Lifetime employment patterns of married women


Material Information

Lifetime employment patterns of married women
Physical Description:
xiii, 194 leaves : ; 28 cm.
Van Velsor, Ellen
Publication Date:


Subjects / Keywords:
Married women -- Employment   ( lcsh )
Women -- Employment   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1980.
Includes bibliographical references (leaves 182-193).
Statement of Responsibility:
by Ellen Van Velsor.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000100029
notis - AAL5489
oclc - 07261311
System ID:

This item is only available as the following downloads:

Full Text







To Jack


This research has benefited from the suggestions

and assistance of many.

First of all, I would like to thank my chairman,

Dr. John Henretta, for all the time and effort he spent

assisting me over the past three years. His methodological

assistance and his intellectual concern and enthusiasm

for the project were invaluable to me. I also appreciate

his aid and encouragement in obtaining outside support for

this dissertation. My graduate education has been enhanced

immensely by our association.

This research was supported by National Science

Foundation Dissertation Grant SOC-7913975. Without this

support and the support of the Department of Sociology and

the Center for Gerontological Studies, this research would

not have been possible. The analyses used in this presenta-

tion were performed using the facilities of the Northeast

Regional Data Center and the Center for Instructional and

Research Computing Activities.

I would like to express special appreciation to Dr.

Cynthia Rexroat, who worked diligently with me even before

she was a committee member. This dissertation has been

much improved by her prompt and conscientious reading and


editing of earlier drafts. She is also a wonderful friend

whom I will miss.

Dr. Leonard Beeghley is another friend and advisor

to whom I owe a great debt of gratitude. His support and

encouragement have extended throughout my graduate career.

Above all, his enthusiasm for my work has given me confi-

dence in myself.

I also would like to thank my other committee members,

Drs. Felix Berardo, Benjamin Gorman and Henry Fishkind, for

their support and their comments on earlier drafts.

These remarks would be incomplete without special

thanks to Dr. Angela O'Rand. Although she was a member

of my committee for only a short time, her input to this

project was invaluable. She aided me in conceptualizing

the research and has been a continuing source of encourage-

ment and support.

Finally, there is an additional group of friends

whose unfailing emotional support has made possible the

completion of this work. Among these are Mary Ann Hilker,

Hugh Potter, and Dr. Jose DelaTorre.

To Jack, my husband, I owe the greatest debt of

gratitude. Although he has suffered much throughout my

graduate education, his love and support have been unceasing.



ACKNOWLEDGEMENTS......................... ......... iii

LIST OF TABLES....................................viii

LIST OF FIGURES................................... ix

ABSTRACT.......................................... x


I INTRODUCTION.............................. 1

Measuring Wives' Employment.......... 1
Plan of Study ......................... 4
Note to Chapter I................... 8

II LITERATURE REVIEW......................... 9

Introduction.......................... 9
Review of the Literature............. 10
Family Characteristics and
Wives' Labor Supply............. 10
Fertility and Labor Force
Participation.................... 16
Rural/Urban Differences and
Labor Market Variables.......... 28
Relationships Among Other
Factors.......................... 30
Family Life Cycle ............... 35
Discontinuity of Female Labor
Force Participation............. 40
Patterns of Employment.......... 41
Summary ......................... 46
Research Objective ................... 48
Notes to Chapter II.................. 49

III EMPLOYMENT PATTERNS....................... 50

Introduction.......................... 50
Patterns of Employment................ 51
Employment Pattern Contrasts......... 54
Never-Employed versus
Ever-Employed Wives............. 54
Continuous versus
Discontinuous Workers........... 56
Employment Over the
Life Cycle....................... 57
Notes to Chapter III............. 61

IV DATA AND METHOD ...... ....................... 62

Data.................................. 62
Constructing a Typology of
Employment Patterns................... 66
Objective of Typology............ 66
Identifying Employment
During Three Life Cycle States... 67
Identifying Employment Patterns.. 69
Method of Analysis.................... 77
Objective of Analysis ............ 77
Independent Variables............ 77
Interaction Terms................ 90
Method of Analysis............... 91
Missing Data ..................... 93
Notes to Chapter IV.................... 95


Introduction........................... 99
Typology .............................. 100
Distribution Over Twelve
Employment Patterns.............. 100
Distribution by Cohort............ 102
Collapsing the Typology.......... 105
Review of Contrasts................... 106
Introduction...................... 106
Never and Ever-Employed Wives.... 107
Employment ....................... 112
Life Cycle Employment Patterns... 115
Notes to Chapter V..................... 127

VI RESULTS .................................... 128

Relationships Among the
Independent Variables................. 128
Employment Pattern Contrasts.......... 130
Employed Wives.................... 134
Discontinuous Workers............ 140
Life Cycle Employment Patterns... 145
Summary of Findings................... 158
Notes to Chapter VI.................... 162

VII CONCLUSIONS... ............................. 172

Issues in Wives' Employment........... 172
Policy Implications................... 176
Implications for Future Research....... 178


REFERENCES........................................ 182

BIOGRAPHICAL SKETCH ............. ....... ............ 194



Table Page

5-1 Lifetime Employment Patterns of
Married Women ......................... 125

5-2 Employment Pattern by Cohort........... 126

6-1 Correlations Among Independent
Variables.............................. 165

6-2 Logit Coefficients for the Contrast
of Never-employed and Ever-employed
Wives.................................. 166

6-3 Logit Coefficients for the Contrast
of DOUBLTRK and INANDOUT Wives......... 167

6-4 Logit Coefficients for the Contrast
of DOUBLTRK and DTRKAB1 Wives.......... 168

6-5 Logit Coefficients for the Contrast
of DOUBLTRK and INTERUP Wives.......... 169

6-6 Logit Coefficients for the Contrast
of INTERUP and TRADITNL Wives.......... 170

6-7 Logit Coefficients for the Contrast
of MIDLIFE and NEVEREMP Wives.......... 171



Figure Page

3-1 Possible Employment Patterns........... 60

6-1 Plot of Equations Representing the
Relationship between TOTKIDS and
Employment Continuity at each level
of Birth Spacing, Older Cohort......... 172

6-2 Plot of Equations Representing the
Relationship between TOTKIDS and
Work Force Reentry after Childbearing
at each level of Birth Spacing......... 173

Abstract of Dissertation Presented to
the Graduate Council of the University of Florida
in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy



Ellen Van Velsor

December, 1980

Chairman: Dr. John C. Henretta
Major Department: Sociology

This study deals with the process of wives' employ-

ment over the first three stages of the family life cycle

and the background and fertility characteristics related

to this process. A typology of employment patterns is

constructed by tapping employment during the first three

stages of the family life cycle. Twelve patterns are

found to describe the worklives of ninety-seven percent

of the women in this sample. Employment pattern contrasts

are used to assess the relationships between background

and fertility characteristics and the following: 1) employ-

ment 2) employment,continuity 3) employment before child-

bearing, 4) employment during childbearing,and 5) employ-

ment after childbearing.

The analysis begins by contrasting wives never and

ever-employed over the first three stages of the family

life cycle. The only generalization that can be made

across cohorts is that the number of children a wife

bears has a significant negative impact on her probability

of employment. Older wives tended not to work when they

had a large number of children and/or when their education

(i.e., their potential wage) was low. Younger wives tended

not to work when they had a large number of children, when

their husbands' education (i.e., income) was high, when

marriage (and thus, childbearing) was begun late in life,

and/or when childbirth followed quicklyafter marriage.

Two generalizations can be made across cohorts

regarding employment continuity. One is that the higher

the frequency of geographical mobility, the lower the

probability that wives will remain continuously employed.

The second is that the greater the number of children,

the lower the probability that wives will remain contin-

uously employed. The older wives most likely to have been

employed discontinuously were those who had made frequent

long-distance moves and those who had had several children

an average of two to four years apart. The younger wives

most likely to have been employed discontinuously were

those who had moved frequently, those who had had several

children (regardless of their birth spacing), and those

who had non-farm backgrounds.

The only factor related to employment before child-

bearing across cohorts was the length of the interval

between marriage and first birth. Both a short first

life cycle interval and frequent geographical mobility

were associated with a decrease in younger wives' pro-

bability of first state employment. Among older wives,

a long first life cycle interval was associated with an

increased probability of employment before childbearing,

while having had four or more years of college was associated

with a decreased probability of employment during stage


Employment during childbearing can be attributed

to no one factor or group of factors for all wives. A

highly educated husband, a high frequency of geographical

mobility and close birth spacing were all associated with

a decrease in younger wives' probability of employment

during childbearing, while downward marital mobility was

associated with an increase in younger wives' probability

of employment during this stage. Among older wives, a

short interval between marriage and first birth was asso-

ciated with an increase in the probability of employment

during childbearing.

In both cohorts, marital timing and number of child-

ren are the only factors associated with work force reentry

after childbearing. For younger wives, initial employment


in mid-life is related to these same variables. For

older wives, only wife's education is significant. This

finding confirmed the earlier finding that older wives

with little education tended to remain never-employed.



Measuring Wives' Employment

The extent of labor market activity among various

social groups and the forms it takes are two crucial aspects

of labor force participation. The employment of married

men and women differs vastly in both these regards. For

men, participation in the work force is relatively con-

tinuous and, for the most part, full-time from the curtail-

ment of education to retirement. The employment of married

women is typically discontinuous and is patterned around

life cycle events such as marriage and childbearing. In

addition to its discontinuity, the female labor force is

characterized by intra-group variability in work patterns.

Some women never engage in paid employment, while some work

only before their children are born. Others may be employed

before and after childbearing, but may remain at home while

their children are small. Some wives are employed contin-

uously in every stage, either in pursuit of a career or

out of economic necessity. Others enter the work force

for the first time once their children are in school.

This individual discontinuity and intra-group varia-

bility has hindered attempts to explain female labor force

participation in that the typical ways of measuring this

variable have proven inappropriate for women. For example,

until recently research has relied on cross-sectional

aggregated data and has defined and measured labor force

participation as participation at a particular time rather

than as a process over the life cycle. Most studies used

Census data and measured labor force participation as

presence in the labor force during the survey week. How-

ever, the use of this measure improperly assumes that a

woman's labor force activity in one week adequately por-

trays work patterns over time. When work history data are

available, an alternative measure of labor force attachment

is the number of years worked in the last five or ten years.

Again, this measure is not necessarily indicative of the

current or future labor market activity of women since the

five to ten year span may have fallen during the childbearing

years for some women and afterwards for others.

More recently longitudinal data on female labor market

experiences have become available, facilitating clarification

of the relationship between female labor force participation

and life cycle events. Most current research using longitu-

dinal data measures female employment as either the propor-

tion of years worked since leaving school or the proportion

of years worked during various life cycle stages. Both

measures increase our understanding of women's worklives in

that they tap the discontinuity of female labor force

participation. However, looking only at the proportion

of years worked since leaving school does little to further

our understanding of the variability among women in life-

time work patterns. That is, women may spend equal pro-

portions of their lives in the work force, but the activity

may occur at varying life cycle stages for different women.

This consideration is an important one since time out of

the labor force during the childbearing years has been

shown to have a more detrimental effect on wages than time

out of the labor force in later years (Polachek, 1975).

Measuring labor force participation as the proportion of

years worked during various life cycle stages solves this

problem and points in an interesting direction. That is,

if the data allow one to link labor force participation

with life cycle events, certain lifetime patterns should

emerge. Only two studies identify any such patterns (Elder

and Rockwell, 1967; Young, 1978); and neither focuses on

the classification of lifetime employment patterns among

women in the United States.

If we are to understand the process of wives' employ-

ment, a measure is needed which taps both the discontinuity

of women's worklives and the intra-group variability in

employment timing. A typology of employment patterns

satisfies both these criteria. Most married women enter,

withdraw from, and reenter the work force several times

over the life cycle. Employment patterns tap this dis-

continuity by measuring intervals of both employment and

nonemployment. Wives vary in the way they time their

employment, as well; and employment patterns measure timing

by indicating the life cycle stages at which employment


The present study has two goals. The first is to

construct a comprehensive typology of the lifetime employ-

ment patterns of wives in the United States. This typology

will be an improved measure of wives' employment in that

it will describe both the extent and the timing of wives'

employment over the life cycle. It is assumed that the

timing of employment is associated with socioeconomic and

fertility factors, which also influence wives' employability

and wage rates. Thus, the second goal is to define the

factors related to employment, discontinuous employment,

and employment in each of the three early life cycle stages.

This objective is met by assessing the effects of a set of

family characteristics on several contrasts of the employ-

ment patterns encompassed in the typology.

Plan of Study

The literature on female employment is reviewed in

Chapter II. Early analyses of the relationships between


family characteristics and wives' labor supply are reviewed

first. For the most part, these are the studies which used

cross-sectional data and measured wives' employment as

presence in the labor force during the survey week. Some

of the more recent studies of female employment have attempted

to clarify the nature and direction of the relationship be-

tween fertility and labor force participation; and this large

literature is reviewed next. Studies involving labor market

variables and rural/urban differences in employment follow.

Numerous studies deal with the relationships among factors

usually associated with female employment; and these are

summarized here. Another theme in the labor force litera-

ture stresses the importance of family life cycle stage as

an explanatory variable. In general, these studies indicate

that family and fertility characteristics act to push women

into the labor force during certain stages and to encourage

them to withdraw during others. The next section reviews

studies of wives' employment discontinuity. Although the

discontinuity of female employment is an important factor,

it represents only one aspect of a complex phenomenon.

Another aspect involves the forms this discontinuity takes

among women. Studies of wives' employment patterns comprise

the final section of the second chapter.

Chapter III includes a discussion of both employment

patterns and theoretically interesting contrasts. That is,

if we wish to investigate the factors related to wives'

employment, ever-employed wives can be usefully contrasted

with never-employed wives. Continuous and discontinuous

workers can be contrasted to study the factors related to

employment discontinuity. The factors associated with

employment in each life cycle stage can be analyzed similarly.

The data and methods used in the present study are

discussed in Chapter IV. This chapter describes how longitu-

dinal data were used to tap employment in each of the three

early life cycle stages. It also describes the creation

of fertility timing measures and other variables from.these

data and explains the LOGIT procedure used in this analysis.

Chapter V describes the typology of employment patterns

emergent from the work histories of the wives in these data.

It assesses cohort differences in the distribution of wives

over these patterns and discusses the potential effects of

family characteristics on the employment pattern contrasts

chosen for further analysis.

The results of that analysis are presented in Chapter

VI. Six models are estimated to address several issues

important in studying wives' employment. The first model

addresses why some wives remain never-employed, while most

work for pay at some point in the family life cycle. The

second concerns factors related to employment discontinuity.

The third through sixth models deal with employment before,

during, and after childbearing.


The final chapter summarizes the findings presented

in Chapter VI and discusses both the policy-related and

theoretical implications of the present research.


Note to Chapter I

In the past thirty-five years, wives have increas-
ingly spent more of their adult lives in the work force.
The trend among men has been to delay labor force entry
in favor of prolonged education and to decrease their
age at retirement, resulting in a decrease over time in
the number of years spent working (Neugarten and Hagestad,
1976). Although male/female differences in the extent of
time spent in the labor force may have decreased since
World War II, undeniable differences remain. In addition,
the above trends seem to be leveling off in recent years.



The literature pertaining to female labor force

participation is interdisciplinary and cross-cultural

in nature. All of that which will be reviewed here is

economic or sociological and, with the exception of two

studies of Australian wives, deals with the employment

of women in the United States. Most of the economic

studies, and many of those from a sociological perspec-

tive, used cross-sectional data, as large-scale longitu-

dinal surveys have just recently become available. Some

of the cross-sectional analyses used aggregated data and

studied, for example, the effect of the fertility rate or

industry mix on female labor force participation rates in

Standard Metropolitan Statistical Areas (SMSAS). This

aggregate level of analysis provides a vital understanding

of macro-level labor market and demographic processes over

time, but obscures the dynamics of employment among indivi-

dual females. Studies utilizing individual level data,

especially longitudinal analyses, have begun to fill this

gap by specifying the factors which lead women from different

socioeconomic classes, races, regions and religions to

enter paid employment in varying life cycle stages.

Review of the Literature

Family Characteristics and Wives' Labor Supply

The earliest studies of women's labor market activity

concentrated on the relationships between family character-

istics and wives' labor supply. In response to economic

models which posited leisure as the only alternative to

employment, Mincer (1962, 1968) argued that wives must

choose between leisure, paid work, and housework in deciding

how much time to commit to labor market activity. In his

research, Mincer found an inverse relationship between

husband's income, both permanent and transitory, and wife's

rate of labor force participation, controlling for wife's

wage. His "added worker" hypothesis (i.e., that wives

work primarily to supplement husband's income in times of

financial need) resulted from Mincer's finding that husband's

transitory income was a better predictor of wife's employ-

ment than was his permanent income. He also found that

the effects of both permanent and transitory income on

wife's employment were weaker, the higher husband's educa-

tion. Based on his finding that the relationship between

women's wages and their employment level, controlling for

husband's income, was stronger than the effects of husband's

incomes on wife's employment, controlling for her wage,

Mincer concluded that, overall, the wife's response to ner

market wage was stronger than her response to her husband's


Subsequent research supports these findings. A later

elaboration of Mincer's early work found wives' labor supply

was dependent on their current wage, fertility and other

family members' incomes and wages (Cain, 1966). Rosenfeld

and Perrella (1965) looked at wives who entered the labor

force in 1963 or 1964. They found women with low income

husbands and those with a child under six years old were

most likely to give financial necessity as their main reason

for working. In a more general study of marriage and

divorce, Carter and Glick (1970) found that the greater

husband's income, the less likely were mothers to be em-

ployed. However, among couples with no children wife's

labor force participation was positively related to husband's

income. Sweet's (1970) analysis of the relationship between

family composition and wives' employment provided support

for the "added worker" hypothesis in that income adequacy

(i.e., the ratio of current income/current need) was a

significant predictor of wife's employment probability,

net of wife's age, education, age of youngest child and

number of children. The results of his (Sweet, 1973) study

indicated wife's education was positively related to her

employment, with education having a greater effect on employ-

ment status the lower her husband's income. Oppenheimer

(1970) found that wife's labor force participation depended

on her ability to enhance family income. That is, the

larger the ratio of wife's to husband's income, the more

likely wives were to participate in the labor force. Either

a high potential for earnings by wives or low earnings by

husbands would increase the likelihood of wives' employment.

Bowen and Finegan's (1969) work is probably the most

extensive study of labor force participation. These authors

argued that men's participation in the labor force is inde-

pendent of family economic situation and personal socio-

economic characteristics. Wives' employment was contingent

on family need, relatively high market productivity (vis-a-

vis her home productivity and/or her husband's market produc-

tivity), the availability of viable substitutes for her

home labor, and high demand in the female labor market.

Their findings were consistent with the "added worker" hypo-

thesis, in that there was a positive association between

wives' employment and the unemployment rate of husbands.

Dickinson and Dickinson (1973) questioned the "added

worker" hypothesis based on their finding that the relation-

ship between husband's permanent income and wife's employ-

ment was the strong negative one. Their results indicated

that wives seek employment as a result of long-term economic

disadvantage, rather than short-term financial need. Simi-

larly, Sobol (1973), using an index of wife's labor force

attachment over ten years, found that both husband's abso-

lute income and his income relative to those of significant

others had a greater effect on wife's employment than did

the percent change in husband's income over the survey

years. She also found wife's education and husband's educa-

tion to be positively related to wife's labor force partici-

pation and wife's age at marriage to be negatively related

to her likelihood of employment.

Smith (1973) modeled both husbands' and wives' labor

supply as a function of wife's wage, husband's wage and

the presence of children in the home. He found a rise in

wife's wage reduced husband's labor force participation, as

well as increased husband's earnings reducing wife's employ-

ment. In both the male and female models, a rise in one's

own wage had a greater effect on one's employment than a

similar change in spouse's wage. In addition, he found the

presence of children had a negative effect on wives' employ-

ment and a positive effect on the employment of husbands.

Thus, Lloyd (1975) argues that the presence of children rein-

forces the traditional division of labor that other economic

forces, such as improvements in household technology and the

decreasing male/female differential in education, are breaking


Although the labor force participation of males is

usually seen as independent of family characteristics, the

economic pressures accompanying the establishment of a

family and childrearing may increase the husband's pro-

pensity to change jobs and move in search of higher or more

stable income. The reciprocal effects of wives' employ-

ment and the residential mobility of families was the

subject of study by Long (1974). Controlling for occupa-

tion, Long found men whose wives were working at the survey

date to be less likely to have made a long-distance move

and more likely to have moved locally during the preceding

year than were men whose wives were not employed. Looking

at five-year migration rates, Long found a positive associa-

tion between wife's employment in 1965 and subsequent five-

year local mobility. Wives' continued nonparticipation in

the labor force was correlated with low rates of local

moving and high rates of interstate mobility. As a result,

the author suggests that a working wife's income acts to

supplement family income, thus facilitating the upgrading

of family housing. Yet there were no significant differences

in the rates of interstate migration between 20-24 year

old men with working and non-working wives in 1965. Long

found than an employed wife decreased the probability of

an interstate move only after the husband was 30. In

addition, Long found that the greater the distance moved,

the less the probability that a wife employed in 1965 would

be working in 1970. His research showed that the residential

mobility of families during the formative stages of the

husband's career has a significant negative effect on the

probability of wife's continued employment.

A study by Duncan and Perrucci (1976) reported similar

findings. These authors measured the effects of various

aspects of husband's and wife's occupations on the proba-

bility of family migration. They found the higher the hus-

band's occupational prestige, the greater the demands to

migrate emanating from his occupation and the greater the

opportunities for employment in his field elsewhere in the

country, the greater the probability of family migration.

Family migration was not affected by comparable aspects of

the wife's work role. In terms of the impact of migration

on the wife's employment, Duncan and Perrucci found that

interstate migration facilitates the employment of wives

not in the work force prior to the move, but inhibits con-

tinued employment among wives who worked preceding migration.

Sandell (1976) used the National Longitudinal Survey

data to analyze the effect of migration on the wages of

husbands and wives. His results showed that the earnings

of husbands increased as a result of migration, while wives'

earnings decreased. However, total family income increased,

as well, so the positive effect of migration on husband's

earnings tended to outweigh its negative effect on wife's


These studies of the effects of family characteristics

on wives' labor supply comprised the earliest reserach on

female labor force participation. Mincer pioneered the

field by positing a revised model of the time inputs into

various activities and formulating the "added worker" hypo-

theses. His work led to many additional studies of the

relationships between wives' employment and the various com-

ponents of family income. Thereafter, models of female

labor force participation were expanded to include such

factors as wife's education, the presence of children and

residential migration. Yet Mincer's work is still contro-

versial in that subsequent research has yet to determine

whether most wives work primarily to supplement husband's


Fertility and Labor Force Participation

Some of the more recent studies of female employment

have attempted to clarify the nature and direction of the

relationship between fertility and labor force participa-

tion. The issue of causality is an important one in this

regard. Several authors have argued that the employment/

fertility relation is a spurious one, produced by common

antecedent variables. Mincer's (1963) assertion to this

effect was the first. Terry (1974) argued that controlling

for education, religion, SES, residence and age at marriage

may weaken the negative employment/fertility relation or

may make it positive, and may result in different findings

at varying life cycle stages. Mincer's argument is supported

by Terry's (1975) analysis of the 1960 Growth of American

Families data, and by Scanzoni's (1975) comprehensive work

on sex roles, employment and childbearing. When Terry con-

trolled for a set of factors known to influence both female

employment and fertility (e.g., age, Catholicism, education,

residence, socioeconomic status and length of marriage),

the work/fertility relationship disappeared among blacks.

Scanzoni modeled wife's employment and intended family size

as a function of role norms, socioeconomic status, education,

and fertility timing variables for white non-Catholics,

white Catholics and blacks. For all but the Catholic wives,

he found that both sex role attitudes and number of children

were related to the proportion of a wife's married life

spent in the labor force. However, in predicting intended

family size among employed women, Scanzoni found sex role

attitudes to be the strongest correlates of birth intentions

for all three groups. He concludes that sex role modernity

may be antecedent to both fertility and employment decisions.

Scanzoni found that work motivation was important to fer-

tility patterns, as well, among employed Catholic and

black wives. Those who worked for personal satisfaction

tended to have fewer children than those who worked out of

necessity. Ryder and Westoff (1971) reported a similar

finding; and Dowdall (1974) found the effects of work

attitudes on wives' employment to increase with higher

income. She also reported that structural factors, such

as education, income, and children's ages, were weakly

associated with present female employment, as were the

structural variables and sex role attitudes. However,

the inclusion of sex role attitudes as predictors of female

employment status significantly increased the explained

variance in this dependent variable.

In addition to those just discussed, at least three

studies suggest that the effect of employment on fertility

may depend on the woman's work commitment or motivation.

Kupinsky (1971) used a work commitment index (the number

of years worked during marriage as a percentage of the number

of years married) to investigate whether the effect of

wife's employment on fertility differs by wife's socio-

economic status. He found that, among wives committed to

their work role, high SES wives were more likely to main-

tain a small family size than lower SES wives. This rela-
tionship was strongest among rural women. Kupinsky con-

cluded that future research should assess the proportion

of wives' married lives spent in the labor force, by cohort.

Hoffman (1974, 1975) suggested that if the negative effect

of employment on fertility is not a spurious one, it is

probably a result of the incompatability of the worker

and mother roles since in populations where the extended

family facilitates child care, this negative relationship

does not exist. Hoffman argued that rewarding employment,

rather than employment per se, may be what reduces fertility.

Similarly, Birdsall (1976) argued that the effect of work

on fertility depends on the woman's attitude toward her

work and on her satisfaction with it. When commitment and

satisfaction are high, the opportunity cost of children

is also high, and fertility will decrease as employment

increases. This set of events occurs mainly in urban

areas, according to Birdsall. In rural places, female

employment does not reduce fertility because the opportunity

cost of children is low and the benefit high. In fact,

cross-culturally, the fertility of working women in rural

areas is higher than the fertility of housewives in urban

areas. Birdsall argued further that sociological and

economic research on female employment has assumed a con-

gruence of interest between husband and wife concerning

the value of children and has thus treated the household

as a utility-maximizing unit. She claimed that one result

of this perspective has been that few studies have specified

wife's earnings as a proportion of husband's earnings.

Doing so would tap the potential or actual relative economic

contribution of the wife to the household and would facilitate

investigation of the link between fertility and the economic

importance of wives' employment.

Subsequent studies have sometimes included such a

measure in their models. As described in the second section

of this chapter, Oppenheimer (1976) found wife's labor force

participation depended on her ability to enhance family

status (see p. 12). Weller (1977) found that the greater

the ratio of wife's to husband's income, the smaller the

completed family size. This negative relationship was greater

among persons married less than ten years and among those

with twelve or more years of schooling. Weller's results

supported the argument for spuriousness, as well, by sug-

gesting that both employment and completed fertility are

dependent on education, race and fertility timing variables.

His results indicated that: 1) the work/fertility relation-

ship was strongest among wives with a college education and

among those who had been married less than ten years; 2)

the work/fertility relationship was stronger among whites

than among blacks, controlling for other variables; and

3) the relationship between employment status and number

of children was strongest among women married less than

ten years. Concerning the latter point, Weller suggested

wives' employment may be more closely related to the timing

and spacing of births than it is to family size. Alterna-

tively, wives married more than ten years at the survey

date may have tended to enter the labor force after child-

bearing. Weller argued that if this were the case, the

timing of employment would be the critical factor to study.

Waite and Stolzenberg (1976) reported their findings

were inconsistent with Mincer's hypothesis of spuriousness.

They used the National Longitudinal Survey data to measure

the reciprocal effects of fertility plans and employment

expectations, and controlled for variables affecting fer-

tility. They found the greater impact to be the effect of

labor force participation plans on fertility expectations.

Smith-Lovin and Tickamyer (1978) criticized Waite

and Stolzenberg's use of fertility and employment expecta-

tions as misleading and, instead, used actual fertility

and labor force behavior. They found that the number of

years a woman had been employed since her marriage was best

predicted by her education, her pre-marital work experience

and current fertility, spouse's income, and marital dura-

tion. The number of children she bore, on the other hand,

depended on her religious preference in high school and

marital duration. Labor market activity had no significant

net effect on actual fertility in these data.

A recent paper (Cramer, 1980) examined the contradic-

tory results reported in the Waite and Stolzenberg (1976)

and Smith-Lovin and Tickamyer (1978) studies. Cramer argued

that methodological problems plague both analyses. Although

each set of findings can be supported by the literature,

Cramer pointed out that both studies show the presence of

serious multicollinearity problems. These problems lead

to unreliable estimates of the reciprocal effects of wives'

employment and fertility. Using sensitivity analysis to

detect such instability, Cramer estimated models analagous

to those in both studies and found support only for Waite

and Stolzenberg's conclusions (i.e., that employment plans

reduce expected family size). He found no conclusive

evidence either to support Smith-Lovin and Tickamyer's con-

clusion that completed fertility strongly affects cumula-

tive employment or to suggest reciprocal causation. According

to Cramer, misspecification is another major problem with

the Smith-Lovin and Tickamyer model, in that there was no

temporal relationship specified between the cumulative

employment and fertility variables. Cramer estimated dynamic

models of both cumulative and expected employment and

fertility. Using the former (cumulative) model, he found

wife's initial (i.e., prebirth) employment significantly

depressed fertility, although the coefficient was small.

However, having a baby substantially reduced wife's subse-

quent employment. Thus, Cramer argued that although wife's

cumulative employment and fertility do affect each other,

the effect of fertility on employment is the greater one.

After estimating the latter (expectations) model he concluded

that employment plans depend primarily on recent employ-

ment, while fertility expectations are best predicted by

recent fertility. Neither type of behavior (i.e., employ-

ment or fertility) had much effect on either kind of expec-


Several other researchers have examined the relation-

ships between aspects of female employment and fertility

without dealing extensively with the issue of causality.

Preston and Richards (1975) modeled marriage rates in SMSAS

as a function of a set of labor demand variables. Controlling

for male earnings, population of the SMSA, the proportion

Catholic, the proportion of college educated females, and

the racial composition and sex ratio of the SMSA, they

found the more female the industry mix and the higher female

wage rates, the lower the proportion of married females.

Both percent Catholic and population of the SMSA were nega-

tively related to the proportion marriage, while the sex

ratio exerted a positive influence. These results were

not due to the migration of unmarried women to high demand

cities, as a migration ratio was included in the model to

control for this event. Because the effects of the labor

market variables were stronger than those of the marriage

market factors, the authors concluded the job opportunities

effect on marriage rates dominated the marriage opportunities

effect. Santos (1975) also found that, despite the rise in

real family income over time, the rise in female market

potential was a significant factor in the decline in female

marriage rates. Women are less likely to marry if they reside

in an area of relatively high employment opportunity.

Waite and Spitze (1978), Ross (1974), and Bowers

and Hastings (1970) all studied the relationships between

employment or taste for work, and fertility timing variables.

Waite and Spitze focused on factors affecting the probability

of marriage and first birth at a given age for females.

Variables which indicated a preference for future employ-

ment (e.g., taste for market work, school enrollment and

mother's education) tended to increase ages at marriage

and first birth among the women in their data. Current

employment acted to delay marriage and first birth, as well.

Ross focused on the timing and spacing of births and wives'

labor force participation. She found that women with higher

education had an earlier first birth than did other women,

and spaced their births more closely. A negative relation-

ship was found between education and number of children, as

well. Women with more education had higher labor force

participation rates in all birth intervals; and the higher

their husband's income and education, the lower the wife's

rate of employment. Finally, among wives who worked after

having children, those with higher husband's income, more

children and wider birth spacing were more likely to begin

paid employment during their childbearing years. Bowers

and Hastings looked at fertility timing differences by

wife's employment status among college graduates in Utah.

They found working wives' families were smaller and more

closely spaced than families of non-working wives. How-

ever, non-working wives had had a shorter marriage/first

birth interval than working wives. Wives who worked during

childbearing generally had the most widely spaced births.

Wives who worked before, or before and after, childbearing

spaced their births closer than wives who worked during

childbearing, but wider than wives who worked only after

childbearing was completed.

Two studies focused on the intensity of female employ-

ment. Kushican and Scneffler(1975) analyzed data on married

men and women and found a sex difference in the effect of

number of children on work time. The addition of children

tended to increase male work time; while for women, an in-

crease in the number of children was related to a decrease

in time worked. Their finding parallels that by Smith

(1973) concerning the presence of children and husbands'

and wives' labor supply (see p. 13). In addition, Kushman

and Scheffler found a positive association between the

quantity of labor supplied by husbands and wives, controlling

for number of children. The authors speculate that increased

work time on the part of males allows women to work more

without upsetting the balance of family power by changing

their relative contributions to family income. Another

explanation may be that an increase in financial need caused

increased employment on the part of both spouses. Gramm

(1975) also analyzed data on the intensity of labor force

participation (i.e., worked full-time, part-time, or not at

all in 1969), among a sample of married teachers. Although

husband's wage had a negative effect on time worked by wives

in these data, the most important factors were presence of

the first child and the age of that child. Return to paid

employment was quicker after births subsequent to the first

and as the first child got older.

Two other studies indicate the persisting influence

of fertility behavior on wives' employment. Groat, Workman

and Neal (1976) studied several factors related to the extent

of labor force participation among a sample of working

mothers. They found the longer the marriage/first birth

interval of these wives, the longer their work duration and

the greater their occupational status. Among Protestants,

a long marriage/first birth interval and high occupational

status were related to having fewer children; while among

Catholics, these factors were associated with greater completed

fertility. Finally, premarital work experience was associated

with a longer marriage/first birth interval and also with

continued employment after marriage. Sweet (1970) found

that the probability of being in the labor force for women

aged 50-54 with no children present was negatively related

to the number of children ever born. This analysis suggests

that the amount of time spent in childbearing has a long-

lasting effect on female labor force participation.

Analyses of wives' employment and aspects of their

fertility probably comprise the largest segment of the lit-

erature on female labor force participation. Although the

issue of causality in the work/fertility relationship has

been intensely debated, it has yet to be resolved. Most

research either models one variable as a function of the

other or measures their reciprocal effects. The findings

of these studies are contradictory and a convincing refuta-

tion of Mincer's hypothesis of spuriousness has yet to appear.3

Research has shown that the employment/fertility relation

varies by race, education, religion, marital duration and

the motivation to work. We have also found that employment

opportunities affect marriage rates and that the timing of

marriage and the timing and spacing of births are all parti-

ally dependent on wife's education. The timing of first

birth and the spacing of births have been linked to employ-

ment during various life cycle intervals, as well; and sex

differences have been found in the effect of children on

work time.

Rural/Urban Differences and Labor Market Variables

Other studies have explored rural/urban differences

in wives' labor force participation, or have focused on

the effects of labor market variables on female employment.

Sweet (1972a) compared the differential effects of education

on the labor force participation of rural and urban wives.

He found the greatest employment difference among rural

and urban wives with low levels of education. Farm wives

were less likely to be employed. However, at higher levels

of education, rural wives were more likely to be employed

than urban wives. Sweet speculated that labor demand factors

may be important in that employment opportunities may be

greater for college graduates in rural areas than they are

for college graduates in urban areas. He also argued that

having a low education hinders employment more in rural

areas than it does in cities where lower-skill jobs are

more numerous.

Sweet's research is not the only work to point out

the relevance of demand factors to wives' employment rate.

Blau (1972) argued that female labor supply is highly res-

ponsive to changes in both employment opportunities and the

wage rate. Bowen and Finegan's (1969) work included a chap-

ter on labor market conditions. Controlling for husbands'

income, other family income, education, percentage non-white,

percentage of wives with a preschool child, and net migration

between 1955 and 1960, they found that all the labor market

variables' effects on wives' employment were significant.

The rate of female employment was greater, the lower the

unemployment rate, the more female the industry mix, the

lower the wages of domestics and the greater the female

wage rate.

A more recent study compared the effects of supply

and demand factors on the employment of rural and urban

women. Chenoweth and Maret-Havens (1978) reported that,

overall, rural women were much less likely to be employed

than were urban women. Of those employed, rural women

were the least likely of any to have a permanent career

attachment to the labor force. On the other hand, these

authors found no rural/urban difference in the effect of

education on employment, although urban women were more

likely to have additional job-related training beyond their

formal education. The remaining supply factors tested

included marital status, husband's income, number of child-

ren and sex role attitudes. In general there were no signi-

ficant rural/urban differences in the effects of the supply

factors on female employment. However, a different picture

emerged for the demand factors. Chenoweth and Maret-Havens

found the industry mix in the urban places provided more

job opportunities for women. The least favorable industry

mix was found in rural areas located outside SMSAS. Employed

urban women were also more likely to receive higher wages

than their rural counterparts. Rural women spent less

time in the labor market than urban women due to their

relative disadvantage in terms of both employment oppor-
tunities and wages.

Studies of labor market variables and rural/urban

employment differences have led to a simliar conclusion.

That is, demand factors (e.g., wages and industry mix)

are at least as important as supply factors (e.g., fer-

tility and human capital variables) in predicting wives'

employment. The greater the availability of jobs and the

higher their pay, the more likely is employment among all

types of women.

Relationships Among Other Factors

Numerous studies deal with the relationships among

factors usually associated with female employment. Some

of these look at the determinants of age at marriage or

educational attainment, while others study the factors

related to aggregate fertility rates or various components

of individual fertility.

Voss (1975) studied the social determinants of age

at first marriage; Marini (1978) compared the effect of

age at marriage on the educational attainment of males and

females; and Waite and Moore (1978) analyzed the effect of

age at first birth on the educational attainment of black

and white females. Both Voss and Marini found a strong

positive association between age at marriage and educa-

tional attainment among females. In Voss' data, the effect

of age at marriage on years of schooling was a moderate,

positive one for males. Using follow-up data on 1969 high

school students, Marini found that age at marriage had no

significant effect on educational attainment among males.5

Based on their analysis of the National Longitudinal Survey

data on women aged 14-24, Waite and Moore asserted there

was a direct relationship, however, between age at birth

of first child and female educational attainment, the

effect being twice as great among whites as among blacks.

Age at marriage was related to length of marriage/

first birth interval in a study by Ruzicka (1976), and to

spacing of births by Bumpass and Westoff (1970). Among a

sample of Australian wives, Ruzicka found the earlier the

age at first marriage, the shorter the marriage/first birth

interval. Bumpass and Westoff found that as age at marriage

increased, children tended to be more closely spaced. This

association grew stronger, the greater the number of child-

ren. Among women who married early, those with large

families spaced their births wider than those with fewer

children. Among women who married late, those with large

families spaced their births closer than those with fewer

children. They also found that spacing was negatively

related to number of children, net of marital age, and

that wife's education was negatively related to spacing

of births. They suggest that highly educated wives may

space births more closely than wives with less education

partly because the former tend to marry at a later age.

In an earlier work, Bumpass' (1969) attempted to

specify the role of age at marriage in the differential

relationship between socioeconomic status and fertility

among Catholic and non-Catholic couples. That is, Catholic

couples usually display higher fertility with increased

socioeconomic status, while non-Catholic couples tend to

show reduced fertility, the greater their socioeconomic

status. Bumpass found age at marriage to have an important

interaction effect. When wife's age at marriage was less

than 19, the association between fertility and

socioeconomic status was strong and negative. As wife's

age at marriage increased, the strength of this negative

relationship decreased and became positive when wife's

age at marriage was greater than 23. Thus, the

analysis suggested that the original relationship was due

to the over-representation of Catholics among late-marriers.6

In addition, Bumpass found that, among wives who married

early, the greater their educational attainment, the lower

their fertility. However, among wives who postponed marriage,

higher fertility was more likely for women with higher

levels of education. This interaction of age at marriage,

education and fertility was found among both the rural

and urban, Catholic and non-Catholic wives in his data.

The effect of economic variables on aggregate fer-

tility rates or on individual family size was the subject

of several further analyses. Cain and Weininger (1973)

used cross-sectional aggregate data to study the economic

determinants of fertility rates in SMSAS. They found fer-

tility rates were higher, the higher male wages and the

lower female wages. Gardner (1973) was concerned with pre-

dicting changes in individual family size. Factors which

acted to reduce completed fertility included increases in

both wife's and husband's wages and educations. The nega-

tive effect of husband's education was significantly

greater in families where the wife was not employed. Two

papers by Simon (1975a, 1975b) dealt with the effects of

income and education on the probability of successive births.

In the first, he found that an increase in family income

was associated with a greater probability of subsequent

births among families with one child. However, among fam-

ilies with several children, an increase in family income

decreased the probability of future births. His second

paper is basically an extension of the original model to

include various interaction effects. There, he found the

positive effect of additional income on successive births

to be stronger for more highly educated women than it

was among wives with fewer years of schooling. In addition,

the positive effect of an income increase on successive

births was weaker among blacks than it was for whites.

Presser (1971) argued that most fertility research

has studied racial differences in completed fertility and

ignored the role of timing of first birth. Analysis of

the 1960 Census data has indicated the median age at first

birth was two years younger for black women than it was

for white women (22.2 years vs. 20.1 years). She found

that black women have shorter birth intervals, more timing

failures, more premarital births, earlier premarital births,

a lower probability of marriage following a premarital

birth, and a greater probability of labor force participa-

tion after first birth. Presser argued many of these fac-

tors may largely explain racial differences in completed


By examining the interrelationships among factors

usually predictive of female labor force participation,

the above studies allow one to better understand how several

variables may jointly induce wives' employment patterns.

For example, women who marry early tend to be low on educa-

tion and to have an early first birth. They also tend to

space their births widely. One would expect all of these

factors to inhibit employment, at least through the child-

bearing years. Additionally, the effect of husband's

income on wife's employment may vary according to family

size. In families with one child, a rise in husband's

income increases the probability of a subsequent birth

and probably decreases wife's employment as a result. Among

larger families, husband's income is negatively related

to the probability of future births and may have a lesser,

or different, effect on wife's employment.

Family Life Cycle

Another theme in the labor force literature stresses

the importance of family life cycle stage as an explanatory

variable. Becker (1965) took a somewhat novel approach,

focusing on husband/wife time inputs into home and market

activities. He argued that as factors (e.g., husband's

wage, wife's wage, number of preschool children, etc.)

influencing the spouses' relative productivities change,

wives alter the proportions of time they devote to work at

home and to labor market activity. Oppenheimer (1974)

found that families may experience money shortages as the

demand exerted on family resources changes over the life

cycle. This "life cycle squeeze" usually occurs when child-

ren are in their teens and is especially severe when the

father is in a blue collar occupation, in which income

peaks relatively early. Gove et aL (1973) argued that

families meet the "life cycle squeeze" situation in pat-

terned ways. In working class families, characterized

by a low income profile, an early income ceiling, a large

number of children and early timing of first birth, the

wife is likely to be in the labor force; children may be

working; the husband may work overtime; and relatives may

live with the family. In middle class families, charac-

terized by a late income ceiling, high education, later

marriage and a delayed first birth, the wife may work

from marriage to first birth. Gove's work seems to imply

either that no life cycle squeeze occurs in middle class

families or that it occurs earlier in that group. The

marriage/first birth period is one of relative financial

strain due to expenses which must be incurred to establish

a home and the low earnings of husbands relative to their

permanent incomes.

Several studies suggest that life cycle stage inter-

acts statistically with other independent variables. Terry

(1975) addressed rival explanations in the work/fertility

relationship by testing the common influences of other

factors. In general, she found the effects of these other

variables (e.g., education, socioeconomic status and resi-

dence) on the work/fertility relationship differed across

life cycle stages. For example, among white wives, number

of children was positively related to working only during

the childbearing years. Leibowitz (1975) demonstrated

the importance of life cycle stage when she found that

the employment status of women in the childbearing and

childrearing stages was independent of their educational

level. The education/employment relationship disappeared

in these stages because the time input into child care

increased with the years of schooling of the mother.

Similarly, both Becker (1965) and Gronau (1973a) found that

young children increased the value of (home) time more

for highly educated women than for those having lower levels

of education. Mott (1972) assessed the effects of educa-

tion and life cycle stage on the employment of two cohorts

of women in Rhode Island. At all levels of education, he

found higher rates of premarital labor force participation

among women reaching marriage age by World War Two. Among

this older cohort, the lower a women's education, the less

likely she was to have been employed between marriage and

first birth. Among the cohort reaching marriage age after

World War Two, the greater a woman's education, the more

likely she was to have been employed during this life cycle

stage. The younger cohort was more likely to have worked

during the childbearing stage, while the older cohort was

more likely never to have worked. Finally, Mott found that

employment in one life cycle stage was a good predictor of

employment in later stages, for both cohorts. The later

the life cycle stage, the stronger the association between

work in that stage and in subsequent intervals. Blau (1976)

reported a similar finding concerning the predictive power

of work in one stage for work in the next. In two papers

(1976, 1980), Waite modeled the probability of a wife's

working during each of several life cycle stages as a

function of a set of independent variables. In the earlier

paper, these were length of the interval, age during the

interval, Catholicism, demand for female labor during the

interval, wife's wage potential, husband's income, and con-

traceptive use in the interval. In her later work, the

variables included wife's wage, her education, the demand

for female labor, size of the labor market, and unemploy-

ment rate of her area of residence, age of the youngest

child, the proportion of weeks since leaving school that

the wife worked, her age and age at marriage, number of

children, husband's health, race, southern background, and

husband's and wife's attitudes regarding female employment.

In general, she found the effects of the independent vari-

ables on labor force participation differed across the

various life cycle stages. In the earlier study, for example,

the positive effect of wife's wage on her employment was

greatest in the period before her first birth, declining

steadily afterward. That research was based on a national

sample of married women, 18 to 44 years old in 1960. In

her later research, based on a different cohort of women

(aged 14 to 24 in 1967), wife's wage potential had a

strong positive effect on employment in all stages of

the life cycle. On the other hand, education had a greater

effect on the probability of labor force participation

before intended childbearing was completed. Additionally,

Waite found the timing of early life cycle events, such

as marriage, to be strongly related to later labor market

activity. Taken together, the overall effects of the inde-

pendent variables (education, husband's income, etc.) acted

to push women into the labor force during certain stages

and to encourage them to withdraw during others. Viewed

over individuals' lifetimes, the result would be patterns

of employment and withdrawal from the labor force. Yet

few studies conceptualize female employment as lifetime


Although we know the labor force participation of

married women is typically discontinuous and is related

to life cycle events, until recently, research has had

to rely on cross-sectional data and has tended to define

and measure labor market activity as employment at a parti-

cular time rather than as a process over the life cycle.

The recent availability of longitudinal data has facilitated

investigation of both the discontinuity of female employment

and the various forms this discontinuity takes among married


Discontinuity of Female Labor Force Participation

One study tapped the discontinuity of female labor

force participation by cumulating spells of employment

and attempting to predict the frequency of labor market

withdrawal and reentry. Hill (1977) found the frequency

increased with number of children, number of years the

husband was unemployed and frequency of geographical


Two studies tapped the discontinuity of female employ-

ment by measuring labor force participation as the propor-

tion of time employed during specific life cycle intervals

(Cogan and Berger, 1978; Polachek, 1975). Cogan and Berger

were concerned with the impact of the timing, spacing and

number of children on a married woman's accumulated work

experience and wage growth over the life cycle. They

found that both the number of children a woman had and

the length of her birth intervals were negatively related

to years of work experience. In addition, Cogan and Berger

found significant interaction between number and spacing

of children such that the greater the number of children,

the greater the negative effect of wide spacing on life-

time work experience. Age at first birth had no signifi-

cant net effect on accumulated work experience in their

data. In terms of wage rates, Cogan and Berger found a

direct positive effect for accumulated work experience

and an indirect effect for childrearing through its effect

on years worked. Polachek was concerned with the way in

which varying degrees of life cycle labor force participa-

tion affect female wage rates and contribute to an explana-

tion of male/female wage differentials. His findings

indicate that a significant proportion of the male/female

wage differential could be explained by male/female varia-

tion in employment continuity. That is, women's wages are

generally lower than men's partially as a result of their

less continuous patterns of labor force participation.

Although they did not look at employment by life cycle

stage, Suter and Miller (1973) report similar findings.

They were able to account for approximately two-thirds of

the variance in the male/female income differential using

occupation, age and work experience. Jusenius' (1976)

research further specified the employment continuity/wage

relationship found by Polachek and by Suter and Miller.

She found that the effect of work experience on wages varied

by the skill level of the job. Controlling for job tenure,

one-third of the typically female occupations monetarily

rewarded only the most recent work experience.

Patterns of Employment

Although the discontinuity of female employment is

an important factor, it represents only one aspect of a

complex phenomenon. Another aspect involves the forms

this discontinuity takes among women. Employment patterns

are both role management strategies and indicators of the

need for added income over the life cycle. Sweet (1972b)

discussed several strategies in combining the worker and

mother roles. He argued that the typical means is to work

from marriage to first birth, to remain at home during the

childbearing years, and to return to paid employment when

the children are school-aged. Other strategies suggested

by Sweet include part-time work, continuous employment

(either career-related or economically necessitated),

sporadic withdrawal and reentry, and permanent labor force

withdrawal when the first child is born. Of course, another

strategy is to forego the worker role and be permanently

out of the labor force. In his analysis, Sweet found that

women with less than twelve years of schooling were more

likely to have followed the latter pattern than were women

with higher educations. He also found college educated

women to be more likely than wives with less education to

return to paid employment soon after childbirth. Among

mothers of young children in general, the rate of return to

the labor force was U-shaped, with wives having high and

low educations exhibiting the greatest return rates. Sweet

speculated that the occupations held by poorly educated

wives may be easier to reenter than those of wives with

intermediate levels of education and that inexpensive child

care may be more readily accessible for these women. One

may add that poorly educated wives are likely to be forced

by financial circumstances to return to work soon after

childbirth, as well. Sweet concluded by stressing the

need for a longitudinal study of wives' employment patterns

and their differential use by various population subgroups.

To date, two studies have identified various life-

time employment patterns among married women (Elder and

Rockwell, 1967; Young, 1978). Elder and Rockwell concep-

tualized employment patterns as presence in the labor force

during the various family life cycle stages. They divided

their sample of mothers into three main categories and

looked at the relationship between worklife pattern and

marital timing. They identified four employment patterns

in a 1925-29 birth cohort: conventional, double-track,

interrupted and unstable. One-third of the women either

never worked at all or worked only until their first birth

("conventional"). One-fourth worked prior to having their

first child and returned to paid employment at a later

stage ("interrupted"). Thirty percent reentered the work

force after each birth ("double-track"). The final category,

"unstable," referred to the twelve percent of women with

fluctuating employment. Elder and Rockwell's data indicated

a link between economic disadvantage, early marriage, and

the double-track employment pattern.7 Women who married

early were typically from a lower socioeconomic background,

lived apart from their parents as teenagers, had a greater

than average number of siblings, and were Southern or

rural in origin. In this birth cohort, women who worked

continuously throughout their lifetimes did so mainly out

of necessity.

In a more recent study, Young (1978) identified the

main employment patterns during the first three life cycle

stages in a cross-sectional sample of Australian wives, and

then assessed the influence of several independent variables

on the choice of work pattern. The women she surveyed

were asked whether they worked outside the home between

marriage and first birth (yes/no), while the children were

still too young for school (yes/no), and after the children

were in school (yes/no). These women were generally found

to be concentrated in five of the eight possible employment

patterns. Thirty-five percent of the wives did not work

at all during these first three life cycle stages (NNN);

twenty-four percent worked only between marriage and first

birth (YNN); eleven percent began working when their child-

ren were in school (NNY); eleven percent worked except when

their children were preschoolers (YNY); and ten percent

worked continuously (YYY). Within these patterns, Young

further differentiated the women according to whether they

worked full-time or part-time during each life cycle stage.

Again, a limited range of work patterns was found. Women

following the YNY pattern worked either full-time during

both working stages or full-time in the first stage and

part-time in the third. Those who worked continuously

(YYY) were either consistently full-time workers or worked

full-time during stage one and part-time thereafter. The

majority of women working only during the first stage (YNN)

were full-time workers, while most of those who began

working only after their children were in school (NNY)

worked part-time. The independent variables Young con-

sidered included wife's education, year of marriage, and

timing of first birth. In terms of education, Young found

that women having greater years of schooling (high school

degree or higher education) were most likely to work during

all three stages (YYY), to work continuously except when

preschool children were present (YNY), or to work only

before first birth (YNN). On the other hand, women low

on education were most often found to be never-employed

(NNN) or to begin working only after all children were in

school (NNY). In terms of year of marriage (which is roughly

analagous to cohort), wives married in the 1930's tended

to remain at home until their children were in school.

Women who married in the 1940's, were most likely to work

in the YNY pattern; while those who married since 1950 were

increasingly found to have worked continuously since marriage.

In terms of the timing of first birth, there was evidence

in this Australian data to indicate that women with a

short marriage/first birth interval tended not to work

during that interval.


The present research builds on the implications of

past analyses of female labor force participation. An

early line of research demonstrated that family charac-

teristics are closely related to wives' labor supply. Some

of the more recent research on wives' employment has

attempted to clarify the nature and direction of the work/

fertility relationship. Rural/urban differences have been

found in the effects of supply and demand factors on female

employment; and numerous studies have focused on the rela-

tionships among variables usually predictive of female

employment. Another line of research has demonstrated that

the effects of socioeconomic factors on wives' employment

vary from one life cycle stage to the next. Thus, an

individual with a particular configuration of characteristics

would be more likely to be employed during some stages and

less likely during others. Recent research has shown that

the discontinuity of female employment is an important

aspect of labor market dynamics. Yet the discontinuity

itself is patterned.

Among one cohort of American women, four patterns

of labor force participation and withdrawal have been

tentatively linked to other variables. Yet Elder and

Rockwell's typology of worklife patterns was secondary

to their main concern with analyzing the factors related

to marital timing. Because they classified wives using

only four employment patterns, their typology loses much

valuable information regarding wives' employment. Young's

classification system, on the other hand, was a comprehen-

sive one. She identified the main employment patterns

among a sample of Australian wives and did a limited

analysis of the factors related to the probability of being

employed in a certain patterns.

Although past research has identified and analyzed

several employment patterns, a comprehensive study of the

employment patterns of women in the United States was yet

to be done. Much cross-sectional research has been done

on wives' employment, but we still do not know why some

wives are never employed while most work for pay at some

point in the family life cycle. Although it is clear that

most wives are discontinuously employed and that discon-

tinuous employment is related to low wages, we have not

discovered the factors that distinguish continuously employed

wives from discontinuously employed wives. Finally, little

is known about the timing of employment among married women.

Research Objectives

The objectives of the present research are: 1)

to describe the employment timing patterns of married

women, and 2) to indicate the family characteristics

associated with these patterns. The first objective is

met by identifying the various employment patterns emergent

from the work histories of two cohorts of American wives.

The second goal is met by identifying socioeconomic and

fertility factors related to employment, to discontinuous

employment, and to employment in each of the three early

life cycle stages. These endeavors will add to our knowl-

edge of wives' employment because they will provide a better

understanding of the association between family character-

istics and the synchronization of events in women's lives.

Notes to Chapter II

It should be noted, however, that although Scanzoni's
results supported this hypothesis, his data were cross-
sectional in nature, and thus, did not capture the dynamic,
processual nature of employment/fertility decisions.

2Kupinsky's finding suggests that employment (work
commitment) has a greater effect on family size among high
SES wives. However, it may be that family size has a
greater effect on work commitment among the high SES group.
That is, the causal sequence may be confused in this study.

Waite and Stolzenberg (1976) tested Mincer's hypoth-
esis by including race, education, marital status and hus-
band's income in their model. However, they failed to
include variables such as ethnicity, religion, sex role
attitudes and mother's employment; all of which can be
prior to, and possibly causal for both wifes' employment
and fertility.

This analysis did not include an interaction term
to test Sweet's hypothesis that the effect of education
on job opportunities differs in rural and urban areas.

However, an early marriage is not inconsequential
for males. Hogan (1980) found men who experience a dis-
orderly transition to adulthood (i.e., marriage occurred
before school completion and/or before beginning the first
job) also experience lower earnings returns to their educa-
tion and substantially lower total earnings.

Other studies that found Catholics are late-marriers
include Elder and Rockwell (1967) and Ryder and Westoff

Another study which shows a link between early mar-
riage and later economic disadvantage is one by Coombs
and Freedman (1970).



Married women with children are often faced with

the difficult task of assuming multiple roles. The wife role

usually includes major responsibility for housecleaning,

shopping, laundry, and cooking; while the mother role in-

volves childbearing, medical care, transportation, baby-

sitting, counseling and attendance at various group meetings.

Many married women take on the worker role, as well, either

for the financial and personal rewards of a career or to

provide a necessary supplement to family income. The simul-

taneous assumption of the wife, mother and worker roles is

a difficult task at best. Given no marital disruption, the

wife and mother roles are relatively fixed once the first

child is born. Employment is the only manipulable role

and is usually timed to balance family need with the wife's

desire to work.

Most families face a period of financial strain at

least once over the life cycle. Between marriage and first

birth, many couples are confronted with the expense of esta-

blishing their own home. Major purchases during this period

often include a house, furniture and appliances. For most

families, income is low during the first interval. If

the individuals are young, their earnings are even more

likely to be low, regardless of their occupations. Once

childbearing begins, wives are generally less likely to be

be employed. However this second period may be one of

financial strain, if the husband's income is insufficient

and/or if fertility is high. Another "life cycle squeeze"

often occurs as the children get older and make increased

demands on the family's resources. Among blue-collar workers,

the third stage can be particularly difficult, since income

peaks earlier than family demand in those occupations.

Among higher-income families, the third stage can also be

a difficult one as older children drain resources through

college-related expenses.

Families employ various strategies to meet their needs

over the life cycle. Some go heavily into debt; while others

cut back on expenditures for clothing, entertainment or

vacations, or forfeit savings or a college education for

their children. One strategy utilized by a majority of

families, at some point in the life cycle, is for the wife

to seek employment.

Patterns of Employment

Wives' employment is discontinuous in that they

usually enter and leave the work force several times over

the life cycle. These intervals of employment and

nonemployment form lifetime employment patterns. A few

such patterns have been identified among wives in the United

States (Elder and Rockwell, 1967), while a fairly extensive

classification of Australian wives' employment patterns

has been compiled (Young, 1978). One of the purposes of

the present research was to create an exhaustive typology

of the employment patterns of married women in the United


The employment of married women may be patterned in

several ways. In each of the three early life cycle stages,

wives may be employed either continuously or sporadically,

or they may remain at home. These three options, at each

stage, result in the twenty-seven possible employment patterns

shown in Figure 3-1. Some of these strategies for combining

home and market work are probably more frequently used than

others, while some probably represent empty cells.

Current knowledge of female employment suggests that

eight patterns may be popular ones. 1) While most women

are employed between marriage and first birth, many never

return to paid employment after their first child is born

(COO). This pattern will be called "traditional" (TRADITNL).

2) Others who are employed during this first life cycle

stage choose to remain at home during their childbearing

years and return to the work force when their children are

old enough to be cared for by others or to attend school (COC).

This second pattern is an interrupted one (INTERUP).

3) Some wives work continuously throughout their adult

lives, taking time off only for childbirths (CCC). Contin-

uous lifetime employment will be labeled "double-track"

(DOUBLTRK). 4) Others are employed sporadically entering and

leaving the work force during every life cycle state (SSS).

This is an in-and-out pattern (INANDOUT). 5) Some women do

not work outside of the home between marriage and first birth,

and a portion of these women are never employed (OOO/NEVEREMP).

6) Of those who later enter the work force, some commence

paid employment after their first child is born and proceed

to work on a continuous basis (OCC). These workers will be

labeled "double-track after first birth" (DTRKAB1). 7) An-

other group enters the work force following their first

birth and works sporadically thereafter (OSS). This is an

in-and-out pattern after first birth (INOUTABl). 8) Other

women delay entry until their children are near school age

(OOC), often beginning paid employment in midlife (MIDLIFE).

These eight patterns are labeled in Figure 3-1.

The two existing studies of wives' employment patterns

have investigated the relationships between family socio-

economic characteristics and separate worklife patterns

(Elder and Rockwell, 1967; Young, 1978). That research has

facilitated preliminary understanding of the association

between family characteristics and the employment timing

pattern utilized to balance home and market work. Yet

many of the patterns are closely related; and we need to

further specify the factors related to the probability

that a woman's worklife follows one pattern instead of


Employment Pattern Contrasts

Three issues are important in studying wives' employ-

ment. First and most basic is the question of why some

wives are never-employed, while most work for pay at some

point in the family life cycle. The second issue involves

the discontinuity of most female employment. That is, what

are the factors inducing or facilitating the continuous

employment of a fraction of wives and the discontinuous

employment of the majority? The third question involves

the factors related to employment before, during, and after

childbearing. Each of these issues can be addressed by

studying the effects of a set of independent variables on

contrasts of related employment patterns.

Never-Employed versus Ever-Employed Wives

Assessing the effects of a set of family characteristics

on a contrast of all ever-employed wives with the never-

employed group addresses the most general question of why

some women are never employed outside of the home, while

most work for pay at some point in the family life cycle.

Most research dealing with the factors related to working

or not working has used cross-sectional data on wives'

employment status at the survey date. Factors positively

related to wife's employment in those studies include wife's

education, premarital work experience and wage potential,

market demand for female labor, husband's education and

his rate of unemployment, length of the marriage/first birth

interval and sex role modernity. Factors negatively related

to wife's employment in cross-sectional research include

husband's income and education, family income adequacy,

wife's fertility and the spacing of births, frequency of

long-distance migration and rural residence. However, the

use of survey week employment status as a measure of life-

time employment improperly assumes that a woman's labor

force activity in one week adequately portrays her work

pattern over time. Wives who are employed at the survey

date may have been out of the work force during an earlier

life cycle stage and/or may withdraw for an interval in

the future. Research that has focused on lifetime employ-

ment has found that older cohorts and wives with less than

twelve years of schooling were most likely never to have

been employed (Sweet, 1972a Young, 1978).

Continuous versus Discontinuous Workers

Contrasting the effects of family characteristics

on continuous lifetime workers and sporadic lifetime workers

should indicate some of the inducements and facilitators

of permanent work force attachemnt. Wives who are employed

in every life cycle stage can be divided into two groups.

One group works every year from marriage through child-

bearing, as well as after all their children are in school

(DOUBLTRK). These women leave their jobs only long enough

to give birth and recover. The employment of the second

group also spans the three life cycle stages, but these

wives are not employed year after year. Instead, they are

employed sporadically, working one or several years on and

one or several years off (INANDOUT). The work force entry

and withdrawal of the sporadically employed is not tied in

any obvious way to their life cycle stage, but may be,

instead, a response to periodic increases in the family's

need for income. As such, their employment may be less

due to career aspirations or work commitment than is the

employment of continuous workers. Wives who are employed

sporadically in all three life cycle stages may differ

from continuously employed wives in terms of their frequency

of geographical mobility, as well. Previous research has

indicated that family migration significantly decreases

the probability of wives' continued employment (Long, 1974).

Employment Over the Life Cycle

Ever-employed wives time their employment in different

ways. Both of the patterns involved in the previous contrast

were characterized by at least some employment in every

life cycle stage. Yet many wives are not employed in

every stage. Some are not employed between marriage and

first birth, while many remain at home during the second

(childbearing) stage. Wives employed before, but not during

childbearing may reenter paid employment after childbearing

or may remain at home permanently. Wives who have never

been employed may enter the work force for the first time

in mid-life.

The present analysis uses employment pattern contrasts

to investigate the factors related to employment before,

during, and after childbearing. In general, wives who

work are most likely to be employed during the first life

cycle stage, when role conflict and husband's income are

relatively low. Yet not all ever-employed wives begin the

family life cycle in the work force. The only study with

specific findings relevant to first stage employment is one

by Young (1978). She found that women who had a short inter-

val between marriage and first birth were not likely to be

employed during that stage, but tended to enter the work

force at a later date. To further investigate the factors

related to employment before childbearing, one can construct

a contrast of employment patterns which differ only in

regards to first stage employment. That is, wives employed

continuously since marriage can be contrasted with wives

employed only since first birth.

Previous research has linked work during childbearing

to both high and low wife's education, wide birth spacing,

greater family size, and younger cohorts (Bowers and Hastings,

1970; Mott, 1972; Terry, 1975; Young, 1978). Although some

studies indicate wife's education is independent of employ-

ment during the childbearing years (Leibowitz, 1975; Young,

1978), others have linked employment during this stage

to both very high and low education (Sweet, 1972a; Waite,

1980). To further investigate the factors related to employ-

ment during childbearing, one can construct a contrast of

employment patterns which differ only in regards to second

stage employment. That is, wives who worked before and

after but not during childbearing can be contrasted with

wives who worked continuously in every stage.

Third stage employment may represent a return to the

work force or an initial entry. Two contrasts are needed

to study employment after childbearing, because the factors

related to the work force reentry of previously employed

wives may differ from the factors related to initial employ-

ment in mid-life. The literature provides little information

on either type of third stage employment. Cogan and Berger

(1978) found both length of birth spacing and number of

children to be negatively related to years of work experi-

ence. They also found significant interaction between

the two variables, such that the effect of number of child-

ren on years worked was greater, the wider the spacing of

those births. Since wives who enter or reenter the work

force after childbearing presumably have more work experi-

ence than wives who do not, the above findings would lead

one to expect closer birth spacing and fewer children among

wives who worked in stage three. To further investigate

the factors related to employment after childbearing, one

can construct contrasts of employment patterns which differ

only in regards to employment in stage three. To address

work force reentry, wives employed only before childbearing

can be contrasted with wives employed before and after child-

bearing. To study initial employment in mid-life, never-

employed wives can be contrasted with wives who entered

the work force for the first time in stage three.


Life Cycle Stage

Second Third Pattern Name









Figure 3-1. Possible Employment Patterns
(0 = out, S = sporadic, C = continuous)


Notes to Chapter III

The contrasts chosen here are not the only ones
that could have been used to address the three issues,
but they were the ones most directly relevant to the

In this study, childbearing is assumed to be over
when the youngest child has reached his/her third birth-
day (see Chapter IV for further discussion and justifica-
tion of this definition .

Causality, in this relationship, probably goes in
both directions. That is, some wives willingly with-
draw from the work force at first birth because they
desire a large family; and they space their births widely
to maximize the time spent with each preschooler. Other
wives who withdraw from paid employment at the start of
childbearing may intend to return to work, but may be
prevented from doing so by unplanned additional births
and/or subfecundity which results in wider-than-planned
birth spacing.


The data used in the present research are from the

National Longitudinal Survey of the Labor Market Experi-

ences of Older Women (NLS). This survey was desinged

by the Ohio State University and the United States Bureau

of the Census, and funded by the United States Department

of Labor. The survey of older women is comprised of

8 waves to date, beginning with a comprehensive inter-

view in 1967 and including a mail survey in 1968, three

subsequent interviews in 1969, 1971, and 1972, telephone

surveys in 1974 and 1976 and a final interview in 1977.

However, only the first six waves were available for use

in this study.

The sample of women, aged 30 to 44 in 1967, was drawn

using a multi-stage probability sample of 235 areas

representing every state and the District of Columbia.

The sample areas were selected by grouping all counties

into approximately k,900 primary sampling units from

which 235 strata were formed. These strata were relatively

homogeneous as to socioeconomic characteristics. One

primary sampling unit was selected from each stratum

and a probability sample of households was drawn from

each selected unit to represent the civilian non-institu-

tional population. A sampling ratio of greater than 3:1

was used to insure reliable estimates for blacks.

In the initial survey, 5,083 women were interviewed.

Weights for respondents in 1967 were adjusted to account

for persons not interviewed in that year due to refusal,

absence or nonavailability. In addition, the sample pro-

portions were adjusted with respect to the residence, age,

color and sex characteristics of the population as a whole.

These adjustments were made by carrying the 1960 Census

data forward to account for population aging, mortality

and cross-national migration. Sample attrition was small;

approximately 87 percent of the original respondents were

reinterviewed in 1974.

The data contain fairly complete work history informa-

tion including several measures of current employment and

work experience both prior to and since the initial survey,

as well as background socioeconomic characteristics. The

data also include variables related to marital events and

childbearing, and contain fairly complete household record

information. Overall, the NLS data set on older women is

well-suited to the construction and analysis of lifetime

employment patterns.

The present analysis is concerned with identifying

and analyzing the employment of married women as it is

patterned around the family life cycle. Hence, only

married women whose worklives could be captured within

a family life cycle framework are included. From the

original sample of 5,083 women, all women were deleted

who were not married at all of the six interview dates

and who did not remain married to the same man throughout

the survey years. Those wives who had been married more

than once were deleted, as well, since a past disruption

of the family life cycle may have had some effect on a
woman's employment pattern. In addition, all women were

deleted who did not have at least one child by 1967. Wives

who had their first child after the survey began were not

included in the final sample because their likelihood of

completing childbearing by 1974 was probably low. To

the extent that wives having their first child after age

thirty cluster together in terms of background character-

istics, such as high education, the results will be biased

against these women and the effects of such variables as

college education will be underestimated. Similarly, to

the extent that wives who begin childbearing this late

have very small families, long marriage to first birth

intervals, or close birth spacing, the effects of these

variables will also be underestimated. However, these

biases should be negligible since the number of women

having their first child during the survey years was low

(N = 21). Finally, all women who had given birth less

than nine months after marriage were deleted since a

premarital pregnancy was probably a timing failure and,

thus, an abberation from the "typical" order of life

cycle events assumed in constructing the typology of

employment patterns. This decision necessitated the

deletion of black women from the sample, since approxi-

mately half had had premarital pregnancies. Although

half of all black women did not have premarital preg-

nancies, the inclusion of the latter portion would have

created a censoring problem. That is, black women who

experienced premarital pregnancies were probably different
in other ways from black women who did not. Thus, the

latter group would not have been representative of black

wives as a whole.

The final sample includes 1519 white women, aged

30 to 44 in 1967, who remained in their first

marriage through 1974, and had at least one child by 1967

and no premarital pregnancies. These women are an atypical

group, in that they have followed the normative order of

life cycle events and have had no marital disruption. The

sample was limited to these women in order to simplify

interpretation of the results. Wives' employment is pro-

bably related to the ordering of life cycle events as well

as to their timing. The inclusion of wives whose life

cycles followed different patterns (e.g., wives who had

a premarital births) would have necessitated the inclusion

of an additional variable to tap the order of events.

Marital disruption is probably related to employment

timing as well. Because the sample used here is a restricted

one, the findings should be generalized only to other wives

whose life cycle followed the normative pattern and who

experienced no marital disruption. The employment timing

of other groups of wives is equally important and should

be addressed in future research.

Constructing a Typology of Employment Patterns

Objective of Typology

With the exception of the never-employed group, the

employment patterns discussed in Chapter III are all forms

of wives' (usually discontinuous) employment. Previous

studies of wives' employment have been attentive to its

discontinuity, but few have dealt with the forms this

discontinuity takes among women. Those that have studied

employment patterns either have done so in a limited fashion

or have focused on populations other than the United States.

The typology of employment patterns constructed here

addresses both these concerns, in that it is a near

exhaustive classification of a national sample of American

wives according to their lifetime patterns of employment

and non-employment. As such, it demonstrates, in retro-

spect, how these cohorts of women timed their employment

and indicates the proportions of wives who used the various

employment timing strategies.

Identifying Employment During Three Life Cycle Stages

The typology was constructed by tapping employment

during the first three stages of the family life cycle.3

The first stage begins when a woman marries and ends at her

first birth. Since all of the women in the final sample

were married and had at least one child by 1967, identifi-

cation of this stage was not a problem. The second stage

begins when the first child is born and ends when the last

child is three years old. The youngest child in 1974 is

assumed to be the last child. This assumption seems justified

because by 1974, the women were aged 37 to 51 and most there-

fore had completed childbearing (Ryder and Westoff, 1971)4

The third life cycle stage begins with the end of child-

bearing (i.e., when the last child turned three) and is

open-ended, in that all employment from this point to 1974

is combined, regardless of the age of the youngest child

by 1974. The disadvantage of leaving the third stage open-

ended is that information is lost for those women actually

in later life cycle stages by the end of the survey. Many

of the women surveyed here had entered the fourth or fifth

cycle stage by 1974. However, the literature suggests

that the effects of family characteristics on wives' employ-

ment are greatest in early life cycle stages (Becker, 1965;

Gove, 1973; Gronau, 1973a; Long, 1974; Terry, 1975; Waite,

1976, 1980). During these early stages, the demands on the

woman as mother are at their highest and family resources

(i.e., income and assets) are relatively low. Life cycle

squeezes do occur later on, especially as children begin

to leave home for college; but the extension of the classifi-

cation to encompass later stages would have necessitated

the deletion of wives not yet through those stages. Further

reduction in the sample size was rejected in favor of classi-

fying the employment of a maximum number of wives.

Wives with children in 1967 responded to several ques-

tions regarding their employment during the first life cycle

stage. In the initial interview, women were asked when

their last job ended, relative to their years of marriage

and first birth. Possible responses were "before marriage,"

"between marriage and first birth," or "other" (after first

birth and/or currently employed in 1967). All those who

worked between marriage and first birth were asked in how

many of these years they worked six months or more. Employ-

ment during the second stage is tapped as follows. Women

who had a child by 1967 were asked in how many years follow-

ing their first birth they were employed six months or more.

Since this question was asked of respondents in 1967, the

response covers employment throughout the childbearing years

among women whose last child was three by 1967. For women

whose last child turned three between 1967 and 1974, the

number of years given in the above response is combined

with the number of years worked 26 or more weeks in each

of the years between 1967 and the year the last child was

three. Employment in the third stage is measured from the

year the last child was three. If this occurred prior to

1967, employment is measured both prior to and since 1967.

If childbearing ended in 1967 or later, employment is mea-

sured from whatever year the last child turned three.

Identifying Employment Patterns

The typology was created by making each employment

pattern a dummy variable, coded one if the woman's employ-

ment history fit the pattern and zero otherwise. Following

the definition of each employment category, wives coded one

for that pattern were removed from further classification

procedures. Respondents remaining after all employment

patterns were defined were considered unclassifiable. As

will be seen in Chapter V, the number of such women is

small. The analysis was begun by classifying respondents

using eight of the possible twenty-seven patterns (see

Figure 3-1). These included the TRADITNL, INTERUP, DOUBLTRK,


These patterns were chosen on the assumption that the cell

frequencies would be high.

Wives are classified as "never-employed" (NEVEREMP)

if they reported no appreciable employment during any of

the three life cycle stages discussed above. Wives are

coded one on NEVEREMP if they satisfied all of the following

conditions: 1) they were not employed six or more months

during any year between marriage and first birth; 2) they

were not employed six or more months during any year between

first birth and 1967; and 3) they did not work twenty-six

weeks or more during any of the survey years, from 1967

through 1974. This procedure allows women to work less than

six months in any year and still be classified as never-

employed. The problem had no solution, however, since all

questions concerning employment prior to 1967 were phrased

in terms of years worked six or more months during an interval.

Wives who did not satisfy all of the above conditions are

coded zero, while respondents missing on any of the relevant

variables are missing on NEVEREMP.

Women are classified as "traditional" (TRADITNL) if

they were employed in every year between marriage and first

birth, but reported no employment thereafter. Thus, wives

who worked six months or more in any year between their

first birth and 1974 are coded zero on TRADITNL. Again,

respondents with missing values on any of the relevant

variables are missing for this employment pattern.

Wives who were employed before and after, but not

during childbearing are labeled as "interrupted" (INTERUP).

In order to be included in this category, a woman had to

be employed six months or more in every year from marriage

to first birth. She also must not have reported employment

between her first birth and the year her last child turned

three. Thus, if her youngest was three before 1967, the

year she returned to work after her first birht must have

been later than the year her last child was three. If the

youngest turned three after 1966, no employment could be

reported between 1967 and that year; and the variable tapping

years worked after the first birth had to have a zero value,

since it was measured in 1967. Finally, women classified

as "interrupted" had to be employed on a continuing basis

subsequent to her last child's third birthday. The minimum

criteria in this regard was employment in 1974. That is,

the respondent must have returned to the work force by the

last wave of the survey. Women with children younger than

three in 1974 had not yet completed the childbearing stage,

as defined here. Those women who worked from marriage to

first birth and were employed subsequent to 1974 (i.e.,

after their youngest child turned three) are classified as

"traditional," based on their available employment history.

However, there are few such women overall (N = 16), so mis-

classification of this sort is minimal.

Wives who worked continuously from marriage through

1974 are labeled "double-track" (DOUBLTRK), following the

convention used by Elder and Rockwell (1967). DOUBLTRK

wives were employed six or more months in every year during

each life cycle stage and were out of the work force for

no longer than one year following childbirths. Inclusion

in this category, then, necessitates employment between

marriage and first birth. In addition, the year the respon-

dent returned to paid employment must have been the year of

first birth or the following year. Because the data do not

include the years of birth of children other than the first,

the measurement of inter-birth employment was complex. The

first step involved ascertaining whether wives were employed in

every non-birth year of the childbearing stage. A variable

was created which indicated the number of years between first

birth and 1967. From this variable was subtracted the number

of children present in 1967 minus one (for the year of birth

of the first child), to indicate the number of years available

for work (i.e., the number of years not occupied by child-

bearing) between first birth and 1967. If the actual number

of years worked between first birth and 1967 was greater

than or equal to the number of years available for work,

another variable (PRE67) was coded one. If the woman was

employed fewer years than she had available for work, she

was coded zero on PRE67. For the years beginning with 1967,

a different procedure was used. For each year (1967, 1968,

1969, 1971, 1972, and 1974) a variable was created which

equaled one if the respondent was employed and/or gave birth

in that year, and equaled zero if she did neither. In order

to be classified as "double-track" a woman had to be coded

one on each of these variables (PRE67 and each of the employ-

ment/childbirth variables from 1967 to 1974). If she spent

as little as one year neither working nor childbearing, she

is not considered continuously employed.

The INANDOUT pattern captures the work lives of women

who were employed during each of the three life cycles stages,

but were not continuously employed in any stage. Thus,

these respondents reported employment of six months or more

in at least one year between marriage and first birth, but

not in every year. Similarly, they reported some full-year

employment during childbearing and some subsequently.

Wives who were not employed between marriage and first

birth, but who worked continuously thereafter, are labeled

"DTRKABl" (double-track after first birth). The computations

involved in creating this variable are similar to those

described above for double-track wives. The exception is

that women in the DTRKAB1 pattern could not have reported

employment between their marriage and first birth (i.e.,

they must not have worked six months of more in any year

during this interval). The wives in this category entered

paid employment in the year of their first birth or during

the following year.

Similarly, wives who were not employed between marriage

and first birth, but who worked intermittently thereafter,

are included in the INOUTAB1 group. These wives reported

no full-year employment prior to their first birth, but worked

six months or more in at least one year, but not in all years,

of stages two and three.

Some wives remained at home during both the first and

second stages and entered the work force only after their

last child was three. This employment pattern is labeled

MIDLIFE. In order to be included in the midlife category,

women could not have worked six months or more in any year

during stage one. In addition, they must not have reported

significant employment during any year before their child-

bearing ended. If a woman's youngest child had turned three

by 1967, inclusion in this category requires that the number

of years she worked between her first birth and 1967 be less

than or equal to the difference between 1967 and the year

of that child's third birthday. If her youngest child was

three after 1967, her inclusion requires that she did not

work between first birth and 1967 and that no employment

be reported after 1967 and prior to the year of that child's

third birthday. Subsequent to the end of childbearing,

wives in the MIDLIFE category reported employment, of twenty-

six weeks or more, in every year through 1974.

As will be seen in Chapter V, the employment of approx-

imately 85 percent of the respondents is captured using

the above eight patterns. Most of the remaining respondents

were as yet unclassified either due to the continuous employ-

ment restrictions of the categories (i.e., requiring women

who worked during an interval to work six or more months

in every year), or as a result of the restriction, in the

double-track categories (DOUBLTRK and DTRKAB1), requiring

wives to reenter the work force one year or less after

their first birth. That is, some wives worked continuously

in one interval and intermittently during others; while some

women worked continuously but remained at home more than one

year after their first birth. Except for those restrictions,

most the the unclassified women would have been either


As a result, several other employment patterns were

tested. The addition of four categories resulted in the

classification of almost 97 percent of the sample. One of

these (INOUTINT) captures the worklives of women who worked both

before and after (but not during) childbearing, but were

not continuously employed following childbearing. Thus,

respondents in this seventh pattern reported employment

of six months or more in every year between marriage and

first birth, but not in every year following childbearing.

Except for their sporadic employment in the third life

cycle stage, INOUTINT wives would have been in the "inter-

rupted" group. The eighth and ninth patterns are comprised

of women who would have been classified as DOUBLTRK or

DTRKAB1 did those patterns not require near immediate work

force (re)entry after first birth. Wives who worked con-

tinuously in every stage but remained at home for more than

one year after first birth are labeled DELDTRK (delayed

double-track). Wives who were not employed between marriage

and first birth but who worked continuously beginning more

than one year after first birth (yet still during the child-

bearing stage) are called DDTRKAB1 (delayed double-track

after first birth). A small number of women were employed

six months or more in only one year during the three life

cycle stages covered by the typology. These wives (ONE-

YEAR) are grouped together regardless of the stage at which

the one year of employment occurred.

Overall, twelve employment patterns were identified

in the manner described above. The distribution of wives

over these categories and their frequencies by cohort will

be discussed in Chapter V.

Method of Analysis

Objective of Analysis

The objective of the typology of employment patterns

is descriptive. That is, it seeks to identify the various

employment timing patterns of a sample of women who, for

the most part, were beyond their childbearing years. The

goal of the analysis is to assess the effects of a set of

independent variables on the probability that a woman's

worklife followed one employment pattern rather than another.

The independent variables include background, family and

fertility characterisitcs. The objective here is to gain

further insight into the factors related to the various

ways of timing employment and childbearing.

Independent Variables

Wives' lifetime employment may be associated with

several diverse factors. Economic pressures may encourage

them to work outside of the home on a semi-or fully continuous

basis, while an early and/or rapid succession of births

may induce women to remain at home. Previous research

has indicated that wives' employment is negatively related

to husband's income. Number of children and spacing of

births have been negatively related to wives' employment,

as well. One study found that employment before childbearing

is associated with the length of that stage. Other studies

have shown that frequent migration disrupts wives' employ-

ment continuity, and the discontinuously employed wives

in these data may have experienced frequent geographical

mobility, as well. Wife's education and age at marriage

have been positively related to employment in past research,

and the employment of wives who are highly educated and/or

who married late may have followed a pattern that allowed

them to have both family and career rewards. Background

characteristics, such as race, religious affiliation,

rural/urban background and social class of origin also have

been associated with wives' employment. Aspects of the

labor market surely influence employment timing, as well.

A women who seeks employment during a specific life cycle

stage may be unable to find it, due to a high unemployment

rate or a low availability of typically female jobs (during

that period and/or in her area of residence), and/or due

to a lack of job skills on her part. Low wages in available

jobs may induce her to refuse potential employment, as well.

However, not all of these factors are included in

the present analysis of employment patterns. As discussed

previously, race is not included, since substantive and

methodological factors necessitated the exculsion of black

women. Religious affiliation would have been an interesting

and probably valuable factor, but it is not available in

the data. Social class of origin and parents' ethnicity

were either available or estimable, but are excluded to

simplify the model. Although premarital labor force

participation has been shown, in previous research, to

be a reliable predictor of post-marital employment (Groat

et al., 1976; Smith-Lovin and Tickamyer, 1978; Waite, 1980),

this variable is not included in the present model. Both

previous and subsequent employment are endogenous in that

both are probably determined by similar variables. There-

fore, models which use previous employment status to pre-

dict subsequent employment are misspecified. In addition,

employment prior to marriage is omitted from the present

research in order to minimize multicollinearity problems.

A measure of premarital work experience was tested in

preliminary analysis and was found to be highly correlated

with marital timing (r = .403, p = .0001). Income variables

(husband's income and family income) are excluded, as well.

Although previous research has indicated that husband's

transitory income is a good predictor of wife's employment,

its inclusion in the present study was inhibited by the

lack of necessary data. The NLS data include husband's

annual income for the years 1966-69, 1971-72 and 1974.

Yet in the present research, employment is measured as a

lifetime pattern, beginning as early as 1939 and extending

through 1974. The majority of employment, for most women,

occurred prior to 1966 and income variables for these

earlier years are unavailable. Labor market variables

(unemployment rate and industry mix for the labor market

of current residence) are excluded for similar reasons.

This information is available only for the years 1967

through 1974; and much of a woman's employment pattern

was completed by then.

Cohort. The NLS data on older women are comprised

of women born from 1923-1937. Although this time span

is not a long one, it is divisible into two historically

significant decades. The group born in the twenties grew

up amidst the Depression and married during the years

surrounding World War Two. Many of these women were among

those who entered the labor market during the wartime

shortage of male workers. Their children later comprised

the baby boom of the early fifties, making this cohort,

the group of young mothers most indoctrinated in the ethos

of domesticity of that decade. Women born during the

depression years were still children or adolescents during

the war. They did not marry until the 1950's were well

underway and, for the most part, bore their children during

the prosperity of the early sixties. These two groups of

women are not far apart in terms of age. However, labor

force participation among married women has been increasing

steadily since 1940. Current research shows significant

changes from 1940 to 1960 in the effects of several factors

influencing wives to work (Waite, 1976). These findings

suggest that women born in these two decades may differ

in their aggregate distribution across participation

patterns. It is also possible that the effects of the

other variables on participation may vary across the two

groups. Inclusion of the dichotomous variable COHORT

in the model will facilitate investigation of both types

of change.

Marital timing. Marital timing has been shown in

previous research to have important consequences for wives'

employment patterns and for the timing and extent of

fertility (Elder and Rockwell, 1967). Marital timing,

like the timing of most life events, is normatively defined.

Early and late marriers are distinguishable from those who

marry on time; and these definitions may vary by cohort.

However, in these data, the mean age at marriage did not

differ significantly by cohort (20.86 years old for older

women and 20.14 years old for younger women). The standard

deviation was slightly higher among the older cohort (3.44

versus 3.07). As a result, the same definitions of "early,"

"on time," and "late" marriage were used for both cohorts.

Three years was subtracted from the mean age at marriage

for the younger women to mark early marriage; and

three years was added to the mean age at marriage for older

women to identify late marriage. Thus, early marriers are

defined as women marrying at 17 or younger. Late marriers

are women marrying at 24 or older. Women who married between

the ages of 17 and 24 are considered "on time." For pur-

poses of analysis, two dummy variables were created. Res-

pondents who married early are coded one on EARLYMAR, while

women who married on time are coded zero. Wives who

married late are coded one on LATEMAR, while the on time

group is coded zero.

Timing of first birth. The length of the interval

between marriage and first birth is used, rather than age

at first birth, as the measure of timing of first birth.

This decision was made for two reasons. First, it seems

desirable to include as many aspects of fertility in the

model as possible without creating the problem of multi-

collinearity. Age at first birth is significantly correlated

with both marital timing (r = .69, p = .0001) and length

of the marriage/first birth interval (r = .30, p = .0001).

However, the latter two variables are virtually uncorrelated
(r = -.03, p = .2908). Since marital timing and length

of the marriage/first birth interval seem potentially able

to contribute more information to the model than age at

first birth alone, the former two variables were chosen.

The variable "timing of first birth" was created in much

the same way as marital timing. That is, the number of

months between marriage and first birth was calculated

for each respondent in the sample, and the standard devia-

tion was added to and subtracted from the mean. A short

interval is defined as one of 12 months or less. If more

than 36 months intervened between marriage and first

birth, the interval is a long one. Those women for whom

13 to 36 months elapsed between the two events are defined

as having had an average interval. Again, two dummy vari-

ables were created from the three interval lengths. Wives

who had a short interval are contrasted with wives who

had an average interval, by coding the former group one

on SHORTINT and the latter group zero. Women who had a

long interval are coded one on LONGINT, while those having
an average interval are coded zero.

Number of children. Tne NLS data provide information

on the number of children who have left home before 1967.

The data also contain variables tapping the number of

newborns, toddlers, preschoolers, elementary-aged children,

teens and children over 18 in the home in each wave from

1967 through 1972. These latter variables were summed for

1967 and 1972; and a new variable was created, representing

the maximum number of children in either year. Thus, if

more children were present in 1972 than were present in

1967 (i.e., if family size grew between the two waves), the

1972 value was assigned to the new variable. Conversely,

if the 1967 value was greater than the 1972 value (i.e.,

if children left home between 1967 and 1972), the new

variable represented the number of children present in

1967. If family size was constant from 1967 to 1972, the

1967 value was used. However, in order to measure the

actual maximum family size, the maximum number of children

present during the survey years had to be added to the

number of children who had left the home before 1967. This

computation provided the continuous variable TOTKIDS,

representing the total number of children ever residing

in the respondent's home.

Spacing of births. Spacing is defined here as the

average number of years between births. This conceptuali-

zation is used because the data do not provide the actual

birth years of children other than the first. Average

birth spacing was computed using the ratio, number of years

between first and last births divided by the maximum number

of children in the home (TOTKIDS). Close spacing is defined

as less than 2 years between births. Four years or more

between births is considered wide spacing; while the 2-4

year span is defined as average birth spacing. Spacing of

births has no meaning for wives with only one birth. In

order to incorporate them in the analysis, wives with one

child were assigned a value of "average" on birth spacing.10

Two dummy variables were created for purposes of analysis.

Respondents with close spacing are coded one on CLOSE,

while women having average spacing are coded zero. Similarly,

wives having wide spacing are coded one on WIDE, while

respondents with average spacing are coded zero.

Education. Wife's education has been used as a proxy

for several other phenomena, including propensity for

employment, potential wage and employment opportunity.

Educationprobably reflects propensity for work only under

certain conditions (i.e., when a woman is highly educated

and/or when her family economic situation is stable enough

to leave the work option open). Thus, wife's education is

not intended solely as a measure of the desire to work in

the present model. Rather, it is included as a general

indicator of wage potential and employment opportunity,

since the use of these latter types of variables was pro-

hibited by the absence of early measures in the data. Of

course, education is not a precise measure of wages and

labor market opportunities, but it was the best measure

available for this purpose. In spite of the high correla-

tion between wife's and husband's educations (r = .60,

p = .0001), both variables are included in the model. The

net effects of husband's and wife's education on wife's

employment pattern are assumed to reflect somewhat different

phenomena. While wife's education taps her potential wage

and employment opportunity, husband's education is a proxy
for his permanent income. All of these factors are

important in modeling wives' work patterns, since wife's

wage potential and employment opportunity are positively

related to her employment and husband's income is negatively

related to employment. Each education variable was converted

to a set of six dummy variables, in order to allow for

nonlinearity. The dummy variables are: less than 9

years of schooling, 9-11 years of schooling, some college,

college graduate, and graduate education. High school grad-

uates was the omitted group. The net effects of the husband's

and wife's education variables are believed to be the best

available measures of the various inducements to wife's


Marital mobility. Wives' employment is often dependent

on the relationship between family need and family resources.

However, need is a relative concept and probably has much

to do with individuals' expectations based on both early

socialization and the overall economic situation surrounding

their upbringing (Easterlin, 1968). Wives who marry "up"

may be less likely to work, net of other factors, because

their standard of living expectations are more likely to

be met by available family resources. On the other hand,

wives who are downwardly mobile through marriage may be

more likely to be employed, controlling for other variables,

since their expectations concerning a reasonable standard

of living are higher. Marital mobility is measured here

by comparing the wife's father's occupational prestige

when she was fifteen to her husband's occupational prestige

in 1966 (i.e., when the wife was 29-43). The NLS data

contain the Duncan Socioeconomic Index of Occupation for

both father and husband. This index was divided into five

categories of approximately twenty points each. Typical

occupations in each category are as follows: 1) farm

laborers, truck drivers, and barbers; 2) welders, plumbers

and mechanics; 3) insurance agents, technicians and photo-

graphers; 4) public school teachers, accountants and chiro-

practors; and 5) lawyers, professors and physicians.

Father's occupation was divided by husband's occupation and

a three level variable was created. If the ratio was less

than one, husband's occupational prestige was higher; if

it was greater than one, father's occupational prestige

was higher. A ratio of one indicated equal status occupa-

tions. Two dummy variables were used, with equal status

being the omitted group.

Frequency of geographical mobility. Previous research

has indicated the relevance of migration to wives' employment

(Long, 1974; Sandell, 1976). In general, migration may

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EYCSV1514_J4CQM2 INGEST_TIME 2012-03-08T18:55:20Z PACKAGE AA00003886_00001