First Marriage Dissolution, Divorce, and Remarriage:United States

Matthew D. Bramlett, Ph.D., and William D. Mosher, Ph.D., Division of Vital Statistics

Abstract

Objectives

—This report presents national estimates of the duration of ﬁrst andsecond marriages for women 15–44 years of age in 1995. National estimates of theprobability of divorce given separation and of the probability of remarriage givendivorce are also presented.

Methods

—The lifetable estimates are based on a nationally representativesample of women 15–44 years of age in the United States in 1995 from theNational Survey of Family Growth, Cycle 5.

Results

—One fifth of first marriages end within 5 years and one third endwithin 10 years. First marriages of teenagers disrupt faster than the first marriagesof women who were ages 20 years and older at marriage. First marriages of black non-Hispanic women dissolve at a faster rate: 47 percent end within 10 yearscompared with 34 percent for Hispanic, 32 percent for white non-Hispanic, and20 percent for Asian non-Hispanic women. Virtually all separations among whitenon-Hispanic women (98 percent) end in divorce within 6 years, compared withonly 80 percent of separations among Hispanic women and 72 percent of separations among black non-Hispanic women.Women under age 25 years at divorce are more likely to remarry than womenat least age 25 years at divorce. White non-Hispanic and Hispanic women are muchmore likely to remarry than black non-Hispanic women. White non-Hispanicwomen are slightly more likely than Hispanic women to remarry. The data suggestthat women who remarry before age 25 years are more likely to experience asecond marital disruption than women who remarry at ages older than 25 years,although the difference is only significant at late marital durations. Black non-Hispanic remarriages are more likely to disrupt than Hispanic or white non-Hispanicremarriages.

Keywords:

marriage

separation

divorce

remarriage

Introduction

Marriage is associated with avariety of positive outcomes, anddissolution of marriage is associatedwith negative outcomes for men,women, and their children. A fullanalysis of the beneﬁts of marriage toeither children or spouses is beyond thescope of this report, but this brief introduction should serve to highlightthe importance of the data described inthis report. This report releases estimatesof the patterns of marriage, divorce, andremarriage in the United States as of 1995 by several important demographiccharacteristics. A later report will showmore detailed estimates by a widevariety of other characteristics.Compared with unmarried people,married men and women tend to havelower mortality, less risky behavior,more monitoring of health, morecompliance with medical regimens,higher sexual frequency, more

Acknowledgments

The 1995 National Survey of Family Growth was jointly planned and funded primarily by the National Center for Health Statistics (NCHS), theNational Institute for Child Health and Human Development (NICHD), the Office of Population Affairs (OPA), and the National Center for HIV,STD and TB Prevention (NCHSTP), with additional support from the Children’s Bureau.The authors gratefully acknowledge the technical assistance of Wayne E. Johnson, Ph.D., of the NCHS Office of Research and Methodology(ORM) for assistance in estimating standard errors of the statistics in this report.The authors also gratefully acknowledge the helpful review and comments of Christine Bachrach, Ph.D. (NICHD). This report was edited byKlaudia Cox and typeset by Annette F. Holman of the Publications Branch, Division of Data Services.

Number 323

May 31, 2001

satisfaction with their sexual lives, more

ﬁ

nancial savings, and higher wages(1

–

3). The differences between marriedand unmarried people may re

ﬂ

ect acausal effect of marriage or a selectioneffect (healthier people may be morelikely than others to

ﬁ

nd mates and tomarry). Research has suggested that thebene

ﬁ

ts of marriage may be partiallydue to a selection effect and partiallydue to true bene

ﬁ

ts to be gained frombeing married as opposed to beingunmarried (3,4). A lower mortality risk among those who are married has beenshown to persist even after health inearly adulthood was controlled. Thissuggests that at least part of the bene

ﬁ

tof being married is not the result of selection (4).Compared with married individuals,divorcees exhibit lower levels of psychological well-being, more healthproblems and greater risk of mortality,more social isolation, less satisfying sexlives, more negative life events, greaterlevels of depression and alcohol use,and lower levels of happiness andself-acceptance (5). The economicconsequences of divorce can be severefor women. Most often, children remainwith the mother after divorce; the lossof the ex-husband

’

s income often resultsin a severe loss of income per capita(6,7). For men, the retention of incomecombined with decreased family sizemay actually result in an increase in hisnew household

’

s income per capita(6,8).Adverse outcomes accrue tochildren of divorce and children raisedin single-parent families. Although notall single-parent families are the resultof divorce and not all divorced mothersremain single, virtually all children of divorce spend some time in a single-parent household until the motherremarries. Even when the mother doesremarry, studies suggest that children instepfamilies are similar to children insingle-parent families: both groups of children do worse than children livingwith two parents in terms of academicachievement, depression, and behaviorproblems such as drug and alcoholabuse, premarital sexual intercourse, andbeing arrested (9).Compared with two-parent families,single-parent families demonstrate lowerlevels of parental involvement in schoolactivities and lower student achievement(10). Children raised in single-parentfamilies are more likely to drop out of high school, have lower grades andattendance while in school, and are lesslikely to attend and graduate fromcollege than children raised in two-parent families (11). They are morelikely to be out of school andunemployed and are more likely tobecome single parents than childrenraised in two-parent families (11).Studies have found that compared withchildren in two-parent families, childrenof divorce score lower on measures of self-concept, social competence,conduct, psychological adjustment, andlong-term health (5).The positive health bene

ﬁ

ts of marriage and the negative consequencesof divorce illustrate the importance of examining trends and differentials in thepatterns of marriage and divorce overtime.

Trends and differences inmarriage and divorce

In the United States in the secondhalf of the twentieth century, theproportion of people

’

s lives spent inmarriage declined due to postponementof marriage to later ages, greaterincidence of never marrying, and higherrates of divorce. The increase innonmarital cohabiting has alsocontributed to the decline in theproportion of people

’

s lives spent inmarriage. Increasing rates of cohabitation have largely offsetdecreasing rates of marriage (12,13).The proportion of time spent inmarriage has varied across demographicsubgroups. Since 1950, the maritalpatterns of white and black Americanshave diverged considerably. About91 percent of white women born in the1950

’

s are estimated to marry at sometime in their lives, compared with only75 percent of black women born in the1950

’

s (12). Black married couples aremore likely to break up than whitemarried couples, and black divorcees areless likely to remarry than whitedivorcees (12).The degree of attachment tomarriage among black Americans issimilar to that of white Americans asmeasured by attitudes towards marriage(14,15). One prominent explanationoffered by some researchers for thelower proportion of time spent inmarriage among black Americans is theidea of a

‘‘

marriage squeeze,

’’

in whichthe

‘‘

marriageable pool

’’

of black men islow due to high rates of joblessness,incarceration, and mortality (16

–

18).Employed men are more likely thanunemployed men to marry (19).In addition to race and employmentstatus, other characteristics of individuals that have been found to berelated to the probability of marriageinclude education and earnings (20),intact status of family of origin, andparents

’

educational levels (21). Othercharacteristics of individuals related tothe probability of

divorce

include age atmarriage, education, birth cohort (22),religion, marriage cohort, fertility statusat marriage (23), premarital cohabitation(24), and premarital sexual activity (25).Other characteristics related to theprobability of

remarriage

includeeducation and age at divorce (26) andpresence of children from priormarriages (9,22).The lower economic prospects of less-educated young men has beenhypothesized to decrease the probabilityof marriage. The increasing economicindependence of women has also beenhypothesized to decrease the probabilityof marriage, although recent evidencesuggests that the increasing economicindependence of women may actuallyincrease the probability of marriagebecause earnings and employment maymake either partner an attractivepotential spouse (16,20). Marriagemarket conditions may also play a rolein that the probability of divorce ishigher in areas with large numbers of economically attractive potentialalternate partners (16).A full analysis of all the individualand community-level characteristicsassociated with marriage and divorce isbeyond the scope of this report. Thisreport releases estimates of the patternsof marriage, divorce, and remarriage inthe United States as of 1995 by a fewimportant demographic characteristics.A later report will show more detailed

Advance Data No. 323

May 31, 2001

estimates by a wide variety of othercharacteristics.

Methods

Data

The national estimates of marriageand divorce patterns in this report arebased on data from the 1995 NationalSurvey of Family Growth (NSFG).Cycle 5 of the NSFG, conducted by theCenters for Disease Control andPrevention

’

s National Center for HealthStatistics (NCHS), was based on amulti-stage probability sample of thecivilian noninstitutionalized populationof women in the United States. Thiscycle yielded estimates that arerepresentative of women 15

–

44 years of age in 1995. Between January andOctober 1995, in-home computer-assisted personal interviews (CAPI)were conducted with 10,847 women, of whom 1,553 were Hispanic women,6,483 were white non-Hispanic women,2,446 were black non-Hispanic women,and 365 were women of other races andethnic origins. The overall response ratewas 79 percent (27).The sample list for the 1995 NSFGwas selected from households thatresponded to the 1993 National HealthInterview Survey (NHIS). Black andHispanic women were sampled at higherrates than were other women. Samplingweights account for differentialprobabilities of sample selection and fornonresponse and are adjusted to agreewith control totals by age, race, parity,and marital status provided by the U.S.Census Bureau. The 10,847 women inthe 1995 NSFG represent the 60 millionwomen 15

–

44 years of age in thecivilian noninstitutionalized populationof the United States in 1995. Onaverage, each woman in the 1995 NSFGrepresents about 5,500 women in thepopulation, although sample weightsvary considerably from this averagevalue depending on the respondent

’

srace, age, and Hispanic ethnicity; theresponse rate for similar women; andother factors (27,28). See theTechnicalnotesfor additional information.The 1995 NSFG collected completeretrospective histories of each woman

’

sexperiences with marriage and divorce,including the beginning and endingdates of each marriage and the outcomeof each marriage (separation, divorce,widowhood) (29). Given these data, theprobabilities of

ﬁ

rst marriage disruption,of divorce after separation, of remarriage after divorce, and of secondmarriage disruption can be estimatedusing life table techniques.Previous analyses of marriage anddivorce based on vital statistics havecomputed and presented rates of marriage and divorce (30, 31). Rates aresnapshots of data limited to speci

ﬁ

ctime points. The life table analysis inthis report takes a life-cycle approach toestimate the probabilities that:(a) a

ﬁ

rst marriage will end inseparation or divorce,(b) a separation will result indivorce,(c) a divorce from a

ﬁ

rst marriagewill be followed by remarriage, and(d) a second marriage will end inseparation or divorce.Previous analysis of divorce andremarriage based on cycle 4 of theNational Survey of Family Growth useda measure of the cumulative proportionof marriages disrupted as of interview todescribe the phenomena (32). Animprovement over a rate, this statisticapproximates the estimates that life tableanalysis provides. However, it is only asingle measure of the cumulativeproportion at the time of interview; lifetables provide estimates of cumulativeproportions at every time point in thelife course of a marriage.

Life tables

The life table is a tool thatdemographers most often use to studymortality, but it is often applied to thestudy of marital stability. In studyingmortality, the cohort life table is asummary of the mortality history of agiven cohort from birth to death andrequires data on the longevity of allcohort members, a span of more than100 years. As a result, the period lifetable is typically used as a model of what would happen to a given cohort if the age-speci

ﬁ

c death rates from acertain point in time were to remain

ﬁ

xed for the duration of the cohort

’

s life(33,34).As members of the cohort age, theyare subjected to the age-speci

ﬁ

c deathrates of successive age categories in thelife table. At each interval, the age-speci

ﬁ

c death rate for that interval isused to calculate how many members of the cohort die during that interval. Thatnumber of deaths is subtracted from thecount of cohort members, and the resultis the number of cohort members whosurvive and are counted in the nextinterval. Eventually, the last age intervalis reached and the last cohort membersdie. One overall measure of longevity isthe proportion who survive to speci

ﬁ

cages (33). Survivor curves can beplotted that show the proportion of thecohort surviving to each successive agecategory (34,35).To apply life table analysis to thestudy of marital stability, the cohort of people is replaced with a cohort of marriages; age is replaced by maritalduration, and death is replaced byseparation or divorce. In addition, oneother issue must be addressed. Thesample of women is limited to ages15

–

44 years, so the marriage historiesare incomplete. For respondents whosemarriage has not yet ended as of thetime of the interview, the end date of the marriage is unknown. Therefore, theduration of the marriage is unknown andis referred to in statistical literature as

‘‘

censored.

’’

Life table procedures allowfor the simultaneous analysis of bothcomplete and incomplete marriagehistories (22).Life table analysis can handlecensored cases by keeping such cases inthe analysis as long as they are at risk of disruption and then dropping themout once the risk is unknown (36). Forexample, when calculating theproportion of marriages that dissolve ineach duration interval, a marriage thathas existed for 24 months and is stillintact at the time of the interview wouldremain in the denominator for eachduration interval until 24 months of duration is reached, at which point thecase would no longer be used in thecalculations.Widowhood removes a marriagefrom the risk of dissolution. The lengthof time that the marriage would haveendured intact if the husband had notdied is unknown, so cases of

Advance Data No. 323

May 31, 2001