It is possible in this study to look at both, but most attention will be directed to the first. The in-school boys were asked to name their three best friends in their grade in their school. They were also asked if these boys were their best friends, whether in

their grade and school or not, and to name their three best friends in their grade and school three years earlier. The graduates were asked the first of these questions with reference to the time they were in the twelfth grade, and if those named were

AMBITION AND ATTAINMENT 73

still among their best friends. The responses to these questions form the basis of most of the analysis to be reported here. In addition, however, differences in the broader peer contexts were analyzed by viewing all of the boys in the same grade in the school as the peer group and seeing whether some of the differences at the interpersonal level could be attributed to variations in such contexts. Three questions were dealt with. First, are boys who are friends actually more similar in significant respects than boys who are not friends? Second, is similarity to be interpreted as a function of some process of influence? Finally, does information about the boys’ friends help to explain their expectations? Peer Similarity

The assumption that directs attention to peers as a source of influence is that those who are good friends are more similar than those who are not. If there is no greater similarity between friends than between any two randomly selected individuals from the same population, peer influence is a meaningless idea. Thus the first

. task was to determine whether friends are actually more similar than non-friends. In the population studied there is, of course, considerable variation on a number of dimensions. The dimensions most directly relevant to the analysis here, however, were concerned with the individual boy’s views of the future. As a basic index of

similarity, therefore, the educational expectations of friends were considered. Even , with a single dimension there were numerous ways to use the available data, but for the present purposes a simple method sufficed. The degree of similarity was indexed by the correlation between ego’s EdExp and the educational expectations of the boys he mentioned as his three friends (FrExp).

These correlations are presented in Table 7.1 for the three in-school cohorts. Two features of the findings are noteworthy. First, the size of the correlations diminishes from the older to younger cohorts. Second, the clearest agreement between ego and friend is found in the case of the first- or second-named friend. However, sixth graders show very little agreement between ego and any friends. In that cohort, at least, there remains some doubt as to whether one can comfortably assume that friends actually agree more than non-friends do. Although the coefficients are all statistically significant, they are not very large. It is not at all clear, in fact, whether any usual definition of statistical signifi-

cance is suitable in such a case. The issue is not really whether the extent of agreement between friends is greater than zero but whether it is greater than it would be if friendship and educational expectations were randomly linked. Since there is some variation in the socioeconomic make-up of the several schools, and since socioeconomic level is associated with EdExp, one might wonder how much of the similarity between in-school friends is a function of the homogeneity of the

school population. The greater the homogeneity of the school, the more such structural factors would influence the outcome. To provide a point of comparison, therefore, the intra-class correlation coefficient (Haggard, 1958) was computed for each cohort. Basically, such a coefficient

reports the degree of agreement (in this case, agreement in EdExp) between all possible pairs of boys within each school, summed over all schools in a cohort.’

74 PARENTAL INFLUENCE

Table 7.1

Correlations between Educational Expectations of Friends, In-School Cohorts

Grade l 2 3 Order in Which Friend Was Named

12th .520 (816) 472 (765) .457 (731) 9th 473 (390) 421 (390) .464 (361) 6th .257 (313) 311 (311) .241 (304)

NOTE: Numbers in parentheses are the sample sizes.

This procedure produced coefficients of .080, .130, and .146 for the twelfth, ninth and sixth grades, respectively. At least for the two older cohorts, actual friends are clearly more similar (see Table 7.1) than random pairings. The sixth grade deviates from the others in having both the lowest correlation among actual pairs and the

highest correlation among random pairs. In short, friendship pairs in the older cohorts are very much more similar than one could expect by chance, but the pattern is less clear in the sixth grade.”

The fact that first-named friends agree somewhat more with ego than those named later suggests that perhaps agreement varies with the closeness of the relationship. This possibility was tested by three other measures of closeness which probed (a) whether the person named also named the respondent as a friend; (b) whether the person named was defined as one of the respondent’s best friends, overall; and (c) whether the person named had been identified as a school friend three years earlier. All three measures could be used on the two older in-school cohorts, but only the first two could be used on the sixth graders. Several forms of analysis were used in comparing these different kinds of friends, but the results showed little variation in ego-friend agreement on educational goal or any other variable in the basic ambition model. Although there was some tendency for best friends, friends who reciprocated choices, and long-term friends to agree more than others, there were also reversals, and the differences were never very large. There is thus little evidence that agreement varies by the closeness of the friendship.

Among the in-school boys ego-friend similarity as to educational expectation is

greater than their similarity in regard to any of the other variables in the basic ambition model. This is true even in the sixth grade. The only variables used in this report on which there is as great or greater similarity between friends are Grade and Partic. A parallel pattern is found in the graduates among whom similarity between friends is greatest on EdAtt and Grade. Friends’ similarity of IQ, FaOcc and FaEd are all consistently lower than similarity of Grade and EdExp or Ed Att. Thus the analysis showed that friends do generally agree on educational goals more than do randomly selected boys in the same school, and their goals are also

AMBITION AND ATTAINMENT 75

more alike than are most other of their qualities dealt with in this study. However, there is little variation in agreement on goals (or any other measure studied) according to the several measures of degree of friendship. It therefore seemed reasonable to refer to similarity between friends, but not to differentiate among degrees of friendship. As a way of simplifying the remaining analysis, therefore, only the first-named friend was considered. Evidence of Peer Influence

Previous studies have assumed that similarity between friends is an indication that some process of influence had brought about that similarity; but little if any evidence supports the assumption. This is unfortunate since it is quite possible to interpret the similarity in a very different way. One might argue, for instance, that, rather than the friendship influencing the similarity of two boys’ expectations, the Jatter condition actually makes friendship probable. These are not mutually exclusive assumptions, of course, but similarity can certainly occur in either or both of these ways. (See Newcomb, 1961). The only satisfying method of determining to what extent either process occurs

is to follow a cohort of boys over a number of years, charting their friendship patterns and their educational expectations, or whatever other measures seem appropriate. Such data are not available here nor in most other studies of the subject. In the present study, however, there is some basis for charting longitudinal patterns, and it is worth looking at the data to see what can be learned. The in-school boys were asked who were their best friends at the time the data were collected and who were their best friends three years earlier. They were also asked what their educational expectations were at the time as well as what they had been three years earlier. All the problems of retrospective data are encountered here; nevertheless these data may be cautiously interpreted as providing some indication of changing patterns over time. Because of the lower level of ego-friend agreement among the younger boys and possible doubt (based on earlier analysis) about the meaningfulness of these measures in their case, the analysis was limited to the ninth and twelfth graders. In effect, the analysis concerns changes in agreement among the twelfth graders between the ninth and twelfth grades and among the ninth graders between the sixth and ninth grades. Table 7.2 presents an analysis of the replies to these two sets of questions. The

central question is: Is there any difference between long-term and short-term friends in the extent of their agreement on educational expectations? A meaningful pattern was found which pointed to a positive effect of friendship on agreement. First, the data on short-term friends showed that the earlier expectations of boys who became best friends during the past three years were considerably less similar than they were at the time the data were collected. From those data alone one could argue that the lower level of agreement at the earlier (pre-friendship) period

| was simply due to a greater error in the retrospective measures. The data on longterm friends were not consistent with that argument, however: those who were already friends three years earlier seemed to have agreed on expectations more then

than did those who were not yet friends (compare .439 and .325 in the twelfth grade and .489 and .340 in the ninth). Such established friendships, moreover,

76 PEER INFLUENCE

changed less in the matter of agreement over the intervening three years. In effect,

there was no real difference in current agreement among short- and long-term friends, but among the former evidently the extent of agreement changed over the three years. Table 7.2 Agreement of Long-Term and Short-Term Friends on

_ Educational Expectation at Two Points in Time, | Ninth- and Twelfth-Grade Cohorts

Educational Expectation Long-Term Friends Short-Term Friends

9th Grade 12th Grade 489 439 340 325 9th Grade 28513 20 12th Grade 508

Three years ago

Current

Thus these data do provide some support for the idea that those who become friends move toward greater agreement. The evidence would be more convincing, of : course, if these were true longitudinal rather than retrospective data. Also, it is implicitly assumed that the earlier point of reference (three years ago) was just before or at the time when the boys became friends. This, of course, is not true in general, and it is not known how much of the change toward greater agreement might have occurred before the boys became friends. It seems unlikely, though, that only such pre-friendship change is involved.

To the extent that the correlations may be taken at face value, there is also evidence in Table 7.2 of friendship based on similarity of expectations. None of the coefficients is as low as random pairing of the boys would produce. Even those who later became friends were much more alike than randomly selected pairs of boys in their schools (compare .325 with .080 and .340 with .130 in the twelfth and ninth grades, respectively). The same impression of selectivity due to similarity is suggested in the responses of the graduates on being asked to name the boys who were their best friends when

they were in the twelfth grade and to state whether these were still their best friends. They were also asked about their educational expectations when in the

twelfth grade and their current expectations of ultimate educational attainment : (UltEd). The data derived from their replies are reported in Table 7.3. It is clear from these findings that if the retrospective data are accepted as valid, the continuity of friendship over the intervening six years could not be predicted

from friends’ agreement at the earlier point. In fact, those who have remained

AMBITION AND ATTAINMENT ; 77

Table 7.3 . Agreement of Previous and Current Friends on Educational Expectation in Twelfth Grade and Six Years Later, Graduate Cohort

Educational Expectation Still Friends No Longer Friends

In Twelfth Grade 227 277 Six Years Later .265 032 friends had somewhat less similar expectations in the twelfth grade than those who |

have not, though the difference is not large. On the other hand, those who have remained friends have more similar ultimate educational expectations than those who have not; in fact, there is no agreement at all between those who are no longer friends. The striking difference between agreement on educational expectation in

the twelfth grade and six years later suggests some alteration in the basis of friendship. The best available indication of what this might have been is found in the information on the graduates’ educational and occupational attainments. Pairs who are still friends are more similar on both of these measures than those who are not.° Thus the continuity of friendship is more easily understood by reference to what happened in the intervening years than by reference to the extent of earlier agree-

ment. Rather than showing friendship at an earlier point leading to similarity in expectation and attainment, the data point to the impact of intervening events on friendship. Again, although presumably not a matter of either-or, the direction of change seems more consistent with a process of selectivity rather than of influence. Thus friendship does seem to increase the level of agreement on expectations, but agreement between friends is greater than would be expected by chance, even at a point before they become friends. Similarly, the experiences of friends which make them more or less similar do seem to affect their friendship. These findings do

suggest that similarity between friends results from both interpersonal influence } and selectivity in choosing them, but there is no way, in the present study at least, to separate the two effects. Further analysis must therefore be carried out with the knowledge that either interpretation of similarity is probably both right and wrong. To refer, as many earlier studies do, to “peer influence” or to “‘significant-other influence” is unduly simplistic, but it cannot be corrected in the present analysis. One can do so only in the interpretation of the findings. The Effect of Peer Similarity

Some doubts were expressed earlier about the measure of parental encouragement because it was based on the boy’s view of his parents’ wishes. Thus the similarity between what he expects and what he thinks his parents want, though strong, could be largely a function of the fact that both measures were based solely on his report. In the present case this problem does not arise. As the previous

78 PEER INFLUENCE

section has made clear, it may still not be completely safe to refer to peer influence in this analysis, but at least the similarity is based on two independent measures. Including peer similarity in the analysis again requires a decision about its position in the flow of influence represented by the model. One may reasonably expect that there is some tendency for those of similar social levels, with similar abilities, and whose performance is similar, to choose each other as friends. Therefore the measure of the friend’s educational expectation (FrExp) is placed between Grade and EdExp in the model. This implies that the kind of friend chosen (indexed by

FrExp) can be partially explained by ego’s social and intellectual qualities. Throughout, FrExp is based solely on the boy’s first-named friend, such refine- . ments as length of friendship, reciprocation and whether this is a best friend overall being ignored.*

The correlation matrices for the ambition model using FrExp with the three white in-school cohorts are presented in Table 7.4 and the path coefficients for the model are reported in Table 7.5. The most important fact about Table 7.4 is that all of the other variables in the model are correlated with FrExp at levels that are only slightly lower than those of the correlations involving EdExp; that is, ego’s friend’s educational expectations are associated with ego’s characteristics almost as closely as are ego’s Own expectations.

Turning to Table 7.5, it is clear that FrExp varies with the background and performance characteristics of ego and that FrExp helps explain the variation in EdExp, but that these relationships differ by cohort. Only in the twelfth grade is there a significant direct effect of ego’s background on FrExp. In the ninth grade this effect is mediated by Grade to such an extent that the direct paths are not significant. In the sixth grade neither the direct nor the indirect paths are significant. Although it is difficult to understand, only Sib has a significant effect in the sixth grade. These several variables do explain a sizeable amount of the variance in FrExp in the two older cohorts, over one-third in the case of the ninth graders. In turn, FrExp is clearly a significant contributor to the explanation of EdExp in all three cohorts; in fact, it is the only significant contributor in the sixth grade, beside FaEd. In the twelfth grade, in contrast, all the model variables make a direct contribution to EdExp in addition to indirect effects through Grade and FrExp, and in the ninth grade only FaOcc and Sib fail to show a significant direct effect. If the paths in this model are compared with those in the Grade model (Chapter IV), the inclusion of FrExp is seen as reducing the direct paths from all the previous variables to EdExp in all three cohorts. Most seriously affected, at least in the two older cohorts, is the EdExp-Grade path. In the twelfth grade, is is reduced from .406 (in the Grade model) to .337 (in this model), and the comparable coefficients for the ninth grade are .306 and .215. That path is of insignificant size in both sixth-grade models. The major contribution of FrExp is through the explication of the flow of influence rather than the addition of an independently effective source of explanation of EdExp; in none of the cohorts is the coefficient of determination of EdExp raised appreciably (a maximum of 4% in the ninth grade). Finally, FrExp has only a minor effect on the model so far as OccExp is concerned. Only in the twelfth grade is OccExp-FrExp path significant, its major role being to reduce the

AMBITION AND ATTAINMENT . 719

™ — roe)~ARARGAG NOMNONAO 8nA BAAS SARQRRLSS > BQ nn oeSeat en oN FANN HE HAAR MANN A =H YH e 77g acs CQOOAQ a >

= wl —

00 MNNODWOAON Mm-m~oaonvuodt a HSATARGSAAH SSSCRARQAAR ZSOARRSSA ONN MM CO ON FtOMMNMNMNMOM AW OmMmrommMmma mon ©

Oo t 4 a — jaa) t 1 e

= FY oO w” + ro) ova) ry + -) 00 No)

fo fa On AN MORON onmnnrtanmn NOAnKrm~o +t ea aH OR HO DAmnawoet as NAtanowwtm 3 ANMHAtTSO ANANAtM oO NMaANaAaNaM

A

o | a| SERLSS eggees SANSaR 2o|8 HEERSS SEARS ANRMAAMA a a ASRaG L£eZRes SSSR 7=||5aan AMEAH AQaAA —_

3) iL, ' 1 ' an O . 8 8 « . | . 8 8 —

}=)

aw)

ro)

=

a™cD) S wo NOCH — ™~=— © 0 = Os Ww

NA ON ~ ONAN NOonr =o @& ED =oSNAY tTTNnA AAm~aA

za3 =Pa Soe Ses. —~ON ANA heRB a e egg °0:

Be 3Y s

.|

a Ae) in ak 8 tA 00Ve) ™ Ra\O SON A) ‘© el ze 3¢ 3g fae]

=

8 oO CON NO 29 =) 2 Fon

i)

iL,

o

=nvgnwdc C188 oesreaeSze VCissZ oO , azaeaoe s163 tess Szeeoqoe soseseze —~—Amma ON moe nom~ayo Oo Dmnomemnomwmao —

80 PEER INFLUENCE

Table 7.5 -

Dependent DetermiPath Coefficients, Grade-Friend Model, In-School Cohorts

Independent Variables Coeff. Of

Variables FaOcc FaEd IQ Sib Grade FrExp EdExp nation

Grade (.0040) .016 .126* 525* .039 - - - 344 (.3340) (.2645) (-.1153) 12th Grade

FrExp .135* 158* 065 -.037 344* . - .278 (.0078) (.0969) (.0076) (-.0253) (.0770)

EdExp .124* AS1* .145* -.098* 337* .189* - .499

| (.0072) (.0929) (.0169) (-.0664) (.0779) (.1895) OccExp .094* -.079* -.052 -.026 .128* .092* 561* .466 (.1058) (-.9336) (-.1162) (-.3375) (5733) (1.772) (10.84)

9th Grade |

Grade 155* .160* .439* -.127* - - - 434 (.0385) (.4391) (.2091) (-.3953) FrExp (.0038) 065 (.0700) .106(.0212) .186* -.098 327* - - 350 (-.0732) (.0784) EdExp ~.056 .177* .190* 057 .215* .242* - 425 (.0032) (1137) (.0211) (.0416) (.0504) (.2364) OccExp (.0799) .070(.9985) 078 (.1426) .065 (-.4752) -.033(.1190) .026(.8280) 043(9.436) .478* .410

Grade (.0095) .035 .082 .667* -.048 - - - 533 (.2495) (.3267) (-.1467) 6th Grade

FrExp (.0075) 152(.0281) 051(-.0025) -.028 -.191* .105 - - 118 (-.1069) (.0193)

EdExp .108 .144* .109 . -.029 111 .184* - 212 (.0052) (.0774) (.0094) (-.0156) (.0198) (.1785)

OccExp .222* -,123 139 -.118 014 -.107 .230* 205 | (0761) — (-1.422) (.2590) (-1.367) (.0539) (-.3540) (4.921) / NOTE: Main entries are the standardized path coefficients; those in parentheses are in metric form. Asterisks mark coefficients that are more than twice their standard error.

OccExp-Grade and OccExp-EdExp paths rather than to increase the coefficient of determination of OccExp. The logic of the comparable model for the graduates (Tables 7.6 and 7.7) is somewhat different from that for the in-school cohorts. The conceptual place of peers is clearer in that the immediate dependent variable is EdAtt up to the time of data collection, and the peer measure refers to peers who were meaningful to ego when the boys were in twelfth grade. The model thus poses the question of the extent to which a twelfth-grade friend’s subsequent educational attainment has an effect upon ego’s.

| Table 7.6 is similar to Table 7.4 in that friend’s attainment (FrAtt) is shown as consistently related to all the other model variables. It has a somewhat weaker correlation with EdAtt, relative to the other variables, than FrExp does with EdExp (Table 7.4), but that relationship is still comparatively strong. In Table 7.7, FrAtt is significantly associated with ego’s social background and academic performance.

AMBITION AND ATTAINMENT 81

>t OO OKnR A+ AIENAgnan ~ MN = Om MAN

a or — N COM WOM A OM S| ACNMHASTO

o

= a OotanNnNaAa Oo co Val| —

) ; |s } | = Ot+ONNO < NAN ON ST Ly

©. 2 ~ae)Oem |aw

o ™ ¢ oO ov yo o = Oo >= o sé oA ue =) -3=a x @ — WY oO xma, cn 3?0 —

Sec

=e xs ON As \O SON] [~©~A==Ba =a oy8Ams W— Mm ~ 3oD) Ss©=oO o—~

— e Ke — ~— ~ w 3a”

ew} © a—~ a o—~ om pan 56 e|/SesSO SzHO RSOHA] EA] SE - S(3 =1Tyan BS

= CS _S© Ff Ss ga ~ oo m~ + OO 4 Ns co < < = , O'S cS < < oO =

oO>s O ar LL, a uo) Qa I) Y ©

AMBITION AND ATTAINMENT 83

FrAtt also makes a significant contribution to the explanation of EdAtt but this, much like that of FrExp in Table 7.5, does not increase the coefficient of determination of EdAtt beyond its level in the Grade Model (Chapter IV). The major effect of the addition of FrAtt is to reduce the paths to EdAtt from the other variables, but especially from IQ and Grade.

Given the explicating role of peer characteristics in the models of educational expectation and attainment, and given the suggestion made earlier that ego-peer similarity is probably a function of both selection and influence, it is difficult to evaluate the findings presented in this section. Although this is not the problem of non-independence of measures encountered in the case of parental encouragement, still, caution is necessary in the inferring of interpersonal influence. Although FrExp and FrAtt both reflect the social and performance characteristics of ego and help explain his expectations and attainments, it is unclear how these relationships should be interpreted. The structure of the models suggests that ego chooses his friends, at least in part, according to the fit between his own background and performance, on the one hand, and the friends’ characteristics (including FrExp and potential FrAtt) on the other, and that, once chosen, the friends have an effect on ego’s EdExp and EdAtt. This is, indeed, the way I would like to conceptualize the relationships. However, if the findings are viewed in that way, it is not possible to say simply that FrExp is a measure of the friend’s influence any more than it is to say that FrExp is a measure of ego’s criteria of choice. Certainly FrExp helps explain EdExp, but it seems itself to be both cause and effect of ego’s characteristics. FOOTNOTES

' Given the sharp differences between whites and blacks, this should provide a conservative (that is, higher) estimate of a random-pairs correlation. There should be more homogeneity within each race than within the total school population, and thus the coefficient produced should show more similarity among random pairs than would be the case if blacks and whites were both used in a single analysis. * Technically, in this and all analysis in this chapter, the friend used in the analysis is the first codable friend named. Not all of the names the boys gave us could be found in the sample, and in some cases, although the boy was in the original sample, we had no questionnaire from him. Thus, if such a boy were listed in the first position, for instance, the respondent’s record had to be treated as if he had listed no one in that position. As a result, even in the first-named position in Table 7.1, the frequency is less than the total sample size in each cohort. 3 The correlations between ego and friend on EdAtt are .437 and .386 for those who are and are not still friends, respectively; for OccAtt, the comparable coefficients are .486 and .257. 4 Of the first-named friends, overall, 59% were reciprocating friends, 82% were best friends, overall, and 37% had been friends for at least three years (the last figure being based on ninth

and twelfth graders only). In all cases, these percentages dropped for the second and third named friends.

84 PEER INFLUENCE

CHAPTER VIII

SUMMARY, SYNTHESIS,

AND INTERPRETATION The earlier chapters have been devoted to the analysis of factors believed to be

associated with varying levels of educational and occupational expectation and attainment. The point of departure was the basic model of Duncan. It was shown that the graduate cohort in the present study exhibits patterns of educational and occupational attainment similar to those of Duncan’s national sample of young men, the level of occupational attainment in both studies being largely a function of the

level of educational attainment, with father’s occupation also having some direct influence. Educational attainment in both cases is most strongly influenced by IQ but also by father’s occupation and education. In the case of the older in-school cohorts, the basic ambition model has a very similar form. The boy’s educational expectation is the most powerful source of explanation of his occupational expecta-

tion, but his father’s occupation has some direct influence as well. The boy’s educational expectation is also most strongly influenced by IQ, although his family’s characteristics, and especially his father’s education, have significant effects. Throughout the study there was much greater initial success in explaining the ambitions of older than of younger boys. It seemed most reasonable to conclude that the kinds of dependent variables considered here have too distant a reference

to the younger boys, and the means for the achievement of such goals are too unfamiliar to them to provide meaningful answers to our questions. The problem thus seems to lie in the appropriateness of the dependent measures themselves. Throughout the further analysis, the finding from the basic model is repeated so far as occupational expectation is concerned. In all of the elaborated models, educa-

tional expectation is by far the strongest source of explanation of occupational expectations. In the twelfth grade there is a consistent tendency for father’s occu-

AMBITION AND ATTAINMENT 85

pation and son’s grades to have modest direct effects on occupational expectation, and in the ninth grade father’s education has such an effect, but none of these is

ever as much as one-fourth as great as the effect of educational expectation. Equally striking, in none of the elaborated models does the inclusion of additional variables appreciably alter the paths to OccExp in the basic model. The same general pattern is found in occupational attainment in the graduate cohort. The OccAtt-EdAtt path is by far the most powerful in the basic model, and neither it nor any of the other paths to OccAtt in the basic model is altered very much by the addition of other intervening variables. Moreover, the coefficients of determination of OccExp and OccAtt in the basic models are not altered very much

by the addition of any of the intervening variables.

Because of these general outcomes of the previous analysis, this summary focusses on the two older in-school cohorts and the graduate cohort, and is concerned solely with the findings relevant to educational expectation and attainment. Explanation and Explication Two analytic strategies guided the present study. One was longitudinal, focussing on the differences across cohorts and viewing the changes from youngest to oldest cohort as indicative of probable changes over time. This will be summarized in the next section. The second strategy was explicative. It involved the search for bases of explana-

tion of the relationships between the exogenous variables and educational and occupational outcomes, expected or actual. This was done by means of various _ intervening variables. As suggested in Chapter I and clearly shown in earlier analysis, such attempts at explication have also had the effect of increasing the explanatory power beyond that of the basic model. Table 8.1 summarizes the outcome of the explicative strategy in the ninth and

twelfth grade cohorts; its two panels present the path coefficients and the coefficients of determination for the EdExp portions of the ambition models from the ninth and twelfth grades.’ Only the direct paths to EdExp are presented there; the paths to intervening variables are not reported. The outcomes of the analysis can be viewed as explicating the effects of the exogenous variables and also as explaining EdExp. The last row in each panel, which reports the coefficient of determination for each model, is most relevant to the latter perspective. The results from a purely

explicative point of view, however, may be appreciated by noting in the rows associated with the individual exogenous variables the degree to which the paths have been reduced as one moves from the basic model to any particular elaborated model.

To turn to the ninth-grade panel first: the characteristics of the elaborated models are shown to be significant from the two perspectives, explanation and explication. All the elaborated models have coefficients of determination that are higher than that for the basic model. The improvement in explanation ranges from an increase of 3.3% in the Grade model to 7.6% in the Grade-Friend model. The alteration of direct paths is also apparent, especially in the EdExp-IQ path, which is reduced from .354 in the basic model to as low as .132 in the Grade-Fate model.

All of the elaborated models, in fact, reduce that path by more than one-third.

86 SUMMARY, SYNTHESIS, AND INTERPRETATION

Table 8.1

Summary of Ninth- and Twelfth-Grade Models, EdExp as Dependent Variable MODEL

Independent Variables Grade- Grade- Grade-

| Basic Brade Partic Fate — Friend 9th Grade

IQ 354 219 .219 132 = .190 Sib -.028 025 030 006 38.057 |

FaEd 264 184 ~~ 171 190 = .177

FaOcc 115 089 .070 .056 056 Grade 306 = .228 296 ~~ .215

Partic .169 Fate -.252 FrExp .242

Coeff. of Determination 349 382 402 430 = .425 12th Grade

IQ 370 =3©.160 ~——«.173 129 ~~ .145

Sib 128 -112 -094 -113 ~ -.098

FaEd .223 173 154 157 .151 — FaOcc 167 151 147 3.152 ~—— 124 _ Grade , 406 3.317 .405 337

Partic .198 Fate | -.128 FrExp 189

Coeff. of Determination — 378 484 314 10 499 There are also large reductions in the paths from father’s education and occupation,

perhaps the most impressive being the reduction by more than one-third of the EdExp-FaEd path in the Grade-Partic model.”

In the twelfth-grade panel, these results are also found, but they are even stronger.” The coefficient of determination is increased over the basic model from 10.6% (in the Grade model) to 13.6% (in the Grade-Partic model). The IQ path is also drastically reduced here, as in the ninth-grade cohort; in all models it is reduced by more than one-half and in the Grade-Fate model by almost two-thirds. And, again, the paths from FaOcc and FaEd are smaller in the elaborated models, the largest reduction being in the EdExp-FaEd path. _

It is apparent that the most powerful intervening variable in both cohorts is Grade, although Partic, Fate, and FrExp all have additional independent effects on both the coefficient of determination and the size of the several paths. It is also

AMBITION AND ATTAINMENT 87

true that in both cohorts the exogenous variable most strongly affected by the inclusion of the intervening variables is IQ. In the basic model in both cohorts IQ is the most potent source of explanation of EdExp. The magnitude of the EdExp-IQ path is greatly reduced in both cohorts, especially by the inclusion of Grade. The EdExp-FaEd path is the next largest in the basic model, and it is also considerably reduced in at least some of the elaborated models. The analysis has thus been successful to a notable degree with respect to both goals. As much as one-half of the variance of EdExp is explained by the models,

and the intervening variables do serve to explicate the relationship between the exogenous variables and EdExp. It is equally true and equally important, however, that none of the intervening variables serves to reduce the effect of the exogenous

variables on EdExp below a statistically significant level in either cohort. (The EdExp-Sib and EdExp-FaOcc paths are not significant in the ninth grade in the basic or the elaborated models.) In the ninth-grade models IQ and FaEd continue to

have a direct effect on EdExp, as do all four exogenous variables in the twelfth grade. Thus it cannot be claimed that the analysis has wholly explained the rela-

| tionship between the exogenous variables and EdExp, although it has explaineda considerable portion of it. A similar summary of the analysis of the graduate cohort is presented in Table 8.2, the dependent variable being EdAtt rather than EdExp.* Here again we find evidence of both increased explanation of the variance in EdAtt and the explication of the effects of the exogenous variables. The coefficient of determination is increased by one-half in the Grade-Partic model. Also as in the EdExp analysis, the ‘EdAtt-IQ path is most strongly affected by the introduction of intervening variables, reduction being greatest in the Grade-Friend model. In fact, that is the only path that is appreciably altered by the intervening variables. The EdAtt-FaEd path is non-significant in all models, and the EdAtt-FaOcc path remains equally strong, no matter what intervening variables are introduced.

Table 8.2 Summary of Graduate Models, EdAtt as Dependent Variable MODEL

Grade- Grade-

Independent Variables Basic Grade Partic Friend

IQ .249 225 .215 178 FaEd 070 045 018 011

FaOcc 249 215 .209 .220 Grade 398 333 356

Partic FrEd} 272 135

- Coeff. of Determination 341 446 313 436 7 88 SUMMARY, SYNTHESIS, AND INTERPRETATION

The outcome of the combined analysis summarized here is to move toward a clarification and specification of the factors involved in the educational attainment of young white males. All exogenous variables have been shown to be important in explaining the process, although FaEd appears to influence goal-setting more than attainment, and FaOcc seems to act in the reverse fashion. Given the lack of data for the graduates, it is difficult to comment on Sib with confidence, but the data in , Table 8.1, together with Duncan’s earlier findings with respect to EdAtt, would lead one to believe that both expectation and attainment are influenced by size of family. By far the most powerful exogenous variable, however, is IQ. Although its influence on EdExp is sharply reduced by the introduction of intervening variables, the direct paths involving it remain strong. In addition, the most powerful intervening variable is Grade, whether EdExp or EdAtt is used as a dependent variable. Most simply put, IQ, Grade and FaEd are most effective in explaining EdExp; IQ, Grade and FaOcc are most effective in explaining EdAtt. These patterns are most clearly found in the ninth-grade and graduate cohorts, the twelfth grade being intermediate.

On the basis of these findings, one can offer some general observations on the process of goal-setting and attainment. Evidently the boy’s intelligence exercises a strong influence on his ability to perform in school, which, in turn, affects his goals

and his ability to achieve them. His social origins also influence his goals and attainments, his father’s educational level influencing his educational goal independent of his own ability and performance, and his father’s occupation influencing his ability to reach his goal, independent of all of the other factors. Thus, although

it may be argued that the major source of influence on a boy’s ambition and attainment is his own ability as it is expressed through academic performance, that is far from the whole picture. In the first place, the boy’s academic performance

(i.e., Grade) is not wholly dependent on his IQ, FaEd also making a significant contribution to an explanation of Grade in the twelfth and graduate cohorts, and all four exogenous variables doing so in the ninth grade. Second, even when Grade is taken into account, the other exogenous variables significantly influence EdExp and EdAtt. Finally, participation, fatalism, and friend’s expectations or attainments are important sources of explanation over and above IQ and Grade.

The pattern of relations shifts in the analysis of expectation as the boys grow older. The educational expectations of the sixth-grade boys are not very fully explained by any of the variables (IQ and FaEd explain the most), whereas more variables contribute in the ninth grade, and in the twelfth even more. Thus the models are more effective the older the boys, and this can be seen in both the coefficients of determination and the size of the path coefficients. There seems to be an increasing tendency for the boys to shift from dependence on fathers as models of educational goals to a greater reliance on their own abilities and performance on their families’ capacity to support their educational desires. Such longitudinal patterns may be inferred, although the present study does not contain data wholly appropriate for the purpose of longitudinal analysis. In the next section, however, a form of analysis is presented that goes somewhat beyond inference across cohorts.

AMBITION AND ATTAINMENT _ 89

A Synthetic Cohort Model

Thus far, the discussion has indicated that Grade is the most powerful source of explanation of both EdExp and EdAtt. Also, there is an increasing tendency for a boy in the later years of public school to take his own ability and performance and his family’s economic position into account in setting goals. This seems realistic in that ability, performance, and economic position most fully influence his educational attainment. Such an interpretation assumes, in effect, that in this study the younger cohorts offer a reasonable basis for estimating what the older cohorts were like in the past. That is, the ninth- and twelfth-grade data are assumed to resemble

what would have been collected from the graduates six and nine years earlier. Similarly, it is assumed that the graduate data are like data that could be collected from the in-school cohorts six years after they graduate. The foregoing discussion has thus treated the data from the four cohorts as if they were taken from a single cohort at four points in time. In this section that assumption is more explicit and it is used to generate a longitudinal model of the process of educational attainment. The basic theorem of path analysis (see Duncan, 1966) may be written:

ri = > Pig "iq? q

where i and j refer to two variables in the model and q refers to all variables from which direct paths lead to variable i. Thus the correlations between all pairs of variables in the model may be expressed as a sum of the products of other correlations and paths. Conversely, all paths may be derived from a knowledge of the matrix of correlations among all variables in the model.* This means that all one

needs to construct a model of the type used in this report is a matrix of the

correlations of the relevant variables. To the extent that the samples are comparable, therefore, it is possible to build models of the type we have been using from data taken from different samples. These are called “synthetic cohort”? models. Here the basic problem, of course, is that it is usually quite difficult to demonstrate the comparability of the samples. Even if that can be done, problems of sampling error remain. Consequently, such models must be used with care. On the other hand, they can be very useful in illuminating the implications of interpretations of partial data sets for more extended analysis.

The model to be constructed here again is confined to the ninth- and twelfthgrade and the graduate cohorts because of the weak outcome in the sixth grade. The model must also be limited to those variables for which there are adequate data to complete the correlation matrix, or at least that part of it which represents the relationships to be included in the model. For instance, although fatalism was shown to be a meaningful source of influence on expectation, there is no basis for

, estimating the relationships between Fate measured at one point in time and Fate measured at another point in time, or between Fate and many other variables measured at another point in time. Fortunately, however, most of the variables that

90 SUMMARY, SYNTHESIS, AND INTERPRETATION

have been shown to be most important in the single cohort models are amenable to the kind of analysis proposed here. The network of paths in the model (Figure 8.1) is omitted for simplicity’s sake,

but two dotted lines represent missing paths. That is, the EdAtt-Sib and EdAttJrExp paths cannot be computed in this model, but all other paths can be. The

model thus implies that a boy’s grades in junior high school (JrGrade) depend on | the four exogenous variables; his educational expectation in junior high (JrExp) depends on his grades and the exogenous variables; his senior high-school grades (SrGrade) depend on his junior high-school grades and expectation and the exogenous variables; his senior high-school expectation (SrExp) depends on all preceding variables; and his educational attainment depends on all preceding variables except Sib and JrExp. Although the two exceptions occur because of lack of data, they do not seem unreasonable since they imply that the effects of Sib and JrExp are built into the flow of influence prior to the last step in the model. In order to compute the paths in such a model, it is necessary to obtain correlations between all pairs of variables which are to be connected by straight or curved arrows (paths or correlations). Table 8.3 presents those correlations. It will be recalled that the exogenous variables were measured in all of the in-school cohorts and all but Sib were measured in the graduate cohort. It was thus necessary to decide on the most suitable source of the correlations for present

purposes. A review of the coefficients reported in Chapter III shows that the FaEd-FaOcc coefficients are almost exactly the same, and the FaEd-IQ coefficients are very similar. The FaOcc-IQ coefficient, however, is much higher among the ninth graders than among either the seniors or the graduates. Also, although the measures involving Sib are available for just the two in-school cohorts, the Sib-IQ

coefficient is much higher in the ninth than in the twelfth grade.° In the light of such variations it was decided to use some more stable estimate and one that at the

same time would be more indicative of the characteristics of the whole sample. Since the exogenous variables were viewed as functioning largely as antecedent sources of influence on the variables to the right in Figure 8.1, only data from the in-school cohorts were used, the correlations among the exogenous variables thus being based on the combined ninth- and twelfth-grade samples. The other variables in Table 8.3 are in each case taken from a single cohort. In many cases, in fact, there is only one possible source of a coefficient; for instance, all correlations involving Ed Att can be computed only on the graduate cohort. Most of those coefficients have been used in previous analysis, but there are exceptions: for instance, the EdAtt-JrGrade correlation is based on the graduates’ grades in the ninth grade.

Correlations involving junior and senior high-school measures (columns 4

through 7) often have two or more possible sources. The JrGrade-JrExp correlation, for example, was actually computed from the ninth-grade cohort data, but a similar measure was available from the twelfth grade. In the latter case it is a correlation between the seniors’ recollections of their expectations when in the ninth grade and their ninth-grade grades. The general rule followed was to use contemporary rather than retrospective replies and to choose the most reliable measure possible.’ As a result, in the case in point, the JrGrade-JrExp correlation from the ninth-grade

AMBITION AND ATTAINMENT 91

eH Oy Ont, 0 S T Toe mt < ON Oo ™m™ AN ™m~ M1

aa)

oF iw, mM com~wr On = en Fm mm te WO ON on ONTENTS TONM

© ; o= o

SI Anmooste

a}

ODOn FN Gi nannaost NY i

co) naa a [xSasa NANT

= “Tart ~—



=)

aie

5 18

on Bl EWP RRGS

oe |OTnNAYTS

>> Cae Ra — Suni

a.

Oo ==[1 COOOM sh oe

v7ts =Fig- | ot

a} D

= =)

\S)

7

S| so

oO ©OQ oO ms]

LL, e

2 00“ e2rs . 7p) &

s> 3

7 a)

=

e)

a w © = cas} © Su Bas Oo One Soa o ot ses mt CF) fFasVPRysl| Cy me ee NN *Z oO @

92 SUMMARY, SYNTHESIS, AND INTERPRETATION

|\ ,»5 | \ = | \ 2 \ S | “ :7)3 |||

_ |. | | |

|

OO

AMBITION AND ATTAINMENT 93

cohort was used, instead of the one from the twelfth grade. In some cases, however,

it was necessary to resort to retrospective replies and to one-year grade records rather than those based on two or more years. This was the case with the correlations of JrGrade with SrGrade, SrExp and EdAtt, of JrExp with SrGrade and SrExp, and of SrExp with EdAtt. As a result, those six correlations are the most questionable part of the matrix.®

The path coefficients for the full model (Figure 8.1) were computed by using the coefficients reported in Table 8.3. Those coefficients, together with the coefficients of determination, are reported in Table 8.4. In effect, the first two steps in the model (the paths to JrGrade and JrExp) are simply another version of the ninth-grade cohort’s Grade model for EdExp.”? From that point on, however, a more synthetic quality enters the model and it has some disturbing characteristics, the most disturbing outcome of all being the pair of strong negative paths involving JrGrade (SrExp-JrGrade and EdAtt-JrGrade). The problem in this case seems to be the high JrGrade-SrGrade correlation (.824). When two independent variables are so highly correlated, any multivariate solution using them is unstable, and the resulting coefficients are questionable. Less obvious but equally disturbing are some of the paths from the exogenous variables to SrGrade; in particular, the SrGrade-Sib and SrGrade-FaOcc paths, though seemingly of moderate size, have signs that are the reverse Of what one would expect. Given the strong JrGrade-SrGrade correlation,

and the very strong resulting SrGrade-JrGrade path, the meaning of the other coefficients of paths to SrGrade is at least doubtful. The basic difficulty with the model seems to be that it includes pairs of measures that are almost redundant. This is especially true of the two grade measures and the two measures of expectation. The correlation of the expectation measures has not led to quite so obvious a problem because of the missing EdAtt-JrExp path, but the

conceptual problem is the same. Thus, although the model appears to be very powerful, at least as to the size of the coefficients of determination, its internal structure is not very meaningful. To gain greater clarity, two more delimited models were constructed, based on the same set of data. Both models acknowledge the problems just noted by deleting _

a large number of paths. All paths from the exogenous variables to SrGrade and SrExp were deleted and the SrExp-JrGrade and EdAtt-JrGrade paths were also removed. In effect, this is to argue that, although JrGrade and JrExp are influenced directly by the exogenous variables, SrGrade is affected directly only by JrGrade and JrExp, and SrExp is directly influenced only by JrExp and JrGrade. (The effects of exogenous variables on SrGrade and SrExp are thus seen as flowing through JrGrade and JrExp.) Furthermore, the only effect of JrGrade and JrExp on Ed Att is seen as flowing through the intervening variables. The two models differ in their representation of the sources of EdAtt. The first and more radical model (here called A) views EdAtt as wholly a function of the boy’s characteristics, specifically

SrGrade and SrExp, whereas the exogenous variables and JrGrade and JrExp are viewed as only indirectly affecting it. Model A assumes that the direct effects of the exogenous variables are limited to the early years of the boy’s life. The second model (called B) presents EdAtt as a function of his characteristics (SrGrade and SrExp) and also of the exogenous variables. This is, in effect, to argue

94 SUMMARY, SYNTHESIS, AND INTERPRETATION

GetsCo |

CES or OT MH qm 00 \O oO-[~ a—N ~ +wrhm 4 Oo oO ¢ OQ

or t § 98) is st ry ¥,i WN

0nH mM e © o = On . = = + + = = hm, ; 01S oo 2/3 & oO

(=) ; = < aoY £ — 2 YQ = 2bi 8 o 2. > Ld DN +s > 33 | _—

7

uj +— PY

a2) = 3 7 %S @

= x

= )oO — Cc @ / > rm ; 1.

a

ik = >< c= ~ _— a5)~~ ps5 s PCJ)om 2 o—

=

oO —

¢& ae 2A Ors CD) oO~ S

= = < tr

a2 sc o|3 =re ™ \O oO es D a,aS P| Qo '7‘ x_ ool az 2 18 a— 3) oe, oO “= = D O a Wm ' § §' WY } OJ NAN — O Ly = c

om}

=}

foes

oon a) ro “almy ouc' t | ' 86

OFS oo EN eek —]

Sy a23mN SS].