The Continuing Retreat of Marriage: Figures from Marital Status Life Tables for United States Females, 2000–2005 and 2005–2010

Abstract

The lack of vital statistics data on American marriage and divorce has made it difficult to follow post-1995 changes in marriage behavior. Here, a new approach, Rate Estimation from Adjacent Populations (REAP), is used in conjunction with vital statistics mortality data and recently released divorce data from the American Community Survey to construct marital status life tables that reflect the lifetime implications of observed or inferred rates of marriage, divorce, and mortality. Methodologically, the chapter sets forth the features of the REAP approach. Substantively, the analysis shows that the retreat from marriage is continuing, but unevenly. The probability that a woman ever marries has fallen to 80 %, and the average age at first marriage has risen, slightly, to 27 years. At the same time, the probability of divorce appears to be holding steady at about 43–46 %. The results suggest that the great transformation of the American family has not yet run its course.

Appendix: Data Assembly, Adjustment, and Application

1.1 Mortality Data

The mortality rates used in both the 2000–2005 and 2005–2010 marital status life tables were taken from the published United States Life Tables, 2005 (Arias et al. 2010, Table 3). Death rates for females aged x to x + n, were calculated as life table deaths, _n d _x , divided by life table person-years, _n L _x .

Life table person-year values were used to project the total female population in 2000 to 2005 and then to 2010. That was done to prevent distortions from in and out migration. The projected female populations, by age, were proportionally allocated to the four marital statuses using the data populations described below.

Mortality by marital status was obtained from National Vital Statistics Reports, Vol. 55, No. 19 (August 21, 2007), specifically from Table 25, “Number of deaths, death rates, and age-adjusted death rates for ages 15 and over, by marital status and sex: United States, 2004.” Age 15 was taken as the minimum age at marriage. Husbands were assumed, on average, to be 2.5 years older than their wives, e.g. the death of a husband age 50 produced a widow aged 47.5. Five year age groups up to age 85 were generated from the published 10-year age groups up to age 75 by linear interpolation.

1.2 Divorce Data

The divorce data originated from the American Community Survey (ACS) conducted by the U.S. Bureau of the Census (Elliott et al. 2010). The survey inquired as to whether the survey respondent was divorced within the past 12 months. From that ACS data, Kennedy and Ruggles (2014) calculated age-specific female divorce rates for the years 2008–2010, and those rates are used for the 2005–2010 tables. The divorce rates for the 2000–2005 interval were found by linear interpolation to the year 2002.5 using the 1995 divorce rates underlying Schoen and Standish (2001) and the 2008–2010 KR rates.

In 2006, the Census Bureau conducted an evaluation of the ACS marital history questions (O’Connell et al. 2007). In Table 35, the Evaluation Report stated that 7.8 % of the 176 respondents who reported a divorce in the last 12 months did not actually have a divorce decree finalized during that time. The highest incidence of false reports, 13.5 %, was for persons aged 55–64. The Evaluation Report nonetheless concluded that the ACS produced reasonable estimates when compared to estimates from vital statistics. (The Evaluation Report did not come to a similar conclusion with respect to the retrospective marriage question.)

Elliott, Simmons, and Lewis (2010) also performed a review of results from the ACS retrospective marital history questions, reaching a similar conclusion with respect to the divorce data. Elliott et al. (2010) also reported some state level comparisons, including data for Delaware and New Hampshire, states they described as having high quality vital statistics that did not differ statistically from the ACS numbers. However, in 2007, 3215 divorce decrees were filed in Delaware, while the 2008 ACS estimated 4318 divorces to women. In New Hampshire, there were 4981 divorce decrees filed while the 2008 ACS estimated 5059 women divorcing in the previous 12 months. The ACS overestimate for New Hampshire is only 1.6 %, but for Delaware, it was 34.3 %. There is thus a real possibility that divorce is overreported in the ACS.

1.3 U.S. Population Data by Marital Status for 2000, 2005, and 2010

The Rate Estimation from Adjacent Populations (REAP) procedure requires counts of the female population, by marital status, at ages 0–4, 5–9, …, 80–84, and 85+ for the years 2000, 2005, and 2010. The female population for all marital statuses combined for 2000 came from Census 2000, Summary File 1, Table 2, and for 2010 from the corresponding 2010 Summary File. For 2005, the aggregate populations were found from Current Population Survey, Annual Social and Economic Supplement, 2005, Table 1.1. However, the 2005 and 2010 population data were only used at ages 0–4 in 2005 and 2010, respectively. Populations at higher ages were found by 5-year survival, as described above.

The 2000, 2005, and 2010 female population by marital status, based on U.S. Census Bureau data, were taken from tables titled “Population, 5-year Age Groups, by Sex, Country, Marital Status, Age and Year” that appeared on the United Nations Economic Commission for Europe (UNECE) website, www.unece.org/pxweb/dialog/Print.asp?Matrix=005_GEPOPop5YearMaSta_r&timeid. For 2000, 10-year age groups at ages 55 and above were broken into 5-year age groups using linear interpolation and data from Kreider and Simmons (2003, Table 1). For 2005, linear interpolation was used augmented by data from Table 1, “Marital Status of the Population 15 Years and Over by Age and Sex: 2005”, U.S. Census Bureau, Current Population Survey, Annual Social and Economic Supplement, 2005. For 2010, linear interpolation was again used on the UNECE data.

1.4 Finalizing the Rates of Interstate Transfer

The REAP procedure was applied to the 2000 and survived 2005 and 2010 populations to produce estimates of the 2000–2005 and 2005–2010 marriage rates. The initial estimates, subject to errors in the census enumerations, survey weights, interpolations, marital status reporting, and other assumptions, were not fully satisfactory. The estimated first marriage rates appeared quite accurate up to age 37.5, by which age the great majority of first marriages were entered. For higher ages, m_sm in both years was assumed to be (0.75) times the rates used in the 1995 U.S. Female marital status life tables of Schoen and Standish (2001), a reasonable approximation of the post-1995 decline in marriage (cf. the NCHS state level data in the table “Marriage Rates by State: 1990, 1995, and 1999–2011” downloaded 8/5/2014 from cdc.gov/nchs/data/dvs/ marriage_rates_90_95_99-11.pdf).

For widow remarriage, the REAP estimated rates were unusable. In both years those rates were set equal to (0.75) times the Schoen and Standish (2001) m_wm rates at every age. The error here is hard to assess, but widowhood plays a small role at the ages of principal interest here. The estimated remarriage from divorce rates in both years were quite reasonable up to age 47.5, after which they were assumed to be (0.75) times the m_vm rates underlying Schoen and Standish (2001). The bulk of remarriage from divorce events occur before age 47.5, so the error introduced by the fractional reduction should be minor.

1.5 Constructing the Marital Status Life Tables

The multistate life tables were calculated by the linear method (Schoen 1988, Chap. 4). Under age 15, values followed the U.S. Life Tables for 2005. Special procedures were needed for the open-ended 87.5+ age group. There, modified flow equations implemented matrix Eq. (4.21) of Schoen (1988, p. 70) to find the PP_j(87.5+) values from known mortality and estimated marriage rates.