THE RELATIONSHIP BETWEEN .FAMILY ROLE STRUCTURE AND FERTILITY ' Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY WILLIAM O'HARE 1976 LIBRARY Michigan Stat: University . This is to certify that the thesis entitled THE RELATIONSHIP BETWEEN FAMILY ROLE STRUCTURE AND FERTILITY presented by William P. O'Hare has been accepted towards fulfillment of the requirements for 7/4; Jegree in M70 . Major professor Datefléflo 62; /¢ 7é 0.7639 ' NOW“ ’2‘Ep'1398 E32133 NOV 0-6 2006 ABSTRACT THE RELATIONSHIP BETWEEN FAMILY ROLE STRUCTURE AND FERTILITY By William O'Hare Past research on the relationship between family role structure and fertility suggests that couples who share family and household activities and functions have fewer children than couples who do not share in the performance of family and household functions. This relationship serves as the point of departure for the present study. The relationship between family role structure and fertility is re-examined by employing three indices of family role structure and two fertility measures. The extent to which recent changes in family role structure are related to recent changes in period fertility rates is also investigated. The study is based on data gathered in 1955 and l97l in Detroit, Michigan Standard Metro- politan Statistical Area. Analysis shows that there isaisignificant negative correla- tion between the number of children born to a family and the degree of family role integration in task performance, decision making and general family role structure. Couples who share household tasks and decision making generally have fewer children than couples who William O'Hare do not share task performance and decision making. The relationship between family role structure and actual fertility remains unchanged while controlling for age of the respondent, age at marriage of the wife, and length of marriage, educationtyfwife, and family income. The correlations between desired fertility and family role structure indices were not consistent, therefore no firm conclusions are drawn concerning this relationship. There are strong indications that the family role structure of the typical American family became less segregated between 1955 and l97l. This change in family role structure is offered as one reason for the decrease in aggregate fertility over the period. There is also a close association between the amount of decrease in family role segregation and the amount of decrease in the fertility level of several sub-groups between 1955 and l97l. / THE RELATIONSHIP BETWEEN FAMILY ROLE STRUCTURE AND FERTILITY By Qo“ °'< William O'Hare A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Sociology 1976 ACKNOWLEDGEMENTS I would like to thank many people for their help and support. I would like to thank the people at the Detroit Area Study Group and the Institute for Survey Research at the University of Michigan for their help in securing the data for this study. I would also like to thank Lynn DuBois for her clerical help in pro- ducing earlier drafts of the thesis. I would like to acknowledge the userl comments and guidance provided by Dr. Richard Hill and Dr. Bernard Finifter, members of my guidance committee. I would like to give special thanks to the two people I have worked most closely with in producing this thesis, Dr. James Zurches and Dr. J. Allan Beegle. These two faculty members have provided a very useful source of information and guidance throughout my days as a graduate student. I would also like to acknowledge some of the people who have supported and encouraged me throughout my student career. I would like to thank my wife Maria-Elena, who has given me moral support throughout my graduate student days. Undoubtedly the two pe0ple who have shaped my life most are my mother and father, Martha and Arthur O'Hare. WhateverI achieve and whatever I become is due largely to the love and support I have received from my parents. Thank you. ii TABLE OF CONTENTS LIST OF TABLES . Chapter I. INTRODUCTION II. LITERATURE AND THEORY III. METHODOLOGY. The Data The Variables . The Hypotheses IV. FINDINGS REGARDING THE RELATIONSHIP BETWEEN FAMILY ROLE STRUCTURE AND FERTILITY . . V. FINDINGS RELATED TO RECENT CHANGES IN FERTILITY AND FAMILY ROLE STRUCTURE. . v1. sumnnv AND CONCLUSIONS APPENDIX A--DISTRIBUTION OF RESPONSES FOR QUESTIONS ON FAMILY ROLE STRUCTURE . . APPENDIX B--INTER-RELATIONSHIPS AMONG FAMILY ROLE STRUCTURE ITEMS AND INDICES APPENDIX C--FREQUENCY DISTRIBUTION OF MAJOR VARIABLES . APPENDIX D--INTER-CORRELATIONS AMONG THE MAJOR VARIABLES . APPENDIX E--DESCRIPTION OF SUB- GROUPS USED IN TABLES 15-19. . . . . . . REFERENCES Page iv 73 93 TOG 113 119 I30 T33 T35 LIST OF TABLES TABLE 1. ZERO-ORDER CORRELATIONS BETWEEN FAMILY ROLE STRUCTURE VARIABLES AND FERTILITY VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1955 . ZERO-ORDER CORRELATIONS BETWEEN FAMILY ROLE STRUCTURE VARIABLES AND FERTILITY VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1971 . PARTIAL CORRELATIONS BETWEEN ACTUAL FERTILITY AND FAMILY ROLE STRUCTURE VARIABLES WHILE CONTROLLING FOR SEVERAL EXTRANEOUS VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1955 . PARTIAL CORRELATIONS BETWEEN ACTUAL FERTILITY AND FAMILY ROLE STRUCTURE VARIABLES WHILE CONTROLLING FOR SEVERAL EXTRANEOUS VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1971 . . . MULTIPLE CORRELATION ANALYSIS OF ACTUAL FERTILITY AS A DEPENDENT VARIABLE WITH FRSI AND FOUR OTHER INDEPENDENT VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1955 . . . MULTIPLE CORRELATION ANALYSIS OF ACTUAL FERTILITY AS A DEPENDENT VARIABLE WITH TDI AND FOUR OTHER INDEPENDENT VARIABLES FOR MARRIED WOMEN AGE 21 T0 45 IN 1955 . . MULTIPLE CORRELATION ANALYSIS OF ACTUAL FERTILITY AS A DEPENDENT VARIABLE WITH FDMI AND FOUR OTHER INDEPENDENT VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1955 . . . MULTIPLE CORRELATION ANALYSIS OF ACTUAL FERTILITY AS A DEPENDENT VARIABLE WITH FRSI AND FOUR OTHER INDEPENDENT VARIABLES FOR MARRIED WOMEN AGE 21 TO 45 IN 1971 . . . . . . MULTIPLE CORRELATION ANALYSIS OF ACTUAL FERTILITY AS A DEPENDENT VARIABLE WITH TDI AND FOUR OTHER INDEPENDENT VARIABLES FOR MARRIED WOMEN AGE 21 T0 45 IN 1971 . . . iv Page 48 48 58 6O 63 64 65 66 67 TABLE 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. MULTIPLE CORRELATION ANALYSIS OF ACTUAL FERTILITY AS A DEPENDENT VARIABLE WITH FDMI AND FOUR OTHER INDEPENDENT VARIABLES FOR MARRIED WOMEN AGE 21 T0 45 IN 1971 . . . MEAN ACTUAL FERTILITY BY FIVE YEAR AGE GROUPS FOR 1955 AND 1971 . . MEAN OF FRSI SCORES BY FIVE YEAR AGE GROUPS FOR 1955 AND 1971 . . . . . . . . MEAN, STANDARD, DEVIATION, AND STANDARD ERROR OF THE MEAN FOR PARITY LEVELS OF MARRIED WOMEN AGE 21 TO 29 IN 1955 AND 1971 . . MEAN, STANDARD DEVIATION, AND STANDARD ERROR OF THE MEAN FOR FAMILY ROLE STRUCTURE VARIABLES FOR MARRIED WOMEN AGE 21 T0 29 IN 1955 AND 1971 SUMMARY OF ACTUAL FERTILITY OF MARRIED WOMEN AGE 21 TO 29 IN SELECTED SUB- GROUPS IN 1955 AND 1971 . . . . . . SUMMARY OF FRSI SCORES FOR MARRIED WOMEN AGE 21 TO 29 IN SELECTED SUB-GROUPS IN 1955 AND 1971 I SUMMARY OF FDMI SCORES FOR MARRIED WOMEN AGE 21 TO 29 IN SELECTED SUB-GROUPS IN 1955 AND 1971 RANK ORDERINGS OF SUB-GROUPS BASED ON CHANGES IN MEAN 0F FRSI SCORES AND MEAN ACTUAL FERTILITY BETWEEN 1955 AND 1971 . . . . . RANK ORDERINGS OF SUB- GROUPS BASED ON CHANGES IN MEAN OF FDMI SCORES AND MEAN ACTUAL FERTILITY BETWEEN 1955 AND 1971 . . SUMMARY OF MAJOR HYPOTHESES TESTED . Page 68 74 74' 79 82 85 86 87 90 90 94 CHAPTER I INTRODUCTION Recent fertility trends in the United States have been well documented and are summarized by the U.S. Census Bureau in the following way: "The record of the Nation's fertility since World War II was one of a rapid rise in the birth rate to a high level that persisted for approximatelya dozen years after the end of the war, followed by a decline since the late 1950's. The decline has been especially sharp between December 1970 and November 1973, to levels of fertility which appear to be lower than any previously on record for the United States" (34:1). This change is reflected in the fact that the crude birth rate for the U.S. was 25.2 in 1957 but dropped to 15.6 by 1972 (32 i). In 1973 and 1974 the crude birth rate was 15.0 (36:1). The dr0p in the crude birth rate between 1957 and 1974 amounts to a reduction of about 40%. The movement of the general fertility rate in the U.S. followed a similar trend over the past couple of decades. The general fertility rate dropped by about 40% from 123.0 in 1957 to 73.4 in 1972 (32:2). Despite the substantial changes in the period fertility rates of the U.S. over the past three decades, little has been accomplished by demographers in an effort to determine what social forces may have caused, or contributed to these changes. In a recent article, Ryder commented on the fertility experience of the 1 of the U.S. since the Second World War and stated that "...the level of fertility has gone up for every sub-group, and then it has gone down for every sub-group, and we are far from an explanation of why that happened and whether it will happen again" (28:505). The recent decline in the period fertility is made more intriguing by the fact that the large birth cohorts of the post war "baby boom" have recently begun entering the prime childbearing ages. Under a constant age-specific fertility schedule this event would result in an increase in the number of births and crude measures of fertility. The present study, which focuses on the situation outlined above by Ryder, is undertaken with the understanding that a complete explanation of the recent fertility decline is not likely to come from a single study but from numerous studies each of which provides a partial explanation. A simple model involving family role structure and.ferti1ity will be developed and tested as one explanation for the decline in the period fertility rate of the U.S. between the mid 1950's and the early 1970's. The study is undertaken with the idea expressed by Ford and DeJong that, "changes in the birth rate which cannot be attributed to changes in the population structure often provide clues to significant changes in the working of the family system" (10:155). Although some authors as O'Neill and O'Neill (24), Glick and Norton (12), and Bernard (1) have suggested that family life in America has changed substantially over the past 20 years, there have been few empirical studies relating these changes in family life to the recent changes in period fertility. The study conducted here will use some survey data gathered in the Detroit, Michigan area in 1955 and 1971 to examine the family role structure-fertility re- lationship and the relationship between recent changes in family role structure and period fertility. The analysis of the collected data will involve two distinct stages. First, I will test a model Concerning family role structure and fertility which has already received some empirical support from other research. The model will be tested with data collected at two different points in time. In addition to testing the model as it has been used in previous studies, I will be examining the concept of family role structure in more detail than has been the case in past studies. The second stage of the analysis will employ the family role structure-fertility model to help explain the decline in period fertility between the mid 1950's and the early 1970's. I will show that there have been changes in the internal structure of the typical American family which would lead one to expect the period fertility decline which has been observed. Changes in period fertility be- tween the mid l950's and the early 1970's will be further analyzed by examining the relationship between changes in family role struc- ture and fertility in selected sub-groups of the total population. In the following chapters of this thesis, I review relevant literature and theory, describe the source of the data used, ex- plain how the key variables of family role structure and fertility are operationalized, state the major hypotheses of the study, present the analysis and findings, and conclude with a summary of major findings and conclusions. CHAPTER II LITERATURE AND THEORY Many studies have examined the relationships between various characteristics of families and the fertility of those families. Since most of these studies found that one or more dimensions of family life are closely related to fertility, it is reasonable to suspect that changes in the typical family unit in a society are likely to be associated with, and perhaps cause, changes in the level of fertility of the society. 0f the fertility studies that incorporate the concept of family, only a few are directly related to the study conducted here. The studies which focus on the interaction between the spouses in a family will be examined closely while other studies which examine dimensions of the family in relation to fertility will be references only briefly in developing the theoretical background for the present study. In several studies, researchers have conceptualized the family as a distinct unit within the social structure. This approach was used by Blake (2) in her study of fertility in Jamaica, Freedman and Takeshita (11) in their study of family planning in Taiwan, and by Yaukey (40) in his study of population in Lebanon. These researchers examined the predominant type of family unit in in the society and considered its relationship to the fertility level of the society. While these studies demonstrate that the type of family unit present in a society has a direct bearing on the fertility level of the society, none involved analysis related to the internal structure of the family roles. Another set of studies focus on the relationship between certain characteristics or experiences of one spouse in the family and the fertility of the family. Goldberg (13), Ryder and Westoff (27), Westoff (37) and Kupinsky (18), among others, have examined the work experience of the wife in relation to her childbearing. Generally wives who work outside the home have fewer children than wives who do not work outside the home. Ryder and Westoff summarize the relationship this way: "Women who are working or who have worked expect (and to a lesser extent want) fewer children than those who have never worked. Those who are working because they 'like to work' expect the smallest families of all" (27:90). Kupinsky (18) suggests that working outside the home may increase fertility by weakening the extent to which the woman identifies with the wife-mother role, and increase the extent to which the women's self-concept is based on a non-familial role. Through Kupinsky's argument, a good deal of research on fertility and the work experience of the wife can be tied to the wife-mother role which is a role that is instrumental in the present study. Michel and Fayerabend (21) as well as Kiser and Whelpton (l6) fbund that marital satisfaction and marital adjustment of the wife are related to fertility. Families where the wife is satisfied and adjusts well to married life have fewer children than families where the wife is unsatisfied and does not adjust well. Although norms associated with the wife-mother role were used to measure the satisfaction and adjustment of the wife, the extent to which family role interaction contributed to the satisfaction or adjustment of the wife was not discussed in these studies. It has also been shown that the value structure of one spouse may affect the childbearing of the couple. Clifford (8) found that the general value orientation of the wife is associated with the number of children the wife desires and actually bears. A per- son who is oriented towards the future and who believes that man- kind has a good deal of control over nature is categorized as having a "modern value orientation" anda person who is oriented toward the present and believes that mankind has little control over nature is categorized as having a "traditional value orientation." .Clifford found that women with a Amodern value orientation" desired and have fewer children than women who have a "traditional value orientation. The studies mentioned thus far have shown that many aspects of family life are associated with fertility, but none has dealt directly with interaction between the spouses in a family. The extent of marital role interaction in family life will be a crucial concept in this study, and it is the focus of the studies described next. Hill's (14) study of family and fertility in Puerto Rico and Rainwater's (25,26) studies of poor families in the U.S. examined communication patterns between spouses and the couples contraceptive effectiveness. Both authors conclude that couples who do not communicate well with each other usually do not experience effective contraception. The authors suggest that communication is better between spouses who share family activities and responsibili- ties than spouses who do not share family and household functions. The sharing of thoughts and feelings through communication may be seen as one indicator of the general sharing of family and house- hold activities. Kiser and Whelpton (16) conclude that couples who share the responsibility for controlling births are more successful in con- trolling the number and spacing of their children than couples who do not share this responsibility. This seems reasonable since the probability of both spouses simultaneously neglecting the responsi- bility for using contraceptives is likely to be lower than the probability of a single spouse neglecting the responsibility for using a contraceptive. It is also plausible that the extent to which the responsibility for contraception is shared may be a mediating variable in the communication-fertility relationship noted by Hill (14) and Rainwater (25,26). The studies just cited show that the extent to which spouses share household and family duties and activities, particularly communication, is important in determining the contraceptive efficacy and consequently the fertility of the couple. The extent to which the husband and wife share the tasks and activities re- quired to maintain a household and family is an important component of the theoretical framework proposed by Bott (5). Since the conceptualization of the family used in the present study is taken directly from Bott, a short discussion of Bott's ideas are in order. Bott's theory, derived from her observational study of several English families, is centered on the social roles and role- relationships in the nuclear family and the network of family re- lationships. A role is described as "behavior that is expected of any individual occupying a particular social position" (5:3), and a "role-relationship is defined as those aspects of a relationship that consist (If reciprocal role expectations of each person con- cerning the other" (5:3). Bott's theoretical approach is derived from Lewin's (19,20) theory that behavior of an individual in a social situation is a product of the social environment of the situa- tion and the personality characteristics brought to the situation by the individual. Bott, however, is concerned only with the behavior that is representative of a family role rather than the behavior that may be idiosyncratic or attributable to the personality of the person who happens to be occupying the role. Likewise, in analyz- ing role-relationships within the family, Bott attempts to focus on the behavior related to the roles themselves rather than behavior tied to the personality of the individuals occupying the roles. In analyzing the roles and role-relationships within the nuclear family, Bott concludes that: "There was considerable variation in the way husbands and wives performed their conjugal roles. At one extreme was a family in which the husband and wife carried out as many tasks as possible seperately and independently of each other. . . at the other extreme was a family in which the husband and wife shared as many activities and spent as , much time together as possible" (5:52). 10 Based on these observations, Bott developed two labels to describe the types of role-relationships f0und in families. According to Bott: "...a highly segregated conjugal role-relationship is defined as one in which husband and wife have a relatively large proportion of complementary and in- dependent activities and a relatively small proportion of joint activities. In a joint conjugal role- relationship the proportion of complementary and independent activities is relatively small and the proportion of joint activities is relatively large" (5:55). The type of role-relationship between spouses in a family is re- ferred to as family role structure. Bott's theoretical orientation is consistent with a funda- mental assumption of sociology which holds that patterns of inter- action, often referred to as social structure, emerge in any group which persists over any length of time. While sociology is generally oriented toward large groups, studies show that members of a small group also develop patterns of behavior and interaction which can be described as a group structure. In examining a particular type of small group, a marital dyad, Bott focuses on the patterns of role behavior and role interaction exhibited by members of the dyad and constructs a typology based on the type of role interaction observed. The typology developed by Bott will be altered slightly in the present study, but Bott's fundamental framework will be the basis for developing a measure of family role structure in the present study. Although Bott generally speaks in terms of a family having either a "joint" or a "segregated" family role structure, one can 11 easily use Bott's idea of conjugal role-relationships to measure family role structure on a continuum of role segregation rather than the simple dichotomy used by Bott. This idea will be developed further in the next chapter. The studies described next directly examine the relationship between family role structure, as defined by Bott, and fertility. The three stage study generally referred to as the "Princeton Study," or Family Growth in Metropolitan America (38, 37, 6) examined several areas of family life including the extent to which power in decision making was shared by the husband and the wife. Although the authors of the Princeton Study found no support for a family role structure-fertility relationship, and conclude that "there seems to be little point to further investigation in these areas of family relationships and fertility" (37:197), there are a couple of factors which should be considered when assessing the outcome of the Princeton Study. First, the authors are not clear in describing exactly how family interaction was measured. It is possible that the instrument used to measure family interaction was inadequate. The second factor which may have affected the outcome of the study is the sample. All the respondents in the study were from very large cities and the respondents were homogeneous on many other important charac- teristics. Considering the outcome of other studies which use dimensions of family role structure, it seems likely that one or more of the factors mentioned above influenced the study in such a 12 way that the predicted relationship between family role structure and fertility was not adequately tested. A recent study of fertility in Metropolitan Latin America (7) used the concept of family role structure as independent variable. The extent to which the spouses in a nuclear family shared power and authority in the family is the dimension of family role structure used in the study. Although there was little amplification of their finding, the authors conclude that "the greater the male authority, the greater the desire for more children" (7:88). This suggests that families where the power is shared by both spouses desire fewer children than families where the power rests primarily in one spouse, the male. More generally, it supports the idea that the less interwoven the marital roles are, the more children the couple will desire and have. The study which is most relevant to the study being con- ducted here was done by Stokes (30). Stokes, who builds on the theoretical insights of Bott, operationalized the concept of family role structure by asking wives how much they and their spouses share several specific duties and activities involving the family and household. The 17 specific items used to derive an index of family role segregation included questions on decisions as well as activities. Stokes showed that as role segregation increased, desired and expected fertility increased; moreover, an inverse relationship was also found between role integration and actual fertility. Al- though in a path model, family role structure was not found to be an 13 intervening variable between socio-economic status and fertility, Stokes' findings are provocative and demonstrate the utility of using Bott's conceptualization in further studies of the relationship be- tween the interaction of family members and fertility. Summary of Literature and Theory Rainwater (25,26) and Stokes (30) specifically cited Bott's conceptualization of the family in their fertility studies which showed that joint role-relationship families have fewer children than segregated role-relationship families. The study by Hill (14) and the study conducted by CELADE (7) also suggest that joint role-relationship families have lower fertility than segregated role-relationship families, although Bott's framework was not specifically mentioned in either of these studies. All the studies cited above support Rainwaters'contention that: "...the more interwoven are the interests and concerns of the husband and wife the more likely the wife is to want a small or medium-sized family; the more separate are their interests and concerns, the less they see their marital roles as interpenetrating, the more likely the wife is to want a large family" (26:193). The idea expressed by Rainwater will serve as the underlying hypo- thesis for the expectations concerning family role structure and fertility in this study. A couple of other points regarding the theoretical base of this study need clarification. Although the association between family role structure and fertility has been observed in several studies, the direction of the causal link between these two variables has seldom been questioned. One can easily imagine 14 particular situations where the family role structure of a couple affects the couples' childbearing, and one can also imagine a situation where the addition of a child, particularly the first child, affects the interaction of the couple. In short, either variable could be the causal variable in a particular situation. The problem is one of determining which variable is typically the causal variable when a large number of cases are considered, or in other words, the causal variable in the aggregate situation. The determination of which variable is the major causal variable cannot be fully assessed in this study. In a determina- tion of causality it must be shown that the independent variable occurs prior in time to the dependent variable. In the present study, both variables are measured at the same point in time, the time of the survey, consequently, the temporal sequence of occurance cannot be adequately demonstrated from the data in this study, the direction of causality which will be assumed in the model is not derived from any empirical analysis. The assumed direction of causality used in the present study is based on the following argument. Couples seldom have children in the first nine months of marriage. In one major study, only 5% of the couples studied had their first birth during the first 8 months of marriage, and only 29% of the couples ex- perienced their first birth during the first 11 months of marriage (38:117). Therefore, most couples have an opportunity to develop patterns of interaction or role-relationships free from the effects of actual fertility. Since the role-relationship is developed 15 prior to any actual childbearing, the initial family role structure of a couple cannot be caused by actual fertility. Additionally, marital role behavior and the role-relationship between the husband and the wife are shaped by a long process of socialization which occurs prior to marriage, and long before any actual childbearing. 0n the basis of this temporal sequence, it 'hs assumed that family role structure is the major causal variable in the family role structure-fertility relationship. Two more points regarding the theoretical base of this study demand comment. The first point concerns the larger social context surrounding the family role structure-fertility relation- ship. Some thought should be given to the factors which influence family role structure and those variables affected by changing fertility levels. An empirical analysis of the larger social con- text surrounding the family role structure-fertility relationship is beyond the scope of the present study. However,aishort dis- cussion on some aspects of this topic is included in Chapter VI. The second point demanding comment concerns the intervening variables which constitute the causal link or links between family role structure and fertility. In other words, specification of the variables which explain how differences in family role structure cause differences in fertility. The survey data used for the present study werelwrtintended to investigate the family role structure-fertility relationship and consequently it is not sur- prising that the data do not include adequate measures for examin- ing the details of the causal path between family role structure 16 and fertility. Although this topic is not subjected to empirical investigation, a short discussion regarding some possible inter- vening variables is included in Chapter VI. It may be useful to add a few comments here about the broader context of this research. It is widely held in sociology and demography that changes in aggregate fertility are the result of changes in the social structure of a society, and that fertility differences at the family level are related to the social structural environment surrounding the family. In the present study we will be examining the institutions of family and marriage as one aspect of social structure. Although it is assumed that changes in the family system and differences in family life are likely to be re- lated to fertility, it is by no means assumed that these are the only factors affecting fertility. There are a number of factors which not only have a direct bearing on fertility, but are inter- related with the family system to create a social structural en- vironment in which fertility behavior occurs. Withhithiscontext, I believe family role structure is likely to act as an intervening mechanism between many macro-level social structural variables and fertility. At the aggregate level it has been suggested, primarily through the demographic transition model, that changes in the family unit is a mechanism by which socio-economic develOpment resulted in lower fertility rates. By tying differences in family role structure to broader structural variables, the present research may serve to fill the conceptual gap between macro-level sociological variables and fertility 17 behavior. At the micro-level the relationship between family role structure and fertility may serve as a link tying structural charac- teristics such as religion, race, and socio-economic status to fertility behavior of couples. Although the present research will focus on the family role structure-fertility relationship, there will be some evidence presented regarding the relationship between family role structure and some other sociological variables. It is hoped that the present research can be combined with future research to provide a fuller understanding of socio-cultural con- text of fertility behavior. In summary, the relationship between family role structure and fertility derived from the cited studies will serve as the point of departure for the present study. The relationship between family role structure and fertility will be re-examined, then changes in family role structure will be measured and analyzed in the search for some explanation of the decline in the U.S. period fertility rate from the mid 1950's to the early 1970's. CHAPTER III METHODOLOGY The Data The data used in this study are taken from two surveys of the Detroit, Michigan area conducted by the Survey Research Center of the University of Michigan. One survey, conducted in 1955, will be referred to as the 1955 DAS (Detroit Area Study). The second survey, conducted in 1971 as a partial replication of the 1955 DAS, will be referred to as the 1971 DAS. Respondents in the 1955 survey were selected in the following way. The sampling area was established as those parts of the Detroit SMSA which were used as census tracts in the 1950 U.S. Census of Population. Briefly, the selection procedure for the 1955 DAS involved the random selection of primary sampling units in the fbrm of census tracts, the random selection of sample blocks within the selected tracts, the complete listing of all dwelling units in the selected blocks, and a random selection of dwelling units from the block lists. Ultimately, nearly 900 dwell- ing units were selected from about 300 different blocks. A cluster of three or four blocks was selected from each census tract selected in the initial sampling stage. In each selected dwelling unit the wife of the husband-wife household was asked selected questions regarding fertility and 18 19 family interaction. Although households which did not include a husband and wife were included in the 1955 DAS sample, the data from these households will not be included in the analysis conducted in the present study. A more detailed description of the sampling procedure is given by Takeshita (31). The 1971 study was conducted to replicate parts of nine different studies conducted by the Detroit Area Study group from 1953 to 1959 and in 1968 and 1969. An important concern of the 1971 DAS study was identifying a sampling area which would be com- parable to the sampling area used in the studies done in the 1950's. In an effort to make the sampling area of the 1971 DAS comparable to the sampling area of the studies done in the 1950's, several areas north of Detroit were added to the sampling area used in the 1950's studies. The geographic area used in the 1971 DAS included about 85.3 percent of the 1970 population of Wayne, Oakland and Macomb counties. These three counties comprised the Detroit SMSA at the time of the 1970 Census. Although the Survey Research Center in the University of Michigan feels that the 85.3 percent of the 1970 population of the Detroit SMSA used in the 1971 DAS is very similar to the portion of the Detroit area population used in the 1955 DAS, it should be recognized that the two studies employed slightly different sampling areas. Respondents in the 1971 DAS survey were selected in the following way. Census tracts in the sampling area were stratified into four strata based on their racial composition in 1960 and were 20 selected with probabilities proportional to size at the time of the 1960 Census. Within each selected tract, blocks were selected with probabilities proportional to size at the time of the 1960 Census. Within each selected block, all dwelling units were divided into clusters of approximately 6 dwelling units each and two such clusters were randomly selected from each block. Ultimately, nearly 2400 dwelling units were selected from 204 different blocks. In each selected dwelling unit, all adults (persons over 21 years of age) were eligible subjects. One adults per household was randomly selected from all adults in the household to be a respondent.. A more detailed description of the sampling procedure is given by Fisher (9). , Note that only the wife in a selected family was asked the questions about family relationships in the 1955 study but either the husband or the wife was asked the same questions in 1971. Although Safalios-Rothschild (29) has pointed out that one spouses' perception of the situation in a family may differ considerably from the perception of the other spouse and both may differ from some independent description of the situation, unfortunately we cannot return to the middle 1950's, when the country was experienc- ing high levels of fertility to ask both husbands and wives about family interaction. Since the 1955 DAS used only married females to record interaction of family members, only the married females perception of family interaction in the 1971 DAS will be used in the analysis. It should be kept in mind that references to family 21 role structure in this study refer to the wife's perception of family role structure. Since family role structure is being measured exclusively through the wife's perception, factors that may affect the wive's perception of the family situation should be considered. Two such factors, among others, are the desires and expectations about family relationships which the wife brings into the family situation. Although there is little that can be done to measure the effect of these and other such factors in the present study, the two factors mentioned above will be discussed briefly in Chapter VI. Let me also add that from the perspective of the wife, the perceived situation is the real situation. In terms of the effect of family role structure on fertility, the wife's perception of family role structure is probably more important than some in- dependent measure of family role structure, for it is the perceived situation that the wife will react to or base her behavior on. Only women who reported that they were presently married to their first husband will be included in the analysis. Women who do not have a spouse present could not answer the questions about family role structure which will be important for this study. Understandably, the dissolution of a marriage is likely to have an affect on fertility, and since fertility is a major variable in the study, women who have been married more than once cannot be used in the analysis. Since the 1971 DAS sample included only those persons over age 21, all respondents in the 1955 DAS sample who are under age 21 will be excluded from the analysis. 22 Because fertility is a major variable in the study, the analysis will include only those women in their childbearing ages. Since very few women have children after age 45, and women below 21 have already been eliminated from the analysis in order to make the two samples comparable, the analysis will include only those women who were age 21 to 45 at the time of the survey. It should be noted that the sampling procedures used to gather data for the present study do not match the requirements of a simple random sample model. The sampling design for the two surveys employed in this study involved the use of a technique known as cluster sampling. In short, this means that respondents were selected from dwelling units that were located in small geo- graphical clusters. It has been pointed out that people who live in the same geographical area, particularly an area as small as a block, tend to share many characteristics. The degree to which the respondents in a cluster share a common characteristic is often referred to as the inter-class correlation, or the rate of homo- geneuity (roh). Most estimates of sampling error which are used in tests of significance are based on the simple random sample model. Since the sampling designs used to gather data for the present study deviate from the simple random sample model, there is some question about the extent to which it is appropriate to use error estimates and subsequent tests of significance which are based on a simple random sample model. From a theoretical perspective, the most appropriate course would involve calculation of the rate of 23 homogeneity for each variable for each survey. However, from a practical perspective these calculations would be exceedingly time consuming and costly, with little hope of much payoff based on Kish's statement that "in practical survey clusters, roh tends to be greater than zero, sometimes by much often,tnrlittle"(l7:163). The fact that the number of clusters used was large and the size of each cluster was relatively small are factors that would lead one to expect little effect from the interclass correlation. If the calculation of roh is ruled out for the reasons cited above, we are left with two options; (1) not to incorporate any tests of significance in the analysis, or (2) to use tests of significance based on a simple random sample model. The first option means that the reader would have little guidance in deter- mining which statistics or differences are likely to be important and which are easily due to chance. The second option means that the results of significance tests should be viewed more cautiously than usual. I have selected the second option outlined above, that is, I have included standard tests of significance which are based on a simple random sample model. Since the level of significance used in tests of significance is a somewhat arbitrary choice to begin with, small errors introduced by the clustering effect should not be crucial in interpreting the results of the analysis. In summary, the reader is advised that the tests of significance used in the analysis sections are based on a sampling model which is slightly different than sampling model employed in this study. The exact impact of those differences will remain unknown in this study. 24 Since the phenomena that instigated this study and one for which I hope to give a partial explanation is the period decline in fertility at the national level, it may be worth considering the extent to which the data used here are representative of the national population. The population of the Detroit area is cer- tainly very similar to the national population on some character- istics and probably dissimilar on other characteristics. It will be shown later in this paper that women in the Detroit Standard Metropolitan Statistical Area have had fertility experience similar to the fertility experience of the national population in recent years. The extent to which the pattern of family role structure in the Detroit area families is representative of family role structure in the U.S. population is unknown. Although there is no reason to expect wide disparities between family role structure in the Detroit area families and family role structure in families in other parts of the country, there are also no assurances of their similarity. ' In short, there is only one thing that can be said with any certainty concerning generalizations from the Detroit data to the national population. Hypotheses supported by the Detroit data are worth testing with data known to be representative of the national population. Until such analysis is conducted, generaliz- ing the results of this study to the national population should be undertaken with some caution. 25 The Variables Fertility There are many different ways of measuring fertility, both at the individual level and the aggregate level. The particular measure used in a study generally depends on the nature of the study and the data available. When a questionnaire is designed for a specific study, the questions can be designed to build very detailed and pointed measures. However, when a study involves a secondary analysis of previously collected data, which is the case in the present study, one must use whatever measures can be de- rived from the data. In order to test the hypotheses of the study, it will be necessary to measure fertility for individual units as well as for selected groups of families. In addition to examining fer- tility at the individual and the aggregate level, two distinct dimensions of fertility will be used. One dimension is the actual number of children the woman wants or feels is ideal. The number of children a family wants or feels is ideal will be labeled "desired" fertility, for lack of a better term, and the number of children born to a family will be called "actual" fertility. Unfortunately, questions concerning desired fertility are not identical for both surveys used in this study. While Ryder and Westoff (27) have shown that the way a question about desired fertility is phrased may affect the response received, this should not produce a serious problem in this study because the responses 26 to the two forms of the questions dealing with desired fertility will not be directly compared. Respondents in the 1955 DAS sample were asked how many children they "wanted" to have by the time they were 45 years old. Respondents in the 1971 DAS sample were asked what would be the "ideal" number of children for a young couple with a standard of living similar to the respondent's standard of living. The re- sponses to these two questions will serve as the measure of de- sired fertility for each family. The relationship between family role structure and desired fertility will be examined at both points in time, 1955 and 1971, but I will not use desired fertility as a variable in analyzing changes in fertility and family role structure over time. Both questions on desired fertility tap the respondent's desired fer- tility, but not exactly the same aspect of desired fertility. Consequently, analysis of the changes in desired fertility over time would be confbunded by the lack of uniformity in the questions used to measure desired fertility. In short, it would be impossible to determine if any changes in the measured level of desired fer- tility between 1955 and 1971 should be attributed to actual changes in the desired fertility of the population, or to the phrasing of the questions. Unfortunately, the data required for calculating age- specific fertility rates is unavailable. However, the measurement of the actual fertility of a family or a group of families is very simple and straightforward in this study. The number of live 27 births experienced by a woman will serve as the measure of fertility at the individual family level. The total number of births ex- perienced by a group of women divided by the number of women in the group will serve as the measure of actual fertility at the aggregate level. The measure of actual fertility at the aggregate level can be described as the mean number of births per woman, or mean parity level. Actual fertility measured at the individual level will be used to examine the relationship between family role structure and fertility at both survey dates, that is, 1955 and 1971. Each sur- vey will serve as a separate test of predicted relationship between family role structure and fertility. Although the measures of fertility used in this study are dictated by the data available, the measures are adequate to test the major hypotheses of the study. (The movement of the period fertility rates for the popula- tion of the entire country was presented in the introduction of this paper. Changes in the period fertility level of the Detroit SMSA have been similar to changes in the U.S. fertility rates. The 1955 crude birth rate of the Detroit SMSA was 27.0,1 and the crude lThis crude birth rate was derived by dividing the esti- mated total population of Macomb, Oakland and Wayne counties into the total births for these three counties. Both the population figure and the birth figure are taken from the Michigan Department of Health (22). 28 birth rate of the Detroit SMSA in 1971 was 17.9.2 The change in the crude birth rate amounts to a drop of about 34 percent over the 16 year period. . If the fertility of the two samples, the 1955 DAS sample and the 1971 DAS sample, is representative of the people of the Detroit area during those years, then the actual fertility of the 1971 DAS respondents should be lower than the actual fertility of the 1955 DAS respondents. The above statement is a simplification of the problem because the fertility of the population is based only on births occurring in 1955 or 1971 while the fertility of the' sample includes all births occuring to women in the sample, however, the data shows that the young married woman in the 1971 DAS sample have lower actual fertility than the young married women in the 1955 DAS sample. Family Role Structure As mentioned earlier, the conceptualization of family role structure used in this study is derived from Bott. The main tenet of Bott's orientation is that families vary on the extent to which individuals occupying the husband-father role and the wife-mother role interact. In some families the husband and wife share many activities and duties while in other families the spouses do not share many activities and household duties. 2This crude birth rate was derived by taking the number of live births in the three counties of the SMSA given by the Michigan Center for Health Statistics (23) and dividing by the estimated population of the three counties which was taken from the Bureau of the Census (33). 29 Although Bott labels families as either joint conjugal role families, if they share activities and duties, or segregated con- jugal role families if the duties and activities are not shared, the extent of segregation in the family roles can be measured more precisely than Bott's simple dichotomy. Rather than assigning fami- lies into one of two classes of families, the extent of role segre- gation can be measured along a continuum varying from no role interaction or sharing by the spouses, to total interaction and complete sharing of activities and duties. In the present study, family role structure will be measured along a continuum from no role interaction to total role interaction by the spouses. The end point of the continuum characterized by no role interaction between the spouses will be described as family role segregation and the other end of the continuum, characterized by complete sharing of family roles, will be defined as family role integration. Intermediate points on the continuum can be referenced in terms of the degree of family role segregation or family role integration. The operationalization of family role structure used in the present study is patterned after the measure of family role structure employed by Stokes (30). Stokes' measure is based on responses to 17 items asking wives the extent to which certain family and household functions are shared by herself and her spouse. The questions used by Stokes included items involving decision making as well as items pertaining to the actual performance of 3O household tasks. The responses to the 17 items were averaged to derive a general measure of family role structure. In the present study family role structure is measured using wives' responses to twelve specific questions about how family and household functions are performed. The twelve questions used are listed below. 1. Who 2. Who 3. Who 4. Who 5. Who 6. Who 7. Who 8. Who- buy 9. Who Twelve Family Role Structure Items does the grocery shopping? gets your husband's breakfast on workdays? does the evening dishes? straightens the living room when company is coming? repairs things around the house? keeps track of the money and the bills? usually makes the final decision about what car to get? usually makes the final decision about whether or not to some Life Insurance? usually makes the final decision about what house or apartment to take? 10. Who usually flakes the final decision about what job your husband should take? 11. Who usually makes the final decision about whether or not you (wife) should go to work or quit work? 12. Who usually makes the final decision about how much money your family can afford to spend per week on food? For each of these questions the respondents were offered the five response choices listed in Scale 1. 31 Scale 1 l. Husband always. Husband more than wife. Husband and wife exactly the same Wife more than husband. 01-wa Wife always. A frequency distribution of responses to the twelve items for the subjects included in the present analysis is given in Appendix A, Tables A1 to A12. It should be noted that the questions and re- sponse possibilities were identical for the 1955 sample and the 1971 sample. Scale 1 as it is presently structured is designed for measuring which spouse performs a task or makes a decision rather than the extent to which household activities are shared by the spouses. The extent to which one spouse usually makes household decisions, often referred to as dominance, has been used in other fertility studies with little success. In the Princeton Study (38) dominance measured in three areas of family life, was not found to be related to any of the fertility variables used in the study. The Indianapolis study (16) and Hill's study (14) found little evidence of a relationship between male dominance in the household and fertility. Blood and Wolfe (4:130) found little difference in the fertility of families with different dominance patterns. Since there has been little success in relating measures of family dominance to fertility, and in view of the fact that Stokes, whose operationalization of family role structure was based 32 on Bott's concepts, did find a relationship between family role structure and fertility, the responses to Scale 1 will be recoded to construct an index of family role structure similar to the one used by Stokes. In joint role relationship families, the spouses share most of the household tasks and decisions. Therefore, joint role families, or families with an integrated family role structure, will be identified by many responses that fall near the middle of Scale 1. In contrast, families with a segregated family role structure will be identified by many responses which fall at or near the end points of Scale 1. For measuring family role structure as it has been defined here, it is not important which spouse makes a decision or performs a task, but rather whether the task is generally performed or the decision generally made by one spouse more than the other, or shared about equally by both spouses. Therefore, Scale 1 will be transformed into a three point scale to measure family role struc- ture as it will be used in this study. In constructing the new scale, response number 5 of Scale 1 is recoded to a "l" in the new scale, and response number 1 on Scale 1 is unchanged. This means that a "l" in the new scale in- dicates that the decision is always made or the task is always performed by the same spouse, without specifying which spouse. Response number 4 of Scale 1 is recoded as a "2" in the new scale, and a 2 in Scale 1 is unchanged. Therefore a "2" in the new scale indicates a specific task is usually performed or a decision usually 33 made by one Spouse more than the other, but may occasionally be undertaken by the other spouse. Response number 3 of Scale 1 re- mains unchanged and indicates that a task is performed or a decision made equally by both spouses. The new scale, labeled Scale 1A is given below. Scale 1A 1. A task or decision is never shared by the spouses. 2. A task or decision is sometimes shared by the spouses. 3. A task or decision is always shared by the spouses. The twelve questions used here are obviously only a small sample of possible questions regarding the sharing of family acti- vity. Consequently the measure of family role structure derived from these questions will be only a crude measure of the actual family structure of a couple. However, the particular questions used here involve tasks and decisions that are relatively im- portant and common to all families, and examining the extent to which these twelve tasks and decisions are shared should give a good indication of the general extent of role segregation in the family. Since the questions used to develop a measure of family role structure in this study are dictated by the data available, only questions on decision making and task performance are in- cluded here. However, future studies may find it useful to measure family role structure in other areas of family life such as child- rearing activities and income producing activities. 34 In general, there are two bases for combining single items into broader indices. One method is based on an examination of the content of the items. Items which appear to be measuring the same general concept or dimension are combined to construct an index or measure of the concept. The second method involves the use of quantitative analysis. The statistical relationships among the items are studied by examining inter-correlation matrices, per- ‘forming factor analyses or other similar techniques. Both of these approaches are used in building measures of family role structure to be used in the present study. Although each of the twelve questions listed as family role structure items focus on a specific event or situation, they all share a common dimension by the fact that they all examine the performance of household functions. More specifically, they examine who generally performs the functions. Since all twelve questions have a common focus, it seems reasonable from this analysis of their content to combine the twelve items into a single index designed to measure the extent to which the performance of household functions are shared. In order to examine the inter-relationships of the responses to the twelve items, correlations were calculated and inter- correlation matrices were constructed. Tables Bl and B2 in Appendix B show the inter-correlation matrices involving the twelve family role structure items, all recoded to Scale 1A. Table Bl is based on 1955 data and Table B2 is based on 1971 data. Although many of the correlations in Tables B1 and B2 are small, nearly all the 35 correlations in both tables are positive. The few correlations that are negative are all small and there is no correlation that is negative in both tables. The fact that nearly all the correla- tions in both matrices are positive suggests that spouses who share in the performance of one family function share in the performance of many other family functions, and conversely, those spouses who do not share in the performance of a particular family function are likely not to share in the performance of other family and household functions. This in turn suggests that all twelve items share a common dimension, and at least to some extent are measuring the same concept. Since the twelve family role structure items appear to be measuring the same dimension of family life, and due to the fact that they are all positively correlated, it seems reasonable to combine the twelve items into a single index to measure family role structUre. Once the response for each of the twelve questions has been recoded into the three point scale denoted as Scale 1A, the responses to all twelve questions can be combined to build a general index of family role structure for each family. A score representing the extent to which tasks and decisions are undertaken jointly is derived by calculating the simple unweighted average of the responses to all twelve items. This average will be referred to as the Family Role Structure Index, or FRSI. The possible values of FRSI range from 1.0 to 3.0 with a FRSI score of 1.0 meaning that the spouses share none of the twelve tasks and decisions. An FRSI 36 score of 1.0 is indicative of extreme family role segregation. An FRSI score of 3.0 on the other hand, means that the spouses com- pletely shared all twelve tasks and decisions, which is indicative of families with extreme family role integration. In general, the higher the FRSI score, the less role segregation there is in the family, or stated conversely, the higher the FRSI score the more role integration in the family. The Family Role Structure Index, FRSI, will be used as the principle measure of family role structure in this study. The reader might also note the first six questions in the list of family role structure items are six of the eight questions used by Blood and Wolfe (4:29) to build a measure of the division of labor in the family. Questions 7 through 12 in the family role structure items are six of the eight questions used by Blood and Wolfe (4:19) to construct a measure of decision making or the power structure of the family. The six questions on decision making and task performance used here are the only questions on these two topics that were used in the 1955 DAS and repeated in the 1971 DAS. In other words, two questions on decision making and two questions on division of labor which appeared on the questionnaire for the 1955 survey were eliminated from the questionnaire for the 1971 survey. It is clear that the first six items of the twelve family role structure items all pertain to the performance of household tasks while the last six items of the twelve items ask questions about decision making in the family. From this observation, it 37 seems logical to construct separate indices using the first six items for one index and the last six items for another index, as was done by Blood and Wolfe. However, it may be useful to quantitatively analyze the responses‘to each set of questions. Visual inspection of the inter-correlations among the items given in Tables Bl and B2 show that items 7 through 12 have re- latively higher inter-correlations with each other than with items 1 through 6. This suggests that items 7 through 12 may be measur- ing something not being measured by items 1 through 6. The appro- priateness of separating items 7 through 12 into a cluster or subgroup is shown more clearly by the results of a principal com- ponent oblique rotation factor analysis performed on the twelve items. The results of the factor analyses are given in Tables B3 and B4 in Appendix B. Both tables contain the results of the principal factor solution as well as the factor structure matrix after the oblique rotation. The oblique rotation factor structure matrix is best suited for determining the clustering of variables, for the factor loadings describe the relationship between each factor and each item. Those items which load high on one factor can be clustered into a more global index representing that factor. Tables 83 and B4 show that items 7 through 11 all load highest on the first factor extracted by the factor analysis. Item 12 loads highest on the second factor, but it also loads relatively high on the first factor. Further analysis showed that the correlation between an index composed of family structure items 7 through 11 and index composed of family role structure items 7 through 12 was 38 .95 using the 1955 data and .96 using the 1971 data. The extremely high correlation between these two indices indicates that one could expect to get nearly the same results using the six item index which is suggested by the analysis of the content of the six items as one would using the five item index suggested by the factor analysis. The evidence justifies the construction of a single index from the six items. The responses to the family role structure items 7 through 12 will be used to construct an index to reflect family decision making. The index will be the simple unweighted average of family role structure items 7 through 12 and will be referred to as the Family Decision Making Index, or FDMI. Scores on FDMI range from 1.0 to 3.0 with a score of 1.0 signifying no family role integra- tion and a score of 3.0 indicating complete family role integration. We now turn our attention to family role structure items 1 through 6 which were used by Blood and Wolfe to measure division of labor in the family. Examination of the face content of all six items reveals that they all pertain to the performance of household tasks, as one would expect. Tables B1 and B2 show that the inter-correlations among items 1 through 6 are nearly all positive, but generally they are lower than the inter-correlations between items 7 through 12. Examination of the factor analysis results in Tables B3 and 84 show that items 1 through 6 do not have consistently high loadings on any one single factor as did items 7 through 12. The case is further complicated by the fact that there is some inconsistency between the results using the 1955 data 39 and the results using the 1971 data in terms of which items load highest on which factors. The evidence regarding the unidimensionality of family role structure items 1 through 6 is somewhat contradictory. Analysis of the content of the six family role structure items suggests they are all measuring the same concept while results of the factor analysis suggests that there are several underlying dimensions running through the items. In light of this problem, the six items were used to construct several indices based on the results of the factor analysis. When these indices were correlated with the major dependent variable in this study, fertility, none of the indices derived from the factor analysis correlated more highly than an index based on all six items used collectively. Consequently, it was decided to use all six items to build a single measure of family role structure in the area of task performance. The index used to measure family role structure in the area of task performance, will consist of the simple unweighted average of family role structure items 1 through 6. This measure will be labeled the Task Differentiation Index, or TDI. TDI may take on values between 1.0 and 3.0 with a score of 1.0 indicating a segregated family role'structure and a score of 3.0 indicating an integrated family role structure. Before leaving this section, it may be informative to examine the relationships among the three indices of family role structure which have been created. Table B5 in Appendix B shows the correlations among the three measures of family role structure 40 for both 1955 and 1971 data. Perhaps the most striking observation is the stability of the correlations over time. The difference between the 1955 correlations and the corresponding l97l correla- tion for any given pair of variables is very small. If the true relationships among the three variables remained constant between 1955 and 1971, the evidence presented in Table 85 suggests that the three measures of family role structure used here are very reliable. The strength and direction of the correlations presented in Table 85 are as one would expect. There is a moderate amount of association between the two indices TDI and FDMI, and high correla- tions between the principle index, FRSI, and the two secondary measures TDI and FDMI. Since TDI and FDMI share a common dimension and each includes a unique dimension, one would expect to find some correlation between them due to their common dimension, but not a high correlation since they each measure a dimension not shared by the other index. TDI and FDMI should correlate highly with FRSI since FRSI is simply a composite of TDI and FDMI. Although it is hoped that the measurement of family role structure can be improved for future studies, the measures developed here will be adequate for testing the hypotheses in the present study. Since the present study was developed with the intention of employing the variable of family role structure as it has been used successfully in past studies, further efforts to develop more refined measures of family role structure go beyond the intent of the current study. 41 Hypotheses Two sets of hypotheses will be tested in this study. The first set of hypotheses deals with the nature of the relationship between family role structure and fertility. These hypotheses are actually a further and more detailed test of the relationship be- tween family role structure and fertility that has been observed in past research. In addition to examining the general association, both family role structure and fertility will be examined in more detail than has been the case in previous studies. The two dimen- sions of family role structure, denoted as TDI and FDMI, will be used separately, and actual fertility as well as desired fertility will be used as two distinct dimensions of fertility. Hypothisis l The more segregated the family role structure of a couple the greater the number of children the couple will desire and will have. This will be reflected in a negative correlation between FRSI scores and the number of children born, as well as, FRSI scores and the number of children desired. To further invetigate the relationship between family role structure and fertility hypotheses 1A and 1B, given below, will be tested. Hypothesis 1A Families that segregate the task performance in the house- hold will have higher desired and actual fertility than families where spouses share the task performance. This will be reflected in a negative correlation between TDI scores and both the actual number of children born to the family and the number of children desired. Hypothesis lB Families where the spouses segregate the decision making in the household will have higher actual and desired fertility 42 than families where household decision making is shared by the spouses. This will be reflected in a negative correla- tion between FDMI scores and both dimensions of fertility measured in this study. The relationship between family role struCture and fertility will be further analyzed by calculating partial correlations between family role structure variables and fertility variables while con- trolling several other variables known to be associated with differential fertility. Multiple correlation analysis will also be used to examine the relationship between the family role structure and fertility while controlling the collective effects of several extraneous variables. The variables controlled are the age at marriage of the wife, the total family income, education level of the wife, length of marriage, and the age of the wife. The second set of hypotheses examined in this study deals with changes in family role structure and the period fertility rate between the mid 1950's and the early 1970's. These hypotheses are based on the idea that some of the decline in period fertility be- tween the mid 1950's and the early 1970's can be attributed to changes in the family role structure of the typical American married couple. If hypotheses 1, 1A and 1B are correct then an integrated family role structure is associated with lower fertility and con- sequently a temporal shift in the average family role structure in the direction of more integrated family role structure would result in a lower aggregate fertility rate, since the proportion of families having an integrated family role structure would increase. This general idea is investigated in the second set of hypotheses labeled 2 and 2A. 43 Hypothesis 2 The decline in the period fertility rate between the mid 1950's and the early 1970's can be attributed, in part, to a change in the family role structure of married couples, from more segre- gated to less segregated role-relationshipsirithe marital dyad. This will be reflected in higher mean scores for the three family role structure variables in 1971 than in 1955. To further examine the relationship between aggregate changes in family role structure and fertility, hypothesis 2A will also be tested. Hypothesis 2A There will be a close association between the amount of change from 1955 to 1971 in family role structure and actual fertility in selected sub-groups of the population. This will be reflected in very similar rank orderings of the sub-groups based on the amount of change in family role structure, and the amount of change in the actual fertility of the sub-groups as well as a high correlation between the amount of increase in the mean FRSI scores and the amount of decrease in the mean actual fertility of the sub-groups. CHAPTER IV FINDINGS REGARDING THE RELATIONSHIP BETWEEN FAMILY ROLE STRUCTURE AND FERTILITY There are two groups of hypotheses being examined in this study. Hypotheses 1, 1A and 1B concern the basic relationship between fertility and family role structure. Hypotheses 2 and 2A deal with recent changes in family role structure and fertility in the United States. The first group of hypotheses, 1, 1A and 1B, will be thoroughly examined before giving consideration to the second set of hypotheses. Since hypotheses 1, 1A and 1B will be tested using simple cor- relations, a few comments about correlation coefficients are in order. There are essentially three important components of any correlation coefficient. Those components are (l) the direction of the associa- tion, that is, positive or negative, (2) the magnitude of the correla- tion, varying from 0.0 to 1.0 and (3) the significance of the correlation, which may take any value between 0.0 and 1.0. While the hypotheses tested here make no specifictwedictionconcerning the size of any of the correlations, the size is implicitly considered when looking at significance, since the significance of a correlation co- efficient is dependent on the strength of the coefficient and the number of observations used in calculating the coefficient. The reader is reminded that the significance tests used here are based 44 45 on a random sample model even though the sampling procedures did not meet the criteria of a simple random sample. I might add that the analysis presented here is based solely on a search for linear re- lationships. Although no specific analysis is used in testing for curvilinear relationships, preliminary inspection of the contingency tables did not reveal any evidence that suggest such relationships. ' Preliminary consideration of hypotheses 1, 1A and 1B will involve the use of simple correlation coefficients. Once the hypotheses have been tested using zero-order correlations, the correlations will be re-examined while controlling selected third variables known to be associated with differential fertility. The extraneous variables will be controlled separately using partial correlations and then the extraneous variables will be controlled collectively through the use of multiple correlation analysis. Examination of the zero-order correlations will be sufficient to test all the hypotheses of the first group. Partial and multiple correlation are used to assess the effect of selected extraneous variables on the family role structure-fertility relationship. Table Cl and C2 in Appendix C show that there were 416 qualified respondents questioned in 1955 and 255 qualified respon- dents questioned in 1971. Other tables in Appendix C also show that in 1955 approximately 60 of the 416 families did not supply appropriate data on task differentiation and about 60 families did not supply appropriate data on family decision making. Consequently there are 108 cases where an FRSI score could not be computed in the 1955 sample. 46 Of the 255 cases meeting the initial criteria for inclusion in the 1971 sample, 9 cases did not contain responses on task differentiation, and 11 cases did not contain responses on family decision making. Eliminating cases with missing data for actual fertility, task differentiation, or family decision making left 237 cases where an FRSI score was calculated. Finding§_Regarding the RelationshipyBetween Fertility and an Index of General Family Role Structure The first major hypothesis to be tested concerns the relationship between fertility and a general measure of family role structure. This hypothesis is examined by observing the degree of covariance between both measures of fertility and the Family Role Structure Index, FRSI, which represents a general measure of family role structure. Recall that FRSI was constructed so that the higher the FRSI score, the less the segregation in family roles. Hypothesis 1 states that the more segregated the family role structure of a couple, the greater the number of children born and the greater the number of children desired. Since an increase in FRSI represents a decrease in role segregation in a family, hypothesis 1 will be supported if significant negative correlations between the two fertility variables and FRSI scores are observed in the data. The term significant is used here in a statistical sense. Use of the term significant in this study can be interpreted as 47 meaning beyond the .05 level of significance. In other words, if the results are labeled significant, it means the observed results would occur less than one time in twenty by chance. Table 1, which is based on 1955 data, shows the correla- tions between the two measures of fertility and the family role structure variables. In each cell of the correlation matrix there are three items. The top figure is the zero-order correlation coefficient (r), the figure immediately below the correlation coefficient is the number of observations used in calculating the correlation coefficient. The bottom figure in each cell is the level of significance for the correlation coefficient. The level of significance can be interpreted as the likelihood of getting the observed correlation coefficient if there is actually no linear association between the two variables. Since the hypotheses predict the direction of the correlation, the signifiance level given here is one-tailed measure of significance. Table 1 shows the correlation between actual fertility and FRSI scores is -.21 which is significant at the .001 level of significance. This negative correlation indicates that couples who have high FRSI scores tend to have fewer children than couples who have low FRSI scores. In other words, the more segregated the family role structure, the more children the couple have. Table 1 also shows that the correlation between desired fertility and FRSI scores is -.15 which is significant at the .001 level of significance. This negative correlation means that couples who have low FRSI scores tend to desire more children than couples 48 TABLE l.--Zero-0rder Correlations Between Family Role Structure Variables and Fertility Variables for Married Women Age 21 to 45 in 1955. Family Role Structure Variables FRSI Scores TDI Scores FDMI Scores Actual Fertility r = 0.21 r = -.16 r = 0.12 N = 308 N = 353 N = 356 p = ..001 p = .001 p = .001 Desired Fertility r = -.15 r = -.08 r = -.13 N = 308 N = 353 N = 356 p = .001 p = .063 p = .007 r N coefficient = Pearson Product-Moment correlations coefficient = Number of cases used in calculating the correlations p = The level of significance of the correlation coefficient. TABLE 2.--Zero-0rder Correlations Between Family Role Structure Variables and Fertility Variables for Married Women Age 21 to 45 in 1971. Family Role Structure Variables FRSI Scores TDI Scores FDMI Scores Actual Fertility r = -.21 r = -.19 r = -.15 N = 237 N = 246 N = 244 p = .001 p = .001 p = .008 Desired Fertility r = -.02 r = —.00 r = -.02 N = 234 N = 234 N = 240 p = .388 p = .480 p = .390 r = Pearson Product-Moment correlation coefficient N = Number of cases used in calculating the correlation coefficient 1'." II The level of significance of the correlation coefficient. 49 who have high FRSI scores, which is to say that couples who share household tasks and decisions tend to desire fewer children than couples who do not share these tasks and decisions. Table 2, which is based on 1971 data, contains the correlations between the two fertility measures and the family role structure variables. The three figures in each cell of Table 2 correspond to the correlation coefficient, number of observations, and level of significance described in Table 1. The correlation between actual fertility and FRSI scores in 1971 is -.21 which is significant at the .001 level of signifi- cance. This indicates that families with high FRSI scores have relatively fewer children than families with low FRSI scores. Since a high FRSI score represents an integrated family role struc- ture, there is an association between sharing family functions and having a small family. Table 2 also shows that the correlation between number of children desired and FRSI scores in 1971 is -.02. The significance level associated with this correlation coefficient indicates that we could expect a coefficient of this magnitude approximately 40% of the time by chance alone. While this correlation is in the pre- dicted direction, it does not meet the criteria for concluding significance. As expected, there is a significant negative correlation between actual fertility and FRSI scores at both points in time. The correlation between desired fertility and FRSI scores calculated from the 1955 data was negative and significant, as 50 hypothesized, however, when the same correlation was calculated from the 1971 data, the correlation was not significant at the established level. In summary, hypothesis 1 is only partially supported. There is a significant negative correlation between actual fertility and FRSI scores, but the evidence pertaining to the relationship between desired fertility and FRSI scores is mixed. A significant negative correlation between desired fertility and FRSI scores was found in 1955, but the same correlation calculated from the 1971 data was weak and not significant. (Findings Regarding the Relationship_Between Fertility and Family Role'Structure in the Area of Task’Performance Hypothesis 1A examines the relationship between fertility and the extent to which household task performance is segregated by sex. The Task Differentiation Index, TDI, described earlier, measures the extent of family role segregation in task performance, and is constructed such that high scores on TDI indicate that spouses share task performance. A low TDI score indicates that the performance of tasks is segregated by family role. Hypothesis 1A states, families where the spouses segregate the performance of household tasks will have higher actual and desired fertility than families where the performance of household tasks is shared by the spouses. Hypothesis 1A will be supported if significant negative correlations between measures of fertility and TDI scores are observed in the data. 51 Table 1 shows that in 1955 the correlation between actual fertility and T01 scores was -.16, which is significant at the .001 level. This negative correlation indicates that couples who have high TDI scores generally tend to have lower fertility than couples who have low TDI scores. The correlation between desired fertility and T01 scores, calaculated from the 1955 data, is -.08, which is significant at the .063 level. Although the correlation is in the predicted direction, it does not meet the criteria established for conclusion of significance. In 1971 the correlation between actual fertility and T01 scores was -.l9, which is significant at the .001 level of signi- ficance. Table 2 also shows the correlation between desired fer- tility and TDI scores, calculated from the 1971 data, is .00. This correlation coefficient is obviously not significant. In summary, Hypothesis 1A is only partially supported. There is a significant negative correlation between actual fer- tility and the extent to which the spouses share the performance of household tasks, however, the nature of the association between desired fertility and family role segregation in the performance of household tasks is unclear. No significant correlation was found in 1955 or 1971, although the correlation coefficient calcu- 1ated from the 1955 data was negative, as hypothesized. 52 Findings Regarding the Relationship Between Fertility_andeamilyyRole Structure In the Area of DecisiofirMaking The third hypothesis regarding the relationship between fertility and family role structure is labeled hypothesis 1B. Treatment of this hypothesis will involve analysis of the relation- ship between the extent to which couples share family decision making, and the childbearing of the couple. Recall that decision making patterns are measured by the Family Decision Making Index, labeled FDMI. The higher the FDMI score of a family, the less role segregation in the family decision making. Hypothesis 1B states, families that segregate decision making in the family will have higher desired and actual fertility than families that share family decision making. Hypothesis 18 will be supported if significant negative correlations between fertility measures and FDMI scores are observed in the data. Table 1 shows that the correlation between actual fertility and FDMI scores calculated from the 1955 data is -.12, which is significant at the .011 level of significance. This indicates that the more role segregation there is in family decision making, the higher the actual fertility of the family. The correlation between desired fertility and FDMI scores in 1955, also taken from Table 1, is -.13, which is significant at the .007 level. Indicating the more segregation there is in the family decision making, the higher the desired fertility. Table 2 shows the corresponding correlation coefficients derived from the 1971 data. The correlation between actual fertility 53 and FDMI scores is -.15, which is significant at the .008 level of significance. The correlation between desired fertility and FDMI scores is -.02, which is significant at the .390 level of significance. In summary, hypothesis 1B is only partially supported. There is a significant negative correlation between actual fertility and FDMI scores, but the existence of a correlation between desired fertility and FDMI scores has not been demonstrated. In 1955 there was a significant negative correlation between desired fer- tility and FDMI scores, but in 1971 the significance level of the same correlation did not meet the criteria established for con- cluding significance. It should be noted at this point that actual fertility is more closely associated with the family role structure variables than desired fertility. All correlations between actual fertility and the family role structure variables are negative and signifi- cant, while four of the six correlations between desired fertility and the family role structure variables are not significant. In searching for an explanation of why the correlations between actual fertility and family role structure variables are stronger than the correlations between desired fertility and family role structure variables, it is useful to look at the correlation between actual fertility and desired fertility. In 1955 the correlation between these two measures of fertility was .20. The same correlation was .29 in 1971. 54 The disparity between various measures of fertility has been noted in other research. Ryder and Westoff (26:281), for example, found that women in their sample had a mean current parity of 2.76 and a mean desired parity of 3.29. Since desired fertility generally refers to future fertility intentions while actual fer- tility generally refers to past fertility experience, it is not altogether surprising that these two measures of fertility are not highly correlated. Additionally, it is not surprising that some third variable may be correlated with one of these measures of fertility and not the other. The reader is also reminded of the particular questions employed to measure desired fertility for the two samples used in this study. Recall that desired fertility in 1971 was measured by asking the respondents about the appr0priate fertility level for a hypothetical family similar to the respondents family on one dimension, standard of living. The question used to measure de- sired fertility in the 1955 sample asks specifically about the future fertility of the respondent's family. Also note that two out of three correlations involving desired fertility and family role structure variables calculated from the 1955 data were significant, while all three correlations involving desired fer- tility and the family role structure variables calculated from the 1971 data were not significant. Given the fact that the measure- ment of family role structure for each family is based on inter-- action within that family, rather than a hypothetical one, one would expect these measurescrffamily role structure to correlate 55 more highly with measures of desired fertility regarding the re- spondent's family, the 1955 measure, than they do with measures of desired fertility regarding a hypothetical family, the 1971 measure. In retrospect, it appears that the desired fertility of the 1971 respondents was not measured adequately. This line of reasoning is particularly useful in explaining why there was no significant correlation between desired fertility and the family role structure variables in 1971, but two out of three correlations in 1955 were significant. Tables 1 and 2 also show that the correlation between actual fertility and FRSI scores are higher than the correlations between actual fertility and T01 scores which in turn are higher than the correlations between actual fertility and FDMI scores. The fact that the rank order and the relative magnitude of the correlations are stable for both time periods examined is striking, however, statistical tests described by Blalock (3:405) show that the three correlations in each table are not statistically dif- ferent from one another at any critical level of significance. Therefore, any attempt to provide a detailed theoretical explana- tion for the observed rank order of the correlations is likely to be futile since there is a strong probability that the observed rank orders are due solely to random chance. The examination of hypotheses 1, 1A and 1B, has shown that there is a significant negative correlation between actual fertility and three variables representing family role structure. The relationship between the three measures of family role structure 56 and actual fertility will be investigated further by examining partial correlation coefficients and standardized regression co- efficients. The partial correlation coefficients will reveal the effect of each extraneous variable on the basic relationship be- tween actual fertility and family role structure. Examination of the standardized regression coefficients associated with each family role structure variable will show the collective effect of all selected extraneous variables on the family role structure- fertility relationship. While partial and multiple correlation could be used to develop a complex causal path model involving family role structure, fertility and the five extraneous variables described below, the development of such a path model is not the intent of this study. The sole purpose for employing the partial and multiple correlation techniques will be to determine if the extraneous variables taken separately, or all five used collectively alter the family role structure-fertility relationship observed in the zero-order situa- tion. Using this perspective, we will be comparing each partial correlation and each standardized regression coefficient associated with a family role structure variable to the corresponding zero- order correlation coefficient. Only those situations which produce coefficients which are not negative and significant will be dis- cussed. Since the relationships between desired fertility and family role structure variables are unclear, partial and multiple correla- tions involving desired fertility are not presented or discussed here. 57 Two of the primary dimensions of socio-economic status, income and education, are among the variables controlled in this study. By using income and education as distinct dimensions of the general concept of socio-economic status, we hope to gain better insight into the effect of socio-economic status on the relationship between family role structure and actual fertility. Dissaggregation of this type has been informative in other studies. Three other variables, which can all be described as demographic variables, will also be controlled. These variables are, the age at marriage of the wife, the length of marriage, and the age of the wife at the time of the interview. It is reasonable to expect, and other research has shown, that these variables are associated with differential fertility. These three variables along with the two variables representing socio- economic status will be referred to as the extraneous variables. Table 3 shows that all partial correlations calculated from the 1955 data remain negative and significant while control- ling for each of the five extraneous variables. This indicates that none of the five extraneous variables has any substantial effect on the family role structure-fertility relationship observed in the zero-order situation. Table 4 shows that the partial correlations while control- ling for education of the wife, family income, age at marriage, and age of the wife, calculated from the 1971 data are negative and significant. None of the partial correlations between actual fertility and the family role structure variables while controlling 58 TABLE 3.--Partia1 Correlations Between Actual Fertility and Family Role Structure Variables while Controlling for Several Extraneous Variables for Married Women Age 20 to 45 in 1955. Variable Being Family Role Structure Variable C°"t'°"9d FRSI Scores TDI Scores FDMI Scores Education of the Wife r = -.19 r = -.19 r = -.12 (N = 307) p = .001 p = .001 p = .016 Family Income r = -.22 r = -.22 r = -.14 (N = 297). p = .001 p = .001 p = .008 Age at Marriage r = -.19 r = -.19 r = -.12 (N = 307) p = .001 p = .001 p = .016 Length of Marriage r = -.19 r = -.18 r = -.12 (N = 307) p = .001 p = .001 p = .015 Age of Wife r = -.20 r = -.20 r = -.14 (N = 307) p = .001 p = .001 p = .009 r = Pearson Product-Moment Correlation Coefficient p = The level of significance of the correlation coefficient. Note - A list-wise deletion of cases with missing data was used for each set of partial correlations calculated. This means that all partial correlations in a given set are calculated on exactly the same group of observations. The number of cases included in each set of partial correlations follows the specification of the control variable in the table. 59 for length of marriage are significant at the .05 level. This indicates that the length of marriage does have an effect on the relationship between family role structure and actual fertility, for the couples in the 1971 Sample. The reason the 1971 partial correlations between family role structure variables and actual fertility while controlling for length of marriage are not significant rests with the strength of the correlation between length of marriage and actual fertility. TableS'hiAppendix 0 show that the correlation between length of marriage and actual fertility is .52 in 1971, but only .29 in 1955. Also note that the correlations between length of marriage and the family role structure variables are stronger in the 1971 data, and that all these correlations are negative. The numerator of a partial correlation is equal to the zero-order correlation between the two initial variables minus the cross-product of the correlation between the control variable and one initial variable times the correlation between the control variable and the other initial variable. When the cross-product is large and has the same sign as the zero-order correlation, the partial correlation will be substantially smaller than the zero- order correlation. Of course, there is an adjustment factor in the denominator of the partial correlation which may also affect the final value of a given partial correlation coefficient, but this has little impact in the present situation. In the present discussion, actualgfertility and family role structure variables are the intial variables and length of marriage 60 TABLE 4.--Partia1 Correlations Between Actual Fertility and Family Role Structure Variables while Controlling for Several Extraneous Variables for Married Women Age 21 to 45 in 1971. Controlled Family Role Structure Variable var'ab'e FRSI Scores TDI Scores FDMI stores Education of the Wife r = -.18 r = -.19 r = -.12 (N = 233) p = .003 p = .002 p = .035 Family Income r = -.20 r = -.19 r = —.15 (N = 220) p = .001 p = .002 p = .014 Age at Marriage r = -.18 r = -.17 r = -.13 (N = 233) p = .003 p = .006 p = .024 Length of Marriage r = -.10 r = -.08 r = -.09 (N = 233) p = .059 p = .124 p = .093 Age of Wife r = -.16 r = -.14 r'= -.12 (N = 233) p = .008 p = .017 p = .034 r = Pearson Product-Moment Correlation Coefficient. p = The level of significance of the correlation coefficient. Note 4 A list-wise deletion of cases with missing data was used for each set of partial correlations calculated. This means that all partial correlations in a given set are calculated on exactly the same group of observations. The number of cases included in each set of partial correlations follows the specification of the control variable in the table. 61 is the control variable. Since the correlations between length of marriage and actual fertility as well as the correlations between length of marriage and the family role structure variables are higher in 1971 than in 1955, their cross-products will also be larger. Also notice that these cross-products will have the same sign as the zero-order correlation between family role structure variables and actual fertility. In 1971 these cross-products were sufficiently large to reduce the partial correlations to a size which is not significantly different from zero at the .05 level of significance. While one would normally expect to find a strong positive correlation between actual fertility and the length of marriage, there is a good reason why this was not observed in the 1955 data. Women who were in the older age cohorts in 1955, particularly age 35 to 44, were women who experienced their prime childbearing years during the decade of the 1930's, a decade of low period fertility. Much of the childbearing postponed by these women during the 1930's and early 1940's was never made up. Decisions to delay childbearing became decisions to forego additional childbearing. Consequently, by 1955 older women, who are presumably women who have been married longer, had only slightly higher actual fertility than younger women, who had been experiencing their prime childbearing years during the post war "baby boom." This idea is supported by the fact that the correlation between age of wife and actual fertility was only .08 in 1955. In contrast, the correlation between age of wife and actual fertility was .37 in 1971. The topic of cohort fertility will be discussed further in the next chapter. 62 As stated earlier, multiple correlation will be used in this study to examine the relationship between a given independent variable and the dependent variable while all other independent variables are controlled. A standardized regression coefficient for a given independent variable can be interpreted as a partial correlation between the independent variable and the dependent variable while controlling the effects of all other independent variables in the equation. In the multiple correlation analysis used here, we will focus on the standardized regression coefficient for each family role structure variable. For each time period, 1955 and 1971, three multiple correlation analyses were conducted. Problems of multi-collinear- ity precluded the use of all three family role structure variables as independent variables in the same equation. Another problem precluded the use of the three variables, age at marriage, length of marriage and age of wife as independent variables in the same equation. If one knows any two of these three variables, the third can be derived. Therefore, the age of wife was eliminated as an independent variable in the multiple correlation analyses. In each multiple correlation analysis one family role structure variable was added to the four remaining extraneous variables to create a set of independent variables while actual fertility re- mained as the dependent variable for all the analyses. The standardized regression coefficient, the unstandardized regression coefficient, the standard error of the unstandardized regression coefficient and the significance of each regression coefficient for each of the six analyses are given in Tables 5 to 10. 63 TABLE 5.--Multip1e Correlation Analysis of Actual Fertility as a Dependent Variable with FRSI and Four Other Independent Variables for Married Women Age 21 to 45 in 1955. Standardized Unstandardized Standard Significance Regression Regression Error of of the Coefficient Coefficient B Regression Variable (B) Coefficient FRSI -.172 -.O76 .024 .002 Family Income -.134 -.093 .040 .021 Age at Marriage of the Wife -.225 -.O93 .023 .000 Education of Wife -.019 -.023 .071 .749 Length of Marriage .189 .048 .015 .002 Constant 5.488 .743 .000 R2 = .177 N = 298 64 TABLE 6.--Multip1e Correlation Analysis of Actual Fertility as a Dependent Variable with TDI and Four Other Independent Variables for Married Women Age 21 to 45 in 1955. Standardized Unstandardized Standard Significance Regression Regression Error of of the Independent Coefficient Coefficient B Regression Variable (B) Coefficient TDI -.182 .073 .022 .001 Family Income -.157 .108 .040 .008 Age at Marriage of the Wife -.225 .094 .023 .000 Education of Wife -.014 .018 .071 .806 Length of Marriage .192 .048 .015 .001 Constant .401 .723 .000 R2 = .180 N = 298 65 TABLE 7.--Multiple Correlation Analysis of Actual Fertility as a Dependent Variable with FDMI and Four Other Independent Variables for Married Women Age 21 to 45 in 1955. Standardized Unstandardized Standard Significance Regression Regression Error of of the Independent Coefficient Coefficient 8 Regression Variable (B) Coefficient FDMI -.103 .033 .017 .057 Family Income -.119 .082 .040 .042 Age at Marriage of the Wife -.233 .097 .024 .000 Education of Wife -.O31 .037 .072 .599 Length of Marriage .192 .049 .015 .002 Constant .889 .715 .000 R2 = .159 N = 298 66 TABLE 8.--Multiple Correlation Analysis of Actual Fertility as a Dependent Variable with FRSI and Four Other Independent Variables for Married Women Age 21 to 45 in 1971. Standardized Unstandardized Standard Significance Regression Regression Error of of the Independent Coefficient Coefficient 8 Regression Variable (B) Coefficient FRSI -.053 .029 .031 .356 Family Income .020 .016 .052 .758 Age at Marriage of the Wife -.l69 .094 .034 .006 Education of the Wife -.146 .141 .061 .022 Length of Marriage .460 .127 .017 .000 Constant .020 1.019 .000 .344 221 67 TABLE 9.--Multiple Correlation Analysis of Actual Fertility as a Dependent Variable with TDI and Four Other Independent Variables for Married Women Age 21 to 45 in 1971. Standardized Unstandardized Standard Significance Regression Regression Error of of the Independent Coefficient Coefficient B Regression Variable (B) Coefficient TDI -.035 -.020 .032 .544 Family Income .019 .015 .052 .776 Age at Marriage of the Wife -.l69 -.095 .034 .006 Education of the Wife -.l49 -.l43 .061 .020 Length of Marriage .464 .128 .017 .000 Constant 4.878 1.026 .000 R2 = .342 N = 221 68 TABLE lO.--Multip1e Correlation Analysis of Actual Fertility as a Dependent Variable with FDMI and Four Other Independent Variables for Married Women Age 21 to 45 in 1971. Standardized Unstandardized Standard Significance Regression Regression Error of of the Independent Coefficient Coefficient B Regression Variable (B) Coefficient FDMI -.047 -.017 .021 .405 Family Income .023 .019 .051 .716 Age at Marriage of the Wife -.172 -.096 .034 .005 Education of the Wife -.l47 -.l42 .061 .022 Length of Marriage .465 .129 .017 .000 Constant 4.843 .955 .000 R = .344 N = 221 69 Tables 5, 6 and 7, which are based on 1955 data, show that the standardized regression coefficients for FRSI, TDI and FDMI are all negative and significant. This indicates that even when the collective effects of all five extraneous variables are controlled, the relationships between family role structure variables and actual fertility remain similar to the relationship observed in the zero-order situation. Tables 8, 9 and 10, which are based on 1971 data, show that the standardized regression coefficient for each family role structure variable is negative, but none are significant at the .05 level of significance. This indicates that collectively con- trolling the five extraneous variables does alter the statistical relationship between the family role structure variables and actual fertility. The reader may also note that the regression coefficient for family income was significant in 1955 but not in 1971. This finding is consistent with the notion of a diffusion model of contra- ceptive methods which holds that knowledge about and use of effec- tive modern contraceptive technique is initially more accessible to the wealthy but over time becomes accessible to people of all income levels. In 1955 when access to modern contraception was limited to the wealthy, there was a relationship between family income and fertility. By 1971, partly through government programs contraceptive methods were nearly as accessible to poor people as to wealthy people and therefore the relationship between family income and actual fertility was diminished. 70 The regression coefficient for education of the wife was not significant in 1955, but was significant in 1971. This situa- tion may be explained by looking at the relationship between opportunities to work outside the home, education level, and actual fertility. In 1971 there were more attractive non-familial roles and activity available to women than there was in 1955. However, the availability of many of these non-familial roles and activities was related to education level. The more education a woman had, the more attractive alternatives she had to childbearing and childrearing activities. In 1955 there were few acceptable alternatives to the mother-wife role regardless of education. Since these non-familial roles competed with childbearing activity, the group of women who participated most in the non-familial roles are likely to have lower fertility. Since education of the wife is related to opportunities outside the home in 1971, it is also related to actual fertility. The regression coefficient for both age at marriage and length of marriage were significant in 1955 and 1971. These findings support a widely held belief that these two variables are major factors in differential fertility. The earlier couples get married, the longer the period they spend at risk of pregenancy, consequently the more children they are likely to have. Obviously length of marriage is a very direct measure of the amount of time spent at risk of pregnancy. While it is easy to see the direct effect of these variables on fertility, the reader is reminded that age at marriage and length of marriage are often related to 71 many social structural variables and therefore may act as an in- direct link between social structure and fertility. The reason the regression coefficient for family role structure variables were not significant in 1971 can be traced to the zero order correlations. Tables in Appendix 0 show that the correlation of marriage and actual fertility was .52 in 1971 but only .29 in 1955. Appendix 0 also shows that the correlation between length of marriage and the family role structure variables was much higher in 1971 than in 1955. The higher the correlation in 1971 acted to diminish the size of the regression coefficient for the family role structure variables to a size not significant at the .05 level. The relationship between length of marriage and actual fertility is somewhat accentuated in 1971 due to the fact that older women who had been married longer, had experienced their prime childbearing years during the post World War II baby boom, while the younger women had experienced their prime child- bearing years during the late 1960's and earlier 1970's when the period fertility level was low. Consequently, older women had much higher fertility than younger women. The relationship between family role structure and fertility observed in 1971 is partly due to the fact that there is a strong relationship between family role structure and length of marriage as well as a strong relationship between length of marriage and fertility. In 1971 when length of marriage was closely associated with fertility, variations of family role structure added little to the explanation of fertility levels. However, in 1955 when 72 length of marriage was less closely related to fertility, family role structure differences did help explain fertility differences. In 1971 those women who had been married the longest also had the most segregated family role structure and also had the highest fertility. However, in this three variable relationship, it appears that length of marriage is closely related to fertility and that the relationship between family role structure and fer- tility is somewhat dependent on the relationship between family role structure and length of marriage. The inter-relationship among these three variables was not as strong in 1955. _ This concludes the analysis generated by hypotheses 1, 1A and 18, involving the basic relationship between family role structure and fertility. It has been shown that there is a significant negative correlation between the actual fertility of a couple and the extent to which the couple share household task performance and decision making. This relationship will be em- ployed in the next chapter which examines recent changes in family role structure and actual fertility in the U.S. CHAPTER V FINDINGS RELATED TO RECENT CHANGES IN FERTILITY AND FAMILY ROLE STRUCTURE In the previous chapter, it was shown that there is a signficant association between actual fertility and family role structure, namely, that the more spouses share family functions, the fewer children the couple have. In this section, recent period changes in fertility and family role structure are examined. Essentially, hypotheses 2 predicts that the recent decline in period fertility rates can be attributed to changes in the family role structure of the typical American family. A more general underlying assumption is that changes in the family role structure of a group are closely associated with, and causally linked to, changes in aggregate level fertility of the group. Hypothesis 2 will be investigated by noting the direction and significance of recent shifts in measures of family role structure and fertility. Testing hypothesis 2A will involve an analysis of the extent of concommitant variation in recent changes in fertility and family role structure for several selected sub- groups of the population. One of the problems in examining hypothesis 2 and 2A con- cerns the relationship between period fertility rates and the 73 74 TABLE ll.--Mean Actual Fertility by Five Year Age Groups for 1955 and 1971. Mean Actual Fertility Mean Actual Fertility Age G'°”p in 1955 in 1971 21 - 24 1.69 (N = 65) 1.24 (N = 38) 25 - 29 2.11 (N = 88) 1.93 (N = 53) 30 - 34 2.52 (N = 104) 3.00 (N = 55) 35 - 39 2.00 (N = 100) 3.35 (N = 49) 40 - 44 2.19 (N = 59) 3.18 (N = 60) TABLE 12.--Mean of FRSI Scores by Five Year Age Groups for 1955 and 1971. Mean Actual Fertility Mean Actual Fertility A99.G'°“P in 1955 in 1971 21 - 24 1.69 1.78 25 - 29 1.62 1.65 30 - 34 1.69 1.61 35 - 39 1.63 1.55 40 - 44 1.61 1.60 75 measure of fertility used in this study. The aggregate period fertility changes presented earlier in the paper are based on the fertility that occurs in a single year, while the variable of actual fertility, themeasurement of fertility for the respondents in this study, incorporates all the fertility experienced by the respondent during her entire marriage. This issue will be clari- fied by examining some age-specific measures of fertility. Table 11 shows the mean actual fertility by five year age cohorts for the 1955 and 1971 samples. Note that some of the older cohorts in 1955, age 35-39 and 40-44, have lower mean actual fertility than some younger cohorts, such as, age 25-29 and 30-34. This can be explained by the fact that the older age cohorts, age 35-44, experienced their prime childbearing years during the 1930's, a decade of low fertility, while the younger cohorts experienced their prime childbearing years in the decade following World War II, a period labeled the "baby boom" due to high period fertility rates. The major point here is that there is little variation in the mean actual fertility of the five age corhorts in 1955 and no discernable pattern. In contrast, the mean actual fertility of the age cohorts in 1971 increases systemtically as one moves up the age scale from 21-24 to 35-39. The mean actual fertility of the 40-44 age cohort is slightly lower than the 35-39 age cohort, but higher than all other cohorts. This pattern is explained by the fact that those women in the older cohorts in 1971 experienced their prime child- bearing years during the post war "baby boom" era, while the women 76 in the younger age cohorts experienced their prime childbearing years during the late 1960's and early 1970's when period fertility does appear to be related to the age of the cohort, that is, the older the cohort, the higher the actual fertility. Although the 1971 period fertility rate is lower than the 1955 period fertility rate, comparison of the 1971 age-specific figures with the 1955 age-specific figures show it is only in the two youngest age cohorts, age 21-24 and 25-29, that the mean actual fertility is lower in 1971. The difference between the changes in period fertility rates and the age-specific figures is reconciled by the fact that most childbearing in any given year occurs to women in their twenties. Therefore, even though three out of the five age cohorts have higher mean actual fertility in 1971 than they did in 1955, the cohorts that were instrumental in producing the period fertility rates had lower mean actual fertility in 1971 than they did in 1955. Since this chapter is based on recent changes in period fertility, the women most closely responsible for those changes should be employed in the analysis. The data presented here show that if the women age 25-29 in 1955 and 1971 have childbearing patterns similar to older cohorts, they will have completed most of their total childbearing by age 29. For example, in 1955 the women age 25 to 29 already have had an average of 2.11 births while the women age 40 to 44, who have nearly completed their childbearing, only have an average of 2.19 births per woman. In 1971 women age 25 to 29 have averaged 1.92 births compared to the 77 3.18 births per woman experienced by the women age 40 to 44. Other research (30) has also shown that it is married women in their twenties who have experienced the majority of all childbearing for any year in the past 25 years. Therefore analysis in this section will involve only married women age 21 to 29. Before moving on to a test of hypothesis 2, it will be informative to examine age-specific family role structure measures in comparison to age-specific fertility measures. Table 12 shows the mean of FRSI scores for five age cohorts in 1955 and 1971. Note that the age-specific FRSI figures for 1955 show little variation among the age groups, much like the age-specific mean actual fer- tility of the same age groups. In contrast, the mean of FRSI scores for the age cohorts in 1971 decrease systematically as one moves up the age scale from 21-24 to 35-39. The mean FRSI score for the 40-44 age cohort is slightly higher than the mean FRSI score for the 35-39 age cohort. From this description of age-specific mean actual fertility figures and mean FRSI scores, a striking pattern emerges. In 1955 where there is little distinction among the mean actual fer- _ tility figures of the age cohorts, there is little distinction among the mean of FRSI scores of the age cohorts. In 1971, where the mean actual fertility showed a marked progression over the age cohorts, the mean of FRSI scores show a similar pattern with changes occurring in the opposite direction of the changes in mean actual fertility. Additionally, in each age cohort where fertility figures were lower in 1971 than 1955, the mean of FRSI 78 scores was higher, and in each age cohort where the fertility figure was higher in 1971 than in 1955, the mean FRSI score was lower in 1971 than in 1955. While these observations do not directly test any of the hypotheses of the study, they are noted as one more piece of evidence suggesting an association between family role structure and fertility. Findings Regarding Recent Changes in Family Role Structure and Fertility For the Total Population The first hypothesis to be tested in this section in- volves aggregate changes in fertility and family role structure between 1955 and 1971. Hypothesis 2 states, the decline in the period fertility rate between the mid-1950's and the early 1970's can be attributed, in part, to changes in the family role struc- ture of married couples, from more segregated to less segregated role relationships in the marital dyad. Testing this hypothesis will involve two steps. First it will be necessary to establish the fact that the young married women in the 1955 sample have a higher level of fertility than the young married women in the 1971 sample. This merely insures that the samples are representative of their populations on this important characteristic. The second step will focus on the family role structure variables measured in 1955 and 1971, and an examination of changes in these variables over the period. Since we are looking for aggregate changes, we will be looking for changes in a measure 79 TABLE 13.--Mean, Standard Deviation and Standard Error of the Mean of Actual Fertility for Married Women Age 21 to 29 in 1955 and 1971. 1955 Sample 1971 Sample Mean = 1.935 . Mean = 1.637 Standard Deviation = 1.255 Standard Deviation = 1.234 Standard Error = .101 Standard Error = .129 N = 153 I N = 91 80 of central tendency, the mean. Hypothesis 2 will be accepted if the mean of any of the family role structure measures, FRSI, TDI or FDMI, shows a significant increase between 1955 and 1971. This will indicate that the family role structure was less segregated in 1971 than it was in 1955. The data regarding actual fertility of young married Detroit women in 1955 is presented in Table 13. Table 13 includes several measures of fertility such as the mean and standard devia- tion of the distribution, and the standard error of the mean. Since the 1955 data is only a sample of the relevant population, a certain amount of sampling error can be expected. Due to sampling error, it is often preferable to establish a confidence interval around a sample mean to estimate the mean of the popula- tion, rather than using the sample mean as a point estimate. The 95% confidence interval for estimating the mean of the actual fertility of young married women in Detroit in 1955, runs from approximately two standard errors below the sample mean to two standard errors above the sample mean. In this situation, the 95% confidence interval for the mean actual fertility in 1955 is 1.734 to 2.135. Table 13 also shows fertility data from the 1971 sample. Table 13 shows several measures of fertility including the mean and standard deviation of the distribution, and the standard error of the mean. Since the 1971 data is also a sample from a larger population, it will be useful to establish an interval estimate of the mean actual fertility of the relevant population. The 95% 81 confidence interval for estimating the mean actual fertility of young married women in Detroit in 1971 is 1.380 to 1.894. It is worth noting that the mean actual fertility of the 1971 sample lies outside the 95% confidence interval established for 1955, and the mean actual fertility of the 1955 sample lies outside the 95% con- fidence interval established for 1971. A T-test of the difference between the mean actual fer- tility for 1955 and the mean actual fertility for 1971 shows the difference to be significant at the .03 level of significance. The level of significance is based on a T-statistic (If 1.81 and over 120 degrees of freedom. The fact that the means of the actual fertility of these two samples are significantly different, simply indicates that the samples are representative of their respective populations. We will now see if these two samples, and their populations, have patterns of family role structure that are significantly different from one another. Table 14 shows the means, standard deviations, and standard error of the mean of family role structure variables in 1955 and 1971. Column 3 of Table 14 shows the difference between the means of the variables for the 1955 sample and the 1971 sample. Column 4 contains the T-statistic for the difference shown in Column 3, and the significance of each T-statistic is given in Column 5. The figures in Column 5 can be interpreted as the probability of getting the observed difference if, in fact, the two samples are random samples from the same population. The T-statistic for the difference 82 .mmgoum Hoe mg» cw mmcmgu com cmumpaupmu no: mew: mucmuwewcmwm any age upumvuaum-h on» .cowuumgmu emporvmga use e? no: we: mucusu an» mucwm .1 Fm - 2 mm, - z .Fmo. u .m.m emeo. - .u.m mmme. - .o.m mpme. - .o.m Fmo. mem.p ewe. mmm.. - gem: _.N._ - cam: meeoem “zed .8 u 2 mm. - z ammo. - .m.m mono. - .m.m 2%. - 5:5. .23. n A; 4-- 4-- Poo.- oem.F - cam: som.P - ewe: meeeem Hoe _a u 2 mm, - z ammo. - .m.m ero. - .m.m comm. - .o.m 88mm. - .o.m mop. m-.P mmo. mo~.P - cam: Nee._ - cam: moeeem Hmma eeeaeaeeeo Ammm_..Feapv e_naeee> we u—umwuuum meow: we weapoagum eeeaeeepemem » eueeeaeeea ePaEem Fem, m_asem mmmp «Pom »FPEeL Amy Aev Amv , ANV APV .pnmp can mmmp cm mm on pm mm< :wEo: worsen: Low mmpnmwgm> mgsuuzspm «Pom aPVEmm so» cam: ecu mo Logs“ ugmucmpm use .copumw>oo ugmucuum .cmmz-.ep mgmFs meaeemFeeem .e o.F NF. mF. mmszo mmsmmz .m o.F oF. pmmFxmmgm mmxmz .N o.F mchaosm zgwuoco .F NF FF 2 m m F o m c m m F .mmmF :F meuFN om< cmsoz vmwggmz Loy mswaF mFmom mo comumFmgLougmucFuu.Fm m4mFS memeemFeepm .e o.F mm. FF. mmano magma: .m o.F no. ummmxmmgm mmxmz .N o.F mchaocm xgmuogw .F FF oF m _. m . V N . , o . , m e m N F .anF :F mquN wm< case: umFsgmz Low msqu meum mo :oFumFmssooLmucFut.Nm m4m Lowe: to xFeFez eoFFeFeeeoeeeeeF-.Fa msm 2000: 20 00200: 00200F0220020000-.~0 000