Honey I Love You,but … Investigating the Causes of the Earnings Penalty of Being a Tied-migrant Military Spouse

Introduction

Mincer identified a tied-mover as the spouse who “moves along with the other even though his (or her) private calculus dictates staying.” ¹ While the underlying theory on tied-migrants is gender neutral, empirical studies generally have identified wives as the tied-movers even though the dynamics within civilian families makes such an association problematic. ² Rotation policies within the military, contrariwise, leave little doubt about which spouse is the tied-mover. Thus, the civilian wives of servicemen (military wives) have proved to be a useful, quasi-experimental sample for studying the labor market outcomes of tied-movers. Over the last three-and-half decades women have entered the military in sufficient numbers so that the civilian husbands of servicewomen (military husbands) now provide a similar, quasi-experimental sample for analyzing the labor force characteristics and outcomes of men who are tied-movers. ³ Using the 2000 Census Sample Edited Detail File (SEDF), we analyzed data on a sample of men married to women on active-duty (military husbands) and compare them to the husbands of working women not on active-duty (civilian husbands). ⁴ Furthermore, we corroborate the findings from earlier studies on military wives, which found that these women faced a shortfall relative to their civilian counterparts. ⁵

Our analysis extends these earlier studies in two ways. First, the data provide more robust estimates of the differences observed in the labor force characteristics and outcomes for the spouses of active-duty military personnel and their civilian counterparts. Second, the methodology used expands on earlier studies by accounting for selection bias (or selectivity) when investigating the gap in annual personal earnings between military and civilian spouses.

Overall, our findings suggest a similar earnings gap exists for military husbands as has been found for military wives. This not only supports Mincer’s hypothesis that the cost of being a tied-mover tends to be gender neutral but also adds a new dimension to Sandell’s discussion on women and family migration decisions. ⁶ We do not dismiss, however, the possibility that differences in labor market outcomes experienced by military spouses may be due to factors other than their tied-mover status, such as motivations arising from assortative mating ⁷ or a person’s sense of self. ⁸ Rather, we concur with the broad argument put forth by Hosek et al. that the lower earnings experienced by military spouses result from an interplay between tastes, preferences, and opportunities as well as a confluence of latent factors associated with the notion of tied migration. ⁹

The article proceeds as follows. The relevant tied-migration literature is reviewed, focusing on the earlier mentioned studies and a brief discussion of the literature on decomposition. Next, the data source and variables used in this study are presented. This is followed by a discussion of the relevant summary statistics of the husbands and wives of active-duty military personnel compared to their civilian counterparts. We then present our multivariate findings for husbands and wives. A discussion of the results from our decomposition analysis is next. A concluding section discusses our findings within the context of the earlier literature as well as suggests possible policy implications.

For the remainder of this article, our discussion will rely on the following definitions for the respective spouse categories. ¹⁰

Military husband: A man, not in the military, married to a woman between 18–46 years old who reported being on active-duty in the United States Armed Forces in 1999.

Civilian husband: A man, not in the military, married to a woman between the 18–45 years old who reported not being on active-duty in the United States Armed Forces and was working full-time in 1999.

Military wife: A woman, not in the military, married to a man between the ages of 18–46 years old who reported being on active-duty in the United States Armed Forces in 1999.

Civilian wife: A woman, not in the military, married to a man between the ages of 18–45 years old who reported not being on active-duty in the United States Armed Forces and working full-time in 1999.

Literature Review

Recognizing the impact that family migration had on the earnings of wives, Grossman and Hayghe called attention to some differences in the labor market participation of military and civilian wives. ¹¹ These authors did not include earnings comparisons. Similarly, Schwartz, Wood, and Griffith studied Army wives and, although they provided no direct comparison to civilian wives, concluded that their findings were consistent with previous studies for the usual set of determinants of labor force outcomes for married women. ¹²

Empirical work on earnings issues began in 1983 when Jacobson’s research article for the Center for Naval Analyses utilized a one-period job search model that led him to conclude that “high work-intensity” wives would have earned 45 percent more over three years had they not been involved in a permanent change-of-station move. ¹³ He found that 90 percent of the earnings loss occurred in the first year due to a reduction in weeks worked and hypothesized a negative earnings effect due to lower access to specific training. In 1990, Hogan presented an empirical model that focused on how spousal labor market outcomes affected the likelihood of reenlistment. ¹⁴ He concluded, based on a simulation, that increasing the average tour length by 12 months reduced the expected costs imposed on the nonmember spouse’s labor market opportunities by about 35 percent and increased the average reenlistment probability by about 3 percent. A couple years later, further empirical analysis by Payne, Warner, and Little, lead them to suggest that differences in labor market outcomes for military and civilian wives can address several important, testable human capital consequences of tied migration. In particular, they argued that, by providing disincentives for human capital investment and by inhibiting the full use of previously acquired skills, the frequent, regular moves of military wives might reduce returns to education, experience, and occupational choice. They concluded that the military’s rotation policies placed severe restrictions on the labor supply of military wives and lowered their returns to human capital investment. Moreover, they concluded that a three-year rotation policy would reduce a military wife’s earnings by 40 percent compared to a six-year rotation policy.

A decade later, Booth argued that the earnings differentials observed for military wives could not be completely attributed to human capital penalties, but are best understood as occurring from a confluence of factors that includes not only their status as tied-migrants but also the effects of structural factors within the local labor markets. ¹⁵ He estimated that women workers pay an earnings penalty of about 5 percent per each increase of 10 percent in the local concentration of active-duty military employees. Cooke and Speirs provided an analysis using 1990 census data of the effects of migration on the employment and hours worked for both civilian men and women. While they did not focus on the annual personal earnings for either gender, they concluded that the wives of active-duty military personnel experienced reductions of employment and hours worked per week of 10 percent and 4 percent, respectively. For the husbands of active-duty military personnel, while the results were not statistically significant, both the direction and the magnitude of the change were similar to those of wives, a 6 percent decline in employment and a five-hour decline in hours worked per week.

Over the last decade and half, the Reseach ANd Development Corporation (RAND) has researched labor market outcomes for military spouses. Hosek et al. presented a comprehensive study of military wives, including some insights that underlie arguments of the present analysis. Their analysis “purposely compares military wives with civilian wives … to see how military wives contribute to family income and to learn whether military wives’ labor participation and wage trends lagged and deviated from those of civilian wives.” ¹⁶ Although they noted some widening at that time, they found that military families earned on average about $10,500 less annually than civilian families over the 1987–1999 period (about 21 percent) and argued that about half the difference ($5,643) could be attributed to the lower earnings of military wives. ¹⁷ They traced this difference to several factors, factors that we believe also apply to military husbands and ones that we elaborate upon in our concluding discussion.

In another RAND study, Harrell et al. conducted extensive personal interviews with the civilian wives of men on active-duty and analyzed the 1990 census data using the propensity scoring technique (or look-alike analysis) to develop a rather complete employment picture of military wives and concluded that being a military spouse negatively impacted employment outcomes. Lim et al. updated the Harrell et al. study and provided a comparison of the demographic and labor force characteristics of military and civilian spouses based upon the 2000 Census Public Use Microdata Sample (PUMS) data. As one would expect, these authors’ findings are similar to those found in the Harrell et al., with the exception that they estimated an hourly wage difference based upon the propensity scoring methodology.

Little and Hisnanick extended the analysis of tied-mover effects on military spouses within the context of family and spousal characteristics in order to examine annual earnings losses. Although their focus was primarily on military husbands, these authors found that the annual earnings for military families, on average, are less than their civilian counterparts and that the family’s civilian spouse accounts for roughly 50 percent of the family’s lower earnings. ¹⁸ Military wives and husbands earned about 50 percent and 70 percent, of what their civilian counterparts earned. The authors concluded that, although the age-earnings trends for both male and female tied-movers are similar, military husbands experienced a lower earnings equilibrium than their civilian peers. This shortfall, they argued, could be attributed to less attachment to the work force because of frequent moves, the unpredictable, yet rigid, work schedules of their active-duty spouse, and the greater likelihood of young children in the household.

To investigate the implications of the above findings in a more formal manner, we employ a Blinder and Oaxaca decomposition method. ¹⁹ This method involves separately estimating the (natural logarithm) wage equation for two groups of workers and then decomposing their earnings difference into an explained portion due to human capital endowments and an unexplained portion, attributable to labor market conditions and/or employer preferences. The standard approach to decomposition is to adopt one of the groups as the standard or norm. Differences in the mean characteristics of the (two) groups are weighted by the estimated coefficients for the standard wage equation and summed to obtain the explained and unexplained components of the observed earnings differential.

In order to estimate a labor market outcome, such as wages or hours worked, the accepted practice is to account for sample selection bias to better understand differences in the wages and/or earnings between groups of individuals. Selectivity bias can exist at several stages of the employment process. For example, over their life-course individuals must decide if and when to seek employment and, if they do, what occupation to pursue. If sample selection of either type is present, the ordinary least squares estimation of the wage equation model yields biased and inconsistent estimators. As discussed in the literature, selectivity can complicate the application and interpretation of the observed gaps in earning when the decomposition technique is used. ²⁰ This compounds the usual issue of whether or not part of the unexplained portion in the wage gap should be labeled as employer discrimination or if it is, as argued by Booth, due to overcrowding (probably more likely for wives than husbands) in labor markets with a large military presence. The unexplained gap can always be interpreted as an estimate of adverse labor market outcomes, but the question is whether, or not, this estimate is an unbiased and consistent estimate, which in turn depends on how one characterizes the wage structure in a labor market absent tastes for discrimination by economic agents.

Employment Search

The conventional approach associated with modeling the job search process, which we believe illuminates the case of military spouses, stresses a matching process wherein both the potential worker and the employer face imperfect information. The applicant envisions a frequency distribution of opportunities offering different pay together with a self-imposed lower bound for accepting a job, the reservation wage. The reservation wage reflects the value of alternative uses of one’s time, that is, the opportunity cost of taking a job. Searching may alter the applicant’s view of the distribution of pay and the reservation wage. Finding the perfect match, one that offers full utilization of one’s human capital and hence is towards the right-hand portion of the distribution, has a relatively low probability and may require a costly and time-consuming search as well as high opportunity cost due to foregone earnings.

If expected job tenure is to be short, both the actual and the opportunity cost may recommend the quick acceptance of a low-paying offer, especially when jobs are scarce in the area of one’s expertise. Moreover, a downward revision of the reservation wage may result or, if offers are below the reservation amount, a decision may be made not to work. Those with larger amounts of occupation-specific human capital, frequently acquired through longer job tenure or additional degrees, may find fewer satisfactory opportunities in some locations, resulting in underemployment thus lowering expectations and/or the atrophying of those skills if employment is not found. Should work not be found in a reasonable length of time, expectations of another move may discourage further search causing the spouse to drop out of the labor market altogether.

An employer’s information on a potential employee is always imperfect but for the military spouse it is compounded by the expectations of short job tenure suggesting the distribution of offers to them will reflect these circumstances and cause the distribution to have a leftward bias. Further, networks that normally provide job information to both the employer and the employee may be lacking for both parties when a military spouse is involved in job search. So, from both the supply and demand sides of the matching process, military spouses may be relegated to lower earnings and weaker attachment to the job market than otherwise would be the norm.

Data and Definitions

To investigate the earnings effect of tied migration as it relates to military husbands and wives, data from the 2000 Census of population and housing, the Sample Edited Detail File (SEDF) were analyzed. ²¹ The SEDF consists of social, economic, and housing characteristics compiled from a sample of approximately nineteen million housing units (about one in six households) that received the census 2000 long-form questionnaire. These data allowed identification of subsamples of civilian males and females whose spouses were in the military (our tied-movers), along with families of civilian males and females whose spouse worked full-time and was not in the military. To obtain the military spouse samples, we selected those married men and women between the ages of eighteen and forty-six with a detailed industry code of Armed Forces, excluding those in the Reserves or National Guard. In identifying our sample comparison groups, the same methods in constructing our military samples were used. For example, our comparison sample for military wives consisted of those civilian women married to men, between the ages of eighteen and forty-six, working full-time, and were also civilians (not on active-duty). Similarly, the comparison sample for military husbands consisted of those civilian men married to women, between the ages of eighteen and forty-six, working full-time, and were also civilians (not on active duty). ²²

Descriptive Statistics

Before commenting on the selected demographic and labor force characteristics of military husbands and wives compared to their civilian counterparts, we present Figure 1, which compares their annual personal earnings. The age-earnings profiles by gender are similar in that they have the expected shape and are smooth except for an earnings spike for military husbands around age forty. One plausible explanation for this observation is that husbands married to women service members are different from husbands married to civilian women in that around half of military husbands are former military. The dips seen for military husbands in Figure 1 could be result of a newly entered group of civilian workers who recently completed the end of active service and start the transition to the civilian labor force by looking for a job. According to the data available from the Office of the Deputy Undersecretary of Defense (DUSD), 47.3 percent of married military women are married to other service members. ²⁵ So, unlike military men, many military women meet their current and future spouses in the military while the men are serving also.

OPEN IN VIEWER

Figure 1.

The average annual earnings by age groups of military husbands and wives compared to civilian husbands and wives.

The patterns observed in Figure 1 capture the essence of, but do not provide an explanation for, the differences observed between the husbands and wives by the respective age groupings. These patterns, however, do provide the rationale for validating whether the magnitude of the differences from the decomposition components is similar across gender.

Selected Characteristics of Military and Civilian Husbands

Compared to civilian husbands, military husbands have less work experience due to their younger age. ²⁶ They are more likely to be a minority, have moved recently, and be the father of young children. Military husbands are more likely to be employed by the federal government, but less likely to be self-employed or employed in the private sector. Hours worked by military husbands, on average, are comparable to those of their civilian counterparts (see Table 1).

OPEN IN VIEWER

Table 1.

Characteristics of Military and Civilian Husbands and Wives: 1999.

	Husbands		Wives
Characteristics of respondent	Militarya	Civilianb	Militarya	Civilianb
Total # (weighted)	33,700	12,120,299	445,082	19,841,564
Labor market outcomes
Average personal earningsc	$30,658 ($24,661–$36,655)	$43,588 ($43,170–$44,007)	$14,166 ($14,081–$14,250)	$22,298 ($22,281–$22,315)
Average years of work experience	14.5 years. (10.33–18.67)	19.0 years. (14.40–23.60)	12.0 years. (11.95–12.05)	15.4 years. (15.39–15.41)
Average usual hours worked per weekd	44.4 (38.34–50.46)	45.2 (39.80–50.60)	36.1 (30.78–41.42)	36.7 (31.45–41.95)
% Unemployed	4.0 (3.61–4.39)	2.0 (1.99–2.01)	4.8 (4.68–4.92)	2.1 (2.09–2.11)
Demographic characteristics (% of total)
Age
18–21 years	1.1 (0.83–1.37)	0.5 (0.49–0.51)	11.4 (11.17–11.63)	2.9 (2.88–2.92)
22–26 years	15.0 (14.07–15.93)	5.4 (5.37–5.43)	15.8 (15.54–16.06)	8.4 (8.37–8.43)
27–31 years	22.6 (21.51–23.69)	14.9 (14.85–14.95)	23.3 (23.00–23.60)	19.4 (19.36–19.44)
32–36 years	20.5 (19.45–21.55)	18.9 (18.85–18.95)	23.5 (23.20–23.80)	24.1 (24.05–24.15)
37–41 years	18.8 (17.78–19.82)	22.4 (22.34–22.50)	16.8 (16.53–17.07)	25.5 (24.45–25.55)
42–46 years	13.1 (12.22–13.98)	21.9 (21.84–21.96)	7.3 (7.11–7.49)	16.1 (16.06–16.14)
47+ years	8.9 (8.16–9.67)	16.0 (16.05–16.15)	1.8 (1.70–1.90)	3.9 (3.88–3.92)
Race and ethnicity
White	64.2 (62.57–65.83)	82.3 (82.58–82.62)	74.1 (73.69–74.51	84.6 (84.55–84.65)
White, non-Hispanic	60.3 (58.63–61.97)	77.7 (77.68–77.72)	69.1 (68.67–69.53)	79.3 (79.24–79.36
Black	29.3 (27.75–30.85)	9.6 (9.58–9.62)	13.6 (13.28–13.92)	6.2 (6.17–6.23)
American Indian/Alaska Native	2.0 (1.52–2.48)	1.2 (1.19–1.21)	1.7 (1.58–1.82)	1.1 (1.0–1.1)
Asian	2.9 (2.33–3.47)	4.1 (4.09–4.11)	7.5 (7.25–7.75)	4.7 (4.67–4.73)
Hispanic	7.9 (6.98–7.82)	8.3 (8.28–8.32)	10.2 (9.92–10.48)	9.7 (9.66–9.74)
Educational attainment
Less than high school graduate	4.3 (3.89–4.71)	10.2 (10.19–10.21)	5.8 (5.67–5.93)	8.9 (8.88–8.92)
High school graduate	28.7 (27.79–29.61)	28.2 (28.18–28.22)	25.2 (24.96–25.44)	25.5 (25. 46–25.54)
Some college or associates’ degree	48.4 (47.40–49.40)	32.9 (32.88–32.92)	45.1 (44.83–45.37)	34.3 (34.26–34.34)
College graduate	14.8 (14.09–15.51)	19.3 (19.29–19.31)	18.6 (18.39–18.81)	22.5 (22.47–22.53)
Postgraduate degree	8.1 (7.55–8.65)	9.3 (9.29–9.31)	5.3 (5.18–5.42)	8.8 (8.78–8.82)
Did you live here five years ago?
Yes, same house/apartment	15.2 (14.27–16.15)	47.8 (47.72–47.86)	11.9 (11.65–12.11)	44.8 (44.75–44.93)
No, lived in a different house/apartment in the United States	74.8 (73.66–75.94)	50.6 (50.52–50.66)	77.4 (77.10–77.70)	52.4 (52.35–52.53)
No, lived in a different house/apartment outside the United States	10.0 (9.20–10.78)	1.6 (1.60–1.64)	10.7 (10.50–10.94)	2.7 (2.69–2.75)
Labor force characteristics (% of total)
Class of worker
Reported not working in the last five yearse	2.6 (2.25–2.95)	1.6 (1.59–1.61)	11.4 (11.21–11.59)	11.6 (11.59–11.63)
Private sector	61.3 (60.23–62.37)	72.0 (71.98–72.02)	63.7 (63.41–63.99)	67.4 (67.36–67.44)
State/local government	8.1 (7.50–8.70)	10.7 (10.69–11.71)	9.6 (9.42–9.78)	11.8 (11.71–11.83)
Federal government	22.9 (21.97–23.83)	4.1 (4.09–4.11)	9.7 (9.52–9.88)	2.0 (1.99–2.01)
Self-employed	5.0 (4.52–5.48)	11.3 (11.29–11.31)	5.1 (4.97–5.23)	6.7 (6.68–6.72)
Work without pay	0.1 (0.03–0.17)	0.1 (0.1–0.1)	0.2 (0.17–0.23)	0.4 (0.39–0.41)
Occupations
Managerial/Professional	32.7 (31.67–33.73)	33.3 (33.28–33.32)	29.4 (29.12–29.68)	36.1 (36.06–36.14)
Service occupations	13.1 (12.36–13.84)	9.4 (9.39–9.41)	17.8 (17.57–18.03)	13.3 (13.27–13.33)
Sales occupations	8.1 (7.50–8.70)	10.1 (10.09–10.11)	14.7 (14.49–14.91)	9.5 (9.47–9.53)
Clerical	10.5 (9.82–11.18)	6.4 (6.39–6.41)	22.1 (21.85–22.35)	22.2 (22.16–22.24)
Farming/fishing	0.3 (0.18–0.42)	0.6 (0.60–0.60)	0.1 (0.08–0.12)	0.3 (0.30–0.30)
Construction/extraction	17.7 (16.86–18.54)	17.9 (17.89–17.91)	0.6 (0.55–0.65)	0.6 (0.59–0.61)
Production/transportation occupations	15.0 (14.21–15.79)	20.2 (20.19–20.21)	3.6 (3.49–3.71)	6.2 (9.18–6.22)
Military occupationsf	0.2 (0.1–0.3)	0.3 (0.30–0.30)	0.03 (0.02–0.04)	0.02 (0.20–0.20)
Household characteristic (% of total)
Children in family
Under 6 years only	27.6 (26.61–28.59)	14.9 (14.89–14.91)	27.1 (26.83–27.37)	21.5 (21.46–21.54)
6–17 years only	24.0 (23.06–24.94)	38.1 (38.08–38.12)	29.0 (28.72–29.28)	35.8 (35.76–35.84)
Under 6 and 6–17	16.0 (15.19–16.81)	12.4 (12.39–12.41)	19.2 (18.96–19.44)	19.1 (19.06–19.14)
No own children	32.3 (31.27–33.33)	34.5 (34.48–34.52)	24.7 (24.44–24.96)	23.6 (23.65–23.73)

Source: US Census Bureau, 2000 Census of Population and Housing, (SEDF).

Note: 90% confidence intervals are reported in parentheses.

Unweighted counts: Military husbands: 4,464; Civilian husbands of full-time working civilian wives: 1,872,924; Military wives: 60,489; Civilian wives of full-time working civilian husbands: 3,074,492.

^aA military husband (wife) is a married man (woman) not in the military (a civilian) whose wife (husband) is enlisted in the military.

^bA civilian husband (wife) is someone married to a civilian woman (man) who worked at least fifty weeks and thirty-five hours per week in 1999 and currently working (or looking for work) in 2000. This definition also includes full-time elementary, middle, and secondary teachers.

^cAverage personal earnings are for husbands (wives) who were working in 1999.

^dAverage hours worked weekly for those husbands (wives) who were working in 1999.

^eHusband (wife) reported that he (she) had not work in the last five years.

^fThe last job held by the husband (wife) (up to five years ago) was in the military, but he (she) was not in the military in 2000.

Thus, for those husbands who work, the demographic and labor force data suggest that military husbands might earn somewhat less but the earnings data reveal substantial earnings differences, about 30 percent annually. Stated otherwise, military husbands would need to earn 42 percent more in order to match civilian husbands. If one includes all husbands with no earned income at a point in time, it would require over a 46 percent increase in average annual earnings for them to equal the earnings of their civilian counterparts.

This 4 percent difference arises because military husbands report substantially less attachment to the labor market than their civilian counterparts. They are twice as likely to be unemployed and one percentage point more likely to report not working in the last five years. Further, the “without pay” category is only a tenth of a percentage for each group. If, however, these categories are added together (those with no earnings in 1999), the data suggest that military husbands are twice as likely to be in a non-earning category at a given point in time. For military husbands, the categories add to 6.7 percent, a full three percentage points higher than the civilian reference group. While three percentage points may not seem like a dramatic difference, if one reflects on what a three-percentage point difference implies in discussions of the unemployment rate, the magnitude is substantial. Because there were about 34,000 married women in the military in 1999, the data suggest that about 6,000 of their husbands had no earnings. This is about 18 percent above what would be expected if their labor force attachment were like that of civilian husbands, thus illustrating the selectivity issue.

Multivariate Analysis

A pooled regression model, accounting for sample selection, ²⁸ was estimated, first for the husbands of full-time working wives and then for the wives of full-time working husbands. The results from these models not only validate our previous discussion, but complement the conclusions of earlier studies on military wives as tied-movers, ²⁹ while also lending empirical support to Mincer’s hypothesis that the costs of being a tied-mover extends to both sexes. In order to facilitate our discussion, the models’ explanatory variables are put in three groups: variables controlling for the husband’s (wife’s) characteristics, variables controlling for household characteristics, and variables controlling for spousal characteristics.

Husband’s Pooled Regression Model Results

The individual independent variables within each of the aforementioned categories in Table 2, Husband’s, Household’s, and Wife’s Characteristics, have the expected signs and magnitudes. Note that the estimated correlation coefficient between the error terms of the selection and annual personal earnings equation (rho) is positive and significant indicating that selectivity is a concern in the model. Further, a working civilian husband’s annual personal earnings are inversely affected by his wife being in the military. Specifically, a working husband whose wife was in the military could expect to earn, on average, 42.7 percent less annually compared to his civilian counterparts, a result that is in line with our descriptive statistics. ³⁰

OPEN IN VIEWER

Table 2.

Pooled Regression Results for the Husbands of Working Wives, 1999. For the earnings equation for husbands, the dependent variable is—the natural logarithm of annual personal earnings. For the probit selection of husbands, the dependent variable is—1—was employed, 0—otherwise).

Variable	Earnings model results			Probit selection equation results
	Estimated coefficient	Standard error	Significance level	Estimated coefficient	Standard error	Significance level
Intercept	6.802	0.0267	***	1.078	0.0226	***
Rho value	0.962	0.0004	***
Husband's characteristics
Age	0.114	0.0011	***	−0.006	0.0031	***
Age squared	−0.001	0.0000	***
Race and ethnicity
White	0.100	0.0056	***	0.048	0.0081	***
Black	−0.101	0.0072	**	−0.075	0.0095	***
Hispanic	−0.045	0.0057	***	0.004	0.0083
Educational attainment
High school	0.316	0.0049	***	0.049	0.0056	***
High school plus some college	0.433	0.0049	***	0.104	0.0061	***
College	0.663	0.0057	***	0.279	0.0090	***
Postgraduate degree	0.866	0.0068	***	0.432	0.0119	***
Employer
Private sector	0.413	0.0037	***
State or local government	0.408	0.0054	***
Federal government	0.509	0.0076	***
Occupation
Managerial/professional	0.564	0.0113	***
Services	0.160	0.0118	***
Sales	0.520	0.0117	***
Clerical	0.267	0.0121	***
Construction	0.355	0.0113	***
Production/manufacturing	0.353	0.0113	***
Migration
Lived elsewhere in the United States five years ago	−0.037	0.0029	***	0.037	0.0044	***
Lived elsewhere outside the United States five years Ago	−0.328	0.0115	***	0.039	0.0151	***
Household’s characteristics
Log of non-spousal income				0.006	0.0007	***
Household lives in the midwest	−0.041	0.0039	***	0.053	0.0062	***
Household lives in the south	−0.051	0.0038	***	0.014	0.0056	**
Household lives in the west	−0.028	0.0045	***	0.073	0.0069	***
Young children present, aged 6 and under				0.021	0.0056	***
Wife's characteristics
Log of wife's annual earnings				0.021	0.0013	***
Wife's age				0.001	0.0005
Wife's educational attainment
High school				0.045	0.0064	***
High school plus some college				0.078	0.0033	***
College				0.157	0.0085	***
Postgraduate degree				0.100	0.0104	***
Wife is in the military	−0.557	0.0270	***	−0.049	0.0317	*
Sigma	1.787	0.0009	***

Source: US Census Bureau, Census 2000 (SEDF).

Note: Total number of (unweighted) observations: 1,840,573. Censored observations: 32,351. Uncensored observations: 1,840,573. Wald χ²: 65,286.9 (model goodness of fit measure). Characteristics of the representative reference husband: a non-Hispanic, other race (Asian/American Indian/Alaska Native), less than a high school education, self-employed or working without pay, in the farming and/or fishing industries, have not live elsewhere in the United States five years ago, not lived outside the United States five years ago, lived in the northeast region of the United States, did not have young children (less than six years old), and his wife has less than a high school degree and she is not active-duty military.

***Significance level, .01. **Significance level, .05. *Significance level, .10.

Wife’s Pooled Regression Model Results

Again, the signs and magnitudes of the independent variables in the wife’s equations (Table 3) are in line with expectations. As with the husband’s findings, the estimated correlation coefficient between the error terms of the selection and (rho) is positive suggesting that selectivity is a concern. Note that a civilian wife’s annual personal earnings are negatively affected by her husband being in the military. Specifically, a working wife whose husband is in the military could expect, on average, to earn 53.1 percent less than her civilian counterparts. ³¹ This negative effect on her earnings may be attributable to child care issues associated with the presence of young children and the resulting lower labor force participation of mothers. ³² These findings suggest that, while both military husbands and wives experience a gap in their annual earnings relative to their civilian counterparts, military wives experience a larger shortfall.

OPEN IN VIEWER

Table 3.

Pooled Regression Results for the Wives of Working Husbands, 1999. For the earnings equation for husbands, the dependent variable is—the natural logarithm of annual personal earnings. For the probit selection of husbands, the dependent variable is—1 = was employed, 0 = otherwise.).

Variable	Earnings model results			Probit selection equation
	Estimated coefficient	Standard error	Significance level	Estimated coefficient	Standard error	Significance level
Intercept	6.598	0.0585	***	0.760	0.0126	***
Rho value	0.974	0.0001	***
Wife's characteristics
Age	0.086	0.0027	***	−0.009	0.0002	***
Age squared	−0.001	0.0000	***
Race and ethnicity
White	−0.291	0.0137	***	0.095	0.0045	***
Black	0.156	0.0195	***	0.057	0.0072	***
Hispanic	0.196	0.0140	***	−0.080	0.0046	***
Educational attainment
High school	0.007	0.0140		0.086	0.0046	***
High school plus some college	0.001	0.0138		0.156	0.0046	***
College	0.217	0.0150	***	0.194	0.0052	***
Postgraduate degree	0.449	0.0180	***	0.265	0.0069	***
Employer
Private sector	0.046	0.0072	***
State or local government	0.007	0.0098
Federal government	0.251	0.0166	***
Occupation
Managerial/professional	0.953	0.0245	***
Services	0.238	0.0242	***
Sales	0.584	0.0246	***
Clerical	0.639	0.0243	***
Construction	0.084	0.0345	***
Production/manufacturing	0.557	0.0248	***
Migration
Lived elsewhere in the United States five years ago	−0.276	0.0075	***	0.102	0.0027	***
Lived elsewhere outside the United States five years ago	−0.484	0.0223	***	0.013	0.0070	*
Household's characteristics
Log of non-spousal income				0.016	0.0004	***
Household lives in the Midwest	−0.034	0.0102	***	0.009	0.0037	**
Household lives in the south	−0.107	0.0100	***	0.010	0.0035	**
Household lives in the west	−0.115	0.0115	***	0.043	0.0040	***
Young children, aged six and under				−0.085	0.0018	***
Husband's characteristics
Log of husband's personal earnings				0.008	0.0007	***
Husband's age				−0.003	0.0002	***
Husband's educational attainment
High school				0.015	0.0028	***
High school plus some college				0.022	0.0029	***
College				−0.005	0.0033
Postgraduate degree				−0.050	0.0038	***
Husband is in the military	−0.758	0.0154	***	−0.117	0.0054	***
Sigma	3.458	0.0026	***

Source: US Census Bureau, 2000 Census of Population and Housing, (SEDF).

Note: Total number of (unweighted) observations: 1,085,158—Because of software constraints, in order to more efficiently run the wage equation model for wives, every third observation was selected from the original sample of over three million observations. Censored observations: 123,547. Uncensored observations: 961,611. Wald χ²: 104,086 (model goodness-of-fit measure). Characteristics of the representative reference wife: a non-Hispanic, other race (Asian/American Indian/Alaska Native), less than a high school education, self-employed or working without pay, in the farming and/or fishing industries, have not live elsewhere in the United States five years ago, not lived outside the United States five years ago, lived in the northeast region of the United States, did not have young children (less than 6 years old), and her husband has less than a high school degree and he is not active-duty military.

***Significance level, .01. **Significance level, .05. *Significance level, .10.

Decomposing the Earnings Difference

Our samples consisted of two groups of earners, men married to full-time working wives and women married to full-time working husbands, with a further distinction made between those with a spouse in the military. Our findings that military husbands (Table 2) and wives (Table 3) could expect to have annual earnings that are 42.7 percent and 53.1 percent less than their civilian counterparts suggest an attempt to identify the underlying causes for these differences. The relevant literature has argued that any attempt to identify (or decompose) factors that contribute to an earnings difference should correct for selectivity, in this case an individual’s decision to be employed. ³³

Before discussing our decomposition results, a brief review of the decomposition components is in order. The selectivity component accounts for that portion of the annual earnings gap associated with differences between the wage setting determinants of the two groups. Such differences weigh heavily on an individual’s decision regarding whether or not working (for pay) is worth the forgone opportunity. ³⁴ The human capital component accounts for that portion of the annual earnings gap due to differences in the observed characteristics between the two groups. These characteristics include things such as differences in their level of educational attainment as well as selected demographic control variables. The labor market conditions component accounts for that portion of the annual earnings gap due to the difference in the distribution of unobserved characteristics between the two groups. That is, those non-quantifiable institutional factors, such as the employer’s tastes and preferences and/or their willingness to pay for those characteristics and traits, result in differences in the rate of return that an individual receives for their human capital.

The results presented in Table 4 are the estimated decomposition components based on the standard, Blinder–Oaxaca, decomposition approach and decomposition correcting for selectivity. Using standard methods, the observed earnings gap between civilian and military husbands is equally attributed to differences in the labor market conditions and human capital components. ³⁵ For civilian and military wives, the standard approach suggests that 38.1 percent of the annual earnings gap can be attributed to differences in the labor market conditions component, while 61.9 percent is attributed to differences in the human capital component. Standard decomposition methods allow us to argue that the earnings gap observed between civilian and military husbands was equally due to military husbands having lower levels of human capital and confronting less than favorable labor market conditions relative to civilian husbands. For wives, this approach suggests that the majority of the observed earnings gap, about three-fifths could be attributed to military wives having lower levels of human capital; experiencing less than favorable labor market conditions accounts for about two-fifths of the gap.

OPEN IN VIEWER

Table 4.

Decomposing the Difference in Annual Earnings between Civilian and Military Spouses.

		Component value and the proportion of the earning gap associated with:
	Difference in (natural log) earnings	Selectivity	Human capital	Labor market conditions
Blinder–Oaxaca (Standard) decomposition
Husbands	0.253	—	0.126 (49.9%)	0.127 (50.1%)
Wives	1.035	—	0.641 (61.9%)	0.395 (38.1%)
Decomposition, correcting for selectivity
Husbands	0.649	0.134 (20.6%)	0.241 (37.1%)	0.274 (42.2%)
Wives	1.147	0.284 (24.8%)	0.587 (51.1%)	0.277 (24.1%)

Source: US Census Bureau, Census 2000 (SEDF).

Note: The proportion of the difference in (the natural logarithm of) annual personal earnings relative to the decomposition component are reported in parentheses under the estimated difference.

By comparison, correcting for selectivity in the earnings models for both husbands and wives provides a way to estimate how military spouses assess their labor force participation decision within the context of their tied-migrant status. Correcting for selectivity provides a better insight into how the labor market conditions and human capital components contribute to explain the gap in annual earnings. More specifically, correcting for selectivity results in 42 percent of the annual earnings gap between civilian and military husbands being attributed to labor market conditions, 37 percent of the difference attributed to the human capital differences, and about 21 percent of the difference attributed to selectivity. Similarly, for the earnings gap between civilian and military wives, 24 percent is now attributed to labor market conditions, 51 percent attributed to human capital differences, and 25 percent to the selectivity component. Correcting for selectivity leads to different conclusions than those based on the standard decomposition approach. Selectivity is of similar magnitude for both sexes, but now human capital for wives is twice as important as labor market conditions. Clearly, as noted by Neuman and Oaxaca, the magnitudes of the estimates accounting for labor market conditions and human capital differences can vary greatly from the standard estimates.

Comparing the decomposition results for husbands, not correcting for selectivity could lead one to argue that both components equally explain the difference in the earnings gap. For husbands correcting for selectivity, however, results in a decline in the impact of other components, with labor market conditions becoming proportionately larger in explaining the earnings gap compared to either the impact of human capital differences or the selectivity component. Based on the results from the standard decomposition approach, one could conclude that differences in human capital attributes were a substantial contributing factor in the observed earnings gap between civilian and military wives. Accounting for selectivity, however, results in a clearer, more focused view of the percentage contribution the human capital component plays in explaining the earnings gap. While the percentage contributions of both labor market conditions and human capital components declined, just over half (51 percent) of the observed earnings gap is now attributed to the difference in human capital characteristics that military wives possess relative to civilian wives. Moreover, the percentage contribution of the human capital component is twice that of either the labor market conditions component or the selectivity component.

From the results presented in Table 4, correcting for selectivity provides further support for Mincer’s argument that the penalty for being a tied-migrant is gender neutral. This support comes by way of the proportional contribution of the selectivity component in explaining the observed earnings gap. For both military husbands and wives, this component reflects perceptions regarding human capital endowments and the rate of return on their human capital investment, which together with their tied-migrant status, results in their assessment that their participation in the labor market is less worthwhile than that of their civilian counterparts.

Military spouses, however, are adversely impacted by less favorable labor market conditions and practices such as employer preferences for hiring individuals not married to active-duty personnel over those individuals married to active-duty personnel. For military wives, the proportional contribution of the selectivity component leads us to argue that these women believe that working is less “worthwhile” than civilian wives believe based in part to the perceived rate of return they expect on their accumulated human capital when they do work. For military husbands, while the difference in their annual earnings relative to their civilian counterparts was about one-third of what was observed between civilian and military wives, we found that the effect of less than favorable labor market conditions was substantial compared to the impact experienced by military wives. These findings from the decomposition analysis suggests that military husbands and wives could have the perception that their work will be undervalued in the labor market, which we argue is reflected through the proportional contribution of the selectivity component in explaining the observed earnings differences.

Conclusion

The literature holds that human capital endowments and other supply-side factors explain a large part of the earnings gap between civilian women married to civilian husbands and civilian women married to military husbands, the latter, a group of “tied-migrants.” Speculation that employer preferences and other hiring obstacles also play roles has been mentioned, but not empirically addressed. Further, to date, there has been little research on the earnings of civilian husbands of women on active-duty; thus, there is limited speculation on the causes of their earnings shortfall. The research presented here validates earlier findings regarding the lower earnings of military wives and it explores demand and supply-side causes. Perhaps more importantly, these same issues are addressed for civilian men who are married to women in the military, both confirming that there is an earnings shortfall for this group, but also allowing comparisons of demand and supply-side impacts. Moreover, the methods employed in this analysis allow us to assess the role selectivity plays in explaining the earnings gap experienced by both civilian husbands and wives of active-duty military personnel. We believe this analysis provides new insights into findings previously advanced regarding being a tied-migrant spouse.

Our findings reveal an earnings penalty is 42.7 percent for military husbands and 53.1 percent for military wives. For military husbands, this shortfall is accounted for as follows: 20.6 percent due to selectivity, 37.1 percent due to supply-side issues, and 42.2 percent due to demand-side issues. For military wives, the corresponding percentages are 24.8, 51.1, and 24.1. So, the decision to work or not, selectivity, effects both groups similarly. Human capital is about two times more pronounced than the other components for women but labor market conditions and demand-side issues are more pronounced for men. The magnitudes of the percentages suggest, however, that neither “side” can be ignored in speculation about the root causes of the earnings shortfall.

Our empirical findings validate the phrase used earlier that earnings penalties appear to be due to a combination of factors. Unfavorable labor market conditions caused by employer preferences for the civilians husbands and wives married to spouses not on active-duty in the military, lesser ability of the civilian husbands and wives married to military spouses to convert their human capital into earnings, as well as their weaker attachment to the labor market all appear to play a role. Speculations as to the underlying reasons for these generalizations spread across several areas. Influencing the supply-side percentages is the not-unexpected evidence and frequent speculation that military wives are more likely than military husbands to assume child care responsibilities either for themselves or for other families, devote more time to shutdown and start-up of their household when they move, and take on more volunteering activities, depending on their husband’s rank.

Prior research has also argued that the unpredictable and rigid work schedule of active-duty military personnel is one factor that can explain the lower earnings of their spouses. ³⁶ We believe that such scheduling problems, together with the greater likelihood of the presence of young children, contribute to the lower earnings observed for military spouses, in particular military wives. Additionally, job flexibility is an issue in the short term but one that likely will also affect employer preferences and have demand-side impacts in the future. Many of the jobs that military wives tend to find likely allow flexible hours and are ones that can be started and stopped without much investment by either them or the employer. Military husbands would seem to be little different than military wives in these regards. Ultimately, these factors may transition into demand-side issues as employment experiences and the resulting conditioning fosters behavior wherein there is less investment in job search and fewer opportunities for job-specific training. These interactions are consistent with the usual textbook model that longer and less fruitful job search lowers one’s reservation wage.

The findings of a lower age-earnings path for both groups of military spouses (see Figure 1) supports the argument that both civilian husbands and wives of military personnel face similar labor markets. Admittedly, because there are fewer military husbands, there may be less of a market oversupply issue for these men than has been suggested in the literature for military wives. Military husbands as well as wives, however, were found to work less and earn less annually than their civilian counterparts, leading us to conclude that there is similar evidence for both genders regarding work effort and lower pay. Such evidence for military husbands reinforces the speculation that traditional male job markets offer fewer regular jobs with flexible hours, a demand-side issue, and another reason for the lower earnings and weak attachment of military husbands to the labor market, which fits with our selectivity finding. For military husbands, we believe this is particularly interesting, given the historical data show a higher labor force participation rate for prime working-age men relative to women. ³⁷ In any case, these speculations are consistent with a lower reservation wage and lower wage equilibrium.

On the issue of how moves affect the annual earnings of military wives, prior research has used “tied-migration,” a supply-side factor, to capture relocation as the major reason that the earnings of military wives are lower at every age even though earnings trends over time show increases for both military and civilian wives (see Figure 1). Because military husbands exhibit the same age-earnings pattern relative to their civilian counterparts even though their labor market experiences in general must have evolved in somewhat different ways than military wives, their earnings shortfall, while it may be attributed to their tied-mover status, certainly must have somewhat different root causes and, hence, explanations. One possible explanation for the lower annual earnings of the military husbands could be the availability of part-time employment or their willingness to take part-time employment, as discussed in the earlier univariate analysis. While our empirical findings parse the data in ways we find useful, the dynamic interactions among them is complex and begs for additional research, speculation, and analysis.

More specifically, one avenue of future research could focus on the interaction of marital formation and the earnings prospects of individuals whose betrothed are military personnel or have intentions of becoming military personnel. One could argue that there is the issue of self-selection at the times these marriages are formed and that the future civilian spouse recognizes what they were getting involved in. That is, the prospects of earnings penalties, underemployment, and even unemployment, for duration of their future spouse’s military career. The nature of such research would involve treating the covariates of future civilian’s spouse as endogenous to their labor market behavior and outcomes. However, attempting such an investigation depends on availability of an appropriate data source that would allow for assessing the interaction between the marriage market and the labor market.

The motivation for this article was to better analyze and more clearly explain the earlier findings that military spouses experience lower earnings that their civilian peers. ³⁸ By decomposing the earnings penalty observed between civilian and military spouses into three components (selectivity, human capital/endowment differences, and labor market conditions) for wives we found that the proportions attributable to human capital endowments substantially overshadow the proportions due to selectivity (their decision to work) and to labor market conditions they face. For husbands, the proportions are relatively equal, which are somewhat at odds with earlier referenced research. The decomposition findings accounting for selectivity allow new and useful insights into the earnings penalty experienced by military spouses. For both husbands and wives, their tied-mover status affects two areas: their decision to enter the labor market (selectivity) and the perceived rate of return they expect on their accumulated human capital when they do work (labor market conditions).

Government policy such as allowing civilian spouses to use accumulated G.I. Bill education and training benefits, essentially a supply-side opportunity for the family, may not fully succeed if employers find they can pay these spouses less when they work, lowering their reservation wage, and discouraging job seeking as the traditional model of employment search suggests. Based on our findings, we believe that, when compared to their civilian counterparts, the civilian husbands and wives of active-duty military personnel have a greater obstacle to overcome in the form of less than favorable labor market conditions relative to any education or skills disadvantage, than previously thought.