UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Education, orderly work careers, and organizational participation: a replication and extension of Wilensky's… Wiebe , Peter Michael 1972

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


831-UBC_1972_A8 W53.pdf [ 9.3MB ]
JSON: 831-1.0101717.json
JSON-LD: 831-1.0101717-ld.json
RDF/XML (Pretty): 831-1.0101717-rdf.xml
RDF/JSON: 831-1.0101717-rdf.json
Turtle: 831-1.0101717-turtle.txt
N-Triples: 831-1.0101717-rdf-ntriples.txt
Original Record: 831-1.0101717-source.json
Full Text

Full Text

pcuojfi Hi cio©& no) <=uti>' EDUCATION, ORDERLY WORK CAREERS, AND ORGANIZATIONAL PARTICIPATION: A REPLICATION AND EXTENSION OF WILENSKY'S MODEL by PETER MICHAEL WIEBE B. Chr. Ed. Canadian Mennonite Bible College 1967 B.A. University of British Columbia 1969 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Arts in the Department of Anthropology and Sociology We accept this thesis as conforming to the required standard, THE UNIVERSITY OF BRITISH COLUMBIA Ap r i l , 1972 In presenting t h i s thesis in p a r t i a l f u l f i l m e n t of the requirements for an advanced degree at the University of B r i t i s h Columbia, I agree that the Library shall make i t f r e e l y a v a i l a b l e for reference and study. I further agree that permission for extensive copying of t h i s thesis for s c h o l a r l y purposes may be granted by the Head of my Department or by his representatives. It is understood that copying or p u b l i c a t i o n of t h i s thesis f o r f i n a n c i a l gain shall not be allowed without my written permission. Depa rtment The University of B r i t i s h Columbia Vancouver 8, Canada Date ?y /97£ 7 1 ABSTRACT This thesis treats two problems simultaneously. It performs a replication test and extension of a given model of differential social participation, and a test for differing conclusions at three measurement levels. This involves secondary analyses of a two-cluster random sample. The Wilensky causal-sequence model of education, orderliness of work careers, and organizational participation is tested under several controlled conditions, i . e . , several subsamples, and i n a random sample of a small one-Company industrial city. That particular linear, independent-effects model i s not supported in any of the samples. Education is found to explain directly some of the participation. Furthermore, Wilensky's central relation, the primacy of orderliness as a predictor of participation, is not supported except in a sample very similar to his "middle-mass" sample. A cumulative inter-action model, involving education, orderliness of work careers, intra-generational mobility, and length of residence is proposed. Although the new model per se does not explain any more variation in social participation when compared to the model of the additive effects of those factors, the (cross-products) interaction terms do add a s t a t i s t i c a l l y significant amount of explained variation when combined with the additive effects in a single model of relationships. The analyses close with suggestions for further exploring new interaction arrangements and with the conclusion that the best f i t to the present data i s the model i i of combined additive and interactive effects. The dual assumptions of Davis, with regard to (1) non-differing conclusions about people at measurements levels, and (2) crudeness in statis t i c s being the same as conservatism, are directly challenged by the analyses. Under the rigi d application of the c r i t e r i a (1) of magnitude in correlation values, and (2) of consistency in the signs of relations, interpretative models di f f e r somewhat from measurement level to level, especially when interpreting the empirical relations for each indicator of organizational participation. Under-lying interaction among variables, tested by a linear model, enhances a crude s t a t i s t i c . l i t TABLE OF CONTENTS Page LIST OF TABLES v i i LIST OF FIGURES v i i i Chapter I INTRODUCTION: GENERAL PROBLEMS AND SPECIFIC TASKS 1 An Overview 1 The Substantive Problem 2 The Methodological Problem 4 Multivariate Techniques, Statistics and Decision Criter i a 10 II EXAMINATION OF THE WILENSKY CAUSAL-SEQUENCE MODEL 15 Work History and Differential Social Participation . . 15 The Wilensky Model 17 III THE CONCEPTUALIZATION AND THE MEASUREMENT OF THE VARIABLES 23 Education . . . . . 23 Orderliness of t'ork Careers 28 Occupational Mobility 35 Length of Residence 37 Indicators of Organizational Participation 39 IV DATA CHARACTERISTICS AND THE THREE SAMPLES 4 4 Source of the Data 4 4 The Three Samples 4 6 The Middle-Mass Sample 4 6 The Whole Sample 4 9 iv Chapter Page The Working-Class Sample 49 The Overlap in the Samples 49 V THE THEORY, CAUSAL IIODELS AND THE IIULTI-STAGE ANALYSIS . . 51 The Theory 51 The Middle-Mass Sample 59 The Whole Sample 72 Stage One: The Wilensky I'odel Tested on the Whole Sample 73 Substantive Interpretations 73 Dichotomous level 73 Ordinal level 77 Interval level 77 Methodological Interpretations 80 Stage Two: The Interaction Model Tested on the Whole Sample 82 The Theory of Interaction on Differential Organizational Participation 83 Substantive Interpretations 86 Affiliation 87 Involvement 89 The Working-Class Sample 90 The Scales 90 Stage One: The Wilensky Model Tested on the Working-Class Sanple 91 Substantive Interpretations . 91 Dichototnous level 93 Ordinal level 93 V Chapter Page Interval level 93 Methodological Interpretations 96 Stage Twos The Interaction Model Tested on the Korking-Class Sample 97 A f f i l i a t i o n 97 Involvement 98 Comparisons Across the Three Samples 100 VI SUMMARIES, CONCLUSIONS, AND LIMITATIONS 102 Summaries 102 Substantive Summaries 102 Methodological Summaries 104 Conclusions 104 Substantive Conclusions 104 Methodological Conclusions 108 Limitations I l l BIBLIOGRAPHY 114 APPENDICES 126 I RAW DATA FOR DICH0T0II0US CROSS-CLASSIFICATIONS . . 127 II VARIANCES AMD C0VARIANCES FOR ORDINAL CR0SS-CLASSIFICATIOITS 131 III UNIVARIATE DISTRIBUTIONS, POINT ESTINATES AMD MEASURES OF SOCIALITY 133 v l Page IV THE GENERAL DISTRIBUTION OF CASES IN THE THREE SAMPLES OVER EDUCATION AND OCCUPATIONAL POSITIONS . 145 V DATA TRANSFOPliATIOMS: SKEWNESS AND KURTOSIS, TRANSFORMATIONAL FORMULAE, AND THE EFFECTS ON PEARSON'S P. 147 VI "ORDERLY CAREERS Al© SOCIAL PARTICIPATION", AN ARTICLE BY HAROLD L. WILENSKY 153 VII THE DISPOSITION OF THE POPULATION AND THE SAMPLE . 173 v i i LIST OF TABLES Table Page I The Pearsonian Correlations Between Orderliness and Two Kinds of Education Measures 26 II The Number of Jobs in Recorded Work Histories 30 III Orderliness and Mobility in Detroit and Millport 34 IV Per Cent Persons with Orderly Careers by Occupational Mobility and Education 38 V Weighting Procedure and the Three Samples 45 VI The Test of the Wilensky Model in the Middle-Mass Sample at the Dichotomous Level of Measurement 60 VII Education as a Cause of the Spurious Relation Between Orderliness of Work Careers and Organizational Participation 64 VIII The Test of the Relation Between Orderliness of Work Careers and Organizational Participation: with Occupational Mobility and Length of Residence Controlled 67 IX The Test for Interaction Between Orderliness of Work Careers and Education on Organizational Participation: Somers D on Dichotomies 69 X The Test of the Wilensky Model in the Whole Sample at the Three Levels of Measurement: Phi, Tau(b), and Pearson's r 74 XI * • The Cumulative Interaction Model Tested in the Whole Sample for A f f i l i a t i o n and Involvement: Coefficient of Multiple Correlation. . . . . 88 XII The Test of the Wilensky Modal i n the Working-Class Sample at the Three Levels of Measurement: Phi, Tau(b), and Pearson's r 92 XIII The Cumulative Interaction Model Tested i n the Working-Class Sample for A f f i l i a t i o n and Involvement. Coefficient of Multiple Correlation . . . . . 99 v i i i LIST OF FIGURES Page Wilensky's Temporal Causal-Sequence Model 20 21 Alternative Causal Models The Ad Hoc Regression Through Means of Education for , Five-year Periods 30 The 'Average' Work History The Effect of Truncating an Independent Variable at i t s Extremes; Education and Orderliness of Work Careers . . . 48 Wilensky's Temporal Causal-Sequence Model 57 Occupational Mobility as a Cause of the Spurious Relation Between Orderliness of Work Careers and Organizational Participation 8^ Length of Residence as an Explanation of the Relation Between Orderliness of Work Careers and Organizational C O Participation J O The Interpretative Causal Model for the Whole Sample. 76 Path Diagram of the Four Causal Factors and Organizational A f f i l i a t i o n and Involvement: The Whole Sample ^9 The Cumulative Interaction Model 4^ The Interpretative Causal Model for the Working-Class Sample at the Dichotomous Level of Measurement The Interpretative Causal Model for the Working-Class Sample at the Ordinal and Interval Levels of Measurement. Path Diagram of the Four Causal Factors and Organizational A f f i l i a t i o n and Involvement: The Working-Class Sample . . ACKNOWLEDGEMENT I would like to express my appreciation to Dr. iiartin Meissner for the use of his data in the analyses for this thesis. CHAPTER I INTRODUCTION: GENERAL PROBLEMS AND SPECIFIC TASKS An Overview We shall deal with two problems simultaneously. The f i r s t problem i s differential organizational participation: Why do some people participate more in community organizations than others? We w i l l identify several factors. The second problem i s the compara-b i l i t y of results obtained from different levels of measurement. Will an analysis at the dichotomous level of measurement demand the same conclusions as analyses at the ordinal and interval levels? The study i s a two-stage, multi-sample, multi-level analysis involving: a replication test of a given causal model on a "middle-mass" sample; tests of the same model on a random sample and on a working-class subsample; and f i n a l l y tests of an alternative, inter-action model on the latter two samples. The methodological comparison w i l l be handled in the latter sections of the analysis, i.e., on the second and third sample. This comparison provides a fu l l e r exercise in the process of judgement. The study involves two causal models (stages), three samples, and three measurement levels. To explicate more fully the procedure to be followed, the study examines the Wilensky causal-sequence model - to be discussed presently - on a similar "middle-mass" sample at the dichotomous level only. Assuming that the causal stiucture has broader application, the study tests the same model on a random sample of an industrial community, Millport, and on a more homogeneous working-class subsample from that community. Alongside each of the latter two sample-analyses, an alternative causal model w i l l be introduced and tested. This model entails an expansion and a structural rearrangement of the Wilensky model. The latter two analyses shall be handled at three levels of measurement, through symmetric measures of relations between dichotomous, ordinal and interval scales. The Substantive Problem The sociological work done in the broad substantive area, usually designated social participation, is manifold. Some of the factors shown to affect the various dimensions of social p a r t i c i -pation or leisure-time activities are: occupational mobility (Curtis, 1960), social mobility and status of destination (Vorwaller, 1970), occupational prestige (Clarke, 1956; Hagedorn and Labovitz, 1968a), social class (Reissman, 1954), socio-economic status (Hodge and Treiman, 1968; Cramer, 1950), status crystallization or inconsistency (Lenski, 1956), occupational career patterns (Wilensky, 1960, 1961a), occupational or job characteristics (Hagedorn and Labovitz, 1968b; Meissner, 1971), job lay-off (Pope, 1964), race and religion (Lazerwitz, 1961), age (Axelrod, 1956), sex (Teele, 1962; Stuchert, 1963), family l i f e cycle (Harry, 1970), family participation (i.e., socialization: Anderson, 1943), the cumulativeness of leisure activities (Allart, 1958), the formal-3 informal participation correlation (Hay, 1950), length of residence or geographical mobility (Caplow and Forman, 1950; Scott, 1957; Zimmer, 1955), type of residential area (Zimmer and Hawley, 1959), spatial arrangement of residences and the environment (Whyte, 1953), the degree of urbanization and suburbanization (Sutcliffe and Crabbe, 1963), neighbourhood characteristics (Bell and Boat, 1957), parental aspirations (Hilander, 1969; Rehberg and Westby, 196 7), size of organizations (Indik, 1965), the functionality of participation (Gross, 1S61), social intelligence (Chapin, 1939), and one's self-conception and self-evaluation (Coombs, 1969). The above studies include a l l conceivable dimensions of the general term, social participation. Among the various factors shown to affect social p a r t i c i -pation, a large and heavily-emphasized portion of these can be sub-sumed under the labour-leisure problem (Wilensky, 1960), sometimes entitled the relationship of work and work-related activities (and status) to non-work ac t i v i t i e s . Within the context of the labour-leisure problem, the nature of a person's work history - as opposed to his status or class - is of interest in our study. And since our chosen area is the orderliness of a person's work career, for which l i t t l e literature is yet available, this study is essentially a replication and extension of the work of Wilensky (1960, 1961a, 1961b). Numerous are the dimensions of social participation and broad i s the range of causal factors, hence limits must be set. This 4 study w i l l limit i t s e l f to five indicators of participation in formal organizations. It includes the variables found in 'vilensky's study, namely education and orderliness of work careers, and w i l l be extended to encompass intra-generational occupational mobility and length of residence. Chapter II provides an extensive treatment of Wilensky's theory of and causal model for differential social participation. The Methodological Problem In the discussion of measurement levels a confusion exists over two related dimensions of the problem: (1) the collapsing or categorizing of scales, and (2) measurement per se. The collapsing of an interval scale, X, into a series of categories, X', often coincides with the presumed necessary reduction from an interval to an ordinal scale. Obviously however, there are instances where the reduction to lower measurement need not be accompanied by cate-gorization, or vice versa. An example of such would be a debate over whether a given 4-point scale is interval, ordinal or nominal. Although the extent of the relationship between the two dimensions has not been documented - whether i t be integral or not - i t i s precisely the combination of these two dimensions i n a substantive task that is of interest in the present study. And i t is our assumption that i t is precisely this combined procedure of cate-gorization and reduction that precedes much of empirical sociology. That i s , this procedure is very much a part of the "behind-the-5 scenes" and "before-the-presentation" activity of many studies. There-fore, i t s effect on substantive conclusions needs to be documented. The second problem to be examined in this study, then, is the comparability of results obtained from different levels of measurement, meaning thereby the combination of the dimensions of cate-gorization and reduction. This effort shall aid in our judgement regarding the substantive problem of differential organizational participation, but provides a self-contained problem. The problem is crucial for several reasons. F i r s t l y , with respect to measurement, the procedure taken in the study is typical of some work done in sociology, and the various statistics for measuring empirical relations in those studies are tied to measurement levels. Secondly, i f conclusions differ from level to level - when explicit decision c r i t e r i a are rigidly followed - then the choice of measurement level can result in different substantive conclusions. This is true even after the problem of different results obtained from statistics of varying degrees of "strictness" (cf. Wilson, 1970) for any one measurement level has been solved or controlled for. And thirdly, the examination of the problem can aid in the on-going debate between higher-powered, parametric statistics and lower-level, nonparametric s t a t i s t i c s . Davis (1971, p.3, 12f) suggests that the use of nominal and ordinal dichotomies and the test factor procedure"'" can obtain results as accurate as the higher-powered analyses. 1. Davis employs Yule's Q, which is a less s t r i c t s t a t i s t i c than phi ( 0 ) or per cent difference. Yule's Q does not ta l l y , against the hypothesis, pairs of observations which are tied on the dependent variable and untied on the independent variable: e.g., X^> and Y^ = Y^. Hence, the low values found i n this thesis must be seen in that lxght. Generally speaking, Q = 20 at the lower magnitudes. 6 ... i t also means that we can apply our tools to any set of data without having to assume sophisticated levels of measurement... iiany research experts X\?ould challenge this approach, but we believe that i t i s quite appropriate for beginners and indeed we would hazard the hypothesis that our procedure w i l l usually give the same substantive  conclusions as more sophisticated ones; i.e., you w i l l generally draw the same inferences about "people" with these techniques as with fancier ones. (Davis, 19 71, p.3) It is Davis' expressed intention to educate beginners in the tools of elementary survey analysis and thereby "to provide a sound foundation for advanced study" (1971, p . l ) . However, the argument i s precisely with such seeming soundness, and therefore also with his assurance that, i f anything, "crudeness i s the same as conservativism" in s t a t i s t i c a l analyses and conclusions. F i r s t l y , although we may not go "off half-cocked", we may also seldom find anything empirically important with such crude or conservative techniques. The difference between crude and less crude measures, for sociology, may just be the difference between negligible and non-negligible findings. Secondly, i t is by no means certain that estimates w i l l be conservative. Sonquist (1970, p.8) demonstrates that a curvilinear relationship in a bivariate case is usually an indication of interaction in the multivariate case. That i s , multicollinearity and interaction between two independent variables on a third variable w i l l appear as a curvilinear relation between the dependent variable and either independent variable taken singly. Therefore, a linear, additive s t a t i s t i c a l model applied to an underlying 'true' model involving interaction between some variables w i l l enhance a crude, and 2 especially less s t r i c t , s t a t i s t i c . In such a situation, phi and 2. Davis does deal with interaction and provides for a s t a t i s t i c a l method to handle i t , namely, the "differential minus partial" value. (Davis, 1971, p.95ff) 7 especially Yule's Q computed on dichotomies w i l l appear larger or less conservative than the Pearsonian r computed on the original distributions. Therefore, although Davis' intention is admirable, and possibly the best approach for beginners yet devised, this study may demonstrate some of the unintended deceptions implicit in i t . Our study comprises a warning not to rest with crude measurement and s t a t i s t i c s . Blalock (1964, p.34f, 94) argues that what is needed is a reconsidered weighing of disadvantagest distortions stemming from crude categories, arbitrary cutting points, and the depletion of cases versus distortions stemming from an inadequate attainment of the usual regression-correlation assumptions (see Nygreen, 1971 for such assumptions). While i t is true that the use of Pearsonian correlations and regression coefficients is not s t r i c t l y legitimate unless one has at least an interval scale, i t may turn out that i t is no more misleading to make use of dubious assumptions about level of measurement than i t is to make use of data involving arbitrary cutpoints or ordinal scales that obscure differences i n amount of variation. (Blalock, 1964, p.34) Further on in that work, he .writes: In most instances i t w i l l not be possible to categorize X and s t i l l retain differences i n variation in X which are commensurate with the differences that might have been obtained with less crude measurement techniques. Generally speaking, the greater the number of categories that can be retained, the more likely that one can measure such d i f f e r -ences accurately. The simplicity and other obvious advan- tages of the 2 X 2 table should not blixd us to its defects. (1964, p.124) There are several studies which relate to our problem. For instance, i t has been shown (Blalock, 1964, ch.IV) that the cate-8 g o r i z a t i o n of a v a r i a b l e reduces i t s \'ariation and therefore minimises the Pearsonian c o r r e l a t i o n based on i t . This i s of course l i m i t e d to l i n e a r s t a t i s t i c a l models calculated f or l i n e a r r e l a -t i o n s ; or to c u r v i l i n e a r s t a t i s t i c a l models f c r c u r v i l i n e a r r e l a t i o n s , and so on. I t has been shown (Boyle, 1970) that path analysis on o r d i n a l data, using dummy v a r i a b l e s , varies l i t t l e from path analysis on i n t e r v a l data. I t has been shown (Baker, et. a l . , 1966) that the sampling d i s t r i b u t i o n of the t s t a t i s t i c varies l i t t l e under gross o r d i n a l transformations of i n t e r v a l scales,but that t as a d e s c r i p t i v e s t a t i s t i c , does va -y considerably. The behaviour of many of the standard s t a t i s t i c s lias been documented (Rutherford, 1971) under the following c o n t r o l l e d tabular character-i s t i c s : underlying covariance, matched or mismatched marginals, and 3 equal or non-equal row and column 'n's. However, i n none of these studies are d i f f e r e n t s t a t i s t i c s d i r e c t l y compared, as these s t a t i s t i c s are thought to be applicable to the presumed d i f f e r e n t l e v e l s of measurement. Only Rutherford provides f o r such a comparison but he deals only with the number-of-categories dimension of the "measurement problem" and does not r e l a t e i t to a substantive problem. 3 . These tabular c h a r a c t e r i s t i c s need elaboration. The matching of marginals r e f e r s to the number of categories i n the v a r i a b l e s : the case of matched marginals ex i s t s when R = C (the number of rows equals the number of columns). The second c h a r a c t e r i s t i c deals with the si z e of each category and two comparisons are p o s s i b l e : (1) n.^ = n.~ = n.^ = etc., where the categories of a s i n g l e v a r i a b l e j contain equal frequencies; and (2) n.^ = n^ , n ^ = n , etc., where the f i r s t category of X i s matched i n s i z e to the f l r s X category of Y, and so on. Unity or perfect c o r r e l a t i o n - a s s o c i a t i o n i s optimally possible only i n the case where marginals are matched (R = C) and where they are equal ((2) above). What i s at s take i s : (1) How much i n f o r m a t i o n about the v a r i a b l e s i s l o s t or gained by s h i f t i n g measurement? and (2) How are the c o n c l u s i o n s a f f e c t e d by that l o s s or gain? Put i n other words , can one r e a l l y t r u s t r e g r e s s i o n a n a l y s i s on o r d i n a l data as some have b o l d l y attempted? ( B o y l e , 1970) And can one r e a l l y t r u s t o r . " i n a l a n a l y s i s on o r d i n a l data? I n one, the problem i s a seeming mismatch of mathematical p r o p e r t i e s of a s c a l e to s t a t i s t i c a l o p e r a t i o n s ; i n the o t h e r , the l o s s of i n f o r m a t i o n r e s u l t i n g from a r e d u c t i o n to l e s s s t r i n g e n t measurement assumptions and the a p p r o p r i a t e s t a t i s t i c . The purpose o f t h i s paper i s to examine t h i s i s s u e on the rocky ground of a c t u a l d a t a , i n the context of the subs tan-t i v e problem of d i f f e r e n t i a l o r g a n i z a t i o n a l p a r t i c i p a t i o n . I t deal s w i t h the methodo log i ca l problem at the locus where i t i s most c r u c i a l . I t i s most c r u c i a l to s u b s t a n t i v e c o n c l u s i o n s to know how m e t h o d o l o g i c a l choices a f f e c t e d those c o n c l u s i o n s . A dilemma a r i s e s from combining m e t h o d o l o g i c a l and s u b s t a n t i v e problems. Can one ever be r e s o l v e d w h i l e the o ther i s prob lemat ic? Given the presence of a m e t h o d o l o g i c a l problem, the s u b s t a n t i v e problem may not be r e s o l v e d due to the f a c t that d i f f e r i n g conc lu s ions from d i f f e r e n t measurement l e v e l s , per se , cannot answer the q u e s t i o n : Which i s the r i g h t s u b s t a n t i v e conc lu s ion? The r e s o l u t i o n must then l i e i n the support of measurement assumptions. However, i t i s worth d i s c o v e r i n g i f conc lus ions "about peop le " d i f f e r . 10 Multivariate Techniques, Statistics and Decision Criteria A problem to be examined is the comparability of results obtained from different levels of measurement. Will an analysis at the nominal-dichotomous level of measurement demand the same conclusions as analyses at the ordinal or interval levels? The three s t a t i s t i c s : Pearson's r, Kendall's tau(b) or T^, and phi or 0 w i l l be compared in the conclusions they yield about a substantive problem. Anderson and Zelditch (1968, p.l60f), Hays (1963, p.651f), and others have demonstrated that, for the 2 x 2 case, r = = 0. This argues for the use of these particular statistics in a comparative study. The appropriate multivariate techniques employed in the analyses, applied to the Wilensky causal-sequence model of differential organizational participation (1960; 1961a), w i l l than be correlation (or path) analysis and structural equations (Blalock, 1964; Nygreen, 1971), ordinal association and s t a t i s t i c a l partialiing (Hawkes, 1971; Somers, 1962; Coleman, 1964), and nominal association and s t a t i s t i c a l 4 partialiing , respectively for the three s t a t i s t i c s . Each s t a t i s t i c w i l l be employed on data which according to Stevens (1946) are at the appropriate measurement level. 4. For the 2 x 2 x 2 cases, instead of test factor partitioning, (Lazarsfeld, 1955; Davis, 1971), we used s t a t i s t i c a l partialiing, which can be shown to parallel partitioning closely except where the skewness of the respective marginals differ considerably — that i s , where the skewness of trie two variables of a bivariate distribution differ considerably. The choice of partialiing over partitioning i s desirable because i t eliminates the possibility of discrepancies in results stemming from loss of information (i.e., variation) due to the partitioning multi variate technique (rather than loss of information due to measurement level). Furthermore, i t has been shown (Rutherford, 1971, figures 5 and 6) that r„ T^, and 0 behave in a parallel fashion when marginals are increasingly sVewed. Finally, the problem of sta-t i s t i c a l partialiing versus physical partitioning needs to be more full y explored. 11 The present study raises many related methodological problems - some of which are footnoted in the flow of the discussion. Many of these however must simply be skirted and the reader must be directed to those f u l l e r and better treatments of the problems. However, a few of these demand brief treatment, so that the reader becomes directly aware of them and of the fact that they imp:' age directly on our study. Specifically, the present study acknowledges the problems of the assumptions to 5 path analysis or correlational analysis, the problem of the determination and s t a t i s t i c a l handling of interaction effects, and the problem of s t a t i s t i c a l partialling for ordinal data. Nygreen (1971, p.41) has created an extended l i s t of the assumptions necessary to path analysis. In order to meet the most stringent assumption in some reasonable way, that of Itomoscedasticity or equi-variance, we normalized our data by permissible interval transformations. The formulae employed towards that end, the measures of normality (skewness and kurtosis) derived, and the effect on the correlation coefficients are a l l given in Appendix V. The other assumptions of causal p r i o r i t i e s , interval level measurement, low multicollinearity, r e l i a b i l i t y , validity, linearity, and additivity must simply be made. 5. A path coefficient is a partial correlation coefficient. (Nygreen, 1971) Path analysis, for our purposes, is identical with the strategy of recursive systems of structural equations. 12 Sonquist (1970, chapter 2) elaborates on the pervasiveness of interaction in the social sciences. He also presents a technique for introducing interaction terms into an otherwise linear and additive s t a t i s t i c a l model. (Also confer with Stinchcombe, 1968, p.46) The strategy of performing multiplications between independent variables thought to interact in the data w i l l be employed in this study when the alternative, interaction causal model is introduced in the analyses. These interaction terms can then be fi t t e d into the linear model of path analysis. Furthermore, the identification problem and the determination of the "true'' interaction, so thoroughly discussed by Blalock (1965; 1966a; 1966b; 1967a; 1967b; 1967c: 1968, p.l78ff) on the status inconsistency problem, must be recognized but mostly bypassed. We w i l l proceed in the typical manner of assuming the "low-low" c e l l to be the base-line, calculating both main effects from that base-line, and deducing the interaction as a "remainder" — or in other words, interaction is " a perfect mathematical function of the difference between two effects". (Blalock, 1967b, p.305) There is a sense in which the very crucial problem of identification does not affect the present effort, Since we are basically concerned with comparing a given additive model to a new interaction model, we need not deal, at this point, with the exact nature or degree of that interaction. Blalock challenges the nature of interaction, not the existence of i t . Hawkes (1971) presents a sound argument for the use of s t a t i s t i c a l partialiing on ordinal s t a t i s t i c s . In this he follows 13 Coleman (1964) and Somers (1959; 1962). His article also supports the use of T, as the ordinal analog to the Pearsonian r. The latter D argument is based on the calculation of r from paired comparisons, bypassing the use of means entirely. In the following analyses we shall employ a simple set of cri t e r i a for making assessments very much like those implicitly employed in many empirical studies and, more to the point, almost exactly like those explicitly suggested by Davi* (1971, p.49) and Wilensky (1961). The two c r i t e r i a — magnitude and consistency — create a new temporary definition of "significance". In any set of relations, i.e., those involving five participation indicators, both c r i t e r i a must be met before significance can be declared. In any single relation only the absolute magnitude dimension can be applied. The c r i t e r i a are: (1) Magnitude: Any relation with a correlation or association value of greater than or equal to ± .10 w i l l be considered signi-ficant. ^  This absolute magnitude dimension, creating three possible relations: positive (+), negligible (0) , and negative (-), must 6. We deliberately choose the term "significant", because so much of the subsequent discussion hinges•around the use of the terminology. We wish to dispel confusion by openly stating this new definition of "significant". It can also be read, synonymously, as "at least minimally important" - in our concerned effort to maximize relations in the data. And since cur argument is with Davis 1 (and similar) arguments, we have chosen his level of magnitude (±.10) as our arbitrary significance level, alongside Wflensky's consistency dimension. Using Hay's formula (1963, p.651) for the 5% significance test for T^, we find that the significance levels are: for N = 177, >±.10; and for N = 498, >+,06. This s t a t i s t i c approaches normality quickly, making i t and the magnitude of ±.10 a very reasonable choice. Phi for N = 498 i s > ±.086; r for N = 498 is >, ±,085. 14 be p a i r e d w i t h another diir.+ n s i o n . Among p a r t i c i p a t i o n i n d i c a t o r s , we w i l l l ook f o r m a j o r i t i e s , i . e . , three out of f i v e measures above t h i s magnitude. (2) C o n s i s t e n c y : Summarizing down the f i v e i n d i c a t o r s of o r g a n i z a t i o n a l p a r t i c i p a t i o n , c o n s i s -tency i n s i g n ( i r r e s p e c t i v e of magnitude) i n four out of f i v e r e l a t i o n s w i l l be our second and c o n t r i b u t o r y c r i t e r i o n . A few examples of va lues can he lp to c l a r i f y the o p e r a t i o n of these c r i t e r i a . A n e g l i g i b l e r e l a t i o n i s one w i t h a va lue between +.09 and - . 0 9 ; such a n e g l i g i b l e f i n d i n g can however add to the c o n s i s t e n c y dimension i n our judgements. Under these two c r i t e r i a , the very minimum of support f o r a p o s i t i v e h y p o t h e s i s , i n v o l v i n g f i v e p a r t i c i p a t i o n i n d i c a t o r s , would be the f o l l o w i n g f i v e v a l u e s : +.10 ,+ .10 +.10, +.01, and - . 0 9 . A l t e r n a t i v e l y , e i t h e r of the f o l l o w i n g would i n d i c a t e non-suppor t : +.10, +.10, +.09, +.01, and - . 0 9 : or +.10, +.10, +.10, - . 0 1 , and - . 0 9 . In the f i r s t , the magnitude c r i t e r i o n i s not met; i n the second, the c o n s i s t e n c y c r i t e r i a . 15 CHAPTER II EXAMINATION OF THE WILENSKY CAUSAL-SEQUENCE MODEL7 Work History and Differential Social Participation In the various attempts to explain differential social participation, many researchers employed arguments about status and class, as was noted above. However, Form and Miller (1949), Lipset and Bendix (1952; 1967), Litwak (1960), and Wilensky (1960, 1961a) have suggested that certain dimensions of the work history could better explain differential social participation. In the case of Wilensky, the orderliness of a person's work history, or the degree to which his worklife could be said to be a career, was hypothesized to affect social participation, in the areas of secondary and primary social relations. To summarize his general theory, Wilensky stated that Durkheim's view was crucial, namely that: the central problem of the sociology of work [is] the effect of the division c f labour on social integration, or more precisely the conditions under -which work role, occupational association, and career are most and least effective as bonds of solidarity either within the workplace or between workplace and larger units, or both. (1961a, p.522) Behind this hypothesized relation of work careers to social participation l i e the theoretical arguments that an orderly career provides a cumulative experience of predictability of rewards - more important than immediate rewards and status - of economic sta b i l i t y 7. A copy of the a r t i c l e containing this model is found in Appendix VI. The relevant a r t i c l e i s : Harold L. Wilensky, 1961a, Orderly Careers and Social Participation, American Sociological Review 26: 521-539. 16 and opportunity, of motivation, and of the willingness to defer gratification and to invest time, energy and money i n the community. Orderliness has a spill-over effect; an economically secure person makes a career out of both work and social l i f e - a " l i f e plan" (Mannheim, 1940, p.56). Participation i n community l i f e is a natural extension of participation in the labour market; orderly and pleasant experiences provide motive and opportunity for the former. (Wilensky, 1961a, p.522) Perhaps i t is the case that those persons with solid work careers, who w i l l tend to be those with high statuses, attempt to further crystallize their status dimensions by the appropriate involvement in public l i f e . Leisure as a growing source of status is not a new idea. De Grazia (1962) gives some indication of the growing importance of leisure as status. Whatever the explanation for the relationships, Wilensky discovered support for his main hypotheses that orderliness of work careers explains differential social participation (secondary and primary) and that i t explains more than indicators of social status: income, occupation of father, and in part, education (1961a, p.539). More specifically, an orderly experience is associated with organizational activity, variety of informal contacts, and coherent participation patterns. (The latter refers to the fact that the friends form a tight c i r c l e , include work contacts and range across occupations.) In this thesis we are interested in the f i r s t of these many findings. Since we are interested in organizational participation .1 17 and w i l l deal with that area only, an earlier statement sums up our focal point. With respect to associational ties, educational background and work history seem to constitute a development sequence. The more education, the more motivation and opportunity for a career both at work and in voluntary organizations — and the higher educated men with orderly careers are apparently making a l i f e plan for both. (1961a, p.535) Wilensky then goes on to conclude with a discussion of the reciprocal relation between orderliness and participation, and the possible interaction between education and orderliness in their effects on social participation. He summarizes with an intriguing kind of interpretative model. Three considerations, however., suggest that the causal sequence i s : education •» orderly careers •* participation, whatever the interaction of these variables once the process is underway. (1961a, p.539) These three considerations are: (1) pre-occupational leisure training l i t t l e affects participation; (2) early socialization is unrelated to work history and so personality pre-disposition does not affect orderliness or participation; and (3) structural forces such as family l i f e cycle and economic crises explain variations by age and income in the central relation under consideration. Finally, the reason for a causal chain is based on the temporal sequencing of the variable; education precedes the work career, etc. The Wilensky Model It is not clear what Wilensky's causal model really i s . Even though he does plot a temporal sequence, he strongly hints at inter-18 action effects (p.533, 535, 539), and suggests reciprocal effects between orderliness and participation. He even suggests a further model— education as a cause of both orderly careers and participation (p.535). These several models are unique causal models about the nature of the. relationships. Because of this ambiguity, we w i l l cut this discussion short by simply assuming a temporal-sequence model (see figure 1) as the model to be tested i n i t i a l l y , recog-nizing that we cannot justifiably attribute i t to Wilensky as his sole conclusion. The reason for the arbitrary solution to a theoretical ambiguity i s , in part, simply an attempt on our part to begin somewhere reasonable in these analyses. It is true that, in his multivariate analysis, he must have made the implicit assump-tion of additive effects on social participation. That assumption is built into his technique of multivariate analysis. According to the model, orderliness of work careers intervenes between education and social participation. The structure of the model, i.e., the "intervention" of orderliness, becomes in point of fact a crucial test of the degrees of f i t of the model v i s - a-vis other linear additive models (Blalock, 1964, chapter 3). The causal-sequence model specifies positive relations between education and orderliness of work careers, and between orderliness and social participation (figure 1) and the relation betv/een education and participation is only indirect. The latter relation is mediated by orderliness; orderliness is an intervening variable, which once s t a t i s t i c a l l y controlled would eliminate the empirical 19 relation between education and social participation. An intervening variable does not theoretically invalidate the indirect relation. Such a causal sequence is sometimes called an explanation or inter-pretation model. On the theoretical level, education creates expectations of and opportunities for an orderly work career, and an orderly career yields the steady rewards for social participation. If the expectations created by education are not met by an orderly work career, then education w i l l not lead to social participation. We are differentiating this model from other alternatives pictured in figure 2 below. In the additive model, (a), education and orderliness have independent effects on participation. A theoretically different but empirically similar model is the join t -result model, (b). The crucial difference between i t and Wilensky's chain i s the theoretical validity of the orderliness-participation relationship. The specification model, (c), states that a certain level of education is necessary before the relation takes hold between orderliness and participation. The difference between this and g model (d), suppression, i s not totally clear. Davis (1971, p.100) insists on the distinction. . . . specification involves a difference between conditional relationships while suppression or inter-action is a general accentuation or reversal of a correlation. 8. Sonquist (1970, p.73) says that specification is one type of interaction model, hence our models in figure 2 do not entail an "interaction" model per se. We consider both specification and suppression as interaction models. 20 Figure 1. Wilensky's Temporal Causal-Sequence Model v ORDEPvLINESS OF WORK CAREERS 9. Harold L. Wilensky, 1961, Orderly Careers and Social . Participation, American Sociological Review 26; 521-539; see especially page 539. James Davis (1971, p.107) gives an excellent brief description of causal models. "Causal models are assertions ^hypotheses) about the presence, sign, and direction of influence for relations among a l l pairs of variables in a set." 21 Figure 2. Alternative Causal Models (a) ADDITIVE MODEL: Education + Orderliness .— Participation (b) JOINT-RESULT or SPURIOUS MODEL: +^___^.Education — Orderliness - - - - - - >-- Participation (c) SPECIFICATION MODEL: Education Orderliness > Participation (d) SUPPRESSION MODEL: Education- — ^ . (e) RECIPROCAL MODEL Orderliness. Education Participation Participation Orderliness (f) A l l relations: ^Education Orderliness (g) Any combination of these: Participation Legend: > a direct.and independent causal link __. _ _ _ _ _ > » a spurious link 'n' designates a necessary condition for another relation ^•^r-—interaction effect of two or more variables on / another ^\ a reciprocal or mutual causation loop 22 The reciprocal model, (e), has the appearance of the typical functional-historical logical model (cf. Stinchcombe, 1968, p.89f), but we w i l l not explore the possible similarities. Seyond these several theoretically-pure models are a l l the possible combinations of these, i.e., sociological data or Davis' twilight zone (Davis, 1971, p.96f). Wilensky confused several: he tested relations via an additive s t a t i s t i c a l model ((a), (b) or (a) of figure 2 ) ; dis-covered a temporal-causal chain; and speculated about both interaction and reciprocity ((c), (d) or (f) above). These various models w i l l not a l l be applied to the data. The main difference between Wilensky's chain and the others is the relation between education and participation, that is whether i t exists as a direct link or is negligible, and that fact provides a crucial test of the model. We need to remind the reader that education is a common measure of social status. Tentatively then, i f the relation between education and participation is found to exist as a direct link in our data despite controls, then we must at least re-open the question about the ascendancy of career patterns over status as an explanation of differential social participation. For instance, i f education interacts with orderliness (as in model (a) abo' r2), then perhaps we should at least, conclude that immediate social status interacts with work history to bring (.lout social participation. Or, in other words, that the long term effects of a stable work career and the immediate effects of social status (i.e., education) interact. 23 CHAPTER III THE CONCEPTUALIZATION AND THE MEASUREMENT OF THE VARIABLES Education Education is a mode of achievement in the Western societies. Education is the institutional process whereby persons are trained both in social s k i l l s and in highly-specialized occupational s k i l l s . Education is the training in the capacity to perform well in social contexts. It is the training in basic social s k i l l s such as know-ledge of human behaviour, tolerance, the manipulation of persons, the a b i l i t y to i n i t i a t e (from a wealth of diverse knowledge) and to perpetuate ideas, etc. Education is here measured as the total number of years in formal training, standardized by age. (The standardization process w i l l be explicated presently.) On this basic scale of years, a bachelors degree shall be designated as 16 years, or four years after grade 12. The master's degree w i l l be counted as 18 years; the doctor's degree as 21 years. Most of the interview schedules gave the actual number of years in attendance, hence this scalar transformation was quite reliable. For the interval level of measurement, we must assume that every additional year has the same value despite the usual .and somewhat defensible notion of educational plateaux around successive diplomas. Hence our interval scale is the 24 number of years in attendance (see Appendix III (A))?"^ The ordinal scale for education is elementary schooling (3-8 years), junior high (9-10 years), senior high (11-12 years), and some university (13 and more years). The above categories were chosen partly to achieve a reasonable balance in marginals. The dichotomy for education had to be made by cutting between years 10 and 11. It would have been preferable to make the dichotomy between high school and university but this creates a highly skewed distribution. Besides, in the measurement of social s k i l l s , we need not be too concerned about the traditional scaling categories. Consideration was given to the use of a l l information about educational background. Under the argument that the relation between education and orderliness of work careers ought to consider a l l kinds of training, since these add directly to chances in the occupational realm, we investigated the original interview schedules for additional information. The schedule requested only formal education, thereby making unsolicited or volunteered information about other training quite unreliable. A quick check of the relation between orderliness and both measures of education supported 10. Interval distributions of a l l variables, in the form of bar graphs, can be found in Appendix III, one graph for each variable. Each graph shows both ths whole sample and the working-class sub-sample distributions. Each graph contains a few summary s t a t i s t i c s : means, standard deviation, skewness, and kurtosis. In addition to that, Appendix I provides the complete dichotomous cross-classifications for a l l relations for a l l three samples. This Appendix includes that middle-mass sample similar to Wilensky vs sample, and some variance and covariance s t a t i s t i c s . Appendix II provides the basic statistics of variance and covariance for our ordinal scales for the two main samples. The complete cross-classifications of the ordinal scales are not given, but univariate frequency distributions for these ordinal scales can be obtained from the interval scale distributions in Appendix III. 25 our belief about incomplete information (see table I). Adding a l i t t l e disproportionate information about extra educational training reduces the Pearsonian correlation very slightly. The results are insignificantly different by the usual standards but so easily attributable to the system-atic volunteering of such information by lowly educated persons with dis-orderly careers. For these reasons, we settled on formal education. A. more interesting, accessible and crucial problem was the standardization process. There is considerable support for the argument that age is related to education. Sraith and Pawls (1965) demonstrate a procedure for adjusting education by age to remove the effects of age. Since we are contending with interval distributions, i t was deemed satisfactory for our purposes to enploy a crude approximation as a practical and efficient way of handling the problem. Consequently, the visual assessment of the joint distribution of age and education as well as of the ad-hoc regression (figure 3) allotted a rough sense of the deviation around z-scores and of the approximation of the above method. Since that regression line i s education = 9.0 + .05 age , our task of standardizing could be completed ef f i c i e n t l y . As a technique, we added one year of education for every 20-year period. This rough method parallels rather well the Smith and Rawls" method. Using the "natural kinks" in our regression line as demarcations for our age periods, we added two years of education to the education of a l l of those persons born on or before lr<10 and one year to those born between 1910 and 1935. The two World Wars are the demarcation points. That i s , those persons born in or just before 1910 were entering high school during W. W. I. We Education i s defined relative to oneis age cohorts not in absolute terms over time, and becomes deviations from the average of each age cohort. 26 argue that the wars truncated the educational efforts of many persons either by forcing them or their teachers to work or fight, and by cutting short the supply of teachers for those coming years after the wars. Also, with the coming of the wars, more jobs became available to younger persons, jobs with less of a demand for education, hence persons would be aarly attracted to the economic realm. Summarily then, our scales for education, displayed in the Appendices and used in this thesis, are the standardized scales of formal education. Table I. The Pearsonian Correlations Between^Qrderliness and Two Kinds of Education Measures Orderliness Formal education only .237 A l l reported education .225 (498) 11. The Pearsonian correlation coefficient is calculated from the untransformed whole sample. Figure 3. The Ad hoc Regression Through Means of Education for Five-year Periods vO CM as cc to IT) o + C o •H *J to o 3 W » r --a-o> —< .-4 i n <r -a-SC O vC C CM CO r - ) UO CM CM Cr« i vC O CM 0\ i CM + CO C 0) t—i — I w e x> o i o 1 •H u •H 3 4J M > cfl H cd o O o a) o f-» CL, UO vo CM uo oo i n CO CO vo o >—V o t—1 1—1 1 vO I—1 *—t m / — ^ o o co ! CM o o 4-1 o ( CM a. <-• V / 3 J3 •H PQ 1+4 O u 28 Orderliness of Work Careers Orderliness means consistency in work experience and types of s k i l l s used. It is a matter of the relations between jobs. Two jobs stand in an orderly progression i f the s k i l l s used or needed and the experiences entailed in them are consistent from one to the other. For example, i f a man uses the manual s k i l l s of driving a tractor on one job and then moves to an administrative job where he handles people, then his jobs are not consistent or orderly. Whereas i f another person supervises men in a factory in some submanagement position and then moves to a bank manager's job, he is said to be consistent with respect to his supervisory s k i l l s but not his know-ledge of the job content. This second case is somewhat unlikely due to the specialized s k i l l s involved. The complex decisions involved in attributing consistency demanded an open and flexible set of c r i t e r i a , which would be d i f f i c u l t to reproduce. The consistency criterion applies to endlessly different job-combinations. However, the objection, that more complex job histories would create decision problems, is more apparent than real. Orderliness of work careers was measured by the ratio of the number of years of consistency over the total number of years of recorded work — i.e., consistency years over work history. The interval distribution in Appendix I I I (B) is simply the ratio as calculated above. The ordinal scale was created by choosing the two modal categories (0.00 and 1.00) as separate points and by separating 29 those persons in-between at a point close to the mean (.65 for the whole sample), but not giving either of these two categories a disproportionate amount of variance. The resulting ordinal categories were: orderly (1.00), somewhat orderly (0.60 - 0.99), somewhat disorderly (0.01 - 0.59), and disorderly (0.00). The dichotomy then became the categories about this 'mid-point'; orderly (0.60 - 1.00) and disorderly (0.00 - 0.59). Our reasoning with respect to creating a dichotomy around 0.60 was twofold. For one thing, this strategy provided reasonably distributed marginals for the nonparametric scales. For another, incomplete work histories demanded some sort of shift of the theoretical mid-point between orderly and disorderly with respect to the "mid-point" that Wilensky chose (.20). In comparison to Wilensky's data on work histories, our histories were incomplete, involving a gap between a man's f i r s t job and his later jobs. Our data includes a record of the f i r s t job and the last three jobs. The great majority of the interviews, 85% of the 515 cases, had information on four jobs (see table II). (In the whole sample, 87% had four-job histories.) Now, since the mean work history covers 13.7 years and the mean age is 41 years — from a good b e l l -shaped curve —' we can calculate the size of the gap in the average work history. Knowing that the mean education was 10.8 years and assuming therefore that the typical person begins his work career at about 18 or 19 years of age, we can see that the gap l e f t by recording only four jobs is about 9.3 years on the average (figure 4). Of course, there are numerous minor deviations from this typical pattern. 30 Table II. The Number of Jobs in Recorded Work Histories Number of Jobs Percent 1 3% 1 2 2 7 3 6 4 85 N (100%) (515) Figure 4. The 'Average'Work History JOBS # 1 g a p | i #2 j #3 j # 4 l i f e span — 9.3 years — 18 years 41 years of age of age mean age mean recorded history estimated starting age . calculated mean gap . . 41.0 ) 13.7 ) 18.0 ) 9.3 ) 18 + gap + 13.7 = 41 12. The cases with one-job histories are excluded from subsequent analysis, as well as four unreliable interviews, resulting in a total of 498 cases for the whole sample. 31 The respondents varied on the number of jobs recorded, in length o-f work history, in size of gap, and in breaks between jobs or continuity, but this very rough summary, i.e. , the 'typical' person, gives us a good feel for the data. One more comparison with Wilensky's data involves the amount of orderliness in the sample and t h e extrapolation of this fact to other segments of the population. In order to give even a l i t t l e bit of orderliness a chance to show i t s e l f positively, Wilensky s e t his criterion for t h e dichotomy at .20, or at least one-fifth of the worklife orderly. This placed about 30% under "orderly" (note: not his f i n a l category of "orderly") and 33% under "borderline orderly". Together these became his fi n a l group o" orderly persons (63% of 643). Our sample, having a mean ratio of .65 and an incredible distribution, would contain 74% orderly persons i f we followed his criterion of including anyone above a ratio of .20. This fact is not as impressive as the fact that 67% of our respondents have more than .50 in ratio of orderliness, where Wilensky found only 30%. Our data seems most unlike Wilensky's in precisely the nature of the univariate, distribution of orderli-ness. However, incomplete job histories would conceivably mean a disproportionate sampling of the later, more stable job experiences and therefore a bias in t h e direction of orderliness. To this question, Form and Miller (1949, see their many charts) demonstrate the relative disorderliness of the middle or " t r i a l " period of an occupational career. Furthermore, Lipset and Bendix (1952) saw the f i r s t job as a good predictor of the f i n a l occupation. In 32 effect, then, our schedules omit that t r i a l period — at least for the older respondents. These facts are commensurate with the raising of the criterion for orderliness in the scales. It is l e f t to further analysis to judge the significance of 37% difference in amount of orderliness betv/een the samples (at .20), and 11% at .50. The bias could account for the differences. In connection with the discussion of the amount of orderliness, Wilensky's discussion cf orderliness in the white middle-mass sample extrapolated to the lower class (1961a, p. 526). He concluded that a vast majority of the labor force is going nowhere in an unordered way or can expect a worklife of thoroughly-unpredictable ups and downs, (p.526) Now, the hypothesis that less orderliness w i l l be found among the labor force is upheld by our data. Roughly a working-class sample, our homogeneous subsample, has a mean ratio of 0-46 for orderliness — somewhat less than a whole sample (0.65) which also contains this subsample (Appendix IV). However, the conclusion that very few lower class persons, in absolute terms, have orderly careers is tentatively not upheld (55% above 0.50). Of course, the same reservations stemming from the incomplete joh histories apply here. The construction of Wilensky's scale for orderliness is not without problems. He defines the relation between jobs as "orderly" i f the s k i l l s and experience gained on one jcb are directly functional to performance in subsequent jobs 33 and jobs are arranged in a hierarchy of prestige. (1961a, p.525: underlining ours) Is this measure anything other than occupational mobility? He does, in his preliminary discusc-ion, speak of the orderliness dimens ion 3,s s. kind of occupational mobility. It is surprising then that his scale for orderliness — "combining direction with orderliness" (p.525) — i s empirically unrelated to "worklife mobility" as fo;ind in table II,p.527 of his a r t i c l e (1961a). Our scale for orderliness does not include the prestige dimension. We distinguish movement between s k i l l types from movement between major s k i l l levels. The former is orderliness; the. latter i s mobility. However, in our data (table III), these measures are related. They ought to be even more related in Wilensky's sample. Furthermore, i f occupational mobility is related to social participation (as we have seen from other studies) , and is related to the orderliness of work careers, then i t may be that the systematic operation of the mobility dimension is responsible for the positive relation between orderliness and social participation in his study. (We w i l l test our empirical relation between orderliness and mobility, presently.) We close this discussion with the reminder that our measure of orderliness is somewhat different conceptually and that mobility must be included in our testing and model building. 34 Table III. Orderliness and Mobility in Detroit and Millport Detroit"*""5 14 Millport 15 Mobility Orderly Disorderly Mobility Orderly Disorderly Up Stable 35% Fluctuating 44 and down, and Unstable 34% Clear up, Slight up 16 Horizontal 50 Down, and Up and down 55% 39 6 37% 41 22 N (100%) (472) (230) N (100%) (282) (233) 13. H.L. Wilensky, 1961, Orderly Careers and Social Participation, American Sociological Review, 26: 521-539: see p.527, table II. To make alignments possible, we had to collapse some of his categories as well as our own. 14. Criterion for the dichotomy was between .19 and .20 in the ratio, "orderly" being those with a ratio between 0.20 and 1.00. 15. Criterion for the dichotomy was between .65 and .66 in the ratio, "orderly" being those with a ratio between 0.66 and 1.00. 35 Occupational mobility Occupational mobility or intra-generational mobility i s defined as movement between major levels of s k i l l , authority, and proprietorship. Conceptually, this differs from orderliness in that we are describing a person's movement between major levels of occupational prestige. Movement between occupational prestige levels and between s k i l l types are conceptually independent of each other. The question of their empirical relation remains open, and of course then the theoretical explanation of that possible relation. Our data (table III) challenge Wilensky 1s assumption that they are unrelated. The question then becomes: Is our relation spurious? or Is his negligible relation only apparent? Confounding influences are more lik e l y to occur in his "negligible" relation because of the mixed nature of his measure of orderliness. Our "spuriousness" w i l l be tested presently. Mobility w i l l be measured as "upward" and "not upward" at the dichotomous level. The second or "low" category contains both horizontal types of movement as well as fluctuations. The interval scale taHes this dichotomy as a dummy variable (Boyle, 1970). "Upward" becomes I and "not upward" becomes 0. The ordinal scale i s : "clear up," "slight up," "horizontal," and "fluctuating." Those persons with a slight but clear downward movement were put into "fluctuating." Occupational mobility must be included in our study for two reasons. The obvious reason is i t s known relation with 36 participation. Mobility tends to be positively related to social participation, as indicated in the studies by Curtis (1960), F e l l i n and Litwak (1963), MacFarland (1970), Stuchert (1963), and Vorwaller (1970). But these studies express some conflicting results and this d i f f i c u l t y must be recognized. As a crucial example, Vorwaller finds the socialization hypothesis supported, and feels that status of destination has a great effect by i t s e l f . Socialization processes, i.e., continuous adult or anticipatory socialization, operate to mediate behaviour at expected levels. We do not wish to extend, at this point, the discussion of vertical mobility and it s pathological effects (see Vorwaller, 1970) versus statv . 3 of destination and origin. We recognize that problem and simply assert that occupational mobility be taken into consideration. Another reason is the empirical relation between mobility and orderliness in our data (table III). It is possible that the relation between orderliness and mobility in our data i s due to education. Education could provide s k i l l s for both types of movement and could then be the cause of the relation. Table IV contains a test of this assumption. Since the value for phi does shift some but the relation does not disappear in either partial, we conclude that there is some interaction and some additivity among these three variables. Whether mobility causes orderliness, as our table format suggests, or vice versa, is of no consequence at this moment, but we do conclude that mobility and orderliness are directly related in our data. 37 Length of Residence Residential mobility has been found to be negatively related to social participation (e.g., Scott, 1957). That i s , the more one moves about, geographically, the less one can or w i l l participate in organizations unique to any one community. As an ecological variable, length of residence i s the converse to residential mobility. It is argued that every person needs exposure to the organizational opportunities in the community in order to acquaint himself, f i r s t l y with their existence, secondly with their location in space and time, and thirdly with their entrance requirements and memberships duties. In order to become more involved a person needs to insert himself into the social networks already structured in these sociaj organizations. That i s , in order to hold a position, he has to be known by and to know well some of the other members. Wilensky does not handle any kind of exposure notion, but we w i l l use this variable in the extension of the basie thtory of orderliness and participation — as a test factor and as a variable in the more complex model. Residence or exposure shall be measured by the number of years of li v i n g in the community. The interval scale employs the year as the basic unit. The ordinal scale is built around successively increasing blocks of time: 0-5 years, 6-9 years, 10-19 years and 20 years and up. It is f e l t that after 10 years in the community, the individual years take on less importance as an in d i -cation of exposure. The dichotomous scale for residence i s created 38 Table Iv: Per Cent Persons with Orderly Careers by Occupational Mobility and Education Education Mobility Low High Not Upward 45 58 (119) (142) Upward 59 (108) 78 (129) Phi for association betxvfeen mobility and orderliness control-ling for education. ,15 .21 Phi without the control for education .18 39 by placing the cutting point between 9 and 10 years. Indicators of Organizational Participation The scope of this study is limited to participation in organize -'.ons. These organizations include churches, unions, clubs of a l l kinds ( p o l i t i c a l , fraternal, occupational, etc.), and sports groups. These associations have a formal structure, i f not a constitution, meet regularly, and have consistent membership requirements. Not included are ad hoc sports gatherings or any temporary associations. And we shall use five indicators of participation: the number of memberships in a l l organizations; the number of memberships in voluntary organizations (i.e., excluding churches and unions); the number of meetings attended over the last sir: months; the number of hours spent in these organizations in the last four weeks; and the number of positions held or holding and committee memberships in these organizations. The replication portion of this thesis then does not cover nearly a l l of Wilensky's dependent variables or measures of social participation. We do not tap v i s i t i n g patterns, friendship networks, or any other indicators of primary relations. The five indicators can be subdivided into measures of a f f i l i a t i o n and measures of involvement. The two membership indicators speak only to a f f i l i a t i o n ; the other three comprise, together, a kind of separated scale cf the degree of involvement in organizations. The subdivision into a f f i l i a t i o n and involvement AO contains theoretical advantages. For example, the Curtis study (19t'J) showed that the attendance of occupationally-mobile persons w i l l differ from that of immobile others, even though the proportions of actual memberships are the same. As a matter of strategy, we shall be;*:in our analyses with a l l five indicators of participation. At a convenient point, after their individual fluctuations can be documented and their r e l i a b i l i t y perceived, we w i l l reduce the analysis to two indicators, namely. memberships excluding the churches and unions ( a f f i l i a t i o n ) , and the number of positions (involve-ment) . Trends w i l l be evident within the indicators and readers can make the necessary extrapolations. The f i r s t measure of participation i s the number of memberships including churches and unions. The interval scale w i l l be the number of such memberships; the ordinal scale w i l l entail a steady progression: no memberships, one, two, and three or more. The dichotomy w i l l be set between one and two member-ships since each person can be expected to be a member of at least one organization — u s u a l l y a church or union. The frequency distribution of the scale (Appendix III (E)) bears us out. Only 6% have no memberships whatsoever. Then, we dropped churches and unions from the l i s t of organizations under the assumption that these were only debatably voluntary. This new measure then becomes memberships excluding church and union. Union membership is a condition of employment 41 in nearly a l l the union contracts in the area. The question of voluntarism in church memberships is more d i f f i c u l t to answer. We are fortunate in that in our data there are few churches and few members from the Free Church tradition where membership is more voluntary — an adult decision. To match the reasoning in the other membership scale, the dichotomy is between none and some. The ordinal scale w i l l be a three-point scale (Appendix I I I (F)) of none, one, and two or more memberships. Again, the heavily skewed distribution contributes to our decision here. The interval scale is simply the number of memberships. The third indicator cf participation is the number of times in attendance in the last six months at any meetings of any organizations. Memberships in the unions may not be voluntary, but attendance at their meetings i s . Since attendance is measured over a 3ix months period, we f e l t that as a dichotomy, a break between 8 and S times was theoretically reasonable. This amounts to church attendance once a month plus two club meetings — or some such combination or extent of attendance. The interviewing was done in early summer when attendance would be lower than usual, hence we chose a low dividing point for the dichotomy. The ordinal scale becomes: no attendance, 1-8 times, 9-24 times, and 25 times or more often. And again, the interval scale i s self-evident: the number of times in attendance. Fourthly, a more stringent measure of actual involvement (as opposed to a f f i l i a t i o n ) is the mrsber of hours spent in the last four weeks in organizational ac t i v i t i e s . This measure also includes unions and churches. The dichotomy has i t s cutting point between four and five hours. The reason i s obvious; an hour per week for a church member was a reasonable (stringent) measure. The ordinal scale i s : no hours spent, 1-4 hours, 5-12 hours, and 13 hours or more. Number of hours made up the interval scale. Finally, as a good indicator of involvement, the number of positions held or presently holding became our f i f t h measure. For i t , we had to adopt another three-point ordinal scale; i t became: no positions, 1-2 positions, and 3 positions or more. The interval scale of the number of positions i s a very positively skewed distribution (Appendix III (I)). The dichotomy was none and some. The frequency distributions of a l l of these scales w i l l be affected by the reduction of the number of cases as we move into the working-class subsample. Consequently, in testing our causal models on this sample, we shall have to re-scale some variables to attain reascrable marginal distributions. These changes w i l l be explicated when the respective sample is being analysed in chapter V. Essentially, the working-class sample entails a disproportionate selection of cases from the lower half of the scale for education, where the middle-mass sample selects from the middle section of that variable (see Appendix IV). Since education is related to almost every variable in this analysis, i t is obvious why scalar categories need to be adjusted to the lower ends of the scales. The middle-mass sample retains the scaling properties of the whole sample. 44 CHAPTER IV DATA CHARACTERISTICS AND THE THREE SAMPLES Source of the Data The data for the following analyses came from an "interview survey of presently working adults in an industrial community of about 20,000 population on Vancouver Island" (Meissner, 1971, p.245). "Millport," as the name suggests, is a manufacturing and seaport city, half of whose residents are employed by one "Company" in a number of integrated operations: two logging areas, two sawmills, a shingle m i l l , a plywood m i l l , and a pulp and paper plant. A "two-cluster" sample was drawn from a newly published city directory (Appendix VII). Because of special interests in the technical environment of the Company employees, one "cluster" was sampled disproportionately. In order to make inferences to the community population and therefore to those undersampled white-collar workers, we instituted the weighting procedure describee in table V below. A weighting of "41 makes the proportions approximately equal for the Millport community; each cluster becomes a 6% sample (see Appendix VIT.) . Table V describes the make-up of our three samples as well. The reduction in N between the original data and the (whole) sample actually used in this thesis resulted from the removal of a few incomplete or unsatisfactory interviews, and those seven persons with only one job to their occupational 45 1 Table V. Weighting Procedure and the Three Samples Weighted Samples Cluster Weight Unweighted^ Whole Middle-mass Working-class sample sample sample sample Company '1' 230 230 46% 105 59% 198 Other '4' 67 268 54% 72 41% N (100%) 297 493 177 198 16. The total unweighted sample was 308 respondents. Several one-job persons were removed (K=7j weighted, N=13); four inter-views were considered unreliable or incomplete. 46 histories. Of necessity, these latter persons did not meet the criterion of "relations between jobs" for the orderliness measure. The Three Samples The notion of a middle-mass sample derives from Wilensky. Wilensky's sample consisted of white, male members of the labour force who were between 21 and 55 years of age and ever married, occupying dwelling units in Detroit (1961a, p.529). The middle-mass notion is a residual category obtained by restricting the sample to a range of family income between $5,000 and $13,000 (in 1958-59 wage scales), and to certain occupational-educational categories, with consideration of authority level, i.e., numbers of persons supervised. The source of our middle-mass sample i s cf course the same as the others, i.e., Millport, however i t s characteris-tics d i f f e r from the others. We approximated Wilensky's sample by restricting our f i r s t sample to married male home-owners between 21 and 54 years of age, below the occupational level of professionals and managers, and between 8 and 14 years of formal schooling. We could not ascertain the racial origins nor the precise income levels. However, by the combination of education and authority level, we assume a f a i r l y close approximation to his residual middle-mass sample. The effects of the truncation of a variable need to be discussed. Both Wilensky's and our middle-mass sample have the feature of limiting the sample to persons within a certain range 47 of education. The effect of truncating one's data according to a relevant variable, in this case the independent variable education, should be to reduce correlations between variables no matter what the level of measurement. This would be especially true of dichotomies, but also of interval level correlations i f the joint distributions were something approaching linearity. In the case of dichotomies, for example, this procedure is very li k e l y to remove a great proportion of cases from precisely those cells supporting the hypothesis (see figure 5 ) . That i s , i f the (monotonic) hypothesis between education and orderliness is positive, we can expect that those with high education w i l l , at least, have above median orderliness. Otherwise, i t is meaningless to make such a hypothesis! As in the hypothetical example of education and orderliness (figure 5), where only supportive cases are removed, the measure of association w i l l decrease. It is also obvious that the correlation coefficient w i l l likely decrease under that distribution. The gist of this important digression from the descrip-tion of our samples is that our whole sample, which includes more variation in the education variable, should evidence a better association or correlation between education and orderliness, a l l other things being equal. Therefore, i f our whole sample does not support the Wilensky hypothesis about that relation, we have even more reason to doubt or eliminate that part of his theory. 48 Figure 5 . The Effect of Truncating an Independent Variable at i t s Extremes: Education and Orderliness high E d u c a t i o n low low Orderliness high Cases eliminated by the truncation of education for the middle-mass sample. Demarcation lines for a hypothetical dichotomy. Concentration of jointly-distributed cases. 49 The whole sample need not be dealt with extensively. The discussion of Millport, above, is a good description of the data, and we need only add a few broad characterizations. This sample i s comprised of persons between 19 and 69 years of age (figure 3 above); of both sexes (17% are female); with 11% at the management or profes-sional level of authority; and with an education range of 3 to 22 years. And these are precisely those variables v/hich the middle-mass sample truncated. The working-class sample is a more homogeneous subsample of Company employees. It consists of those Company employees who are also male, wage-earners, below foremen in level of authority, and union members. Generally, these persons work under Jbroadly-similar technical conditions, though geographically scattered, confront the same employing Company in the economic world with (probably) similar orientations, and have a minimum of authority on the job. There are no entrepreneurs or men of supervisory positions in this sample — who might be "organization men" (Whyte, 1956). Such persons would be expected to hold special a f f i n i t i e s for p a r t i c i -pation in voluntary organizations, under the spill-over hypothesis. This subsample should present an interesting substantive comparison with the others. The Overlap in the Samples In the discussion of the samples, i t may have been deduced that these samples were mutually exclusive. The whole sample (Appendix IV) is comprised of a l l reliable interview schedules 50 in the Millport weighted random sample. The working-class subsample and the middle-mass subsample are completely contained in i t . Also, the middle-mass and working-class subsamples overlap by about 50% of their respective respondents (See Appendix IV for a rough configuration of the overlap in our samples). The considerable overlap in the samples is a crucial fact bearing on our subsequent interpretations of the empirical relations and bearing especially on the comparison across samples of those interpretations. Some advantages can be derived from that overlap. The behaviour of working-class individuals with the one-Company experiences has been isolated from that of the other individuals. The nature of the empirical relations in this sample can be employed to deduce the nature of the empirical relations in other segments of the whole sample. 51 CHAPTER V THE THEORY, CAUSAL MODELS AND THE MULTI-STAGE ANALYSES The Theory The theory of differential organizational participation is not well developed in the literature, especially with regard to education and orderliness as causal factors."^ It is therefore necessary that we develop the theoretical framework of this study and point to some of the departures from Wilensky and others. However, the five hypotheses that appear hereafter cannot be said to have been derived from the theory. If anything, the theory was created after the hypotheses were given. That fact is less objectionable in a study with replicatory and exploratory purposes. Explanations of differential organizational participation need to take into account at least a number of factors, which may jointly affect participation. Three types of explanations resort to notions of capacities, opportunities, and personal preferences or early socialization. By capacities we mean social s k i l l s needed in dealing with people (knowledge of human behaviour, tolerance gained by broadened scope, manipulation s k i l l s , oration, the abil i t y and resource to i n i t i a t e and perpetuate ideas, etc.). Under opportunities, we include threei temporal, spatial, and economic opportunities. Are the persons in the right place, at the right time, with enough economic resources? Does a person have adequate resources and the appropriate scheduling of off-work time, and are 17. Avery recent theoretical article (Collins, 1971) deals with the effects of education on str a t i f i c a t i o n and employs some of the same ideas we w i l l employ in the theory. 52 the organizations within geographical reach? The latter are often termed opportunity structures, here seen from the point of view of the individual. By socialization, we mean characteristics acquired by early training (and later re-adjustment)i values, beliefs, and habits. These constitute the make-up of preferences. An explanation of participation which considers only several of these dimensions must of necessity be incomplete and can only account for a portion of the variation in organizational participation. We w i l l consider, at this stage, only capacities and economic opportunities. Let us assume that men and women have a drive to achieve and a corollary need to have this achievement recognized and 18 continually reinforced. Achievement i s attainment of position relative to one's spatial and temporal peers, in some broadly-defined evaluative rank order. Achievement is not the attainment of one's personal goals — since these shift with time and can vary considerably for social contexts — 1 but is attainment relative to 19 visible others. An important mode of achievement in Western society is education — since such evaluative recognition i s not 18. How achievement i s tied to the notion of status w i l l not be explored here — although i t is exp l i c i t l y recognized that education is a well-accepted measure of status. It would not harm our theory i f achievement and acquired status were in fact equated. If any distinction is to be made between status and achievement, i t is this that achievement as measured by education i s a personally acquired status as opposed to ascribed status, father's occupation and education. 19. The difference between achievement relative to personal goals and achievement relative to others in society is most relevant to our "other-directed society" (Riesman, 1950). These may coincide of course. 53 ascribed but acquired. Education provides the necessary social s k i l l s (Smith and Rawls, 1965, p.59), the capacity to maintain 20 performance in organizations. Education also enhances the prestige of the individual and therefore, vicariously, the prestige of the organization to which he belongs. It "makes him a more desirable and potentially effective member from the point of view of others already in the group" (Smith and Rawls, 1965, p.59). Furthermore, attaining one's goal — the achievement of relative acquired status through ed ication — is not self-satisfying. Such achievement must be frequently paraded and exercised in places of authority and decision-making. The v i s i b i l i t y and exercise of achievement satisfies the achievement drive. Organi-zational participation, especially voluntary participation, provides a visible non-occupational arena for the display of achievement and for the continual reinforcement of that achievement by the exercise of decision-making authority. The holding of positions of authority is probably the best measure, although attendance and memberships can also be taken to indicate such an exercise of displaying achievement. Education also arouses expectations of, and provides specialized s k i l l s for, an orderly work career. A higher education is usually associated with a specialized, highly-skilled occupation, but this premise is tenuous for our time. (See also Collins' 20. As w i l l be seen from the accompanying analyses, the theory needs to differentiate between memberships or simple a f f i l i a t i o n and the holding of positions of authority or organizational involvement. The theory is directed more to the latter. 54 dis cussion; 1971, p.l002f. Even manual s k i l l s are becoming highly specialized but these also are demanding formal education now. This argues for the use of a l l types of formal training in the education variable.) Nevertheless we w i l l assume that the more specialized the s k i l l s , the less chance and need to move across s k i l l types due to disruptive crises, i.e., the more orderly the work career. Hence, higher education leads to more orderliness. Continuous organizational participation demands a predictable flow of resources. An orderly work history, with few work-stop gaps, ensures predictability of resources. An orderly career in the economic realm also places the least demands on one's time off-work. A disorderly career demands additional effort in learning new s k i l l s and re-adjustment of experiences. On the acceptable assumption of rigi d limits to everyone's disposable time, extra time and effort needed to adjust one's occupational interests leaves less time and energy for organizational p a r t i c i -pation. Putting these notions together, orderliness of work careers provides st a b i l i t y of resources in time and money for organizational participation. Having to put the pieces of the theoretical framework together into a causal model of additive effects and of the causal-5 5 21 sequence nature i s exceedingly discomforting. The additive effects model i s simply constructed, but the causal-sequence model i s d i f f i c u l t given the above theoretical notions. The explanation or interpretation model, when stated in verbal terms, i s a l l too-easily confused with the interaction model. To proceed, however, in Wilensky !s chain educeition arouses expectations of an orderly career. An orderly career provides the necessary predictableness of resources for organizational participation. In technical terms, orderliness of careers explains the relationship between education and participation. Such a causal chain disserts the notions of achievement, of social s k i l l s , and of achievement display — which tie education directly to participation —- but these are our theoretical notions and w i l l be more relevant to our interaction model. The theory differs from Wilensky Ts in that the focus i s on personal achievement rather than on social integration. In our explanation of the relationships, organizational participation provides an arena for the very personal exercise of displaying individual achievement. That this behaviour also stabilizes the diversity in society is only by default. Although the theory differs from Wilensky's the i n i t i a l causal model remains identical with his. That model is displayed pi c t o r i a l l y in figure 6. The hypotheses that combine to make up 21. It is readily apparent that economic resources may be more than an intervening variable in the system, i.e., part of a suppres-sion or specification model. The reason for the adoption of an additive model stems from the concern for replication of a study which, although i t discusses interaction i n a vague way, looks only at additive effects. Interaction w i l l be treated when an alternative model i s introduced. 56 that causal model are the following ones. : Education i s positively related to the orderliness of work careers; H 2 : Orderliness of work careers is positively related to organizational participation; (a) : Education is positively related to organizational participation; but (b) : Orderliness of work careers explains the relation between education and organizational participation. It is convenient to introduce the remaining variables at this point. Occupational mobility or length of residence may account for some of the empirical relations in our data. It may be that since orderliness and mobility are empirically related (tables III and IV), and that since our conceptual separation of the two may be faulty (i.e., that movement between s k i l l types and s k i l l levels are necessarily related), we can attribute the orderliness-participation relation to mobility. In that sense, mobility i s a cause of both orderliness and participation, making their relation spurious (see figure 7). The relevant hypothesis states that: (a) : Orderliness of xrork careers is positively related to organizational participation; but (b) : Occupational mobility i s the cause of the spurious relation between orderliness of work careers and organiza-tional participation. And our f i f t h hypothesis involves the length of residence or the notion of exposure to opportunities. H,. (a) : Orderliness of work careers i s positively related to organizational participation; but (b) : Length of residence explains the relation between orderliness of work careers and organizational participation. Figure 6. Wilensky's Temporal Causal-Sequence Model EDUCATION ORGANIZATIONAL ORDERLINESS OF WORK CAREERS 22. Harold L. Wilensky, 1961. Orderly Careers and Social Participation, American Sociological Review 26: 521-539; see especially page 539. 58 Figure 7. Occupational Mobility as a Cause of the Spurious Relation Between Orderliness of Work Careers and Organizational Participation OCCUPATIONAL MOBILITY ORDERLINESS OF WORK CAREERS ORGANIZATIONAL PARTICIPATION Figure 8. Length of Residence as an Explanation of the Relation Between Orderliness of Work Careers and Organizational Participation ORDERLINESS OF WORK CAREERS ORGANIZATIONAL PARTICIPATION LENGTH OF RESIDENCE 59 This f i n a l hypothesis, another type of test of Wilensky's central relation, specifies residence as an intervening variable (see figure 8). The more orderly a person's career, the longer he w i l l stay in a community. .And the longer he stays in one area, the more he w i l l get involved in that local community. Hence, i t is not his orderliness per se but his exposure to opportunities that mediates his involvement, even though order-liness provides the predictability of resources for prolonged residence (and participation). The Middle-Mass Sample The assessment of the empirical relations in the middle-mass sample w i l l rely on the magnitude and consistency of values of Somers' D, an asymmetric measure of association (Somers, 1962), appearing in the f i r s t column of tables VI, VII, and VIII. That s t a t i s t i c is algebraically equivalent to per cent difference in the case of dichotomies. It is sometimes called the ordinal analog to the regression coefficient, b or b (Davis, 1971, ° yx xy p.71f; Hawkes, 1971, p.913; Kim, 1971, p.898f; and Somers, 1962, p.807). Having been employed by Wilensky, this s t a t i s t i c allows for a comparison with Wilensky's data. Our assessment for this sample i s , f i r s t l y , that orderli-ness and education are significantly related (criterion: ±.10), and the argument from temporal sequencing would make orderliness a function cf education. Secondly, orderliness and participation 60 Table VI. The Test of the Wilensky Model in the Middle-Mass Sample at the Dichotomous Level of Measurement Relations Hypothi eses Dependent Independent Control Somers' D Phi R 2 3 H l orderliness education .16 .16 + H including orderliness .15 .15 + z excluding orderliness .31 .30 + attendance orderliness .03 .03 0 hours orderliness .08 .08 0 positions orderliness .09 .09 0 H3 (a) including education -.18 -.18 _ j excluding education .05 .05 0 attendance education -.15 -.15 -hours education .09 .09 0 positions education -.02 -.02 0 (b) including education orderliness -.21 -.21 — excluding education orderliness .00 .00 0 attendance education orderliness -.16 -.15 -hours education orderliness .08 .08 0 positions education orderliness -.03 -.03 0 N (177) (177) 23. 'R' means the type of relationship, classified according to the criterion of magnitude: '+' means a non-negligible positive relation; '0' a negligible relation; and '-' a non-negligible negative relation. within each block of five indicators. Consistency i s then determined by 'averaging' 61 are related in a consistent manner ( a l l are positive relations) 24 but only a f f i l i a t i o n appears to be strong. The indicators of 24 active involvement — the last three of each block of five — show negligible relations. Orderliness of work careers in the middle-mass sample affects only a f f i l i a t i o n , especially in the s t r i c t l y voluntary organizations. (This conclusion must be tentative considering the near significance of one other value.) Thirdly, the zero-order relation (H^(a)) between education and participation is surprisingly inconsistent in sign. The s i g n i f i -cant relations are negative; there is no consistency. That relation needs to be. explored more fu l l y . Part of the confusion around education can be explained with a close look at a f f i l i a t i o n . The fact that memberships including churches and unions stands in a significant negative relation with education, and that memberships excluding churches and unions stands in a negligible positive relation, demands a particular interpretation. It appears that the less educated persons of this sample — range of education: 8 to 14 years --belong in a much greater proportion to either churches and/or unions (and seem to attend there as well). This empirical finding seems reasonable; the explanation seems obvious and sound — unions are compulsory for the working class, i.e., the less educated persons. The possibly crucial difference between unions and churches cannot ( be explored here. 24. It is remembered that we distinguish between our five i n d i -cators of participation by grouping attendance, hours spent, and position-holding under measures of involvement and the two member-ship indicators under a f f i l i a t i o n . We conclude, with respect to hypothesis 3, that education within the middle-mass sample makes very l i t t l e difference to memberships outside of the church and union, and 25 to involvement in those organizations. Although more of the less educated persons belong to churches and unions, they do not participate more. The control for orderliness tends to support this. Only two indicators meet significance, but the control makes a l l indicators more negative. That i s , when the consis-tently positive relations between orderliness and participation are controlled, then the edu es. t i.Gn."—p 317 ticipation relations move uniformly in the direction of a non-negligible negative relation. However, the less educated persons s t i l l do not involve themselves at a significantly different level. Since i t is the case that orderliness does account for a f f i l i a t i o n and appears at least consistent among a l l indicators of involvement, and since the Wilensky model does not appear adequate with respect to the test relation (namely, that 25. This conclusion needs to be explained in the light of the significant negative relation with attendance at the dichotomous level. In an extended effort (not shown here) at gauging the effect of arbitrary outpoints, i t was discovered that attendance, with i t s widely-dispersed range of values (see Appendix III (G)), was erratic from one cutting point to another in the middle-mass sample. This makes i t a somewhat unreliable scale for our purposes. 26. Theoretically, i t could also be the other way around. A f f i l i a t i o n with unions could account for an orderly work career. This reversal makes no difference to the empirical test. The completely different and unique theory which presents p a r t i c i -pation as a cause of orderly work careers, or as part of a mutually reciprocal relation, i s extremely alluring but must be bypassed in this whole effort. the hypothesized positive relation between education and participation would disappear), table VII is introduced as an addendum to table VI. It presents the test of the orderliness-participation relation by controlling for education, under the assumption that i t is education which accounts for the spurious relation between the other two (i.e., model (b) of figure 2 above). As expected (from our acquaintance with the data) the relation increases, positively, in a uniform manner. Three out of five indicators are now significant and a l l are s t i l l consistent in sign (table VII, f i r s t column). It is also interesting that the positive relations increase in magnitude with the increasing stringency of our involvement indicators, i.e., positions is highest of i the three. Several conclusions about the Wilensky model, the causal chain, are pending. That model is not confirmed by our data. Although the individual relations pictured in figure 6 are found significant, their combination into a causal model presents a d i f f i c u l t y . Education and orderliness seem to affect participation in a conflicting pattern, even though they are themselves related positively. It is because of the confusion around the education-participation relation — whether that is declared to be direct or indirect — that the model is not confirmed. As we shall shortly see, the model is more inadequate than wrong. The hypothesized positive relation does not appear even though the other two 64 Table VII. Education as a Cause of the Spurious Relation Between Orderliness of Work Careers and Organizational Participation Relations Dependent Independent Control Soraers ^ memb. includ. orderliness education .19 + .19 + memb. exclud. orderliness education .31 + .30 + attendance orderliness education .06 0 .06 0 hours orderliness education .06 0 .06 0 positions orderliness education .10 + .09 0 N (177) (177) 65 27 links in the causal chain appear positive and significant. Education does not seem important to participation, and certainly not in the direction hypothesized. The orderliness-participation relation is found to be crucial and orderliness affects the s t r i c t l y voluntary memberships most. But the conclusions with respect to orderliness need to be tested under further conditions. That theoretically central relation of orderliness to participation may be due to some confounding effects. After a careful look at tables VI and VII, where few non-negligible relations appear, i t seems fruitless to pursue the effort of controls. However, for several reasons, none of them based on the magnitudes of found values, we must pursue our course. .Firstly, i t is just as important to examine a negligible relation as a non-negligible one. Secondly, since the present effort extends to other samples and alter-native causal models, i t is necessary that anticipated negligible relations be empirically determined for further comparisons. Thirdly, the effect of orderliness of work careers, as an alternative to status, class, or socialization, was Wilensky's central concern and needs to be more fully explored and documented. 27. This very important finding clearly impinges on the growing use of path analysis. This finding, where two positive relations in a theoretical chain (E-0; 0-P) do not evidence a positive indirect relation between the ends of the chain (E-P) , clearly challenge the growing pervasiveness of path models. The problem stems from interaction in the data, which elementary path analysis overlooks. 66 Table V I I I presents the t e s t of the f o u r t h and f i f t h hypotheses . Al though the c o n t r o l f o r m o b i l i t y a l t e r s the set of r e l a t i o n s on ly s l i g h t l y , there i s cons i s t ency of s i g n and s u f f i c i e n t magnitude i n three r e l a t i o n s . The c o n t r o l f o r res idence does much the same. S ince n e i t h e r of these c o n t r o l s a f f e c t the o r i g i n a l r e l a t i o n a p p r e c i a b l y , c e r t a i n l y not to the extent of removing i t e n t i r e l y as h y p o t h e s i z e d , we conclude that n e i t h e r hypothes i s i s conf i rmed. The r e l a t i o n of o r d e r l i n e s s to p a r t i c i p a t i o n remains the only n o n - n e g l i g i b l e r e l a t i o n of importance —under the t e s t of a l i n e a r , a d d i t i v e model . Were i t not f o r the ambiguity surrounding the e m p i r i c a l va lues connected w i t h the e d u c a t i o n - p a r t i c i p a t i o n r e l a t i o n s , we cou ld have l e f t the above c o n c l u s i o n s w i t h assurance. W i l e n s k y ' s c e n t r a l hypothes i s about the e f f e c t of o r d e r l i n e s s of work careers would be confirmed even though the g iven model would not be conf i rmed . However, knowing that such c o n c l u s i o n s r e s t upon a l i n e a r , a d d i t i v e c a u s a l model , w i t h i t s a t tendant t h e o r e t i c a l assumptions , t e s t ed by l i n e a r  s t a t i s t i c s , we attempted to a l l a y our doubts . Another type o f m u l t i v a r i a t e a n a l y s i s b o l s t e r s our aroused s u s p i c i o n s about c u r v i l i n e a r r e l a t i o n s i n the d a t a . Such c u r v i i i n e a r i t y , s imply i n s c r u t a b l e to the ( l i n e a r ) s t a t i s t i c a l p a r t i a l l i n g method, can be q u i c k l y and c l e a r l y d i s c o v e r e d by p a r t i t i o n i n g 67 Table VIII. The Test of the Relation Between Orderliness of Work Careers and Organizational P a r t i c i -pation: with Occupational M o b i l i t y and Length of Residence Controlled Relations H, H 4(b) H5(b) Dependent Independent Control Somers' D Phi R including orderliness .15 .15 + excluding orderliness .31 .30 + attendance orderliness .03 .03 0 hours orderliness .08 .08 0 p o s i t i o n s orderliness .09 .09 0 including orderliness m o b i l i t y .15 .15 + excluding orderliness mobility .28 .26 + attendance orderliness m o b i l i t y .03 .03 0 hours orderliness mobility .11 .11 + positions orderliness m o b i l i t y .08 .08 0 including orderliness residence .17 .16 + excluding orderliness residence .34 .32 + attendance orderliness residence .04 .04 0 hours orderliness residence .09 .09 0 positions orderliness residence .07 .07 0 (177) (177) 68 (see table IX; cf. Lazarsfeld, 1955) and is usually due to interaction among variables that are correlated. (Cf. Sonquist, 1970, p.8 and see his chapter 3 oh the AID package). As can be seen from the values in table IX for the orderliness-participation relations under high and low education, there is interaction between education and orderliness on a l l of the participation indicators. Such interaction is not demonstrable 28 by s t a t i s t i c a l partialling. Education specifies the relation between orderliness of work careers and organizational p a r t i c i -pation. A mechanism similar to that behind the effects of status crystallisation seems to be operating: "the more you have, the more you get" ("or the more you want to have"). Where the educational background is high, orderliness has a significant effect on organizational participation. Where education is low there is a suppressed effect or even a significant reversal of behaviour: the disorderly involve themselves more than the orderly. A replication must contend i t s e l f with direct comparability with the original study. In our replication effort \<re have tended to apply a stricter set of c r i t e r i a . On the basis of Wilensky Js criterion of consistency in sign 28. The specification model (see figure 2 (c)) can be and has been seen (Anderson and Zelditch, 1968, chapter 9; Blalock, 1964, p.91f; Davis, 1971, chapter 4) as a special case of interaction just as suppression is another special case. The only difference l i e s in the range of correlation values ever which interaction occurs. Specification is the name given to the case where cor-relation values change signs; suppression where signs remain the same but magnitudes differ. This "only difference" does entail a considerable theoretical difference, but when values scatter close to the zero-point l i t t l e importance can be given to their existence above or below zero. Table IX. The Test for Interaction Between Education and Orderliness of Work Careers on Organizational Participation: Somers D on Dichotomies Education Original Relation Grade 11 Up to and up grade 11 St a t i s t i c a l Partial (from table VII) including orderliness .25 .12 .19 excluding orderliness .41 .19 .31 attendance orderliness .21 -.12 .06 hours orderliness .20 -.10 .06 positions orderliness .17 .01 .10 N (105) (72) (177 29. The comparison of Somers' D for partitions in education is equivalent to the difference of per cent differences test for interaction at the dichotomous level. For example, the f i r s t column represents the per cent difference between orderly and disorderly persons in per cent high participation. 70 among r e l a t i o n s and h i s focus on 'tendencies' i n the data s we would have arri v e d at much the same conclusions as he d i d . By comparison, our r e l a t i o n s evidence stronger tendencies. Had we been s a t i s f i e d with minimal findings and tendencies, then the t h e o r e t i c a l conclusions that we would have deduced, i . e . , the explanations of the r e l a t i o n s , would have been s i m i l a r l y ambiguous i n terms of exactness of causal models. A s t r i c t e r set of c r i t e r i a has forced a c r i t i c a l d i f f e r e n t i a t i o n between causal (and t h e o r e t i c a l ) models. In our study, we did not rest with very minimal r e s u l t s , pursued the ambiguity around education u n t i l we ar r i v e d at an i n t e r a c t i o n model, and now rest t e n t a t i v e l y assured with a new model of r e l a t i o n s . In ad d i t i o n to the problem of the e f f e c t s of a r b i t r a r y cutting points, t h i s dichotomous analysis i s contingent upon another, more c r u c i a l problem. Since t h i s middle-mass sample i s rather small (table V; unweighted, 'N' = 123), the e f f e c t of the weighting procedure could a f f e c t some of the r e l a t i o n s . A singl e weighted respondent can have an e f f e c t of 3.6% to 6.1% i n a zero-order r e l a t i o n given our marginal d i s t r i b u t i o n s (most 30 skewed: 30% - 70%). The most widely-dispersed sc a l e s , attendance and hours, would be the two scales most susceptible to the s h i f t i n g 30. The "3.6% to 6.1%" figu r e i s obtained by c a l c u l a t i n g the e f f e c t of s h i f t i n g a single weighted respondent (n = 4) from one category of a dependent v a r i a b l e to the other category for the largest (111) and the smallest (66) marginals i n our sample. That i s , 4/66 = 6.1% and 4/111 = 3.6%— for hours as dependent v a r i a b l e . 71 of weighted respondents due to a r b i t r a r y c u t t i n g p o i n t s . We are of course u s i n g the terminology of s h i f t i n g c u t t i n g p o i n t s to d e s c r i b e the common procedure of ' g u e s s i n g ' and/or t h e o r e t i c a l l y -de termin ing the best d i c ho t o mi e s , but imprec i s e measurement can a l so be a source of such s h i f t i n g of cases . A l l t o l d , the c o n c l u s i o n s w i t h respect to t h i s sample r e s t on such a tenuous b a s i s . In summary, t h e n , the Wilensky model i s inadequate i f not q u i t e wrong. The r e l a t i o n of o r d e r l i n e s s of work careers to p a r t i c i p a t i o n has found support i n our middle-mass sample i n a dichotomous a n a l y s i s . Al though measures of involvement appear s i g n i f i c a n t p r i m a r i l y under the s p e c i f i c a t i o n model , where l e v e l s of educa t ion must be s p e c i f i e d , the measures of a f f i l i a t i o n evidence s i g n i f i c a n t independent e f f e c t s from o r d e r l i n e s s ( t ab l e I X ) . This core r e l a t i o n appears moderately s t rong and c o n s i s t e n t throughout the i n t r o d u c t i o n of t e s t f a c t o r c o n t r o l s f o r e d u c a t i o n , i n t r a - g e n e r a t i o n a l m o b i l i t y , and l e n g t h of r e s i d e n c e — u n d e r the assumptions of a l i n e a r , a d d i t i v e model . In terms c f a new causa l model , educa t ion s p e c i f i e s tha t core r e l a t i o n . We have a l ready encountered i n t e r p r e t a t i o n problems , stemming from i n t e r a c t i o n e f f e c t s , crude c a t e g o r i e s and a r b i t r a r y c u t t i n g p o i n t s , and i t w i l l be i n t e r e s t i n g to t e s t the above c o n c l u s i o n s under d i f f e r e n t c o n d i t i o n s . The new c o n d i t i o n s a v a i l a b l e a r e : a broader range of educa t ion and o c c u p a t i o n , 72 l a r g e r s a m p l e s , a more homogeneous s a m p l e , t h e e l i m i n a t i o n o f a r b i t r a r y c u t t i n g p o i n t s u n d e r h i g h e r measurement a s s u m p t i o n s , and t e s t s u n d e r new c a u s a l - m o d e l a s s u m p t i o n s . We h a p p i l y t u r n t o t h e l a r g e r s ample and t h e n o n - w e i g h t e d w o r k i n g - c l a s s s ample t o r e t e s t and e x t e n d t h e p r e s e n t a n a l y s i s , and t o make an e x t e n s i v e a s s e s s m e n t o f t h e d i f f e r e n c e s a t measurement l e v e l s . The Whole Sample B e f o r e we go d i r e c t l y t o t h e o t h e r two a n a l y s e s , we need t o g i v e seme more g e n e r a l d i r e c t i o n s as t o p r o c e d u r e s . The a l t e r n a t i v e i n t e r a c t i o n m o d e l w i l l be t e s t e d by o u r two more s t a b l e measures o f p a r t i c i p a t i o n , n a m e l y , member sh ip s e x c l u d i n g c h u r c h e s and u n i o n s , and p o s i t i o n h o l d i n g . F u r t h e r m o r e , t h e c l e a r s e p a r a t i o n o f s u b s t a n t i v e c o n c l u s i o n s f r o m m e t h o d o -l o g i c a l ones w i l l h e l p t h e s t r u c t u r e o f t h e s t u d y . S i n c e anyone o p e r a t i n g a t a s i n g l e measurement l e v e l w o u l d have made s t r a i g h t f o r w a r d i n t e r p r e t a t i o n s , t h e s t r a t e g y f o r s u b s t a n t i v e i n t e r p r e t a t i o n s w i l l be t o t a k e e a c h measurement l e v e l s e p a r a t e l y , i . e . , e a c h c o l u m n i n t a b l e X, and p r o c e e d v i a the two c r i t e r i a o f m a g n i t u d e and c o n s i s t e n c y . The m e t h o d o l o g i c a l i n t e r p r e t a t i o n s w i l l t h e n s o r t o u t any c o n f l i c t s , a m b i g u i t i e s , o r p r o b l e m s , g e n e r a t e d by t h e s u b s t a n t i v e i n t e r p r e t a t i o n s . I f no c o n f l i c t s a r i s e t h e n t h e m e t h o d o l o g i c a l p r o b l e m i s no p r o b l e m . 73 Stage One: The W i l e n s k y \ M r d c l Tested on , the Whole Sample The whole sample has a wider range of e d u c a t i o n , a u t h o r i t y , age, and o c c u p a t i o n , which c o u l d change the p rev ious f i n d i n g s . The q u e s t i o n i s : Where i s the g iven model a p p l i c a b l e ? Table X conta ins i n f o r m a t i o n on the t e s t of the three hypotheses d e r i v e d from W i l e n s k y ' s causa l c h a i n , as w e l l as the two f u r t h e r t e s t s of h i s c e n t r a l r e l a t i o n : o r d e r l i n e s s to p a r t i c i -p a t i o n . Subs tant ive I n t e r p r e t a t i o n s The f i r s t o v e r a l l impres s ion that the minimum of e m p i r i c a l r e l a t i o n e x i s t s i n the data can be r a t h e r d e p r e s s i n g . The h i g h e s t zero-order va lue (.292) i s the c o r r e l a t i o n between educa t ion and p o s i t i o n s . However, i n comparison to W i l e n s k y ' s study (1961a, t a b l e V , p . 5 3 4 ) where the h ighes t percentage d i f f e r e n c e s a r e : 14% f o r a 2 x 2 t a b l e f o r o r d e r l i n e s s and memberships, and 23% f o r a 2 x 3 t a b l e f o r educa t ion and 31 memberships, t h i s f a c t does not appear as d i s t r e s s i n g . On that comparative b a s i s and i n the framework of r e p l i c a t i o n , our e m p i r i c a l r e l a t i o n s are a t l e a s t noteworthy , though not a s tounding . Dichotomcus l e v e l . The f i r s t hypothes i s (column 1 of i n t a b l e X ) , the e d u c a t i o n - o r d e r l i n e s s r e l a t i o n , i s confirmed at e x a c t l y the same magnitude as i n the middle-mass sample ( t a b l e V I ) . The 31. Our s t a t i s t i c s are a l so s t r i c t e r ( W i l s o n , 1970; R u t h e r f o r d , 1971). 74 Table X. The Test of the Wilensky Model in the Whole Sample at the Three Levels of Measurement: Phi, Tau(b), and Pearson's r Relations Level of Measurement H, (b) N Dep. Ind. Control Dichotomous Ordinal Interval 0 R Tb R r R ord. 32 educ. .16 + . 21 + .26 + i n c l . ord. .13 + .09 0 .07 0 excl. ord. .25 + .18 + .20 + att. ord .06 0 .03 0 .03 0 hrs. ord. .06 0 .09 0 .08 0 , pes. ord. .02 0 .05 0 .05 0 i n c l . educ. .01 0 .10 + .21 + excl. educ. .14 + .17 + .28 + att. educ. .09 0 .06 0 .11 + hrs. educ. .13 + .13 + .22 + pos. educ. .19 + .24 + .29 + i n c l . educ. ord. -.01 0 .08 0 .20 + excl. educ. ord. .11 + .14 JL ! .24 + att. educ. ord. .03 0 .05 0 .11 + hrs. educ. ord. .13 + ,12 + .21 + pos. educ. ord. .19 + .24 + .29 + i n c l . ord. mob. .14 + .10 .07 0 excl. ord, mob. . 23 + .17 + ,19 + att. ord. mob. .06 0 .03 Q .04 0 hrs. ord, mob. .07 0 .10 + .08 0 pos. ord. mob. .00 0 .04 0 .04 0 i n c l . ord. res. .14 + .09 0 .07 0 excl. ord., res. .26 + .18 + .20 + att. ord. res. .06 0 .03 0 .04 0 hrs. ord. res. .06 0 .10 + .03 0 pos. ord, res. -.01 0 .06 0 .06 0 (498) (498) (498) 32. The abbreviations stand for: 'ord.' - orderliness; 'educ' -education; 'mob.' - mobility; 'res.' - residence; ' i n c l . ' - memberships including; 'excl.' - excluding; 'att.' - attendance; 'hrs.' -hours; and 'pos.' - positions. 75 truncation of education in the middle-mass sample has not affected that relation in any appreciable way. The second hypothesis, the orderliness-participation relation, finds confirmation only at the level of a f f i l i a t i o n . A l l of the indicators of involvement are negligible by our criterion. Again, as in the f i r s t sample, a l l relations are at least consistent with the positive hypothesis. Therefore, as far as this segment of the model is concerned there is no confirmation. The f i r s t part of the third hypothesis is tentatively confirmed. Education is positively related to participation. That education affects participation is most evident when the organizations are more voluntary (i.e., excluding is greater than including) and when the involvement becomes more stringent ( i . e . , positions > hours > attendance). Furthermore, the control for orderliness, though systematically reducing a l l measures, reduces them only a very l i t t l e . Education is directly and positively linked tc organizational participation. Orderliness of work careers does not mediate between education and participation as the given model expressly states. Mobility and residence have similar negligible effects on the orderliness-participation relation. They neither make negligible relations (with involvement) non-negligible, nor make non-negligible relations (with a f f i l i a t i o n ) negligible. At the dichotomcus level of analysis, our interpretative causal model differs from Wilensky's and takes the form of figure 9 below. Notably, the orderliness-participation relationship is 76 Figure 9. The I n t e r p r e t a t i v e Causal Model for the Whole Sample ORDERLINESS OF absent. The assessr-ent of the spuriousness of the o r d e r l i n e s s - p a r t i c i p a t i o n r e l a t i o n involves a s h i f t i n c r i t e r i a to c o r r e l a t i o n theory. The r e l a t i o n i s spurious but also below ±.10. Given the f a c t that the nesessary s t a t i s t i c a l j u s t i f i c a t i o n f o r the model, namely that r < < r , where r =0, does not hold f o r ' yx.z yx yx.z some in d i c a t o r s at a l l three l e v e l s of measurement (table X), the proposed model needs q u a l i f i c a t i o n . The i n d i c a t o r which most c l e a r l y points to the inadequacy of the proposed model i s also the only i n d i c a t o r which includes only voluntary organizations, namely 'excluding". The i n e q u i t y , r > > r r , i s probably due to t h i s f a c t . The other four i n d i c a t o r s ox oe ex' r J (under Pearson's r) do not evidence t h i s s t a t i s t i c a l inadequacy. I t seems that the model of f i g u r e 9 holds f o r the more compulsory or non-voluntary organizations and a causal model which also purports the o r d e r l i n e s s -p a r t i c i p a t i o n r e l a t i o n holds f o r the more voluntary organizations. This consideration i s outside our immediate task. Our model derives d i r e c t l y from the single-hypothesis approach to model b u i l d i n g . Such an assessment 77 i s made according to the c r i t e r i a of magnitude and consistency, which i n e f f e c t 'average' down the f i v e i n d i c a t o r s of p a r t i c i p a t i o n . Ordinal l e v e l Ue need waste no space i n declaring that the findings as regards the confirmation of the separate hypotheses are i d e n t i c a l to those above and that the general i n t e r p r e t a t i v e nodel i s again as i n fi g u r e 9. That model now gains stronger support i n terms of increased magnitudes f o r the education r e l a t i o n s . I n t e r v a l l e v e l Again, the hypotheses are confirmed and cisconfirraed i n a s i m i l a r pattern and the general i n t e r p r e t a t i v e model remains. The two p o s i t i v e r e l a t i o n s gain i n magnitude. Given the discovery of i n t e r a c t i o n between education and orderliness i n the niddle-mass sample, wben education i s p a r t i t i o n e d (table IK), the c a l l f o r an education p a r t i t i o n f o r the whole sample (tables X and XI) and f o r the working-class sample (tables XII and XIII) must be answered. As a preliminary check f o r s i m i l a r i n t e r a c t i o n , we generated p hi c o r r e l a t i o n s f o r the following models for both samples: 0 i I $ i i $ i s $ / i and 'fi . } where V indicates a p a r t i t i o n po/e pra/e pr/e po/me' po/re r on the subsequent v a r i a b l e s . The tables of values, not shown here, demonstrate several things: (1) the i n d i c a t i o n of i n t e r a c t i o n Is very strong f o r most models i n both sarsples; ( 2 ) memberships excluding i s the more i n t e r e s t i n g caso of i n t e r a c t i o n and needs further study: and (3) mobility and education c l e a r l y i n t e r a c t on a l l f i v e i n d i c a t o r s and i n opposite d i r e c t i o n s f o r the two samples. 77a Furthermore, the exploration of I n t e r a c t i o n v i a the p a r t i t i o n i n g technique i s deemed outside the proposed scope of t h i s t h e s i s . He have resolved to construct .an i n t e r a c t i o n model by the method of cr o s s -products of v a r i a b l e s . Needless to say, the j u x t a p o s i t i o n of the m u l t i p l i c a t i v e technique and the p a r t i t i o n i n g technique poses an I n t e r e s t -ing methodological problem In respect to discovering and t e s t i n g Inter-action among v a r i a b l e s . As regards Wilensky's causal-sequence model then, i t must be concluded that i t i s not confirmed i n the M i l l p o r t random sample. Education d i r e c t l y explains organizational p a r t i c i p a t i o n and explains more of i t than o r d e r l i n e s s does (see also f i u g r e 10), e s p e c i a l l y when Interpreting measures of involvement as against a f f i l i a t i o n . However, the orde r l y persons do hold more memberships i n voluntary organizations than the d i s o r d e r l y . Figure 10, the complete path model, adequately demonstrates these f a c t s on our two c e n t r a l dependent v a r i a b l e s : memberships excluding churches and unions, and position holding. M o b i l i t y Is n e g l i g i b l e i n both path 78 models, when a l l other independent variables are controlled. Orderliness is significant only for a f f i l i a t i o n . Residence is significant for position holding. Two crucial points have become apparent in the preceding analysis. F i r s t l y , in the light of the fact that attendance at meetings is a pervasively used indicator in participation studies, i t should be stressed how poorly i t indicates here (tables VI and X) relative to the other indicators, and how susceptible to cutting points i t appears to be (see footnote no.25 on page 62). There are of course other ways of measuring attendance than number of meetings attended in six months, but i t is one measure that skips about in our data. Secondly, i t has become evident that the indicators of involvement and of a f f i l i a t i o n must be kept quite separate —> as Wilensky's and our consistency criterion does not do. With respect to orderliness they behave quite differently (see tables VI and X). With respect to the previous test for interaction (table IX) , they seem to be best represented by the suppression and s p e c i f i -cation interaction models, respectively. That i s , under conditions of low education, the orderly-careered persons behave quite differently than the disorderly with respect to a f f i l i a t i o n and involvement. More precisely, the relative behaviours of the orderly and the disorderly are reversed. In a f f i l i a t i o n the disorderly lag behind; in involvement the orderly lag behind (see table IX) —when both have low education. 79 Figure 10. Path Diagram of the Four Causal Factors and Organizational A f f i l i a t i o n and Involvement: The Whole Sample (A) A f f i l i a t i o n : Memberships excluding churches and unions: (B) Involvement: Number of positions: 33. The path coefficients, i.e., the straight lines, are fourth-order partials and w i l l not correspond to the respective second-order partials found in table X. For the calculations and the discussion of the path coefficient, see G.T. Nygreen, Inter-active Path Analysis, American Sociologist, 1971, 6:37-43 (No.l). 80 Methodological Interpretations Although our general interpretative models are identical at the three measurement levels (figure 9) — when "averaging" down the five indicators of participation — a more detailed comparison and perhaps more theoretically-sound effort raises interpretative d i f f i c u l t i e s . That i s , i f the interpretative focus shifts from taking the five indicators together to taking them separately, then the differences between measurement levels appear. On the basis of our criterion of magnitude, we would have made 9 cut of 2 5 "errors" in substantive conclusions on single indicators i f we had chosen any particular measure-ment level. That i s , comparing along the rows of values in table X, there are 9 instances where the values are both above  and below our significance level of ±.10 in the same relation. In other words, there are 9 rows which would not evidence uniformity in assessed relations. The others are either consistently above or consistently below that magnitude. For example, in the single-indicator comparison involving memberships including church and union and education, a conclusion at the dichotomous level is radically different from ether levels (0 = .01; T b = .10; r = .21). Several tentative generalizations are in order at this s t a g e — t o help pinpoint problems for future investigation. These generalizations concern themselves with the nature of our univariate distributions. Wherever the underlying interval 81 distributions are relatively non-dispersed, e.g., education, memberships, and positions, there is a steady increase in the magnitude of the correlation values as one moves from dichotomous to interval measurement assumptions. For example, the relation of education to positions shifts from 0 = .19 to = .24 to r = .29. Wherever there is considerable interval dispersion, i.e., a scattering cf outliers along an extensive scale, as in attendance and hours spent (see Appendix III (G) and (H)), there is an incoherent pattern. Wherever such dispersion is accompanied by extremely non-normal frequency distributions, as in the bimodal case of orderliness, there is a reverse order of values. Examples of the last two patterns are orderliness and hours: .06 to .09 to .08; and orderliness and memberships including: .13 to .09 to .07. In 11 out of 26 possible comparisons (table X), 0 appears larger than or r - contrary to Davis' assumption (1971, p.3) about "crudeness" being "the same as conservativism." These seem mainly due to the bimodal distribution of order-liness and to interaction among variables (as shown in table 34 IX). Therefore, the collapsing of scales — either before or after the actual measuremdnt process — does not always bring the expected conservative results. In terms of solutions, a larger sample would f i l l out the gaps in those extensive scales, and probably smooth out the erratic associa-tion measures, but could not, alone, solve the interaction 34. In fact, these factors may be closely related in our data. 82 problem. Therefore, to conclude the f i r s t methodological interpretations, we see that some relative differences exist at the measurement l e v e l s — d u e to categorization, arbitrary cutting points, unique distributions, and interaction. "Conclusions about people" (Davis, 1971, p.3) would have differed at the different measurement levels for the individual indicators of organizational participation. Stage Two: The Interaction Model Tested on the Whole Sample Regardless of the significant and consistent relations found in the data with a minimum of controlling and under the assumptions of a linear, additive model, i t is likely that another theoretical and s t a t i s t i c a l model could better represent the relationships in the data. Thus far i n our analyses we have considered only linear, additive models. By the choice of statistics we have explicitly assumed that a l l zero-order relations were linear rather than curvilinear or.ad hoc relations, and that a i l independent variables had separate effects and could therefore be summed for their total effect on participation. It is possible that this may not be the case, both theoretically and empirically. Sonquist (1970, chapter 2) presents the convincing argument that interaction i s the rule and not the exception in sociology. Following the convincing argument of Sonquist (1970) concerning 83 the pervasiveness of interactions we suggest an alternative model which purports interaction between a l l variables. The Theory of Interaction on Differential Organizational  Participation The following is an attempt to develop a theory of social participation which includes primarily interaction effects. Education arouses expectations of an orderly work career. The pursuit of achievement is reinforced when expectations are met by an orderly career, because the orderliness of a work career makes the availability of resources predictable. This combination of factors w i l l lead to a greater display of achievement in organizational participation than the simple addition of the independent effects of the two factors. Furthermore, when the capacities, the social s k i l l s , gained from a higher education are combined with the anticipatory socialization of intra-generational mobility and the stability of resources, then such social s k i l l s are exercised and such anticipatory socialization finds i t s e l f in greater participation in voluntary organizations. This reasoning makes the assumption that more voluntary organizations are available to members of the higher social strata. (There must be. a greater number of arenas to exercise social s k i l l s , to 'socialize,' and to display achievement for the person who has achieved and is occu-pationally mobile than for him who has not achieved.) Finally, the theory must take into account not only capacities (education and social s k i l l s ) , the predictability cf resources, and preferences 84 35 Figure 11. The Cumulative Interaction Model (a) Orderliness Education Participation (b) Orderliness-education Residence ^ + Participation ( c l > Orderliness Education ^ Mobility 3> Participation (d) Orderliness-education-mobility + -7' Participation (e) Orderliness-e due at. ion-mobility Residence + + ^ Participation 35. The letter-labeling of each successive stage of the cumulative model corresponds to those found in tables XI and XIII in the following pages. For example, '(a)' of the above figure corre-sponds to '(a)' of table XI. 85 (occupational mobility speaks somewhat to this), but also exposure to opportunity structures. The longer a person is exposed to organizational opportunities the more he w i l l get 36 involved and a f f i l i a t e . The combination of a l l of these factors into a l i f e plan of work and leisure would evidence the greatest participation in organizational a c t i v i t i e s . There are several techniques for testing or finding interaction. At the dichotomous level, differences between percentage differences (e.g., table IX) has been a standard way of arriving at interaction. An alternative method is suggested by Sonquist (1970, p.409). Stinchcombe (1969, p.46) also suggests the simple multiplication between observed values for each unit of analysis for the relevant independent variiibles. Our strategy w i l l be to successively build upon our f i r s t interaction variable, education-orderliness, by systematically building up such cross-37 products. A respondent's education is simply multiplied by his orderliness ratio and correlated with a particular dependent variable. That new interaction term is then again multiplied by 36. The problem of an exposure curve, similar to the family l i f e cycle, in participation w i l l be bypassed here. Such a curve may also be complicated by or coincid>; with the family l i f e cycle and the age curve of participation — a l l temporal dimensions. 37. It w i l l be noticed that our cross-products interaction model makes a rather simple assumption about the exact nature of the combinations. It gives equal weight to a l l factors entering into the terms and combines "low" with "low" and "high" with "high." In a more refined study this simplistic combination would not be permissible. Such a simplistic assumption is acceptable for monotonic functions (including curvilinear monotonic) but not for non-monotonic or reversing functions between the said variables. A l l cross-products are based on interval levels scales. 86 the respondent's occupational mobility factor and again correlated with participation. At the various stages, length of residence is introduced as a specification variable. In effect, the new model is tested for short-term and for long-term residents. The cumulative model can then be judged at several stages and the assessment can rest at any stage. Figure 11 and table XI display the order in which the task progresses. Tha analysis of interaction w i l l be handled in two sections, both in the whole sample and in the working-class sample. F i r s t , the number of memberships excluding churches and unions, or a f f i l i a t i o n , w i l l be analyzed. Secondly, the holding of positions, or involvement, w i l l be studied under the same theory. Substantive Interpretations In table XI (and XIII), the judgment concerning the significance of the interaction variables w i l l entail a shift in c r i t e r i a . Since there w i l l be considerable overlap in the variation of the dependent variable that is explained by the interaction term and that by the summation of additive effects, our (essentially zero-order) c r i t e r i a must be deleted. The absolute magnitude of a correlation is no nonger adequate to assess importance. Because of this overlap, the difference between the coefficient of multiple determination for the additive 2 effects, e.g., R ,and the coefficient for the interaction 87 2 38 effect, e.g., R , cannot be equated with the total amount of x.oe M variation explained singly by either. We w i l l , therefore, simply compare such coefficients to see i f the interaction term adds a significant amount (based on a 5%, two-tailed test) beyond that  already explained by the additive effects. Such an assessment continues for long- and short-term residence. And since a l l data hitherto has been given in correlation values, the interaction analysis w i l l use the multiple correlation or the square root of the multiple determination coefficient. A f f i l i a t i o n Several interesting points can be l i f t e d out. Table XI ( f i r s t column) demonstrates that the two interaction terms, oe_ and oem, explain more than any of the independent effects (i.e., 'o', 'e', or 'm' taken singly), but less than the "summation" of the respective effects represented by the multiple co-relation. That i s , R is greater than R or R ; and R is greater than x.oe x.o x.e x.oe R . However, significantly more of the variation in a f f i l i a t i o n x.oe ' & J is explained by the supplemented interaction term (row "0 + E + (0)(E)" of table XI) to warrant a model containing both additive and inter-active effects. Furthermore, the addition of occupational mobility to the interaction term appreciably effects the explained variation only in the case of long-term residence. Finally, length of residence makes considerable difference as a 38. The symbol 'oe' (underlined) refers to the new interaction variable formed by the cross-product of education and orderliness; 'oem,' then also includes occupational mobility in the term. For a good discussion of the coefficients of multiple correlation and determination see Croxton, e_t. a l * (1967, p.473f). -,,2 _ rx o rx e " xo rxe roe a n d R2 - r 2 because oe is a single variable. - I . • • — X • 0 6 X 0 6 — — 1 - r2 — — 88 Table XI. The Cumulative Interaction Model Tested on the Whole Sample for A f f i l i a t i o n and Involvement: Coefficient of Multiple Correlation , r - A f f i l i a t i o n : Memberships Causal factors singly or in combination Uncontrolled Long Residence Short Residence Orderliness (0) Education (E) Mobility (M) .200 .278 .132 .090 -.025 .217 .283 .490 .100 (a) 39 0 + E .308 .101 .512 (b) (0)(E) .296 .070 .464 0 + E + (0) (E) .333 ..102 ' .547 (c) 0 + E + M .321 .232 .514 (d) (0) (E) (M) .306 .119 .473 0 + E + M + (0) (E) (M) .334 .237 .545 Involvement: Positions Orderliness (0) Education (E) Mobility (M) .054 .292 .117 -.048 .248 .223 .138 .340 -.061 (a) 0 + E .293 .270 .343 (b) (0)(E) .167 .041 .268 0 + E + (0) (E) .300 .270 .357 (c) 0 + E + M .307 .347 .356 (d) (0) (E) (M) .185 .106 .227 0 + E + M + (0) (E) (M) .307 .343 .365 N 498 238 260 39. The symbols refer to additive models, e.g., '0 + E'; or to interactive (cross-product) models, e.g., '(0)(E).' 89 specification effect. The additive and interactive effects on a f f i l i a t i o n differ considerably by length of exposure. The effects of orderliness and education on a f f i l i a t i o n are pronounced among short-term residents. This reversal of our prediction, regarding the place of exposure is not astounding since both Gans (1967) and Litwak (1960) point t o the phenomenon of a flurry of involvement by newcomers and the reduced selection of contacts and associations over a longer period of time. The conclusion about the pure interaction model proposed earlier i s that the f i n a l stage of the cumulative interaction model of figure 11 is s t i l l somewhat inadequate for a f f i l i a t i o n . There are combined additive and interactive effects. The shorter the length of exposure to organizational opportunities, the greater the combined additive and interactive effects of education, orderliness, and occupational mobility on a f f i l i a t i o n . Involvement The lower half of table XI contains similar data for the holding of positions. Although a l l of the various additive and interactive components do successively contribute significant amounts to the explanation o f the holding of positions (especially under short-term residence), i t is education which contributes the bulk of explained variation in involvement. And not even length of exposure has the same large specification effect that i t did for a f f i l i a t i o n . Therefore as regards our interaction model, i t 90 is not confirmed. There are additive and interactive effects, but education alone explains most of the variation in involvement. The Working-Class Sample The Scales A l l things remain the same for this last analysis except that several scales needed to be reconstructed to prevent extreme skewness of marginals from affecting measures of association. The resulting scales are: (1) for education: low (2-9 years), and high (10 years and up) for a dichotomy; and 2-7 years, 8-9 years, 10-11 years, and 12 years or more for an ordinal scale. (2) for orderliness: orderly (0.50 - 1.00) and disorderly (0.00 -0.49) for a dichotomy; and disorderly (0.00), somewhat disorderly (0.01 - 0.49), somewhat orderly ( 0.50 - 0.87), and orderly (0.88 - 1.00) for an ordinal scale. (3) for mobility: identical to scales of previous samples. (4) for residence: low (0-13 years), and high (14 years and more) for a dichotomy; and 0-8 years, 9-12 years, 13-19 years, and 20 years or more for the ordinal scale. (5) for memberships including: one or less memberships, and two or more for a dichotortious scale; and none, one, and two or more for an ordinal scale. (6) for memberships excluding: identical to scales of previous samples. (7) for attendance: identical to scales of previous samples. 91 (8) for hours spent: identical to scales of previous samples, and (9) for positions: none, and some for a dichotomy; and none, one position, and 2 or more positions for an ordinal scale. Stage One: The Wilensky Model Tested on the Working-Class Sample It is remembered that Wilensky extrapolated from the nature of his middle-mass sample to the working class. His e s t i -mation that there is much less orderliness in the working-class holds out in our data (See Appendix III (B)). The middle-mass evidences more orderliness than the working-class: 65% as compared with 40% at a ratio of greater than or equal to .60. But there is s t i l l considerable orderliness: 55% of this present sample are equal to or above .50 in our ratio scale. Given the incompleteness of our work histories, this fact needs qualification. However, the comparison between the present sample and our middle-mass sample can be made. These two samples are not mutually exclusive (Appendix IV); they do overlap, making the differences between them even more crucial. Substantive Interpretations The f i r s t overall impression cf the empirical relations found in the sample is not encouraging. As compared with the whole sample where 41 out of a possible 78 values met our magnitude criterion, the working-class sample displays only 31 values which 'meet the criterion (see table XII). Therefore, generally speaking, the empirical relations hypothesized find meagre support. 92 Table XII. The Test of the Wilensky Model in the Working-Class Sample at the Three Levels of Measurement: Phi 3 Tau (b), and Pearson's r Relations Hypo. Dep. Ind. Control H, ord. educ. i n c l . ord. excl. ord. att. ord. hrs. ord. pos. ord. i n c l . educ. excl. educ. att. educ. hrs. educ. pos. educ. Level of Measurement Dichotomous Ordinal Interval 0 R Tb R r R .07 0 .14 + .14 + .01 0 .03 0 .07 0 .12 + .07 0 .10 + -.07 0 -.04 0 -.01 0 .00 0 .01 0 .03 0 .05 0 .05 0 .06 0 .09 0 .11 + .13 + .09 0 .15 + .16 + .12 + .08 0 .15 + .16 + .15 + .18 + .23 + .23 + .25 + i n c l . educ. ord. .09 0 .11 + .13 + excl. educ. ord. .08 0 .14 + .14 + att. educ. ord. .13 + .09 0 .15 + hrs. educ. ord. .16 + .15 + .18 + pos. educ. ord. .23 + .22 + .25 + i n c l . ord. mob. .00 0 .00 0 .03 0 excl. ord. mob. .10 + .05 0 .08 0 att. ord. mob. -.10 - -.05 0 -.03 0 hrs. ord. mob. -.02 0 -.01 o .00 0 pos. ord. mob. .03 0 .04 0 .04 0 i n c l . ord. res. -.01 0 .02 0 .05 0 excl. ord. res. .11 + .07 0 .09 0 att. ord. res. -.08 0 -.04 0 -.02 0 hrs. ord. res. .00 0 .00 0 .01 0 pos. ord. res. .03 0 .03 0 .03 • 0 (198) (198) (193) 40. The abbreviations are those given in table X, the footnote. 93 Dichotomous levol Our assessment of the empirical relations, following rigidly the dictates of our c r i t e r i a , gives, rise to confirmation of only one relation at the dichotomous level. Only the education-participation relation finds confirmation, in that the involvement' indicators a l l reach the necessary magnitude and a l l values are positive. Furthermore, the controls for mobility and residence do nothing to the negligible relation between orderliness and participation. In conclusion, our interpretative model must be as in figure 12 below, where the one relation is important. Orderliness seems to play no significant role whatsoever. Ordinal level Our assessment at the ordinal level gives rise to a new empirical relation and therefore to a new interpretative model. Whereas the education-participation relation s t i l l remains and even gains stronger support — the education-orderliness relation also becomes significant. Again, mobility and residence leave the central relation unaffected. In conclusion at the ordinal level, the interpretative model now resembles those found in the previous sample. It appears in figure 13 below. There is no relation, not even a spurious one, between orderliness and participation. Interval level Our assessment at this level matches the ordinal analysis and so the interpretative model of figure 13 represents the 94 Figure 12. The Interpretative Causal Model for the Working-Class Sample at the Dichotomous Level of Measurement ORDERLINESS OF WORK CAREERS EDUCATION ^ ORGANIZATIONAL PARTICIPATION Figure 13. The Interpretative Causal Model for the Working-Class Sample at the Ordinal and Interval Levels of Measurement ORDERLINESS OF EDUCATION 95 Figure 14. Path Diagram of the Four Causal Factors and Organizational Affiliation^and Involvement: The Working-Class Sample (A) A f f i l i a t i o n : Memberships excluding churches and unions: •Education Memberships (B) Involvement: The number of positions: 41. The path coefficients, i.e., the straight lines, are fourth-order partials and w i l l not correspond to the respective second-order partials as found in table XII. For the calculations and the discussion of the path coefficients, see G. T. Nygreen, Interactive path analysis. American Sociologist, 1971, 6: 37-43 (No. 1). 96 interval level as well. Education takes on a very definite and significant role with regards to a f f i l i a t i o n and involvement, in the working-class sample. Since interpretative models diff e r somewhat from level to level, the conclusions with respect to Wilensky's model reside in the resolution of the measurement assumptions. Although we argue for the interval level and Pearson's r s t a t i s t i c , we s t i l l do not find Wilensky's model confirmed. The central relation involving orderliness and participation nowhere finds confirmation across the five indicators. Figure 14, which employs our two central dependent variables in complete path models, underlines the insignificance of orderliness. Education is important to both a f f i l i a t i o n and involvement. Mobility i s important to a f f i l i a t i o n , underlining the various mobility studies. And r e s i -dence i s again found to be highly important for involvement. Methodological Interpretations As has already been suggested, interpretative conflicts arise in this sample at both levels: at the level of the general interpretative mcdel and at the individual indicator level. The conflict between models directly challenges the Davis assumption about non-differing conclusions. The methodological assessment at the level of single indicators — also a more theoretically-sound procedure — adds considerable support to the challenge. Comparing along the rows of table XI for uniformity among relations, we find that 11 out of 25 comparisons (i.e., rows) 97 evidence non-uniformity. Also, on the conservativism-in-crude-stati s t i c s issue, we find 8 out of 26 .instances where phi i s somewhat larger than t a u ^ or Pearson's r. Stage Two: The Interaction Model Tested on the Working-Class Sample We refer the reader b*.ck to figure 11 and the accom-panying theoretical discussion of our cumulative interaction model for the background to table XIII which contains the test of that model. Again, the assessments w i l l be divided between a f f i l i a t i o n and involvement. A f f i l i a t i o n The block of multiple correlation values for a f f i l i a t i o n generally contains the same set of internal relationships as for the whole sample, but on a reduced scale. That, i s , the interaction terms generally explain more than the separate independent effects and less than the "summated" independent effects (R ). Also, the interaction terms always add a x.oe significant amount of explained variation beyond that explained by the additive model. However, the length of exposure as a specification variable now stands as we had predicted in our theoretical discussion — and not as in the whole sample. Those persons of the working-class who have a longer residential history show the highest correlation between their orderliness, education, and mobility and their a f f i l i a t i o n whether under an additive, an interactive, or a combined model of assumed relation-98 ships. There i s one exception to this pattern. Interestingly, the model of combined interactive and additive effects evidences a sharply increased degree of explanation under short exposure to opportunities. For both dependent variables, the last multiple correlation in the third column (i.e., for short residential histories) is by far the highest in the respective block of values. Involvement The second block of multiple correlation values in table XIII present much the same picture as the test for involve-ment in the whole sample. Although the picture is here somewhat more confusing, education remains the best, independent predictor of involvement among the working class of Millport. A l l other components of the models: orderliness, mobility, and residence, (whether parts of additive, interactive or combined models) do successively contribute to the amount of explained variation. The last model does not predict well for long-term residence but does for short-term residence. Again, the combined additive and interactive model explains the holding of positions remarkably well for short-term residents. We are not prepared to tackle some of the ambiguous inter-relationships among components of models found in table XIII. These are probably related to the high multiple correlation value for R , but we are not prepared to explicate x.oemoem r , short that complicated model any further. 99 Table XIII. The Cumulative Interaction Model Tested on the Working-Class Sample for A f f i l i a t i o n and Involvement: Coefficient of Multiple Correlation Causal factors singly A f f i l i a t i o n : Memberships or in combination Uncontrolled Long Residence. Short Residence Orderliness (0) .105 .176 -.017 Education (E) .157 .174 .139 Mobility (M) .150 .164 .069 (a) 0 + E .179 .247 .154 (b) (0)(E) .154 .225 .055 0 + E + (0)(E) .207 .264 .247 (c) 0 + E + M .219 .268 .178 (d) (0)(E)(M) .212 .233 .142 0 + E + M + (G)(£0(M) .267 .271 .394 Involvement: Positions Orderliness (0) .062 -.033 .117 Education (E) .251 .275 .229 Mobility (M) .116 .131 .010 (a) 0 + E .253 .277 .233 (b) (0)(E) .121 .022 .190 0 + E + (0)(E) .261 . 2 7 7 ; .299 (c) 0 + E + M .270 .291 .233 (d) (0)(E)(M) .148 .051 .213 0 + E + M + (0) (E) (M) .279 .291 .373 N 198 101 97 42. The symbols refer to additive models, e.g., '0 + E'; or to interactive (cross-product) models, e.g., '(0)(E).' 100 Comparisons Across the Three Samples Any comparisons across our samples must incorporate the facts of considerable overlap and the complete containment of the two subsamples in the whole sample. But these facts also provide some advantage. The overlap in samples (Appendix IV) means that any empirical differences found between them can be emphasized because such differences w i l l be due to those respondents that are in one sample and not in the other. The comparison of the working-class to the middle-mass allows us to see the behaviour of non-Company persons of the 'upper-middle' class. The comparison of the working-class to the whole sample allows us to see the behaviour of the persons in the higher socio-economic brackets (i.e., those of higher education and occupation). Only a few cross-sample comparisons w i l l be highlighted here; the summarizing chapter w i l l obviate others. S t i l l others have already been stated. The Wilensky model as such is not confirmed in any sample. Remaining within the confines of an independent-effects model, i t must be said that education remains the best predictor throughout the samples and orderliness appears to be important to participation only in the middle-mass sample which truncates the extremes in education and therefore, mostly the extremes in orderliness. Given the bimodal nature of the frequency distribution for orderliness in the whole sample, such a truncation could explain why orderliness appears important only in that sample. A bimodal distribution became a more rectangular distribution offering possi-b i l i t i e s of a higher correlation with other distributions. 101 In the interaction model, exposure, i.e., length of residence, operates somewhat differently for Company than for non-Company persons of Millport. For the Company workers (table XIII), exposure tends to stand as a positive specification between a f f i l i a t i o n and the causal factors. By deduction, exposure stands as a negative specification between a f f i l i a t i o n and the causal factors for the non-Company persons of Millport, i.e., for the higher socio-economic person (table XI). Intra-generational occupational mobility and orderliness of work careers tend to behave in reverse order on a f f i l i a t i o n with organizations for working-class or Company members and for non-Company persons. That i s , for the working-class persons (figure 14), mobility has a significant independent effect on a f f i l i a t i o n but not so orderliness; for non-Company or higher socio-economic persons (figure 10), orderliness has a significant independent effect on a f f i l i a t i o n but not so mobility. Again, this deduction is based on a rigi d application of the magnitude criterion and on the differences between a homogeneous subsample and the whole sample containing that subsample. 102 CHAPTER VI SUMMARIES, CONCLUSIONS AND LIMITATIONS Sumataries, The summaries to this present study are most eff i c i e n t l y and legibly given in point form, under the relevant substantive and methodological subcategories. Substantive Summaries 1) The amount of orderliness in our middle-mass subsample (65% - .60) and our working-class subsample (40% - .60) is very high compared to that in the Detroit sample. The working-class has a relatively lower degree of orderliness. Both statements must be qualified by the fact that our job histories are incomplete, probably mostly in terms of the disorderly years of many older respondents. An extrapolation, based on this qualification, w i l l be attempted shortly. 2) Wilensky's causal-sequence model — the particular linear, additive model we attributed to him — does not find support anywhere in our three samples as a "best f i t t i n g " model of the empirical relations. 3) Education in the middle-mass subsample behaved in an unexpected manner. A l l non-negligible zero-order relations with participation were negative. Interaction was posited, tested and confirmed for that sample. Education was found to specify the relation between orderliness and involvement, and to suppress the 103 relations between orderliness and a f f i l i a t i o n . Education, as a speci-fication variable, was not explored further. 4) In the Millport sample and the working-class subsample, education is found to stand in a direct link with participation and as a good predictor of both a f f i l i a t i o n and involvement. 5) For the working-class (Company) subsample, intra-generational mobility is meaningful with respect to a f f i l i a t i o n . For the higher socio-economic non-Company persons, orderliness of work careers has the significant effect. Both of these summaries pertain to the model of the additive effects of causal factors (figures 10 and 14). ?•) Attendance, a pervasive measure in the participation literature, behaved in an erratic manner and also somewhat unexpectedly. It stands in some non-negligible negative relations in the two subsamples. 7) Empirically, organizational involvement behaves differently from organizational a f f i l i a t i o n . The interaction model supports their theoretical separation. 8) Our cumulative interaction model, per se, does not find support in the Millport sample or in the working-class subsample. The addition of an interaction term to an additive model Is supported. Not explored f u l l y , the interaction between education and length of residence on involvement finds empirical support. 104 9) Under the interaction model assumptions, exposure to opportunities or length of residence operates in reverse order on a f f i l i a t i o n from working-class persons to nonworking-class persons. For the Company working-class persons, lengthy exposure to opportunities brings out a strong relation between a f f i l i a t i o n and the interaction of the causal factors. For the non-Company others, shorter exposure brings out the same effect. For involvement in organizations, both groups of persons evidence a similar positive effect under short-term residence. Methodological Summaries 1) Davis' dual assumptions about the non-differing results from measurement level to level and about crudeness as con-servativism i n st a t i s t i c s are both directly challenged by some of the relationships, especially at the more theoretically-sound single-indicator level of analysis. K 2) Of the two sample-analyses dealing with the measure-ment problem, the whole sample deviates less from level to level both for the general interpretative models and for the single-indicator interpretations. Conclusions Several conclusions can be drawn from the summary statements. Substantive Conclusions 1) With respect to the incomplete job histories and the amount of orderliness, some acceptable projections can be 105 made to cover the gap. We know that the mean recorded work history is 13J years long and that the mean gap is 9 years long. On the basis of the literature on work careers, we can assume that about two-thirds of that gap is disorderly for the average person. Knowing that the mean ratio, of orderliness in the Millport sample is .65, we can calculate the number of orderly years per respondent (.65 x 13.7 = 9 years orderly). Therefore, one-third of the 9 years (the gap) plus 9 years orderly totals to 12 years orderly over a projected work history of (13.7 + 9 =) 22.7 years. Accepting the assumption of two-thirds disorderliness in the gap in the records, then i t follows that the projected mean ratio of orderliness for the Millport sample is approximately (12/22.7 = ) .53. This s t i l l seems to be more orderliness than Wilensky found, although he did not work with means but with per cent of persons above the .50 ratio (which was 30%). 2) Since Wilensky's linear causal-sequence model as such is nowhere supported and since the cumulative interaction model, per se, is also not supported relative to the explanatory power of the additive model, we propose a model combining additive and interactive effects for a more refined study of differential organizational participation. However, for education and orderliness, the linear, independent-effects model also found some support in the case of a f f i l i a t i o n (whole sample, figure 10), therefore some crucial comparative work is needed. We have tested only a particular linear, additive model as opposed to a particular 106 interaction model. There exist many other p o s s i b i l i t i e s . 3) Why does a theoretical (interaction) model, expressly designed to explain involvement, explain more of a f f i l i a t i o n than of involvement? Two answers are possible and must be studied. Either the interaction terms are not refined enough in the proper weighting of factors for involvement, i.e., the wrong s t a t i s t i c a l model; or both the theoretical and the s t a t i s t i c a l model are wrong for involvement in organizations. The interaction between achievement and exposure on involvement is worth exploring as a new theoretical and s t a t i s t i c a l model. Furthermore, in future studies of differential organizational participation, indicators of a f f i l i a t i o n and of involvement must be kept separate both empirically and theoretically. This strategy may also help to explain some of the erratic and unexpected behaviour of the relations between education and attendance. 4) In this conclusion we wish to combine several findings and draw a more general conclusion. F i r s t l y , why is orderliness important in the middle-mass subsample and education important in the whole sample and in the working-class sample? Truncation of those cases in the extremes of education, and therefore orderliness, can account for some of this. Since education, a status measure, predicts well in these data, the question of the ascendency of orderliness of work careers over social status must at least be re-opened. Orderliness simply is not a factor in participation among the one-industry workers. Secondly, in respect to the 10=7 shifting of the relative roles of mobility and orderliness, for the working-class and the nonworking-class persons, a more complete exposition must be demanded. For the working-class person in the Company's employ, i t is the movement to new and more prestigious s k i l l s and experiences that lead him to a f f i l i a t e himself with voluntary organizations. For the non-Company person of Millport, i t is the consistency in s k i l l types and experiences that have this effect. These statements could be tentatively generalized to social classes for some further interesting study. To tie these loose ends together, suggestions in terms of new interaction models, which attempt to combine the central factors for each roughly-defined segment of Millport, would be as follows: i) For the xrorking-class persons, the interaction between education and occupational mobility under long exposure to organizational opportunities (i.e., long residence) should explain considerable participation in organizations. i l ) For the "nonworking-class" persons, the interaction between education and orderliness of work careers under short exposure should explain participation in organizations. 6) Finally, inasmuch as the upper socio-economic brackets in the data evidence stronger empirical relations than the working-class bracket, i t seems reasonable to suggest that the theory and hypothesized relations are s t i l l very much of middle- and upper-class orientations. The two theoretical models 103 do not tap working-class reality very well. Methodological Conclusions In the conclusion on the methodological issue, we wish to answer or redirect several questions. (1) What are the reasons for the differences between measurement levels, as we have defined measurement level? (2) Is there any great danger .in the choice of measurement level? (3) Where do we go from here? Because we deliberately chose to combine the several dimensions of the measurement problem, the reasons for the differences in results at measurement levels are also several. Further study should address each of these reasons in turn, although as noted some of this work has been done recently. F i r s t l y , the effect of categorization (Blalock, 1964, chapter IV) or the collapsing of scales causes loss of variation in variables and therefore loss of "explainable" variation. As we now see more clearly this is true only for linear statistics applied to s t r i c t l y linear relationships. Secondly, the incomparability of statistics may explain some of the differences (Rntherford, 1971). By "incomparability" is meant the fact that phi, tau(b), and Pearson's r employ differing types and/or amounts of information from a bivariate distribution. Thirdly, the differences at the measurement levels may be a result of the rigi d application of cri t e r i a including the fact of an identical magnitude applied uni-laterally across levels. However, rigi d i t y is an advantage in 1:Q9> that i t provides for uniformity in interpretation across empirical studies. Perhaps, i t might be suggested, the magnitude criterion should have been adjusted for measurement levels, but most tests of significance pay no attention to size of tables and such tests show the 5% significance levels to be quite similar for phi, tau(b), and r (see footnote, page 12) for samples of equal size. Tau(b) approaches normality very quickly with increasing size of samples (Hays, 1963, p.651f). The above reasons for differences have been directed 45 at the normally-expected pattern of differences between measure-ment levels, namely that phi w i l l be less than tau(b) and tau(b) less than Pearson's r or, for example, a pattern similar to 0 = .08, T, = .14, and r = .20. The above reasons would explain D that pattern of differences. Another type of difference between measurement levels that is also a crucial discovery is the case where phi is larger than tau(b) and tau(b) larger than Pearson's r. This speaks to the "crudeness as conservativisro in s t a t i s t i c s " issue. In the two samples that perform multi-level analyses, vre find 5 instances out of 52 comparisons where phi is greater than both the others and 4 instances where i t is greater than one of them. In some of these 9 instances phi is also significant where 45. Given the knowledge of the separate effects of categorization and of reduction to lower measurement levels, the positive relationship between increasing measurement level and increasing magnitudes of correlation values becomes the "normally-expected pattern." 110 the others are not. This methodological problem is partly due to the bimodal nature of the orderliness scale (especially in the whole sample where i t is pronounced) and to interaction among independent variables showing up, in an independent-effects model, in the form of a J-curve relation with attendance (see table IX). There seems to be a small reversal of relationship with attendance, for education and orderliness each taken singly. That i s , the interaction of low education with disorderliness tends to bring about the same behaviour, namely higher attendance, as the interaction of high education with orderliness (table IX). And the collapsing of bimodal orderliness has the effect of increasing i t s respective relationships with other variables — hence causing the pattern: 0 > T^ > r. Is there any great danger in the choice of measurement level? F i r s t l y , the obvious must be underlined; we need better measurement i n terms of higher levels. Then, aside from imprecise measurements, we have a dilemma. Either we apply rig i d c r i t e r i a , knowing these c r i t e r i a can cut directly through the middle of our tenuous magnitudes, and are assured of uniformity in interpretations. (This is also Davis' concern). Or we employ unstated c r i t e r i a and continue with ambiguity and implied differences in interpretations. Differences in magnitudes do exist, especially for single indicators, and some of them 111. are considerable. Such differences: can be exaggerated or conservative differences, implying entirely different underlying theoretical models. We need to explore the various dimensions of the metho-dological problem but we also need to explore more refined inter-action models — a s even our methodological problem suggests — and to explore varying types of interaction models of differential organizational participation. Limitations 1) A theoretical limitation, one not mentioned unt i l this point, is the adverse mixture of differing types of organi-zations into one block under organizational participation. For instance, the inclusion of sports groups alongside business and p o l i t i c a l groups could be questioned. For the theory of achieve-ment display this needs radical revision. We suggested that different theoretical models be set up for different types of organizationsand also for involvement in them. 2) Blalock's discussion of and the import of the iden-t i f i c a t i o n problem (e.g., 1967b) in status inconsistency, i.e., the nature of the interaction term, is simply bypassed in this study. In some sense, this is not crucial since Blalock nowhere attacks the existence of interaction but only the relative proportions of additive to interactive effects. We have assumed one of the many possibilities (Blalock, 1967b, p.307). 112 3) The measurement of orderliness of work careers has a serious limitation not only in terms of incomplete work histories but also in terms of the underlying concept of st a b i l i t y or predictability of resources. The criterion of consistency between johs, supposedly a f i r s t step to measuring s t a b i l i t y of work careers, runs into the serious validity problem. Does consistency actually help to measure stability? For example, a man with four jobs to his complete work history, entailing 1, 3, 2 and 25 years respectively and evidencing consistency between the f i r s t three jobs only, would have an orderliness ratic of 6/31 years or .19. However, are those 25 years (in the one job) not contributing most to the st a b i l i t y of his resources? Such a measure of stability would conclude with a ratio of 25/31 or .81. Some balance in these dimensions of the notion of s t a b i l i t y is needed. 4) The study does not enter the problem of theoretically assessing and interpreting the effects of logarithmic transfor-mations of the scales. n a ^ II"* dees nor €»ISI 114 BIBLIOGRAPHY Allart , E. 1958 "On the cumulative nature of leisure a c t i v i t i e s , " Acta  Sociologies 3: 165-172. Anderson, N. H. 1961 "Scales and s t a t i s t i c s : parametric and nonparametric," Psychological Bulletin 58: 305-316. Anderson, T. R., and M. Zelditch. 1968 A Basic Course in Statistics, New York: Holt, Rinehart and Winston. Anderson, W. 4. 1943 "Family and individual social participation," American  Sociological Review 8: 420-424. Axelrod, Morris. 1956 "Urban structure and social participation," American  Sociological Review 21: 13-18. Baker, B. 0., C. D. Hardyck, and L. F. Petrinovich. 1966 "Weak measurement versus strong s t a t i s t i c s , " Educational and Psychological Measurement 26: 291-309. Bell, W., and M. Boat. 1957 "Urban neighborhoods and informal social relations," American Journal of Sociology 62: 391-398. Bell, W., and M. T. Force. 1956 "Urban neighborhood types and participation in formal associations," American Sociological Review 21: 25-34. Biakelock, E. 1960 "New look at the new leisure," Administrative Science  Quarterly 4: 446-467. Blalock, H. M. 1962 "jpuriousness versus intervening variables: problem of temporal sequencing," Social Forces 40: 330-1964 Causal Inferences for Non-experimental Research. Chapel H i l l : University of North Carolina Press. 115 1965 "Theory building and the s t a t i s t i c a l concept of inter-action," American Sociological Review 30: 374-380. 1966a "The identification problem and theory building: the case of status inconsistency," American Sociological  Review 31: 52-61. 1966b "Comment: status inconsistency and the identification problem," Public Opinion Quarterly 30: 130-132. 1967a "Status inconsistency and interaction: some alternative models,'1 American Journal of Sociology 73: 305-315. 1967b "Tests of status inconsistency theory: a note of caution," Pacific Sociological Review 10: 69-74. 1967c "Status inconsistency, social mobility, status integration and structural effects," American Sociological Review 32: 790-801. 1967d "Path coefficients versus regression coefficients," American Journal of Sociology 72: 675-676. Blalock, H.M., and A. B. Blalock. 1968 Methodology in Social Research. New York: McGraw-Hill Book Co. Blau, P. M., and 0. D. Duncan. 1967 The American Occupational Structure. New York: Wiley and Sons Co. Boudon, Ray nond. 1968 "A new look at correlation analysis," in Blalock, H.M. and A.B., Methodology in Social Research. New York: McGraw-Hill Co. " Boyle, R. R. 1970 "Path analysis and ordinal data," American Journal of Sociology 75: 461-481. Brown, G. W., and M. Rutter. 1966 "Measurement of family activities and relationships," Human Relations 19: 241-264. Bruce, J. M. 1970 "Intragenerational occupational mobility and vi s i t i n g with kin and friend," Social Forces 49: 117-127. 116 Caplow, T., and R. Forman. 1950 "Neighborhood interaction in a homogeneous community," American Sociological Review 15: 357-366. Chapin, F. Stuart. 1939 "Social participation and social intelligence," American  Sociological Review 4: 157-166. Clarke, A.C. 1956 "Use of leisure and i t s relations to levels of occupations prestige." American Sociological Review 21: 301-307. Coleman, James. 1964 Introduction to Mathematical Sociology. Glencoe, New York: Free Press. Collins, Randall. 1971 "Functional and conflict theories of educational s t r a t i f i -cation," American Sociological Review 36: 1002-1019. Coombs, R. H. 1969 "Social participation, self concept and interpersonal valuation," Sociometry 32: 273-286. Costner, Herbert L. 1965 "Criteria for measures of association," American Sociolo- gical Review 30: 341-353. Cramer, M. W. 1950 "Leisure time activities of economically privileged children," Sociology and Social Research 34: 444-450. Croxton, F. E., D. J. Cowden, and S. Klein. 1967 Applied General Statistics. New Jersey: Prentice-Hall, third edition. Curtis, R. F. 1960 "Occupational mobility and church participation," Social  Forces 38: 315-320. Dahrendorf, Ral 1968 "On the origin of inequality among men," Tn Essays in the Theory of Society. California: Stanford tlni. Press, pp.151-1 Davis, 1964 James A. "A net partial and multiple coefficient for Goodman and Kruskal's gamma," unpublished manuscript. 1967 "A partial coefficient for Goodman and Kruskal's gamma," Journal of the American Statistic Association 62: 189-193. 1 1 7 1971 Elementary Survey Analysis. Hew Jersey: Prentice-Hall Inc. De Grazia, Sebastian. 1962 Of Time, Work, and Leisure. New York: Doubleday and Co. De Tocqueville, Alexis. 1951 Democracy i n America. New,York: Alfred A. Knopf. Duncan, Otis D. 1966 "Path analysis: Sociological examples," American Journal  of Sociology 72: 1-16. Duncan, Otis D., et. a l . 1968 "Power influences on aspirations: reinterpretation," American Journal of Sociology 74: 119-137. Durkheim, Entile. 1933 The Division of Labor in Society. Glencoe: Free Press, translated by George Simpson. Engels, Frledrich. 1892 The Condition of the TJorking-Class In England in 1844. Translated by Florence K. Wischnewetsky, London: Allen and Unwin Ltd. F e l l i n , P., and E. Litwak. 1963 "Neighborhood cohesion under conditions of mobility," American Sociological Review 28: 364-376. Form, W. H., and D. C. M i l l e r . 1949 "Occupational career pattern as a sociological instrument," American Journal of Sociology 54: 317-329. Foskett, J. M. 1955 "Social structure and social participation," American  Sociological Review 20: 431-438. Gans, H. J . 1967 The Levlttowners. New York: Pantheon Books. Gerstl, J. E. 1961 "Leisure, taste and occupational milieu," Social Problems 9: 56-68. Geschwender, J. A. 1968 "Status inconsistency, social isolation, and individual unrest," Social Forces 46: 477-483. Goldberger, A. S. 1970 "On Boudon's method of linear causal analysis," American  Sociological Review 35: 97-117. nr. Goodman, L. A., and W. H. Kruskal. 1954 "Measures of association for cross classifications," Journal of the American Statistic Association 49: 732-764. Gordon, R. A. 1968 "Issues in multiple regression," American Journal of Sociology 73: 592-616. Gross, Edward. 1961 "A functional approach to leisure analysis," Social Problems 9: 2-9. Haberman, P.K., and J. Shcinbergh. 1966 "Education reported in interviews: an aspect of survey content error," Public Opinion Quarterly 30: 295-301. Hagedorn, R., and S. Labovitz. 1968a "Participation in community associations by occupations: a test of three theories," American Sociological Reyiexj 33: 272-283. 1968b "Occupational characteristics and participation in voluntary associations,'.' Social Forces 47: 16-27. Harry, J. 1970 "Family localism and social participation," American Journal of Sociology 75: 821-827. Hauser, Robert M. 1969 "Social participation and social status," American  Sociological Review 34: 549-553. 1970 "Content and consex: cautionary tale," American Journal  of Sociology 75: 645-664. Hawkes, R.K. 1971 "The multivariate analysis of ordinal measures," American Journal of Sociology 76: 908-925. Hay, D. L. 1950 "Social participation of households in selected rural communities of the northwest," Rural Sociology 15: 141-143. Hays, W. L. 1963 Statistics, New York: Holt, Rinehart and Winston. Heise, D. R. 1969 "Problems in path analysis and causal inference," in E. Borgatta, Sociological Methodology, 1969. San Francisco: Jossey-Bass. 119 Hilander, J. S. 1969 "Goal aspirations as ^ related to select measures of social participation," Dissertation Abstracts 30 (3-A) 1242. Hilgendorf, L., et. a l . 1967 "Combined use of linkage and path analysis in the develop-ment of causal models," Human Relations 20: 375-387. Hodge, R. W., and D. J. Treiman. 1968 "Social participation and status," American Sociological  Review 33: 722-740. Indik, B. P. 1965 "Organizational size and member participation," Human  Relations 18: 339-351. Kendall, Maurice G. 1955 Rank Correlation Methods. New York: Hafner Publishing Co. Kim, Jae-On. 1971 "Predictive measures of ordinal association," American  Journal of Sociology 76: 891-907. Knupfer, Genevieve. 1947 "Portrait of an underdog," Public Opinion Quarterly 11: 103-114. Labovitz, Sanford. 1970 "The assignment of numbers to rank order categories," . American Sociological Reviex^'35: 515-523. Land, K. C. 1969 "Principles of path analysis," in E. iiorgatta, Sociological  Methodology. 1969. San Francisco: Jossey-Bass. 1970 "On the estimation of path coefficients for unmeasured variables from correlations among observed variables," Social Forces 48: 506-511. Lazarsfeld, P. F. 1955 "Interpretation of s t a t i s t i c a l relations as a research operation," in P. F. Lazarsfeld and Rosenberg, The Language  of Social Research. Glencoe: Free Press. Lazerwitz, B. 1961 "Some factors associated with variations in church attendance," Social Forces 39: 301-309. Leik, R. K., and W. R. Gove. 1969 "Conception and measurement of asymmetric monotonic relation-ships in sociology," American Journal of Sociology 74: 696-709. 120 Lenski, Gerhard. 1956 "Social participation and status crystallization," American Sociological Review 21: 458-464. Lipset, S. M., and R. Bendix. 1952 "Social mobility and occupational career patterns," American Journal of Sociology 57: 366-374. 1967 Social Mobility in Industrial Society, Berkeley: University of California Press. Lipset, S. M., and F. T. Malm. 1955 "First jobs and career patterns," American Journal of Economics and Sociology 14: 247-261.' Litwak, E. 1960 "Reference group theory, bureaucratic career and neighborhood primary group cohesion," Sociometry 23: 72-84. MacDonald, M, , C. McGuire, and R, J. Havighurst. 1949 "Leisure activities and the socioeconomic status of children," American Journal of Sociology 54: 505-519. MacFarland, D. D. 1970 "Intra-generational social mobility as a Markov process," American Sociological Review 35: 463-476. McGinnis, R. 1958 "Logical status of the concept of association," Midwest  Sociologist 20: 72-77. McNemar, Q. 1962 Psychological Statistics. New York: Wiley and Sons. Mannheim, Karl. 1940 Man and Society in an Age of Reconstruction. Trans-lated by E. A. ShiTs. New York: Harcourt, Brace and Co. Marx, Karl, and Friedrich Engels. 1959 Basic Writings on Politics and Philosophy. Edited by Lewi's S. Feuerl New York: Doubleday and Company. Meissner, Martin. 1971 "The long arm of the job," Industrial Relations 10: 239-260. Nygreen, G. T. 1971 "Interactive path analysis," American Sociologist 6: 37-43. Osterreich, Helgi, 1965 "Geographical mobility and kinship," International Journal  of Comparative Sociology 6: 131-144. 121 Pope, Hallowell. 1964 "Economic deprivation and social participation in a group of 'middle class' factory vrorkers," Social  Problems 11: 290-300. Press, L. I., et_. a l . 1969 "An interactive technique for the analysis of multivariate data," Behavioural Science 14: 364-371. Rehberg, R. A., and D. L. Westby. 1967 "Parental encouragement, occupation, education and family size: artifactual or independent determinants of adolescent educational expectations," Social  Forces 45: 362-373. Reissman, L. 1954 "Class, leisure, and social participation," American Sociological Review 19: 76-84. Reisman, D. 1950 The Lonely Crowd. New Haven: Yale University Press. Rosenberg, M. 1962 "Test factor standardization as a method of inter-pretation," Social Forces 41: 53-61. Ritherford, Brent M. 1971 "Non-metric correlational methods: a sensitivity analysis using Monte Carlo simulation," Department of P o l i t i c a l Science, University of Oregon. A paper presented at the 1971 PSA meetings, Honolulu. Scott, J. C. 1957 "Membership and participation in voluntary associations," American Sociological Review 22: 315-326. Sewell, W. H., e_t. a l . 1969 "Educational and early occupational attainment process," American Sociological Review 34: 32-92. Smith, Joel, and H. D. Rawls. 1965 "Standardization of an educational variable: the need and i t s consequences," Social Forces 44: 57-65. Snedecor, G. W., and W. G. Cochran. 1967 Statistical Methods. Iowa: Iowa State University Press. Somers, R. H. 1959 "The rank analogue of product-moment partial correlation and regression," Biometrika 46: 241-246. 122 1962 "A new asymmetric measure of association for ordinal variables," American Sociological Review 27: 799-811. 1968a "On measurement of association," American Sociological  Review 33: 291, 292. 1968b "An approach to the multivariate analysis of ordinal data," American Sociological Review 33: 971-977. Sonquist, John A. 1970 Multivariate Model Building, Michigan: Braun and Brumfield. Stevens, S. S. 1946 "On the theory of scales of measurement," in A. Danto and S. Morgenbesser, Philosophy of Science, 1960, Stuchert, R. P. 1963 'Occupational mobility and family relationships," Social Forces 41: 301-308. Sutcliffe, J. P., and B. D. Crabbe. 1963 "Incidence and degree of friendship in urban and rural areas," Social Forces 42: 60-68. Teele, James E. 1962 "Measures of social participation," Social Problems 10: 31-39. Thielbar, G., and S. D. Feldman. 1969 "Occupational stereotypes and prestige," Social  Forces 48: 64-72. Turner, M. E., and C. D. Stevens. 1959 "Regression analysis of causal paths," Biometrika 15: 236-258. Turner, R, H. 1960 "Modes of social ascent through education: sponsored and contest mobility," in R. Bendix and S.M. Lipset, Class, Status  and Power, Glencoe: Free Press, second edition. Veblen, Thorstein 1934 The Theory of the Leisure Class. Hew York: Modern Library. Vorwaller, D. J. 1970 "Social mobility and membership in voluntary associations," American Journal of Sociology 75: 481-496. Whyte, W. H„ Jr. 1953 "How the new suburbia socializes," Fortune (August):120-122, 186-190. 123 1956 The Organization Man. New York: Doubleday Co. Wilensky, H. L. 1960 "Work, careers and social integration," International Social Science Journal 12: 543-560. 1961a "Orderly careers and social participation," American  Sociological Review 26: 521-539. 1961b "Life cycle, work situation, and participation in formal associations," in R. W. Kleemeier, Aging and  Leisure, New York: Oxford University Press. 1962 "Labor and leisure: intellectual traditions," Industrial Relations 1: 1-12. Wilson, Thomas P. 1969 "A proportional-reduction-in-error interpretation for Kendall's Tau-B," Social Forces 47: 340-342. 1970 "Measures of association for bivariate ordinal hypotheses," unpublished paper, Santa Barbara, California. Wright, C. B., and H. H. Hyman. 1958 "Voluntary association memberships of American adults," American Sociological Review 23: 284-294. Wright, Sewall. 1960 "Path coefficients and path regressions," Biometrika 16: 189-202. 1960 "Treatment of reciprocal interaction, with or without lag, in path analysis," Biometrika 16: 423-445. Wylie, J. 1953 "Survey of 504 families to determine the relationships between certain factors and the nature of the family recreation program ," Research Quarterly 24: 229-243. Young, J. N., and S. C. Mayo. 1959 "Manifest and latent participators in a rural community action program," Social Forces 38: 140-145. Zimmer, B. G. 1955 "Participation of migrants in urban structure," American Sociological Review 20: 2.18-224. Zimmer, B. G., and A. H. Hawley. 1959 "Suburbanization and church participation," Social Forces 37: 343-354. 1-26 A P P E N D I C E S I RAW DATA FOR DICEOTOMOUS CROSSrCLASSIFICATIONS (A) Oichotoinous Cross-Classifications for the Middle-!"ass Sample (B) Dichototaous Cross-Classifications for the Whole Sample (C) Dichotomous Cross-Classifications for the Uorking-Class Sample II VARIANCES AND COVARIANCES FOR ORDINAL CROSS-CLASSIFICATIONS (A) Variances and Covariances for the "hole Sample Ordinal Cross-Classifications (B) Variances and Covariances- for the Working-Class Sample Ordinal Cross-Classifications III UNIVARIATE DISTRIBUTIONS, POINT ESTIMATES AND MEASURES OF NORMALITY (A) Univariate Distribution of the Number of Years of Education (B) Univariate Distribution of the Ratio of Orderliness of Work Careers (C) Ordinal Distribution of Intragenerational Occupational Mobility (D) Univariate Distribution of the Number of Years in Residence (E) Univariate Distribution of the Number of Memberships Including Church and Union (F) Univariate Distribution of the Number of Memberships Excluding Church and Union (G) Univariate Distribution of the Number of Meetings Attended (H) Univariate Distribution of the Number of Hours Spent in Organizational Activities (I) Univariate Distribution of the Number of Positions Held or Presently Holding IV THE GENERAL DISTRIBUTION OF CASES IN THE THREE SAMPLES OVER EDUCATION AND OCCUPATIONAL POSITION V DATA TRANSFORMATIONS: SKEWNESS AND KURTOSIS, TRANSFORMATION FORMULAE, AND EFFECTS Ob PEARSON'S R VI "ORDERLY CAREERS AND SOCIAL PARTICIPATION", AN ARTICLE BY HAROLD L. WILENSKY VII THE DISPOSITION GF THE POPULATION AND SAMPLE 127 AFPEiaiX I RAW DATA FOE DICHOTOMOUS CKOSS-CLASSIFICA.TIONS (A) Dichotomous Cross-Classifications For The Middle-Haas Sample VARIABLES! Memberships: Including Excluding Attendance Hours Positions Residenoe Mobility Orderliness Education hi lo hi lo hi lo hi lo hi lo hi lo hi lo hi lo Educ*tion: high (N-105) 57 48 59 46 40 65 43 62 58 47 75 30 58 47 75 30 • var-.241 low («=. 72) 52 20 37 35 38 34 23 49 41 31 49 23 35 37 40 32 Orderliness: high (N-115) 77 38 75 40 52 63 46 69 68 47 87 28 73 42 var-,228 • oov=.038 low (N- 62) 32 30 21 4t 26 36 20 42 31 31 37 25 20 42 Mobility: high (N- 93) 59 34 59 34 42 51 31 62 54 39 72 21 var-.249 cov-.071 cov».016 low (N- 84) 50 34 37 47 36 48 35 49 45 39 52 32 Residenoe: high (H-124) 74 50 64 60 54 70 43 81 74 50 var-,210 oov-,039 oov-,036 oov-,008 Ion (a- 53) 35 18 32 21 24 29 23 30 25 28 Tar-.237 var».248 var=.2.47 var».234 var=.247 177 Memberships, including, high (jMOjjj c o v „ _ . 0 1 3 c o v „ . 0 1 0 c o v„.o 35 oov~.043 Memberships, excluding: high JN- 96) c o v . . < 0 1 8 C O V . . 0 4 8 0 O V . . 0 V 1 o o v . . 0 i 2 low (N- 81) Attendance: high (N» 78) _-.N ..... i /„> n „ \ cov—.004 OOT».006 cov-.008 oov»-.035 low \N= 99J Hours: Positions: hi«h low high low (N- 66) (N-111) (»- 99) (K= 76) cov—.018 COT—.021 cov-.018 cov-.022 cov-,026 COT-.011 cov-.021 cov— .004 Variance and Covariance measures are taken from: Roland K. Havkes, "Multivariate Analysis of Ordinal ^ Measures." American Journal of Sooiology. 76:919, table 1, 1971. Optimum values for Tarlanoe and covariance •re .250 in the 2 X 2 table. (fl) Dichotomous Cross-Classifications For The Whole Sample VAHIABLKS: Memberships, Including Excluding Attendance Hours Positions Residence Mobility Orderliness Education hi lo hi lo hi lo hi lo hi lo hi lo hi lo hi lo Education: high (N-271) 183 88 169 102 157 114 130 141 171 100 165 106 129 142 182 89 • vara.248 low (K-227) 151 76 109 118 111 116 79 148 100 127 137 90 108 119 117 110 Orderliness: high (N-299) 216 83 197 102 168 131 133 166 165 134 197 102 164 135 var-.240 . cov-,039 low (K-199) 118 81 81 118 100 99 76 123 106 93 105 94 73 126 Mobility: high (N-237) 157 80 145 92 130 107 92 145 142 95 171 66 var=.249 cov*.044 oov-,0001 low (H-261) 177 84 133 128 138 123 117 144 129 132 131 130 Residence: high (M-302) 202 100 164 138 164 138 127 175 192 110 var-.239 oov-.055 C O Y - , 0 3 2 cov-,001 lo * (U-196) 132 64 114 82 104 92 82 114 79 117 var» .221 var=.247 var-,249 var» .244 var-.248 N - 498 Memberships, including: high (N=334) low (K-1 64) oov—.O01 oov—.004 oov.031 oov-.003 Memberships, excluding, high (N=.278) c o v = _ > 0 0 9 O O V . . 0 2 6 cov=.06O cov=.036 low VN=220J Attendance, high (N=268) C O V l l > 0 0 3 o o v , .005 cov-.OH cov=.022 low ^=230; Hours: high (N-209) O O T = # 0 0 , O O V„_. 0 1 5 C O V . . 0 I 5 c ov» .033 low \h-i£ti'j) Positions: high (N=271) low (N=227) cov=.056 oov-.026 cov~.005 oov».047 Variance and covariance measures are taken from: Roland K. Hawkes, "Multivariate Analysis of Ordinal Measures." American Journal of Sooiology. 76,919, table 1, 1971. Optimum values for variance and covariance are .250 for 2 X 2 tables. (C) Dichotomous Cross-Classifications For The Vorking-Class Sample VARIABLES i Memberships: Including Excluding Attendance Hours Positions Residence Mobility Orderliness Education hi lo hi lo hi lo hi lo hi lo hi lo hi lo hi lo Education: high (N- 98) 70 28 47 51 49 49 39 59 61 37 53 45 38 60 57 41 T a r - .250* low (N-lOO) 63 37 39 61 38 62 25 75 39 61 48 52 33 67 51 49 Orderliness: high (N-108) 73 35 53 55 44 64 35 73 57 51 62 46 51 57 var-,248 • OOT».018 low (N- 90) 60 30 33 57 43 47 29 61 43 47 39 51 20 70 Mobility: high (N- 71) 50 21 36 35 36 35 26 45 40 31 49 22 Tar-.230 COT-,062 OOT-.014 low (B-127) 83 44 50 77 51 76 38 89 60 67 52 75 Residence: high (N-101) 1* 27 51 50 47 54 34 67 59 42 Tar-.250 COT-,065 oov-,035 OOT-,015 low (H- 97) 59 38 35 62 40 57 30 67 41 56 var-.221 Tar-.246 Tar-,246 Tar=.219 Tar».250 Memberships, including: high (N-133) ... ... r ' ° ^ cov-.031 oov-.012 cov-,002 cov-.021 Memberships, excluding: high (N= 86) .... ... r 0 l o ° (N-112) oo'".036 COT-,026 cov-.031 cov=.022 H = 198 Attendance: high (N=> 87 low (N=111 Hours: Positions: high (N- 64) low (N-134) oov.013 cov-.024 cov—.017 cov=.030 cov.007 cov».015 cov-.OO! cov-.037 low*1 (N^gsj c o v='°4 0 oov-,021 cov-,012 cov-,058 Variance and Covariance measuruo are takon from: Roland K. Huwkee, "Multivariate Analyoio of Ordinal Measures." American Journal of Sociology. 76:919, table 1, 1971. Optimum values for variance and covariance •re .250 in the 2 X 2 table. 131 APPENDIX II VARIANCES AND COVARIANCES FOR OTOIHAL CROSS-CLASSIFICATIONS (A) Variances And Covariances For The Whole Sample Ordinal Cross-Classificationa* VARIABLES! Memberships Including Excluding Attendance Hours Positions Rosldenoe Mobility Orderliness Education (scale size) (4) (3) (4) Education (4) cov.036 cov».059 oov-.020 Orderliness (4) cov.033 cov.063 cov-.011 Mobility (4) cov-.011 cov.035 cov.005 Residence (4) cov— .011 cov«-.007 oov-.012 var-.350 var-.324 var-.366 (4) (3) cov. 048 00 v . 083 cov.034 cov.019 cov—.008 oov-.032 oov-,027 cov.060 var-,364 var-.321 (4) (4) cov—.010 cov.006 oov—.016 oov-,051 cov. 051 var-.354 var-,366 (4) (4) cov».077 var-.367 var-,367 N - 498 (B) Variances And Covariances For The Working-Class Sample Ordinal Cross-Classifications VARIABLES 1 Memberships Including Excluding Attendance Hours Positions Residence Mobility Ordorlineaa Education (scale alee) (3) (3) (4) (4) (3) (4) (4) (4) (4) Education (4) oov-.030 oov-.049 oov-.030 oov-.0'^ 1 oov-.076 oov—.000 oov-.007 oov-.049 var-.J54 Orderliness (4) oov-.009 cov-.024 cov—.013 cov-.002 cov-,017 cov-.043 oov-.083 var-,364 Mobility (4) cov.028 cov.028 cov.019 cov=.026 cov.012 cov.050 var».311 Residence (4) cov-,028 cov-.022 cov-.005 cov-.017 cov.046 var-.372 N - 198 varw.329 var-.291 var-.357 var-,349 var=.3H A l l scales marked by their respective number of categories. Optimum values for variance and covurianoe shift with the sice of the scale. Optimum values are: ( 1 ) 3 X 3 tables: var » cov - •333j (2) 4 X 4 tables: var - cov - .3751 and (3) 3 X 4 or 4 X 3 tables: var / cov, but cov - .354 and var - as In (1) or (2) above. APPENDIX III UNIVARIATE DISTRIBUTIONS, POINT ESTIMATES AND MEASURES OE JIORMALITY 134 In addition to the basic histograms for each variable, Appendix III supplies several summary s t a t i s t i c s , some measures of dispersion, and the pattern of collapsing for the ordinal and nominal scales. Each histogram — a frequency distribution across the interval scales — contains the whole sample (N = 498), with the working-class subsample (IT = 193) as a subset (see the following legend). The summary s t a t i s t i c s , the dispersion measures and their formulae are: (1) the mean: X = Z X ± (2) the standard deviation: s = / (3) the skewness coefficient: Sk = / (4) the kurtosis coefficient: K = < * i - * ) 2 (* 3) 2 I N 3 2 (* 2 ) 2 where i t j , T T ^ * A N < ^ 7 14 a r e fc^e s e c o n<*» third and fourth moments about the mean respectively. Skewness and kurtosis are measures of relative dis-persion and give an appraisal of the normality of a distribution. The optimum values (i.e., for a normal curve) are Sk =0.00 and K = +3.00, In the above formulae. 135 whole sample frequency distribution working-class subsample frequency distribution marginals for dichotomous scales, showing demarcations along interval scales. marginals for ordinal scales, showing demarcations along inter a l scales the mean of the distribution the standard deviation sk.ewr.ess before the transformation kurtosis before the transformation (A) Univariate Distribution of the dumber of Years of Education 100 -70 60 -50 40 -30.-10-2 3 4 5 6 7 A s S I ; K Whole sar. pie 10.8 2.9 .84 5.02 10 11 12 13 14 15 16 17 18 19 20 21 22 Working-class subsample Whole sample 100 98 27 t 73 66 1 . 32 227 271 96 1 131 172 1 99 Working-class sample 9.6 2.2 .40 4.65 (E) Univariate Distribution of the Ratio of Orderliness of T'ork Careers ~100 ' 90 - m - 70 - 60 - 50 - 40 - 30 ~ 20 - 10 SI: TJhole sanple .65 .398 - .56 1.67 Working-class sample .46 .393 - .01 1.41 180-i 7 d i v -70 " 60 ~ 50 -40 -30 -20 -C O .10 .20 .30 .4(: ,50 • 6( jG4f Uorking-elass subsaaple 90 26 61 .70 108 .80 .90 1.00 47 to* 199 90 124 299 1175 138 (C) Ordinal D i s t r i b u t i o n of Intragenerational Occupational M o b i l i t y 200 4-180 f 160 f 140 f 120 f 100 80 f 60 f 40 20 t Clear up S l i g h t Horizontal Fluctuating Working-class subsample 71 Whole sample 237 127 261 (D) Univariate D i s t r i b u t i o n of the Number of Years of Residence 2 4 ( r^ "^^ |•v-^ T^.^ •^,t^ •^^ •^ ••,•,••^ ,;•.,•,•,.;•;| 10 12 14 10 • 1 •'• * *••••» * * • 18 20 22 24 26 28 97 101 . 45 f 45 1 I 63 1 AS 196 302, 106 1 90 177 1 125 36 38 42 45 48 52 54 Working-class subsample Whole sample SO 1A0 (E) Univariate Distribution of the Number of Memberships Including Church and Union s Sk Whole sample 2.3 1.5 1.10 4.44 Working-class sample 2.2 1.3 1.43 5.47 _65_ 1 65 2 133 3 4 54 5 6 7 8 9 Working-class subsample 164 30 1 134 168 3 3 t Whole sample (F) Univariate Distribution of the Number of Memberships Excluding Church and Union 220 _ 210 140 130 120 110 100 90 80 70 60 50 40 30 20 10 Whole sample Working-class sample X 1.1 0.7 s 1.4 1.0 Sk 1.96 1.82 8.14 6.92 d 1 112 JLUL. J2—L 2 3 4 5 6 7 8 9 3 4 Working-class subsample 220 220 278 134 T 144 Whole sample (G) Univariate Distribution of the dumber of Meetings Attended 20-)5" / • N V V 40" 35-3) 20-15-10" i Whole Horhlng-cl sample sample — 33.3 33.6 s 29.2 29.2 .5 .55 is. 2.01 2.10 " 0 2 4 b 8 10 15 20 351!11 76 | 2 i 25 30 _5j2 35 40 45 50 55 60 65 ' 7C Working-dass subaample 1 n n in F*i X L 75 30 85 90 9 8 230 92 I 138 26 f; 9 4 _L7_4_ Whole sample (H) Univariate D i s t r i b u t i o n of the Number of Hours Spent i n Organizational A c t i v i t i e s 1S5-40-r Whole sar.ple Working-class sample X <\6 9.1 s 15.7 17.9 SI: 2.66 2.86 K 11.02 11.22 U3< w0 2 4 6 0 10 134 -86l 43J 24. 289 1851 104 15 20 40 25 LMJ i-fca-ELiiiJ m \ w\ / 30 35 40 45 50 x 60 Working-class subsample \(J K 75 N30 X'H6 ^ 96 209 85 124 Whole sample (I) Univariate D i s t r i b u t i o n of the Number of Positions held or Presently Holding X s Sk K iihole sample 1.6 2.2 2.06 7.80 Working-class sample 1.2 2.1 2.92 12.77 90 80 -70 -60 " 50 -40 -30 -; 20 -10 -98 98 227 227. 1 52 100 2 3 48 5 6 7 8 9 Working-class subsample 10 11 12 -15JL 27|1 118 Whole sample APPENDIX IV THE GENERAL DISTRIBUTION OF CASES IN THE THREE SAMPLES OVER EDUCATION AND OCCUPATIONAL POSITION 146 The General Distribution of Cases In the Three Samples Over Education and Occupational Position. low OCCUPATIONAL POSITION high The Whole Sample The Middle-Mass Sample The Working-Class Sample APPENDIX V DATA TRANSFORMATIONS: SKEIINESS AND KURTOSIS, TRANSFORMATION FORMULAE, AND EFFECTS ON PEARSON'S 148 The transformation of one's data distributions into normal distributions is executed frequently in some disciplines, eg., geography, where the pursuit of mathematical models is consuming researchers. A good source for this pro-cedure and for the appropriate tables of values for skewness and kurtosis i s Snedecor and Cochran (1967). In a Dominion Bureau of Statistics study, Davis (1969) uses such transformations and also begins to tackle their theoretical import. The theoretical discussion of transformations is quite recent. Basic-al l y , i t i s hoped that valuable and pertinent experience w i l l be gained by actually performing such transformations. There are some preliminary theoretical justifications available. The shifting of frequency curves can be j u s t i f i e d by suggesting that measurement over a longer period of time would have yielded normal or more normal dis-tributions for our participation variables. For example, over a 10-year period (rather than a 6-month or 4-week period) i t i s quite l i k e l y that few individuals would remain ttfithout any memberships in organizations, would have attended no meetings, or spent no time, etc. By performing a transformation, we are in fact changing our sample to match more closely a theoretically normal population. Another reason for transforming concerns the t a i l of skewed d i s t r i -butions or curves. The theoretical impetus for a logarithmic transformation i s the notion of increasing propensities for particiaption. That i s , the more one participates, the more one w i l l participate. Individual Involve-ment increases at an increasing rate. Those who hold a position have a greater probability of holding additional positions. A logarithmic trans-formation would give each additional unit less weight. Similarly, i f the relation between education and participation i s of the same nature (pro-pensity for participation increases at an increasing rate with education), then a logarithmic transformation w i l l make a curvilinear relation a linear one (Croxton, ejt. a l . , 1967, chapter 5 and p.95). S t a t i s t i c a l l y , this transformation allows the (linear) correlation coefficient to reach an optimum value of unity. In transforming our data we used two basic formulae and two tests of normality: namely, skewness and kurtosis. These two tests of normality are based on: (1) the second and third moments about the mean for skewness; and 149 (2) the second and fourth moments for kurtosis (Snedecor and Cochran, 1967; Croxton, jet. a l . , 1967, chapter 10). The two basic transformation formulae are lop, (x) and log (- ). e e 1 - x Essentially, these formulae x^ere used to shift our scales away from zero values and to pull in the t a i l of a positive skew. The common logarithm of zero is undefined and the logarithm of a negative number i s mathematically impossible. Transformations of this sort are order-preserving; they affect only the size of the intervals. The elaborated formulae for calculating our logarithmic values x-jere: (i) for residence, (both) memberships, hours, and positions: l o g e (x) - l o g e ((kj / s)(X i) + ( k 2 - ((X / s)(k 1)))) where 'kj' and ^2* are constants, 's' is the original standard deviation, and 'X' i s the mean of the original distribution £ ( i i ) for orderliness and attendance: l o g e ( l * ) = l o g e (((max - min)(XjL) + min)/(l - ((max - m i n ) ^ ) * min))). where 'max* and'min' are the maximum and minimum of a predetermined range of values. Since the original distributions varied radically from one variable to the next (see Appendix III), several appropriate changes x*ere made in the internal structure of each elaborated formulae. The internal changes refer to k j , IC2, max and min. The k 2 constant is a scaling constant, moving the scale away from zero; and k^ i s , in effect, a new standard deviation such that the new curve meets the requirements of normality. Max and min combine, by creating a new range of values, to perform the same function. These four constants differed for each variable. Table Va l i s t s the respective values of these constants and gives an indication of the success of the attempts to obtain normality. It i s immediately obvious that education has not been transformed. After numerous attempts to improve the measures of normality, i t was found that the original distribution for education retained the best estimates of normality. Identical transformations x/ere retained for both samples: the whole sample and the xtforkinp-class subsample. Table Va Tho Transformation Constants and Measures of Normality Before and After Transformations Measures of normality Constants Whole Sample (N - 498) Working-Class Sample (N •> 198) Before After Before After 1 2 max min Sk X Sk K Sk K Sk K Education .68 5.17 .34 4.42 Orderlinesa .97 .26 -.56 . 1.67 -.14 1.46 -.01 1.41 .47 1.91 Mobility (dummy variable) Residence 1.5 3.0 1.16 4.53 -.07 2.54 1.03 5.34 -.46 3.30 Including 7.0 20.0 1.10. 4.44 .07 4.62 1.43 5.47 .56 5.03 Excluding 6.0 10.0 1.96 8.14 .63 2.22 1.82 6.92 .92 2.75 Attendance .80 .01 .50 2.01 .05 1.84 .58 2.10 .40 2.03 Hours .6 .5 2.66 11.02 .76 2.52 2.86 11.22 1.19 3.37 Positions 7.0 5.0 2.06 7.80 -.04 1.18 2.92 12.77 . .13 1.24 Optimum values for skewness (Sk) and kurtosis (K) are 0.00 and +3.00, respectively. The ranges of significance for the 5 % criterion are: (l ) N - 500: *.179 * Sk ^ -.179 and 3.37 * K » 2.67; (2) N - 200: +.280 » Sk * -.280 and 3.57 » K » 2.51. See Snedecor, G. W., and W. 0. Cochran, 1967. Statistical Methods. Iowa University: Iowa Univeraity Press, for tables of values for skewness and kurtosis. 151 It i s of sor;e. concern to assess how Pearson's correlation i s affected by the above transformations — since that coefficient is a linear measure. The effect of transforming the data seems to be sporadic in our data (table Vb). A quick comparison of those correlations that increased with those that decreased shows a slight trend to the increase. But five of the f i f t y increased by .10 or more. Also, there seems to be no consistency over any particular variable, i.e., over any transformation or combination of transformations. Tentatively, we conclude that there i s no detectable systematic effect on the Pearson correlation in our data. 152 Table Vb Effect of Transformations on Pearsonian Correlation Whole Sample Raw Data Transformed Data E 0 8 E 0 *.? i.i R * 0 11 .24 .16 .26 .13 R -.01 .03 .21 -.02 -.02 .22 I .25 .08 .04 -.01 .21 .07 .02 -.002 X .32 .17 .14 .07 .28 .20 .13 .05 A .07 .01 .05 -.01 .11 .03 • -.03 .05 H .10 .05 .11 .04 .22 .08 -.01 .11 P .18 .05 .04 .09 .29 .05 .12 .25 Working=»class Subample Raw Data Transformed Data E 0 +.* 11 R E 0 0 ti .13 .25 .14 .21 R -.05 .19 .16 -.04 .16 .19 I .14 ,06 .17 .04 .13 .07 .15 .11 X .16 .10 .17 .01 .16 .10 .15 .08 A .16 -.01 .05 -.02 .15 -.01 .07 .05 H .23 .09 .14 -.06 .23 .09 .15 .01 P .23 .13 .11 .01 .27 .07 .10 .16 The various symbols represent our variables as follows: *E' «= education; *0' = orderliness; = mobility; *R S = residence; 'I' = memberships including; 'X' = memberships excluding" "A! = attendance; 'H' = hours spent; and 'P* = positions. APPENDIX VI "ORDERLY CAREERS AKD SOCIAL PARTICIPATION" AN ARTICLE BY HAROLD L. WILENSKY CAREERS AND SOCIAL PARTICIPATION S 2 1 tude are tremendous, and he himself did not bring them fully out. But there is one im-portant conclusion which he did draw. If so-cial conduct is learned conduct, if it arises in and through education and experience, then there is nothing pre-detcrrnined about social life, then the measureless variety of social forms ceases to be a mater for surprise,14 as it must be if it is regarded as a fixed patern like an organism or an equaly fixed patern like an equilbrium system. Spencer moves to the very brink of the realization that a so-ciety is an order underlaid by comparative freedom, not an order dominated by compar-ative necessity. Not all the seeds he has sown 14 Loc. cit., pp. 602 et seq. blossom under his bands; but they are there all the same. In the history of sociological thought, Spencer's third theory has played a far more significant part than cither of the other two. To prove this, it is necessary to mention only one name—Wilam Graham Sumner. Sum-ner's foikway concept is directly inspired by, and based on, Spencer's concept of ego-altruistic sentiments. It was through Sumner that what is most vital in Spencer's sociology entered into American sociology, and through American sociology into world sociology. In this case, the son was probably greater than the father; but we who are the son's sons, do wel to remember our father's father. His in-spiration is still with us, and so it will remain, a treasured possession, even in days to come. ORDERLY CAREERS AND SOCIAL PARTICIPATION: T H E IMPACT OF WORK HISTORY ON SOCIAL INTE-GRATION IN T H E M I D D L E MASS * HAKOLD L . W I L E N S K Y The University of Michigan Durkheim's and Mannheim's ideas about careers as a source of social integration are put lo the test among 678 urban white moles of the upper-working and lower-middle classes, aged 21-55. Data suggest thai chaotic experience in the economic order fosters a retreat from both work and the larger communal life. Even a taste of an orderly career enhances the vitality of participation: compared with men who have chaotic work histories, those who spend at least a fifth of their worktives in functionally-related, hierarchically-ordered jobs hive stronger attachments to formal associations and the community. Their contacts with kin, friend, and neighbor are al once more integrated, wide-ranging, and stable. Their "occupational com-munity" is stronger: These contrasts are especially marked among young, high-income college men—a vanguard population. Although men with work histories that fit the model of "career," comprise only a liny slice of the labor force in modern societies, they may be strategic for social order. T is clear that class cultures (sustained by similar levels of income and educa-tion and common absorption of the mass media) and ethnic-religious cultures * Paper read at the American Sociologic.il Asso-ciation meetings, New Y o r k Ci ty , September, 19C0. Part of a larger study of "Labor and Leisure in the Urban Community" made possible by grants from the National Institute of Mental Health (M-2209), 1958-60, and a fellowship at the Center for Advanced Study in the Behavioral Sciences, 19S6-S7. For description and theoretical background of the study, sec It. L . Wilensky, "Work , Careers, and Social Integration," International Social Science (sustained by common descent and early socialization) significantly shape social rela-tions. It is also clear that occupational cul-tures (rooted in common tasks, work sched-Jottrnal, 12 (Fa l l , I960), pp. 543-560, and Work and Leisure, Glcncoc, 111.: The Free Press, forth-coming. I am indebted to Albert J . Reiss, J r . and Morr is Janowitz for critical readings o( an early draft, and to John C . Scott, Michael T . Aiken, Paul R . K immcl , Betty Burnside and other members of our labor-leisure seminar for research assistance, especially in the difficult task of coding complete work histories. 522 AMERICAN SOCIOLOGICAL REVIEW ulcs, job training, and career paterns) are sometimes beter predictors of behavior th;:n both social class and pre-job experience. The question of. which is most important in determining specific types of social relations remains quite open. The evidence, while scanty, suggests that wherever work ties are severed, there is a decline in community participation—whether we, consider varia-tions over the life span of persons or varia-tions among groups at one point in time. For instance, excluding churches, there appears to be a general curve of participation in for-mal associations which closely paralels a job satisfaction curve—a sharp drop in the early 20's, especialy among hard-pressed married couples, a climb to a peak in the middle years, a slight drop-of and then a final sag in the 60's~-and these cycles seem to be in-terdependent, although good longitudinal data are as usual lacking.1 More important, those persons and groups with the most ten-uous ties to the economic order—from men squeezed out of the labor market (older workers and retirees, the unemployed) to "unemployablcs" who seldom get in (skid row bums, adolescents of the slum)—are also those who are most isolated in commu-nity and society.2 Plainly, Durkhfeim's view that in modern society workplace and occupational group draw the person into the mainstream of so-* Sec H . L . Wilensky, "L i fe Cycle, Work Situa-tion, and Participation in Formal Associal jns," in R . W. Klecmeicr, editor, Aging and, Leisure, New Y o r k : Oxford University Press, 1961, pp. 213-242. On the effect of unemployment, see E . W. Bakkc, The Unemployed Man, New Y o r k : E . P . Dutton & Co., Inc., 1934 and Citizens Without Work, New Haven: Yale University Press, 1940; B. ZawaUski and P . Lazarsfcld, "The Psychological Consequences of Unemployment," Journal of Social Psychology, 6 (May , 193S), pp. 224—51; P . Eisenbcrg and P . Lazarsfcld, "The Psychological Effect of Unem-ployment," Psychological Bulletin, 35 (June, 193S), pp. 358-90; M . Komarovsky, The Unemployed Man and His Family, New Y o r k : Drydcn, 1940. C f . ' W . Kornhauser, The Politics oj Mass Society, Glencoe, III.: The Free Press, 19S9. On the position of youth, see S. N . Eiscnsladt, Prom Generation to Generation: Age Groups and the Social Structure, G'sncoe, 111.: The Free Press, 19S6. On the decline of participation with aging, see citations in Wilensky, op. cit. None of these studies, however, tells us whether re-employmcnt briniis a return to previous participation levels, none deals with the impact of other forms of economic deprivation or discontinuity (e.g., continuous employment, but chaotic "carters"). cial life deserves further elaboration and test. This is perhaps the central problem of the sociology of work—the efect of the division of labor on social integration, or more pre-cisely the conditions under which work role, occupational association, and career are most nd least efective as bonds of solidarity either within workplace or between work-place and larger units, or both.8 The guiding hypothesis of the present study is this: The vitality of social participation, pri-mary and secondary, and the strength of at-tachment to community and to the major institutional spheres of society are in part a function of cumulative experience in the eco-nomic system. Participation in community life is a natural extension of participation in the labor market; orderly and pleasant ex-periences in the latter provide motive and opportunity for the former. Specificaly, where the technical and social organization of work (I) offers much freedom—e.g., dis-cretion in methods, pace or schedule, and opportunity for frequent interaction with felow workers who share common interests and values (2) necessitates sustained and wide-ranging contact with customer or client, making the work role visible to the commu-ity and (3) provides an orderly career in which one job normaly leads to another, related in function and higher in status, then work atachments will be strong, work in-tegrated with the rest of life, and ties to com-munity and society solid. Conversely, if the task ofers little .workplace freedom (assem-bly-line workers, dentists, accountants and many engineers) if the job demands no cus-tomer or client contact and yields no readily visible status claim ("burr knocker" in a job shop, "console operator" in an insurance office), if the work history is punctuated by unexpected periods of unemployment, dis-orderly shifts among jobs, occupations, and industries, then work atachments will be weak, work sharply split from leisure, and ties to community and society uncertain. In short, to the extent that men are ex-posed to disciplined work routines yielding little gratification and have "careers" which s Cf . E . Durkhcim, "Some Notes on Occupational Groups" i n The Division of Labor in Society, trans-lated by George Simpson, Glencoe, III.: The Free Press, 1947, Preface lo 2nd edition. CAREERS AND SOCIAL PARTICIPATION 523 are in no way predictable, their retreat from work will be accompanied by a withdrawal from the larger communal life—and this will apply to the middle class as wel as the work-ing class. In order .o tackle these and similar prob-lems, we have in the past two years inter-viewed about 1500 men in various occupa-tional groups, strata, and workplaces. This paper deals with 678 white males of the "middle mass" and the efect of their career paterns on the kinds and strength of their lies to community and society. "CAREER" AS A SOCIOLOGICAL PROBLEM There is uncommon agreement that the study of types and rates of mobility is crucial to an understanding of modern society. And there are hints that worklife mobility may be more fateful than intergenerational mo-bility.4 It is therefore remarkable that de-tailed work histories which cover a decade or more have been reported in only about a dozen studies.5 These studies have located * H . L . Wilensky and H . Edwards, "The Skiddcr: Ideological Adjustments of Downward Mobi le Workers," American Sociological Review, 24 (Apr i l , 1959), pp. 215-231. Cf . VV. Read, "Some Factors Affecting Upward Communication at M i d d l e - M a n -agement Levels in Industrial Organizations," un-published Ph .D. thesis, University of Michigan, 1959, which reports a correlation of +.41 (p<.01) between upward worklife mobility and holding back "problem" information from the boss among 52 middle-level executives (mean age 37), but no correlation for intergenerational mobili ty. 8 E.g., P . E . Davidson and H . D . Anderson, Oc-cupational Mobility in an American Community, Stanford, C s ' f . : Stanford University Press, 1937; W. H . Form and D . C . Mi l le r , "Occupational Career Pattern as a Sociological Instrument," American Journal of Sociology, 54'(January, 1949), pp. 317-329; S. M . Lipset and R . Bendix, "Soc ia l .Mobi l i ty and Occupational Career Patterns," American Jour-nal of Sociology, 57 (January and M a r c h , 1952), pp. 336-374, 494-504; W . L . Warner and J . C . Abcgglen, Occupational Mobility in American Business and Industry, Minneapolis: University of Minnesota Press, 1955; the U . S. Bureau of Labor Statistics series on work histories of skilled populations, and the earlier Works Progress Administration studies. For the complete list see "Work , Careers, and Social Integration," op. cit. The Six Ci ty Survey of Occu-pational Mobi l i ty , although based on only one decade, contains the most adequate data and most extensive and representative sample. F rom it we can say that in most cases changing jobs means changing both occupation and industry, and pro-jeeting the data, the average worker wi l l hold 12 jobs in a 46-ycar worklife and only one man in types of job histories in various strata, and they leave no doubt that modern adult life imposes frequent shifts between jobs, occu-pations, employers, and workplaces; but they tell us little of the consequences for person and social structure. Let us define career in structural terms. A career is a succession of related jobs, aranged in a hierarchy of prestige, through which persons move in an ordered (more-or-less predictable) sequence. Corolaries are that the job patern is instituted (socialy-recognized and sanctioned within some social unit) and persists (the system is maintained over more than one generation of recruits). It has long been recognized that careers in this sense are a major source of stability. Every group must recruit and maintain its personnel and motivate role performance. Careers have served these functions for or-ganizations, occupational groups, and socie-ties, and in cultures as diverse as those of medieval Europe, Soviet Russia and the United States. Careers also give continuity to the personal experience of the most able and-" skiled segments of the population—men and groups who otherwise would produce a level of rebelion or withdrawal which would threaten the maintenance of the system. By holding out the prospect of continuous, pre-dictable rewards, careers foster a wilngness to train and achieve, to adopt a long time perspective and defer immediate gratifica-tions for the later pay-of. In Mannheim's phrase, they lead to the gradual creation of a "life plan." 6 Most men, however, never experience the joys of a life plan because most work situa-tions do not aford the necessary stable pro-gression over the worklife. There is a good deal of chaos in modern labor markets, chaos five w i l l remain at the same occupational level throughout his worklife—if the decade 1940-50 is typical, if the Census categories are meaningful, etc. G . L . Palmer, Labor Mobility in Six Cities, New Y o r k : Social Science Research Council , 1954; A . J . Reiss, Jr., "Occupational Mob i l i t y of Pro-fessional Workers," American Sociological Reviciv, 20 (December, 1955), pp. 693-700; and A . J . Jaffe and R . 0 . Carleton. Occupational Mobility in the United States, 1930-1960, New Y o r k : King's Crown Press, 1954. 8 K . Mannheim, Man and Society in un Age of Reconstruction, translated by E . A . Shils, New Y o r k : Harcourt Brace and Co., 1940, pp. 56, 104-106, 181. 524 AMERICAN SOCIOLOGICAL REVIEW intrinsic to urban-industrial society. Rapid technological change dilutes old skills, makes others obsolete and creates demand for new onrs; a related decentralization of ii dustry displaces milions, creating the paradox of depressed areas in prosperous economies; metropolitan deconcentration shifts the cli-entele of service establishments, sometimes smashing or restructuring careers; recurrent crises such as wars, depressions, recessions, coupled with the acceleration of fad and fashion in consumption, add a note of un-predictability to the whole. There are many familiar variations on the main theme: in industries such as construction, entertain-ment, maritime, and agricultural harvesting, the employment relationship is typicaly casual. In food processing and the needle trades, drastic seasonal curtailments are common. All this is easy to see among populations that ate depressed, deprived, or marginal—• non-white victims of racial discrimination, underdogs on relief rolls, miners stuck on a played-out coal patch, the chronicaly unem-ployed among the old, sick, or disabled. But what about the vanguard population —the great middle mass around which For-tune magazine's portrait of the "soaring sixties" is drawn? What portion of the grow-ing middle—the lower middle class and upper working class—can be said to have orderly careers? This paper will show that a majority of the middle mass experiences various degrees of disorder in the work his-tory. At the same time it will test rveral hypotheses about the functions of predicta-bility by comparing the behavior of men with disorderly work histories with that of the more fortunate, the men of orderly career. It will give special atention to those at higher education and income levels. As-suming that the middle mass will grow in size and influence, that education and in-come levels will continue to rise, and as-suming (with less confidence) that job his-tories will become more like careers, we may thereby come to some hints of the direction of social change. DIMENSIONS AND MEASURES OF CAREER Although we have many sophisticated discussions in the sociological literature,7 * E . C. Hughes, Men and Their Work, Glencoe, two difficulties have discouraged the full and ystematic exploration of the problem of "career": (1) conceptualizing both career patern and social participation; (2) gather-ing and processing data beyond crude levels. In handling careers as a source of social integration, I have given special atention to one dimension of job history beyond di-rection and amount of movement: the degree of orderliness—i.e., how wel it fits the model of "career." Complete work histories and information on each job, employer, and job change were recorded in a detailed interview. In coding, we first aimed to rank all jobs in four classes according to society-wide judg-ments of occupational prestige: upper-middle (high non-manual), lower-middle (low non-manual), upper working (high manual), lower working (low manual). Detailed in-structions covered major classification prob-lems: e.g., foremen were coded as high manual; all service occupations, farm em-ployment, and highest military rank achieved were distributed among the four categories (e.g., farm owners or managers on farms of 100 acres or more were counted as low non-manual, on farms of 50-99 acres, high manual, all other farm employment, low manual). Where detailed instructions did not cover the case, the folowing additional information was used (in the priority listed) tofit jobs in: (1) the North-Hatt scale (e.g., service jobs like detective were counted low non-manual; playground director, police-man, Greyhound bus driver, railroad con-ductor, high manual; below, low manual); (2) data on education and training—e.g., all ambiguous jobs which might be skiled man-ual (requiring six months or more training on or of the job) were counted low manual if no special training was indicated in a sepa-rate batery on training; (3) job status or training inferred from the Dictionary of Oc-cupational Titles; (4) data on changes in rates of pay arid reasons for leaving (used III.: The Free Press, 1958; H . S. Becker and A. L . Strauss, "Careers, Personality, and Adult Socializa-tion," American Journal of Sociology, 62 (Novem-ber, 1956), pp. 253-263; N . . H . Martin and A. L . Strauss, "Patterns of Mobility within Industrial Organizations," Journal of Business, 29 (April, 1956), pp. 101-110, and the work of Erving Golf-man. See also such texts as R. Dubin, The World of Work, New Jersey: Prentice-Hall, Inc., 1958, ch. 15, and E . Gross, Work and Society, New York: The Thomas Y . Crowcll Co., 1958. CAREERS AND SOCIAL PARTICIPATION 525 as a last resort to indicate direction of move-ment).8 Combining direction with orderliness, the job histories were then coded as follows: 1 . Orderly horizontal progression. Orderly: the skills and experience gained on one job are directly functional to perform-ance in subsequent jobs and jobs are arranged in a hierarchy of prestige. In-dicators: (a) an expert on occupations could defend the sequence as an or-dered, more-or-less predictable pattern or, as a last resort only, (b) R planned the sequence, sees one step leading to another (coder read seven questions that would indicate this). And horizontal progression: some evidence of increased job status within one occupational stratum. E.g., carpenter's apprentice, journeyman, foreman in construction firm. These are functionally related, hierarchically-ordered jobs that do not cross-cut occupational strata. 2. Orderly vertical progression. At least half of the years covered in the work history are in jobs that are functionally related and arranged in hierarchy of prestige, and the mobility pattern cuts across occupational strata. 3. Borderline orderly vertical progression. At least a fifth but less than half of the worklife is in jobs that are functionally related and arranged in a hierarchy of prestige. The mobility pattern cuts across occupational strata. 4. Disorderly horizontal movement. Less than a fifth of the worklife is in func-tionally-related, hierarchially-ordercd 8 We excluded low-status "clerical" jobs held for less than two years when respondent was under 21 (e.g., errand boy), as well as "moonlighting" (cov-ered in a special baltery, which showed that 36 per cent of the middle mass have at some time held secondary jobs). A code for worklife mobility covered all cases: on distance, much (clerk to col-lege-educated engineer, non-manual to manual), some (non-college accountant to sales clerk, semi-skilled operative to foreman) or none (stable high non-manual, stable low manual, etc.); and on total pattern, four types of up, four stable, 11 fluctuating, six unstable or highly fluctuating, and four types of down. Rased on a IS per cent check (102 cases), the code as a whole has 83 per cent reliability; all but four of the disagreements, however, were choices between detailed codes within the major categories, "fluctuating" or "unstable," and could not affect this analysis. jobs—i.e., by no stretch of the imagina-tion could the history be clarified as a fifth or more orderly. Mobility pattern docs not cut across occupational strata. 5. Disorderly vertical movement. For at least four-fifths of the worklife, jobs are neither functionally related nor hier-archically ordered. Cuts across occupa-tional strata. 6. One job for entire worklife. In order to avoid exaggeration of labor-market chaos, coders were instructed to re-solve doubts in favor of orderliness. Simi-larly, "some orderly" was combined with "orderly," on the assumption that a modest amount of job predictability can go a long way to integrate a man into the social sys-tem. From inspection of detailed tables it appears that these borderline cases do oc-cupy a middle ground with respect to the dependent variables.0 This entire coding effort was guided by the assumption that what several students have viewed as "bureaucratic" vs. "entrepre-neurial" or "old middle class" vs. "new" 1 0 is better treated as types of careers that cross-cut organizations large and small, bu-reaucratic and not, as well as diverse age grades and economic strata. For instance: "bureaucracies" come and go. In many sec-tors of the economy mergers, relocations, and shutdowns (rooted in the population shifts and technological changes discussed above) reach big firms and high-seniority workers, and, sometimes, as the business press notes, they take their toll in executive careers, too. Even in the more stable bureaucracies, some layers of personnel (e.g., foremen) have quite chaotic job patterns. In the little 0 A 15 per cent check showed over ft0 per cvnt reliability—-and much of the disagreement, was be-tween orderly vertical and borderline orderly. For various reasons the 21 men in the "one job only" category were eliminated. A separate a:n lysis, com-paring one-job men with the rest of the middle mass shows similar distributions on all participation measures but two: one-jobbers have a wider r:im;e of secondary contacts and a lower rar.w of prhiury contacts. To include them as "orderly horizontal" would slightly strengthen the case. 1 0 See,"e.g.", T>. R."Miller and G ,K. Swamon. The Changing American Parent, New York: John Wiley & Sons, Inc., 195S; C. W. Mills, While Coil.tr, New York: Oxford University Press, 1951; I.. Corey, The Crisis of the Middle Class, New York: Coviei, Fricde, 1935. 526 AM KRrCAN SOCIOLOGICAL REVIEW T A I I L F . 1 . DtsvumunoN o r T J H ; M n m i . B M A S S B Y T Y P E o r W O R K H I S T O R Y Number Per Cent Orderly Orderly Horizontal Prugrcssion Orderly Vertical Progression SS 116 13% 17 Borderline Some Orderly Vertical Progression 223 33 Total Orderly 427 63 Disorderly Disorderly Horizontal Disorderly Vertical 36 194 5 29 Tota l Disorderly 230 34 Onr Job Only 21 3 Grand Tola ! 67S 100% leagues, too, there are great variations: a wel-located, wel-run service station may prosper for years, another may be the scene of recurrent bankruptcies. The concept, "disorderly work history" (counterposed to "career"), takes account of these variations, permits determination of the incidence of real careers, and brings a broad range of cases into view. Table I shows that in the middle mass—a relatively secure population, wel of by American standards—only 30 per cent can by any stretch of the imagination be said to act out half or more of thoir work histories in an orderly career. If wc count the lower class, excluded from this sample, it is apparent that a vast majority of the labor force is going nowhere in an unordered way or can expect a worklife of thoroughly-unpredictable ups and downs. Comparing men with orderly careers and those with disorderly work histories, we see in Table 2 an almost even distribu-tion among a variety of social character-istics — although high-income colege men do have a decided edge among the orderly. Thus, as an independent variable, orderli-ness cross-cuts traditional categories of soci-ological analysis; if it has an impact on social life, that impact is distinctive. DIMENSIONS AND MEASUHKS OF O B J E C T I V E SOCIAL PARTICIPATION The research aimed to analyze social re-ktions with an eye to their capacity for link-ing the person and his immediate family to the larger communal life. Four dimensions were found useful in approaching this prob-lem: (1) the number of roles played (groups p rticipated in, primary, informal, or second-ary), the frequency of contact and amount of time in each; (2) the range of participa-tion; (3) role integration—the degree of coherence of the pattern; and (4) the sta-bility or duration of relationships. Together these constitute an index of the vitality of objective participation, the dimensions and measures of which are discussed below. Un-lss otherwise specified, coding reliability on a 15 per cent check was over 95 per cent. Secondary attachments. From detailed in-formation on each organization to which the respondent belongs plus a batery on organi-zational and community activities, the fol-lowing indexes were developed. 1. Number of memberships. Each organi-zation counts one point for a quartile rank-ing. Church counts one only if respondent reported a current religious preference and atended church services at least once a month. Men in the two highest quartiles (three organizations or more) were scored High. 2. Frequency of contact and time spent. For each organization other than church, we asked when it held regular meetings, how many of the last six the respondent atended, whether this was usual, etc. In the absence of precise information, we counted a meeting as worth two hours. We similarly elicited data on time spent on organizational activi-ties other than regular meetings—commitee work, phone cals, money-raising, etc. We then estimated the usual hours per month for all organizational activity. This measure avoids the common tendency to over-esti-mate levels of participation. Excepting reg-ular church services, we found that almost CANKERS AND SOCIAL PARTICIPATION 527 TABI.K 2. DIRECTION ASI> DISTANCE OP Momuiv , FATHER'S OCCUPATION, WORKING FOR Oi iitss, AM) A'.'.E ARK UKKF-LATED TO OlU»:m.lN'BSS IS T H E WORK HISTORY — ~ Orderly Disorderly Worklife Up 35% 34%. Mobility Stable 21 15 Fluctuating and Down 21 25 Unstable - 23 25 100% (427)* 100% (230)* Inter- Much Up 23% 18% p;neration3l Some Up 11 4 Mobility Stable 43 42 Some Down 2 14 Much Down 18 18 R Lived in Orphanage 2 4 Father's Occup. High Non-manual 1% ~-% . Stratum When R Low Non-manual 37 32 Was a Teen-ager High Manual 30 36 Low Manual 22 21 Living in orphanage or father dead 10 11 Per Cent of Self Employed Worklife More than 20% of Worklife 14% 16% Self Employed 20% or Less of Worklife 9 S Never Self Employed 77 76 Religion Catholic 46% 41% Jewish 3 4 Other 51 54 Age Over 40 46% 39% Under 40 54 61 Occupational While Collar 31% 4l7o Stratum liluc Collar 49 59 Income Under 8000 51% 66% SCOO and More 49 34 Education Less than High School 30% 47% High School Grad 40 35 Some College 30 18 *N's vary slightly because of "not ascertains" on the comparison variables. 21 1-job cases have been deleted. NA's for age, income, and education are cumulative. 4 in 10 of the middle mass, a relatively ac-tive population, spend less than 30 minutes a month on all organizational activity; me-dian time is about two hours. (Compare ' their median estimated time watching TV— about 48 hours per month.) 3. Range of secondary participation. A man may play many roles in many groups, interact at a high rate, and devote much time to it—-al with family and relatives living in a homogeneous neighborhood. If our interest is in the integrative potential of social ties, we must note, too, the participation patern of miners, longshoremen, and others who arc surounded by men like themselves, and who, in lodge and union, at home and at the bar, reinforce their common alienation and iso-lation. Accordingly, special atention has been given to the range of social relations, pri-mary and secondary. "Range" is the varia-tion in interests (concern for or stake in the coure of events), values (afectively-toned, group-shared desires which serve as criteria for choice among ends) and status (prestige levels) represented by all the roles the per-son plays. This aims to permit classification of respondents from "community isolation" through ever-wider circles of involvement. Tw operational criteria were used for sec-52S AMERICAN SOCIOLOGICAL REVIEW ondary range: (a) heterogeneity of member-ship to which exposed; (b) number of major institutional spheres to which exposed. a. Heterogeneity: the degree to which the respondent's organizations provide oppor-tunity for interacting with people who differ from him in important social characteristics. All types of organizations were put in four categories according to heterogeneity of their members in occupation, income, education, ethnicity or religion or race, age or life cycle, and sex—in that order of importance. The categories are: narrow range or least hetero-geneity of membership (e.g., exclusive craft unions, professional groups, property-owners associations, business pressure groups, etc.); medium-restricted range (e.g., minority de-fense organizations, like the NAACP or Polish National Alliance, ambiguous multi-craft unions, church-connected organizations, workplace-based leisure associations); me-dium range (e.g., inclusive labor unions, service clubs and civic groups, PTA, church, youth-serving groups, non-neighborhood card and social clubs, fraternal organizations, etc.); broad range (tolerance organizations, charitable organizations, political parties, etc.). In the scale developed, minimum ex-posure was required; if respondent attends none of the last six meetings or answers "don't know," if he names church preference but never attends, he receives no points. Otherwise, the more range, the more points. Pre-coded lists were made up in difficult cases (e.g., unions). Coding reliability for a IS per cent check of the detailed code was 81 per cent. The lowest two quartilcs (0—11) were counted as low range, the highest two (12-42), high range. b. Number of institutional spheres covered by respondent's membership—economic, po-litical-military, educational, religious, public welfare, re-creation-aesthetic. The major and peripheral functions and purposes of each type of organization were listed on a chart, with three points for a major function in a given sphere (e.g., labor unions are mainly economic), one point for peripheral functions (unions, one for political, and one for wel-fare). Maximum respondent score in any sphere is three. The detailed code has a re-liability of 84 per cent. The scoring system, with a minimum exposure criterion, yields a dichotomy of 0-7 vs. 8-18. 4. Attachment to the community. Except among officials and community elites there are few obvious and direct attachments to the communities constituting a metropolis. On the assumption that schools and churches are major transmitters of the most widely-shared community values, and the private welfare structure is another potential source of contact, we used voting in the most recent school election and size of contribution to church and charity as indicators of com-munity attachment. A high score goes to men who remember the school election and say they voted to increase school taxes (40 per cent of all). A charitable contribution of at least $100 was scored high. (The median amount is S100 to S199; the median slice of annual family income is 1.8 per cent.) Primary and informal relations. The study assumes that "primary" relations vary in their potential for expanding the person's horizons; some ties to kin and friend are more parochial than others. Accordingly, the interview included many questions on the number and group contexts of leisure-time activities, number of persons known well, the last time the respondent got together socially with neighbors, relatives not living with him, fellow-workers, people in the same line of work, and others, a complete battery on each of three best friends, and two ques-tions on income and location of relatives. The major dimensions and measures of objective primary relations used for this analysis follow: 1. Range of primary relations ,(See sec-ondary range above). The minimum expo-sure counted was "got together socially within the past two months." The range score begins with one point for living with wife and children, moves up to four points for recent contact with people in the same line of work who work elsewhere or "other" friends (cross-neighborhood, former workplace, etc.). Contact with often-seen relatives earns from one to four points. (If almost all or about half live in the neighborhood and most are in the same income bracket, it is worth only one point; if few or no relatives contacted are neighbors and some are financially much better or worse off, then the respondent gets three points.) Fellow workers in the same type of work get three points, those in a dif-ferent line get four points. CAREERS AND SOCIAL PARTICIPATION 529 The top two quartiles score 8-18, a high range; the rest, low. 2. The coherence and integration of pri-mary and secondary roles. A segregated role ii one in which the behavior expected and preferred in one group has nothing to do with the behavior expected and preferred in any-other group. For instance, when an occupa-tional role (banker) becomes so elaborated that its rights and obligations come'to em-brace the behavior .expected in other role systems (country club, family, Republican Party), we speak of "integrated roles." It is assumed that those roles that are most [ integrated are most efective in social con-trol; from the viewpoint of the person such roles reduce conflict and choice and make for an easy faling in line. Here are the meas-ures of role integration: : a. Work associates and leisure-time friends. The degree to which social relations at work - overlap those of work was used as the first index. We coded the number of three best friends in the same workplace or occupation or both, how often each is seen socialy out-side of work, and other leisure contacts with other work associates. A strong "occupa-tional community" (at least one best friend from work or in the same line of work who is seen socialy at least once a week) covers 199 cases—29 per cent. b. Friendship circles. The atack on tra-< ditional theories of urbanism and the redis-) covery of Cooley's primary group have di-i verted atention from several crucial con-) trasts betv/een modern and pre-modern pat-terns of friendship—among them, the tend-ency for urban-industrial populations to be linked in open networks rather than circles. Almost one-fifth of the middle mass have two or three best friends, none of whom know one another wel; another six per cent have no friends or only one. The modal case names three best friends, two of whom know each • other well; this or beter was scored as high integration. c. Overlapping friendship and member-ship. Those who have at least one member-ship in common with one best friend are scored high on integration. 3. Duration of friendships. A crude meas-ure of stability of primary ties was the number of best friends who have been known for ten years or more. To score "high vi-tality" the respondent had to have either (1) one or two best friends, each known that long; or (2) three best friends, two of whom arc of ten years' duration. In sum: this study assumes that any social tie integrates. But those relationships that re wide-ranging, most coherent and stable are most efective in binding the person and his immediate family to community and society. Specific hypotheses derived from these ideas about careers and participation are listed below with the findings. THE MIDDLE MASS SAMPLE The analysis is based on a multi-stage probability sample combining the use of city directories with block supplementation and segmentation procedures. The popula-tion sampled consisted of all private dwel-ing units within the tracted areas of the tri-county Detroit Standard Metropolitan Area and, within the selected dweling units, those white, male members of the labor force who ere 21 to SS years old, ever married, who fell into the "middle mass." 1 1 The middle mass is a residual category. We eliminated the lower half of the working class by requiring a minimum family income of S5,000 in any one year of the past five; we cut out the upper-middle class by apply-ing the usual criteria of authority and skill. Upper-middle class authority was indicated by size of firm of owner (e.g., 25 or more employees), number of subordinates of man-ager or official (e.g., 100 or more) or con-trol over client and monopoly of vital service among professionals. Upper-middle class levels of skill involve mastery of an abstract, codified body of knowledge (e.g., minister, architect or professor)—and imply a colege education usualy going beyond the bachelor's degree, or executive training after colege. In order to eliminate upper middles, interviewers were given a list of specific occupation-education categories which were ineligible. Occupational cate-gories which remained—clerks, salesmen, 1 1 Alt primary sampling units were stratified on the basis of population; those which had a popu-lation of less than 25,000 (not self-representing) were further stratified on the basis of geographical and economic categories. 530 AMERICAN SOCIOLOGICAL REVIEW craftsmen, foremen, small proprietors, semi-marginal- or woo Id-be-pro fessionais and tech-nicians, managers and officials with few sub-ordinates, and operatives with high income —constitute the core of the middle mass. In designing the sample we noted that in a 1953-59 Detroit area sample about 10-15 per cent of the eligibles in such middling occupational categories reported family in-comes of $10,000-15,000. Therefore, we set-tled for an arbitrary top income limit for ''lower-middle class" of $13,000—represent-ing two-thirds of those who fit the occupa-tion-education criteria (one standard devia-tion from the mean). Thus, by use of selector questions, we bracketed in white, male members of the labor force, 21 to 55 years old, ever married, whose family incomes in any one of the past five years reached $5,000 or more but never topped §13,000. I think that this procedure yields a reasonable cross-section of the upper-working and lower-middle class —men at the bottom of various ladders lead-ing upward, and some on their way down, too.32 Eligible men of the middle mass were found in 39 per cent of the 2137 occupied dwelling units visited. Of a total of 678 re-spondents, 47 per cent were coded as white collar and 53 per cent as blue collar. Their median age, 39; median family income in 1959, $7,6SO. Nearly three in five live in the suburbs. The interview period was late January through March, 1960. Most interviews were done by graduate students in the Detroit Area Study, the rest by professional inter-viewers.'3 Median interview time: one hour and 48 minutes. 1 2 This highly-stratified sample was dictated by theoretical interest in (1) the effect of types of mobili ty on leisure style, and the general problem of the content and implications of "blurring class lines"; (2) theories of the mass society and their empirical critics. The possibility that this sample represents a vanguard population in terms of its influence on social structure and culture also guided the design. Other samples of the study tap religious-ethnic and underdog subcultures, which remain at the bottom, and occupational subcultures, which flourish at the top. 1 3 The O A S , a research-training facility of the College, The University of Michigan, directed by Har ry M . Sharp, is part of the Master's program in sociology; it supplied various data-collection serv-The analysis concentrates on one dimen-J sion of career, orderliness, as it affects the objective vitality of social participation.11' Critical ratios of the differences in propor-j. tions were computed and treated as one-tail; tests; percentage differences serve as a meas-ure of the strength of relationships. The case rests upon the size and consistency of pre-: dieted differences within a homogeneous sanv: pie in which broad controls for race, family income, occupational stratum, age, marital status, and location were built in by selec-' tion. In testing hypotheses, three social char-acteristics that other studies have shown to be related to participation—education, in-come, and age—will be used as analytical g controls. Then the conditioning effect and the* independent effect of each of these will beg examined. Since the independent and de-M-pendent variables are all dichotomized, sup § port for the positively-stated hypothesis also» supports the obverse.18 % FINDINGS I. Men whose careers are orderly will havtK stronger attachments to formal association* and the community than men whose job tones are disorderly. The data, summarizedg in the top half of Table 3, generally supportg; the hypothesis. Of 72 possible comparison^ ! of the orderly and disorderly (six dimensions^  of participation for 12 education-income| age categories), 24 comparisons shcre| percentage differences in the unexpected direction. The overall net difference ii§ the predicted' direction is eight per cent*. ices for this sample. Before interviewing, the stug dents had had experience in preliminary coding an<? analysis of long interviews carried out in a factor; study by the labor-leisure project, as well as tb usual interviewer training and prc-test field wort 1 4 Data on work milieux—workplace freedom status visibility, and customer-client contact—aw, the intervening variable, work" alienation-indiffer ence-attachment, as well as data on the "quality", or subjective meaning of participation, and tfc effect of such variables on the ways in which per, sons and groups relate to the major institutions B the society wi l l be reported elsewhere. ! 1 5 The obverse holds that chaotic work historic are associated with low vitality of social parlicipfr tion. Because the measure of disorderly work his, tory is better than that for orderly career (whicl, includes borderline cases), i t can be argued tha'. positive results mean stronger support for tb> obverse. Less Than 12 Oracles Under $8000 $8000 and Over H igh School Graduate Under $8000 $8000 and Over Some College or M o r e Under $8000 S80CO and Over Under 40 and Under 40 and . Under 40 and Under 40 and Under 40 and Under 40 and Net A v g . 40 Over 40 Over 40 Over 40 Over 40 Over 40 Over % i n Predicted Direction 1 2 3 4 5 6 7 8 9 10 11 12 . P < * * Formal Associations and Community a IA N o . o l Memberships 22%* - 2 % 2% : 6% 3 % 38% 9% 12% 6% 17% 3 1 % 17% 13% .01 s> ?c 13 Frequency, Time, K —Organizations 12 10 25 " 2 • 4 27 9 22 —9 17 23 . —3 12 .03 ICi Secondary Range —Heterogeneity —4 — 1 IS 35 —7 27 27 3 4 27 —4 —9 10 .03 > I d Secondary Range —Institutional Spheres 7 —9 —2 24 —18 30 —13 - 5 —9 7 26 —9 2 .30 D IDi Community Attachment in —School Tax 12 —I 14 . —16 2 i —S 4 39 3 —13 27 12 . S .05 C IDs Community Attachment —Chari ty 10 12 4 3 —3 —18 7 — ' 13 —11 38 2 5 .10 Net A v g . Percentage in - C* Predicted Direction 10 ' 2 10 9 — 16 7 13 . 2 7 24 1 8 \>* Primary Participation IIA Primary Range 20 3 21 42 S —8 . 15 24 2 - 4 7 22 19 10 .01 (-) I1B Role Integration: z£ Occupational Community 23 11 35 3 1 — 14 25 4 . 20 S IS, 13 .01 t> I I C Friendship l-i Circles 2 18. 9 16 13 —17 1 35 —2 11 17 —5 S .02 o IID Stability of Friendships 14 —4 5 8 2 3 —11 —9 12 4 —1 23 4 .40 H E Overlapping Friendships 17 and Memberships 3 —9 13 —18 IS 18 —12 4 —7 5 — 3 .50 Net A v g . Percentage In Predicted Direction 12 4 17 10 8 —1 Z 16 2 1 10 10 6 N *** (69) (86) (3.1) (46) . (83) (51) (66) (47) (61) (13) (S3) (37) (643) * Per cent in predicted direction (proportion by which orderly exceed disorderly in high vitality or strong attachment). Each figure is based on a four-fold sub-table. ** The method for computing this combined probability is described i n R. A . Fisher, Statistical Methods for Research Workers, Edinburgh: Oliver and B o y d , L t d . , 1944, 9th edition, pp. 99-101. »»* N ' s on which percentages are computed vary slightly within columns due to different response rates for different measures. 532 AMERICAN SOCIOLOGICAL REVIEW The results for each dimension of secondary atachment folow: A. Men with orderly careers will have more memberships in formal associations and attend more meetings (including church serv-ices ) than men with disorderly work histories. All but one of the 12 comparisons conform in direction to the hypothesis—a net average diference of 13 per cent. p<.0l. B. Men with orderly careers will average more hours a month in all activities of formal associations (excluding church services) than men with disorderly work histories. Differ-ences are in the predicted direction in 10 of the 12 comparisons; the net average in pre-dicted direction is 12 per cent. p<.03. Using a measure which sifts out the activists, 29 per cent of the orderly appear in the highest quartile (10 or more hours per month) com-pared to 19 per cent of the disorderly. C. Men with orderly careers will range widely in their secondary attachments more often than men with disorderly work his-tories: the participants among the former will be exposed to organizations representing a greater variety of values, interests, and status levels. Using the two indexes—hetero-geneity of membership and number of insti-tutional spheres covered by the respondent's memberships—the hypothesis is restated: 1. The memberships of the orderly will offer greater opportunity for interacting with persons who differ from themselves in impor-tant social characteristics. Diferences are in the predicted direction in seven of the 12 comparisons. p<.03. The net average in the right direction is 10 per cent. 2. The memberships oj the orderly will ex-pose them to more of the major institutional spheres of the society. Only five of the 12 comparisons conform in direction; the net average diference is +2 per cent. The results do not even approach significance. D. Men with orderly careers will have stronger attachments to the community than men with disorderly work histories. Two in-dicators: 1. The orderly will more often support local schools; they will remember the last school election and report that they voted in favor of increased school taxes. Eight of the 12 comparisons conform in direction—a net average of 8 per cent. p<.05. 2. The orderly will more often report a contribution of $100 or more to churches and .• charity in the past year. Seven of the com- ' parisons are in the right direction. p<.10. The net average is +5 per cent. Using more stringent measures of community atachment ;• —$200 or more or more than l.S per cent of • family income—this smal net average is wiped out and the patern is insignificant. Orderliness predicts voluntary giving of mod-est amounts; it docs not predict the donation of large sums or (taking account of ability-to-contribute) eforts going beyond the me-dian. .' II. Men whose careers arc orderly will ev-idence greater vitality of primary relations (a pattern of contact with kin, friend, and neighbor which ranges into the wider com-munity) than men whose work histories art disorderly. (It is assumed that the latter often retreat to family-home localism, a lei-sure style which is relatively isolated, al-though it may include many relationships and frequent interaction.) The data, pre-: sented in the lower half of Table 3, generaly support the hypothesis. Of 60 possible com-parisons of orderly and disorderly (five di-mensions of primary relations for 12 educa-ion-income-age categories), only 13 show percentage diferences in the unexpected di-rection. The net average diference in the j predicted direction is eight per cent. Below is an analysis of sub-tables for each hy-pothesis specifying each dimension of pri-mary group vitality. A- Compared with men whose work his-tories are disorderly, men whose careers are orderly will have a wider range of primary contacts—a pattern generally going beyond family and relatives in the same neighbor-hood and income bracket, and beyond others in the same neighborhood or type oj work. Ten of the 12 comparisons conform in direc-tion to the hypothesis—a net average differ-ence of 10 per cent. p<.01. B. Men whose careers are orderly will in-tegrate work and non-work roles to a greater extent than men whose work histories are dis-orderly; their best friends will more often be persons in the same workplace or satne line of work and they will more often see work associates socially. All but one of the 12 diferences arc in the right direction. p<.01. The net average diference is -f-13 per cent. When a looser definition is applied CAREERS AND SOCIAL PARTICIPATION S33 TABLE 4. PER C E N T DIVFERKNCE JN PREDICTED DIRKCTICW FOR EACH DmF. s s ioN OF T H E VITALITY OF PARTICIPATION, CONTROLUKO SINCI.Y FOR KDUCATIOST, INCOME, AND A G E • Grades of School Completed Total F a n ily Income Age Without I t or 12 ( U S More 'Under $3,000 Under 40 & Controls Fewer Grad.) Than 12 $3,000 & Over' 40 Older Formal Associations and Community l\ N o . of Memberships 14%» 5% 14% 17% 13% 15% 14% 145? JB Frequency, Time, —Organizations 13 10 13 3 11 14 12 14 jCi Secondary Range —Heterogeneity 6 6 11 — 1 14 2 12 . ICa Secondary Range —Institutional Spheres — 2 -S 4 —4 4 •—4 6 I D , Community Attachment —School Tax 11 3 14 12 7 16 IS 3 IDs Community Attachment * —Chari ty 11 14 —1 17 8 10 IS 2 Primary Participation IIA Primary Range 13 IS 9 5 4 2S 14 12 IIB Role Integration Occupational Community 7 9 7 4 4 10 9 4 IIC Friendship Circles 6 9 6 S 4 12 4 10 I1D Stability of Friendships 6 12 —2 12 6 3 S 4 H E Overlapping Friendships and Memberships 2 - 3 7 — 4 —1 3 —1 N (643)** (232) (247) (164) (363) (280) (363) (2S0) * A l l percentages refer to how much the men with orderly careers exceeded those with disorderly work histories in strong attachment or high score. * * N ' s on which percentages arc computed vary slightly within column due to different response rates. (no best friends from work, but sees people from work at least once a week) the net dif-ference goes down to 7 per cent. The comple-mentary hypothesis that (/^orderliness leads to a sharp split between work and leisure (no felow-worker friends, no contacts) is sup-ported but not as strongly as the above. C. Men with orderly careers will have two or three best friends who know each other well more often than those with disorderly work histories. Nine of the 12 comparisons are in the right direction. p<.02. Net aver-age diference: 8 per cent. D. Men with orderly careers will have more long-lasting friendships than men with disorderly work histories. Diferences are in the predicted direction in eight of the 12 comparisons. p<.40. Net average in the right direction: only 4 per cent. A complete history of the picking up and dropping of friends would be beter than number of years each best friend was known, which is too much a function of age.ia 1 8 As in the case of the occupational community, however, college men with high incomes and or-E. Men whose careers are orderly will have best friends in clubs and organizations more often than those with disorderly work histories. Seven of the 12 comparisons con-form in direction. p<.50. The net average ' diference is only 3 per cent. The conditioning eject of major social characteristics. What about education, in-come, and age (variables known to be related to social participation) ? How do they condi-tion the impact of an orderly career on the itality of participation? Table 4, controling for these variables one at a time, shows that education, income, and age for the most part.have weak and some-what erratic conditioning efects. Education is most consistent: in five of six measures of secondary atachment a high school diploma or some colege experience strengthens the tendency of the orderly to participate most; derly careers score strong; not one such person in 71 cases falls in the fate-gory "short-term friends or no best friends." Contrast low-income older men with less than high school, and disorderly work his-tories: 14 per cent fit the category (five in 35 cases). 534 AMERICAN SOCIOLOGICAL REVIEW TABLE 5. COLLEGE EDUCATION AKD AN ORDERLY CAREER AUK BETTER PREDICTORS OP STROKC ATTACHMENTS TO FORMAL ASSOCIATIONS AND T H E COMMUNITY - THAN INCOME AND ACE Grades of School Completed To ta l Work 'His tory ; Fami ly Income Age 11 12 M o r e Dis - or (HS Than Under ?3,O0O Under 40 & Ordvrly orderly Fewer Grad.) 12 SS.000 & O v e r 40 Over I A N o . of Memberships 5S%* 41% 37% 54% 60% 47% 54% 50% 50% I B Frequency, Time, —Organizations . 54 41 38 • 54 60 47 52 53 46 I C j Secondary Range —Heterogeneity 58 52 57 5 f 58 57 55 55 S7 I C , Secondary Range —Institutional Spheres 55 55 S4 56 SS 54 57 54 S6 I D . Community Attachment —School Tax 44 33 34 42 46 39 41 39 41 I D , Community Attachment —Chari ty 78 67 69 75 80 69 81 69 81 N (420)** (223) (232) (247) (164) (363) (280) (363) (280) * A l l percentages refer to proportions scoring high vitality or strong attachment. ** N ' s on which percentages are computed vary slightly within columns due to different response rates. in three of five measures of primary partici-pation (range, integration, circles), educa-tion weakens this tendency. Income has its strongest effect on the relationship between orderliness and range of social contacts—the more income, the sharper the contrasts. (E.g., controlling for age and education among men with $8,000 or more, the net average differ-ence between orderly and disorderly in per cent whose primary contacts range widely is 24). The only aspect of participation for which age has an appreciable effect on the hypothesized relationship is that for second-ary range. In general, these control variables boost the effect of an orderly career most among younger, high-income college men and with respect to support of schools, contributions to church and charity, and primary range— the measures most clearly reflecting commu-nity attachment. (For instance: on local charity, the young, high-income college men with orderly careers have a 38 per cent edge over their disorderly counterparts—or, by the criterion of a $200 annual contribution, a 58 per cent edge. There are similarly strong relationships between orderliness and a "yes" vote on school taxes.) TABLE 6. A N ORDERLY CAREER IS A BETTER PREDICTOR OP HICII< VITALITY o r PRIMARY PARTICIPATION THAN EDUCATION, INCOME, OR ACB Work History Grades of -School Completed Tota l Family Income Age 11 12 More • — , Dis- or (HS Than Under ss.ooo Under 40 & Orderly orderly Fewer Grad.) 12 . ?S,000 &Over 40 Over 1IA Primary Range 55%* 42% 45% 51% 57% 49% 52% 57% ' 42% 1115 Role Integration Occupational Community 33 23 • 33 29 25 29 30 32 27 1IC Friendship Circles 80 74 79 75 80 79 76 79 76 H D Stability of Friendships 55 49 56 48 56 SO .56 40 69 H E Ovcrlapping F i icndships and Memberships 49 47 48 45 52 SO 46 46 51 N (420)** (22S) (232) (247) (164) (363) (280) (363) (280) * A l l percentages refer to proportions scoring high vitality or strong attachment. **N*s on which percentages are computed vary slightly due to different response rales for different measures. CAREERS AND SOCIAL PARTICIPATION S3S perhaps this is a hint that the stratifica-tion system functions most clearly with re-spect to community-oriented behavior and among those who range informaly across strata. In other contexts—with respect to the more private associations which lie be-tween the nuclear family and community in-stiutions—the stratification order and the criteria on which it is based are reinterpreted and their impact lessened. The relative effect, oj social characteristics anil work history. Tables S and 6 compare, without controls, the efect of career, educa-tion, income, and age. Table 5 shows that orderliness comes close to education in im-portance as a predictor of secondary partici-pation and is clearly more important than income or age. (Note that comparing the "some colege" men with the "less than IIS" brings a 24 per cent diference on number of memberships. If career type were equaly re-fined—e.g., by dropping the "borderline orderly" category—we might achieve mote than 14 per cent, but sample size does not permit it.) In its efect across all measures of primary participation, career type is move im-portant than any of the other variables. (See Table 6. While the diferences here between orderly and disorderly are smal but consist-ent, the diferences by education, income, and age are both smal and erratic.) Tables 7 and S, which pile one variable on top of another and contrast the extremes, per-mit us to examine further the interplay of all four variables. We see that their cumulative impact on most dimensions of participation is quite' strong. SUMMARY AND INTERPRETATION OF FINDINGS Orderly experience in the economic system is associated with many social ties which range broadly and at the same time overlap. Men who have predictable careers for at least a fifth of their worklives belong to more organizations, atend more meetings, and average more hours in organizational activ-ity. Their atachments to the local commu-nity are also stronger—indicated by support of local schools and, to a lesser extent, by contributions to church and charity. In both formal and informal contacts, the men of orderly career, mote than their col-leagues of chaos, arc exposed lo a great vari-ety of people: the felow-members they see in clubs and organizations represent many social and economic levels; frequently-seen relatives and close friends are more scatered in space both social and geographical, cutting across neighborhoods, workplaces, occupa-tions, or income brackets. Finaly, the total participation patern of the orderly is more coherent: close friends tend to form a circle and they overlap work contacts. (The data do not support the view that best friends share voluntary association memberships.) There is some indication that these friend-ships, anchored in workplace, forming a lei-sure-time clique, may also be longer-lasting." The efect of work history, while reinforced by other social positions and biographical facts, seems to be distinct. Orderliness, as Table 2 shows, is unrelated to three indica-tors of early socialization (religion, father's occupation, and intergenerational mobility). It is also unrelated to distance and direction of worklife mobility and portion of worklife self-employed. Finally, work history is clearly' a beter predictor of participation than in-come or age. With respect to associational ties, educa-tional background and work history seem to constitute a developmental sequence. The more education, the more motivation and op-portunity for a career both at work and in voluntary organizations—and the higher edu-cated men with orderly careers are appar-ently making a life plan of both. If we give a man some colege, put him on a stable career ladder, and top it of with a nice family in-come, he will get into the community act. Give a man less than high school, a thor-oughly unpredictable sequence of jobs, a family income of five to eight thousand and it is very likely that his ties to the commu-nity will be few and weak. The sub-tables on which Table 3 is based permit sonic refinement of this rough sketch. Considering both primary and secondary par-In Table .% wlim-ver any sizable difieiviuvs •appear, the young orderly more often have stable friendships th;;n the older orderly—until we come to the college-educated, high-income .croup where he o.'i/cr orderly have a 2-1 per cent ed;..e. Terhaps the young in this category ate O g a n e - a U i i i Men on the mnhc, who have many lightly-held .'.;•,:>.. iuivnts, while their oi ler colleagues are settled into com-fortable high-seniority positions recitming little mov-ing about among residences and workplaces. T A D L E 7. PES C E N T OF MEN- WITH STRONG ATTACHMENTS TO FORMAT. ASSOCIATIONS AND C O M M U N I T Y , AND PER CENT DIFFERENCE IN PREDICTED DIT.ECTION A s W E ADD EDUCATION, INCOME, AND A C E TO A N ORDERLY CAREER I A X o . o: IB Free.., Time —Orgs. I C , Sec. Range —Hetcrog. I C , Sec. Range Inst. Spheres I D , Com. Attachm. •—School Tax I D , • Com. Attachm. —:Charity X * 420 223 2 O 293 Co O o O 6 c o Orderly vs. Disorderly College or K . S . Orderly vs. Less than H.S . Disorderly College or H.S . H igh Income Orderly vs. Less than H . S . L o w Income Disorderly College, H i g h Income Young Orderly vs. Less than K . S . Young , L o w Income Disorderly 1 4 % * 41 p<.001*** 61 65 G7 34 26 30 54% 13% 41 59 32 61 33 69 36 P < - 0 1 27 23 33 S S % 6% 52 5S 54 57 59 SO 64 p < . 2 0 —14 55% 55 S3 57 56 66 57 44% 11% 33 47 32 52 34 63 2S p< .02 15 18 35 7 S % 11% 67 P < . 0 S 79 61 84 56 87 50 18 28 37 105 153 79 29 pj < 44 * N's may vary slightly from column to column due to cases not ascertained on dependent variables. * * P c r cent in predicted direction (proportion by which orderly exceed disorderly in high vitality or strong attachment). *** C h i square test of significance. T A B L E 8. PER CEXT o r M E N WITH H I G H VITALITY o r PRIMARY PARTICIPATION, AND PER CENT DIFFERENCE I N PREDICTED DIRECTION A s W E ADD EDUCATION, INCOME, AND ACE TO AN ORDERLY CAREER I I A Primary Range I I B Role Integ. Occup. C o m . I I C Friendship Circles I I D Stability Friendships H E Overlapping Friendships and Memberships N * 420 > P$ w in > O to C o > o > o Orderly vs. Disorderly College or H . S . Orderly vs. Less than H . S . Disorderly College or H . S . H i g h Income Orderly vs. Less than H . S . L o w Income Disorderly College, High Income, Young Orderly vs. Less than H . S . , Y o u n g L o w Income Disorderly 55% 42 56 37 60 41 67 47 13%** p < . 0 1 * * » 33% 23 10% p < . 0 2 19 19 20 30 25 33 25 36 25 11 S0% 74 78 74 78 76 81 86 6% P < - 1 0 —5 55% 49 52 49 55 49 45 34 P < - 2 0 11 49% 47 49 50 46 52 50 49 2% P < - S 0 —I —6 223 293 105 153 79 29 44 * N ' s may vary slightly from column to column due to cases not ascertained on dependent variables. **Per cent in predicted direction (proportion by which orderly exceed disorderly in high vitali ty or strong attachment), *** C h i so.uarc test of significance. 538 AMERICAN SOCIOLOGICAL REVIEW ticipation, an orderly career shows ils great-est and most consistent efect in columns 1, 3, S, and 11, and little or no efect or one that is inconsistent in columns 2, 9, and 12. Six of these seven columns are at the extremes of education—"colege" vs. "less than high school." But they come in pairs of best and worst at both levels. Orderly careers count among young high-income .colege men, but not among the older high-income colege men.18 Orderliness counts among the young low-income men who never completed high school but not among their older counter-parts. Since "old" in this sample is 40-55, the explanation may be life-cycle diferences in aspirations and generational diferences in experience. In stage of the life cycle, the older men are now in their peak earnings period (whatever the level); their youthful aspira-tions, if unfulfiled, are now being scaled down. Whether education and income are high or low, the balance between rewards and demands is tolerable. In political and economic experience, the older group is a product of the Great Depres-sion. These massive forces apparently over-come the efect of orderliness in the career: What's a little unpredictability in jobs, com-pared to the general chaos of the 1930's? What's a little job chaos compared to the days when they were hard-pressed family men just starting out? Contrast the young: in terms of stage in life cycle, whatever the income or education, their aspirations are likely to soar high in relation to rewards; in terms of generation, their demand for good jobs, stable careers, and general security is as urgent as it is plain. Disorderliness in the work history activates the already established predisposition to pull away from community life—a predisposition rooted in life cycle stage and political generation. Generational diferences may also explain why, among high school graduates, orderli-ness counts in columns 8 and 6 but not in columns 5 and 7. In relative educational op-portunity and atainment, and for their gen-eration, men 40 years and over who grad-uated from high school are like the young i s l n column 12 there arc five exceptions—positive results for two dimensions of secondary attachment and three dimensions of primary relations. Their import is diminished, however, by skewed distri-butions and a small N . high-income men who have had "some col lege"; 19 their behavior should be similar Thus orderliness increases the vitality of botl primary and secondary participation amonis the older high-income high school graduate; and does something for secondary participa^  tio  among the older low-income high school graduates. Orderliness has little or no efect on anything among the younger counterparts SOME IMPLICATIONS 1. If work role, occupational association, and career are central variables in the soci-ology of work, we must develop typologies oi each and put them to sociological use. The r rliness of the work history—i.e., the de-gree to which it fits the model of career—is one useful concept in that effort. When the idea of career is taken seriously and applied to large populations, we see that few men are gripped by careers for the entire worklife. only a minority for as much as half the work-life. We see, too, that in the perspective of decades, the patern and sequence of jobs may have more impact on a man's social life than any one of the many positions he picks up and drops on the way. 2. With regard to participation, the types of social relations covered by "primary" and the types of formal organizations covered by "secondary" (we have tapped everything from a "Marital Therapy Club" to "Joe's Regulars," a semi-formal, bar-connected crinking club)—these are .the stuff of the discipline. The labor-leisure study suggests the theoretical relevance and operational fea-sibility of concepts such as range of partici-pation and role integration. The results support two apparently contradictory as-sumptions that have been made: that role integration and a wide range of atachments are compatible and have similar efects. There may be an upper limit beyond which " It is important to remember that "college men" in the middle mass are typically professionals, tech-nicians, or supervisors of the "semi-" variety—and their education is a neat fit. The modal case had one to three years of college; only 30 per cent of the "some college" have a bachelor's degree. IJ is impressive that merely a taste of college;—even among those who did not quite make it t h r o u g h -sorts out the college men from others. Both self-selection and training are doubtless among the causes. SEP x o ^©9 CAREERS AND SOCIAL .PARTICIPATION 539 strong role integration blocks exposure to diverse vah ;s, interests, and strata, and a broad range of participation produces severe role conflict with a consequent withdrawal, but at the medium levels tapped by our meas-ures, and in the middle mass, coherence and range go together. 3. The data are consistent, with a major assumption of the study: with advancing in-dustrialism and urbanism, traditional indi-cators of social class—present income and occupational category of self or father—no longer discriminate among styles of leisure and degrees of social integration for a grow-ing middle mass. As determinants of social relations, media exposure, and consumption, these "class" variables are becoming less im-portant than career pattern, mobility orienta-tion, and work mileu—and the associated educational experiences. 4. The data are consistent with the guid-ing hypothesis: chaotic experience in the eco-nomic order fosters a retreat from both work and the larger communal life. It may be that the economic and non-economic spheres are so interdependent that this causal formula-tion cannot be tested—that an orderly career facilitates participation, participation boosts chances for an orderly career, and both are afected by underlying processes of personal self-selection and. organizational recruitment. Three considerations, however, suggest that the causal sequence is: education orderly career ~» pa jcipation, whatever the interac-tion of these variables once the process is underway. First, preliminary analysis not here reported shows that "pre-occupational leisure training" in adolescence (e.g., par-ental participation when respondent was a teenger, respondent's own activities at that time) does not afect later participation. Sec-ond, several crude indicators of early sociali-zation are unrelated to work history (see Tble 2), so the argument that personality prc-dispositions cause both orderliness and p rticipation seems weak. Although an inten-sive psychological analysis might explain part of the variance, these men are not merely recapitulating family history. Finally, the structural forces shaping adult life—eco-nomic crises in economy and firm, life cycle stage—seem to account for the reported vari-ations by age, income, and education in the relation of work history to participation. In so far as job paterns in modern econ-omies become more orderly and more of the population achieves the position of the young, highincome, "some colege" men of the mid-dle mass, participation in community life is likely to increase, whatever its quality. It is quite possible, however, that other structural changes—in the content of jobs, the schedule of work, hours of leisure and the agencies that serve it—wil ofset those labor force trends that make for more predictable ca-reers. It is possible, too, that the generations will again be divided by different trauma: most of the men in our sample did not grow up in the Afluent Society; a little orderliness for them may go further in stimulating com-munity participation than a lot of orderliness will do for a generation with its eyes glued to the television screen and its energies de-voted to the ardent consumption of "goodies" pouring from automated industry. APPENDIX VII THE DISPOSITION OF THE POPULATION AND SAMPLE 174 ft The Disposition of the Population and Sample. Company Other Total Listed i n directory as working 3,687 4,523 8,210 Ratio of "Company'7':0thern in directory .815 1.0 Drawn in sample 347 115 462 Ratio of "Cornpany'7"0ther" in sample 3.0 1.0 Per cent of population sampled 9.4 2.5 Interviews completed 239 69 308 Weighting factor applied 1.0 4.0 Number after weighting 239 275 515 Ratio of "Corapany57r'Other" after weighting .865 1.0 Difference between "Company'VOther" ratios in directory and i n weighted interviews .05 Per cent of sampling population interviewed - unweighted - weighted 6.5 6.5 1.5 6.1 3.8 6.3 Taken from Martin ileissner, A Study of Work and Social Participation in an Industrial Coramunity: A Preliminary Report. Vancouver, British Columbia: Department of Anthropology and Sociology, University of British Columbia; 1967, page 7. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items