Printer Friendly

Quality of available mates, education, and household labor supply.

I. INTRODUCTION

This paper examines the effects of local quality sex ratios by metropolitan area and educational attainment on spouses' labor supply and bargaining power, using individual-level data from 2000, 1990, and 1980. There is evidence in the literature that the availability of potential mates affects the labor market decisions of married individuals, with mate availability measured as the raw number of men relative to the number of women in aggregate marriage markets (e.g., Angrist 2002; Chiappori, For tin, and Lacroix 2002; Grossbard and Amuedo-Dorantes 2007). However, the literature on sex ratios emphasizes that both the local dimension of spouse availability and the economic attractiveness of mates play a large role in marital behavior in the United States (Fossett and Kiecolt 1991; Lichter, LeClere, and McLaughlin 1991).

In this study, we further explore the role of sex ratios on bargaining power and spouses' labor supplies by constructing a refined availability measure that reflects both the local nature of marriage market conditions and their quality. We focus on educational attainment as our qualitative indicator. Education is commonly regarded as a valuable trait in marriage by which individuals assortatively match (Qian 1998; Weiss and Willis 1997). We consider local marriage markets at the metropolitan level and construct a sex ratio by three education brackets (high school [HS] graduates, some college [SC], and college-college plus [CC]), within which individuals usually sort. In the framework of a collective labor supply household model, we test whether this quality sex ratio affects the intrahousehold bargaining power of couples in the corresponding education bracket. Specifically, when the sex ratio is favorable to the wife (i.e., there is a relative scarcity of women in her education bracket), the distribution of gains from marriage is shifted in her favor, generating opposite income effects on spouses. In particular, according to collective models, if a higher number of qualified men in the wife's marriage group of reference increases female intrahousehold bargaining power, then one would expect a reduction in wives' labor supply and an increase in husbands' labor supply (Chiappori, Fortin, and Lacroix 2002).

Additionally, we investigate whether the bargaining power effect of such a sex ratio is greater for higher educated couples. Education is positively correlated with important mate attributes such as wealth and income, while marital gains from educational attainment may be more relevant for highly educated individuals. Moreover, their labor supply response to bargaining power shifts may be stronger due to their relatively more flexible labor supplies. Consequently, our contribution is also to test the theoretical prediction by Iyigun and Walsh (2007) that the sex ratio has a stronger impact on intrahousehold allocations as the assortative rank of couples, measured here by educational attainment, increases.

We use Census data at the metropolitan level for the recent decades of 2000, 1990, and 1980 to build our sex ratios. We add data from the March Supplement of the Current Population Survey (CPS) for the same years to test our labor supply prediction on married couples, using unmarried individuals as control group. Our identification strategy consists of estimating the effects of education sex ratios on husbands' and wives' labor supply and comparing changes in their labor supply behavior cross-sectionally across the U.S. metropolitan areas.

Our empirical analysis reveals that married women significantly reduce their supply of market labor, while their husbands increase theirs, as the corresponding education sex ratio becomes more favorable to women. Results are similar across decades, with a stronger impact in 1980. For instance, in 2000, we find that a 10 percentage point increase in the education sex ratio decreases annual hours worked of "SC" and "CC" wives by 10 and 26.3, respectively, and increases their husbands' by 4 and 9.6, respectively, while HS graduates do not exhibit any significant impact. Consistent with the hypothesis of a stronger effect for higher education brackets, we also find that couples with CC wives exhibit a significantly stronger impact of the quality sex ratio on their bargaining power than those with SC wives, whose estimated quality sex ratio coefficient is in turn larger than that for HS graduates. Our bargaining power interpretation is strengthened by the fact that unmarried men and women do not exhibit any significant reaction to the quality sex ratio on their labor supply.

The findings presented here are consistent with theories that predict that higher sex ratios in the marriage market increase female bargaining power. Moreover, this study represents the first empirical support for the bargaining power effect of a local quality sex ratio by education, and for its stronger impact on couples with higher levels of educational attainment. Our results clearly indicate that local area variations in couples' labor supply are directly linked to the relative scarcity of economically attractive mates. Both the local and the quality dimensions of sex ratios are relevant to explaining household behavior.

A number of alternative explanations are considered. The geographical variation in the relative number of men and women by education may capture differences in local labor market opportunities for women, in marital gains from specialization and in welfare programs, or in the prevalence of married and same-sex partners who do not represent available mates. We argue that these phenomena cannot consistently explain our results, given our intrahousehold bargaining predictions and empirical evidence.

The paper is organized as follows. Section II describes the theoretical framework. Section III describes the empirical specification and data. Section IV presents the empirical results. Section V considers alternative explanations for the findings. Section VI concludes the paper.

II. THEORETICAL BACKGROUND

There are two strands of economic literature related to our study. One strand focuses on the impact of the sex ratio on marriage markets, spouses' bargaining power, and labor supply behavior. Specifically, studies such as Chiappori, Fortin, and Lacroix (2002) and Chiappori, Iyigun, and Weiss (2005) develop a collective household model and demonstrate that favorable outside marriage market opportunities increase a spouse's bargaining power through an income effect, measured as a reduction in labor supply. The opposite effect occurs for the other spouse. This is due to the fact that married men and women have the option of seeking a divorce and remarrying. Therefore, more numerous potential mates in the marriage market of reference should affect the bargaining power of those already married, to the extent that this affects their opportunities outside the marriage. It is widely acknowledged in the literature that married people are responsive to shifts in outside factors, which can lead to income and bargaining power redistribution between spouses and changes in labor supply (Chiappori, Fortin, and Lacroix 2002; Grossbard-Shechtman 1993, 1984; Lundberg and Pollak 1996). Our paper specifically refers to this theoretical framework. Furthermore, our investigation of whether the sex ratio bargaining power effect is greater for higher educated couples relies on a relevant theoretical result provided by Iyigun and Walsh (2007). They incorporate assortative spousal matching into the collective household model, argue that the sex ratios in the marriage market are one of the determinants of intrahousehold allocations, and find that the role of sex ratios increases with the assortative rank of couples. The relative importance of sex ratios in the determination of intrahousehold allocations rises with the couple's assortative rank since wives' share of marital surplus, that is, willingness to accept to stay married, also increases with rank. We measure the assortative rank order with educational attainment and test the prediction by Iyigun and Walsh (2007) using the variation across three different education brackets.

The literature also provides empirical evidence about the effects of quantity sex ratios on labor supply. Using data at both household and aggregate level, Grossbard-Shechtman and Neideffer (1997) and Grossbard-Shechtman (1993) show that increases in local and aggregate quantity sex ratios reduce the labor force participation and hours worked of married women. A recent study by Grossbard and Amuedo-Dorantes (2007) documents the effect of cohort-level sex ratios by region on women's labor force participation. In a study about immigrants to the United States, Angrist (2002) finds that higher sex ratios by ethnicity affect female labor market decisions. He argues that these empirical results are consistent with theories that predict that higher sex ratios increase female bargaining power in the marriage market. Chiappori, Fortin, and Lacroix (2002) find that higher sex ratios reduce wives' labor supply and increase the husbands', using 1990 state Census and Panel Study of Income Dynamics data. However, both the local dimension of spouse availability and the economic attractiveness of mates are relevant marriage market conditions (Fossett and Kiecolt 1991; Lichter, LeClere, and McLaughlin 1991). Our analysis takes into account both of these aspects simultaneously.

The second strand of literature that is related to our paper concerns the spousal sorting by educational attainment and the gains to marriage from education. Spouses have increasingly similar educational attainment, especially among highly educated people (Qian 1998). Specifically, sorting has mainly increased from 1960 to late 1980s, when gender roles have become more egalitarian and social distance between education groups increased (Mare and Schwartz 2005). In this respect, the years 1990 and 2000 are similar. Mare and Schwartz (2005) also report that today husbands and wives are roughly four times as likely to have a spouse who shares their educational background as they are to be married to someone who does not, educational homogamy being particularly strong for college graduates. Strong sorting based on educational attainment is also documented by Weiss and Willis (1997), with the additional finding that similarity in schooling increases marriage stability. Schooling also has cross-productivity effects on spouses in the sense that wives' education is found to increase the productivity and wages of their husbands and vice-versa (Benham 1974; Tiefenthaler 1997).

However, none of these studies explores how the distribution of educational attainment of men and women in local marriage markets affects intrahousehold bargaining power. There is also no evidence on whether this impact is increasing with higher educational rank of couples. Analyzing these effects of quality sex ratios by education is the focus of our paper.

III. EMPIRICAL SPECIFICATION AND DATA

A. Identification Strategy

Our main sample consists of married couples, with both spouses between 22 and 60 yr of age. According to the theory, if the scarcity of educated women in the local marriage market enhances women's bargaining power in the household, then the labor supply of wives should decline and that of their husbands should rise. Additionally, couples in higher education categories should experience a stronger impact on their labor supplies relative to other education categories. We also consider unmarried men and women in the same age bracket, focusing on singles as a "control" group. In principle, singles' labor supplies should not be affected by changes in intrahousehold bargaining power, given that they represent a one-decision maker household. They may experience expected marital gains or losses if they plan to marry in the future and the sex ratio remains stable over time. Therefore, consistent with Chiappori, Fortin, and Lacroix (2002), we do not predict any effect on singles' labor supply. We include intact couples only if both spouses are actually present. We exclude widowed and separated couples in order to keep a clear distinction between multiple and one--decision maker households. For the same reason, we exclude singles that are not the head of their own household, even though their sample size is reduced. (1) However, the sample size for a cross-section is comparable to the number of observations of singles in Chiappori, Fortin, and Lacroix (2002). All individuals in our sample have already completed their education.

The following equations for labor supply are estimated separately for wives and husbands:

[h.sup.f] = [[alpha].sub.1] ln [w.sup.f] + [[alpha].sub.2] ln [w.sup.m] + [[alpha].sub.3]y + [[gamma].sub.1]EdR + [[gamma].sub.2] (EdR x SC) + [[gamma].sub.3] (EdR x CC) + [delta]X + [[epsilon].sup.f]

[h.sup.m] = [[beta].sub.1] ln [w.sup.f] + [[beta].sub.2] ln [w.sup.m] + [[beta].sub.3]y + [[lambda].sub.1]EdR + [[lambda].sub.2] (EdR x SC) + [[lambda].sub.3] (EdR x CC) + [phi]X + [[epsilon].sup.m].

We also estimate a corresponding labor supply equation for unmarried women and men, using the same specification, except for spousal variables:

[h.sup.u] = [v.sub.1] ln [w.sup.u] + [v.sub.2]y + [[theta].sub.1]EdR + [[theta].sub.2] (EdR x SC) + [[theta].sub.3] (EdR x CC) + [upsilon]X + [[epsilon].sup.u].

EdR is our sex ratio, which is constructed by three education brackets, two races, and by metropolitan area, in order to capture the economic attractiveness of local mates. We assign to each individual the corresponding ratio of the number of men over the number of women of his/her race, educational category, and that live in his/her metropolitan area. For couples, our sex ratio EdR corresponds to the number of men over women who are of the same race and education category as the wife of each household. As to race, we focus on black and white individuals and on couples where spouses are of the same race, assuming that the relevant marriage market is limited to one's own race. The coefficient of EdR is common to both races. We consider the following education categories: HS graduates, SC, and college graduate-college plus. HS includes people with HS diploma, or equivalent; SC includes individuals with some college, with or without an associate degree; and CC refers to bachelor's degree and above. This means that the corresponding education sex ratio for a white couple living in Tucson, Arizona, with the wife being a HS graduate, is the total number of white HS graduate men living in Tucson divided by the total number of white HS graduate women residing in this same metropolitan area. We exclude HS dropouts from our analysis to keep our sample homogeneous since HS dropouts are characterized by traits, socioeconomic characteristics, and marriage market prospects that are different from those of graduates (Eckstein and Wolpin 1999; Rumberger 1983). We compute our sex ratio including men and women aged 18-64 yr following Fossett and Kiecolt (1991) who find that measures of the sex ratio based on broad age ranges are satisfactory and may be preferable to sex ratios computed for narrower age ranges. (2)

The interactions of EdR with the dummy variables for the education brackets SC and CC capture the differential effect of our sex ratio for higher education categories. The education dummies refer to the education of the wife and our omitted category is HS graduates. Our identification strategy of the bargaining power effect consists of estimating [[gamma].sub.1], [[gamma].sub.2], [[gamma].sub.3] for wives and [[gamma].sub.1], [[gamma].sub.2], [[gamma].sub.3] for husbands. The impact of the education sex ratio on the labor supply of HS graduate wives and their husbands is captured by [[gamma].sub.1] and [[gamma].sub.1] respectively. The terms ([[gamma].sub.1] + [[gamma].sub.2]) and ([[gamma].sub.1] + [[gamma].sub.3] measure the impact of the education sex ratio on the labor supply of SC and CC wives. The corresponding impacts for husbands are measured by the terms ([[gamma].sub.1] + [[gamma].sub.2]) and 0[[gamma].sub.1] + [[gamma].sub.3]), respectively. We consider that individuals are affected only by the sex ratio of their own race and education bracket of reference, within which they usually sort, allowing for a differential impact across education groups. We thus regard the substitution effect with potential spouses from other education categories to be negligible. In this sense, our coefficients represent the partial effects of the education group-specific sex ratios.

The other regressors are the wage rate [w.sup.i] (of spouse i or of unmarried individual u), household nonlabor income, y, and a vector of covariates X. X includes age, experience, education of each spouse, a dummy variable for race, number of household members, and number of (young) children in the family. X also includes state unemployment rate, state total labor force participation, and female labor force participation, to control for the level of economic activity in a state and especially for employment opportunities. We control for local labor market conditions since sex ratios may be correlated with labor demand factors while we want to examine the marriage market effect of sex ratios on spouses' labor supplies. High values of our quality sex ratio by metropolitan area may suggest male workers outnumbering female workers because of a local labor market with gloomy perspectives for women. However, the labor market controls and the comparison with the labor supply effects on husbands and singles should help capture the marriage market impact of our sex ratios. These labor market controls should also mitigate the concern about the potential endogeneity of sex ratios due to endogenous mobility. Our contemporaneous crosssectional sex ratios should not capture the endogenous intertemporal decision of individuals relocating over time to different metropolitan areas. To this end, our approach relates to Chiappori, Fortin, and Lacroix (2002) and the work of Grossbard, for example, Grossbard-Shechtman and Neideffer (1997), who run cross-sectional analyses with sex ratios as exogenous variables. (3) Finally, we add two measures of the prevalence of same-sex unmarried households by metropolitan area, for homosexuals and lesbians in 2000 and 1990 only, as same-sex households were not recorded in 1980. The purpose of including these two variables is to keep our education sex ratio as closely related to the heterosexual marriage markets as possible.

The dependent variable in our labor supply regressions is annual hours worked, defined as total annual hours worked on the longest job held in the previous year. Households in which the wife or the husband does not work are also included in our samples, and we account for a possible selection bias toward working men and women by correcting for sample selection with Heckman maximum likelihood estimator (MLE). (4) As source of identification, we use distributional assumptions on the first step residuals alone or exclusion restrictions. (5) Both procedures yield similar robust results. All female and male labor supply regressions exhibit the same results when estimated without selection correction. We use predicted wages to measure the nonworking spouses' wages and to address the possible endogeneity of individuals' observed wages. To predict individuals' wages, we take a standard human capital approach, also implemented in the collective labor supply literature (e.g., Donni 2005), and consider a wage equation in which wage depends on the individual's age, race, education, education squared, and cubed but does not depend on his/her spouse's characteristics. This equation is then estimated separately for participating wives, husbands, single men, and single women, with a correction for selection bias. (6) The generated fitted values then replace the wage observations of the corresponding individuals in our samples. (7) Finally, Wald tests of overall statistical significance performed on the above-mentioned labor supply regressions do not reject the validity of the framework we use.

We run our labor supply regressions using robust standard errors clustered by metropolitan area, which allows for correlation of household observations within metropolitan areas. Our specifications do not use a differences-in-differences estimator: husbands' and wives' regressions, as well as singles', are estimated separately from one another. As such, they should not suffer from the understated standard errors highlighted by Bertrand, Duflo, and Mullainathan (2004). At any rate, clustering by metropolitan area should rectify such an underestimation, if it is present.

We assume sorting within education brackets. We compute the extent of sorting in our own sample and find that the percentage of couples who have spouses with education levels in the same bracket (HS graduates, SC, and CC) is 58.5% in 2000, 57.5% in 1990, and 55% in 1980. The correlation of spouses' education across education brackets is about .52 for all three decades. These figures are very similar to those reported by the literature acknowledging education assortative mating. Specifically, Chiappori, Iyigun, and Weiss (2007) state that the proportion of spouses who have the same level of education remained fairly constant over time at about 50%. Weiss and Willis (1997) find that the correlation in educational attainments of spouses is on average .57 (around year 1980) and report that this strong correlation is similar in magnitude to the correlations found in many other samples in the United States and other countries.

B. Data

Estimation is carried out on the March Supplements of the CPS, for the years 2000, 1990, and 1980. U.S. Census data for the corresponding years are used to construct our education sex ratio by education brackets, race, and age-groups. Unmarried individuals and husbands and wives from one-family households were extracted from the March CPS into separate files. Records in these files were then matched on the household identification code to create a single observation for each married couple. Data on labor force activity, income, and any variable of interest at the household level are taken from the March Supplements. In particular, the covariate education is derived from the education categories that the CPS provides. (8) CPS weights are used to make the sample representative of the U.S. population and economy. To this sample, we merge data on education sex ratios from the Summary File 4 (SF4) of the Census for 2000 and 1990 and from Chapter C of Volume 1 of the Census for 1980. SF4 and Chapter C contain information compiled from the questions asked to a sample of all people and housing units and is released as individual files for each of the 50 states, the District of Columbia, Puerto Rico, and for the United States overall. We use the cross-tabulations by sex, age, race, and educational attainment to construct separate education sex ratios for the black and white population aged 18-64 yr, by metropolitan area. (9) In 2000, the Census identifies 276 U.S. metropolitan areas excluding Puerto Rico; in 1990 and 1980, the total identified areas are 284 and 288, respectively. We merge these to the CPS data and keep the metropolitan areas present in both data sets. We also exclude the top and bottom 2% metropolitan ratios' outliers. This leaves us with 173 metropolitan areas in 2000, 181 in 1990, and 34 in 1980. (10) The state unemployment rate, state total labor force participation, and female labor force participation are retrieved from the Bureau of Labor Statistics. The two measures of the prevalence of same-sex unmarried households come from tabulations in SF4 and are computed at the metropolitan level. In 2000 and 1990, the Census records a household as a same-sex union if the relationship to the householder is specified as "unmarried partner." We construct two ratios, the number of homosexual unions out of the total number of households and the number of lesbian unions out of the total number of households.

Table 1 presents the descriptive statistics for the main variables by year and demographic group. In our samples, men on average work more annual hours than women and earn a higher hourly wage, while they have similar levels of education. On average, husbands are 2 yr older than wives. As to our education sex ratio by metropolitan area, there are more white women graduating from HS, or having SC education, than white men. On the other hand, there are more white men than women holding a college degree or above, but the gap has been decreasing over time. The pattern is somewhat different for the black population: fewer black women hold a HS diploma relative to black men, but they are more numerous in the SC category.

IV. RESULTS

A. Main Evidence

The main results are shown in Tables 2 and 3. In all the three decades, the estimated effects of our quality sex ratio are positive for husbands and negative for wives, as predicted by the theory. Additionally, couples with CC wives exhibit a stronger response to the quality sex ratio on their bargaining power than those with SC wives. In turn, SC wives-estimated quality sex ratio coefficient is larger than for HS graduates wives. The point estimates in our samples indicate that m 2000 (columns 1 and 2 of Table 2), a 10 percentage point (11) increase in the education sex ratio reduces SC wives' annual labor supply by about 10 h (p = .001 ), while their husbands' is increased by 4 h (p = .06). As to couples with CC wives, their coefficients for the education sex ratio show a decline in wives' labor supply by 26.3 h (p = .001), and an increase in their husbands' by 9.6 h/yr (p = .05). The year 1990 exhibits a similar impact (columns 3 and 4 of Table 2). The evidence clearly shows that in recent years for both husbands and wives, the estimates for the CC are greater than for SC, the coefficients being statistically different from each other for each spouse. This suggests that changes in the sex ratio of one's education group have a stronger effect on bargaining power if one is highly educated. HS graduates do not show any significant response to changing ratios. (12) As reported in columns 5 and 6 of Table 2, the 1980 sample exhibits a similar pattern of effects between spouses and across education brackets, although their magnitude is larger than in 1990 and 2000. Most notable is the strong impact for the highest education bracket. We suggest that for CC women, the availability of suitable mates was really restricted to the highly educated pool, given that it was uncommon for CC women to marry down. Thus, variations in the CC education sex ratio had a larger bargaining power effect back in the 1980s than in recent years, when marrying down became more socially acceptable for women (Chiappori, Iyigun, and Weiss 2007). Also, wives' attachment to market work has increased since 1980, especially for highly educated women (Pencavel 1998b). It may be that in 1980s, wives' labor supply was more responsive to sex ratio changes because their work attachment was weaker. Finally, divorce rates were at a record high around 1980, so that the higher likelihood of divorce would make couples more responsive to outside marriage market opportunities. The findings from 1980, though, have to be interpreted with caution. The very small number of metropolitan areas identified in that year in CPS (resulting in 34 areas after merging CPS to Census data and dropping outliers) and their consequently modest cross-sectional variation in sex ratios are likely to make our 1980 sample not representative. As such, in our following analysis, we focus on the decades of 1990 and 2000.

As to the size of our education sex ratio effects, the changes for 2000 correspond to a 5.7% (15.06%) reduction of the average annual hours worked by SC (CC) married women and to a 1.7% (4.1%) increase for their corresponding husbands'. Similarly, in 1990, the education sex ratio effects amount to a 6.2% (10.69%) reduction of the average annual hours worked by SC (CC) married women and to a 2.53% (5.9%) increase of their husbands'. These effects are not negligible, given the acknowledged rigidities in the husbands' labor supply (e.g.. Donni 2005) and the frequency of the reported labor supply peaking around 40 h of work/wk. In particular, the impact on husbands is remarkable since traditional analyses do not emphasize their response to the sex ratio, let alone their labor supply increasing with it. The direction of those effects is also the same as in the labor supply impact of the quantity sex ratio found by Chiappori, Fortin, and Lacroix (2002). (13)

As to the other covariates in the spouses' labor supply equations, most parameter estimates are comparable to the literature. In particular, the wives' own wage response is always positive significant, while the husbands' own wage coefficient is negative only in 1980 and the effects are sizable (Table 2). Also the husband's negative estimate is in accord with previous empirical findings in the family labor supply literature. In fact, Chiappori, Fortin, and Lacroix (2002) run similar spouses' labor supply equations and show negative own wage estimates for husbands. Furthermore, we find a positive significant cross-wage effect of husbands' wages on wives' labor supply, as documented in Chiappori, Fortin, and Lacroix (2002) and Blundell et al. (2002).

The signal conveyed by the education sex ratio about the quality of outside marriage market opportunities is increasingly more relevant for highly educated wives and husbands. Education is positively related to important mate attributes such as wealth, income, and success in life; thus, marital gains from educational attainment may be more significant for highly educated individuals. In other words, the availability of valuable mates in the marriage market represents a more desirable outside opportunity, and a more credible threat, for spouses who are per se high-quality mates than for the ones in the lower education brackets. Also, low-educated couples are more likely to exhibit rigid labor supplies so that they may be less responsive to bargaining power shifts in terms of labor supply. Hence, our findings of a stronger sex ratio impact for more educated couples are in line with the prediction by Iyigun and Walsh (2007), who show that imbalances in the sex ratios become more relevant for intrahousehold allocations as the rank of couples in the assortative order rises, measured here by educational attainment. Moreover, our results also match evidence in the literature of stronger educational homogamy for highly educated men and women (Qian 1998). The probability of having a spouse with the same educational background is four times higher than the possibility of marrying someone who does not (Mare and Schwartz 2005).

The bargaining power effect is also estimated on unmarried individuals (singles), separately for men and women. The results are detailed in columns 7 through 12 of Table 3. Their labor supply regressions show no significant impact of the education sex ratio in any decade, as theory predicts. Both men and women exhibit economically negligible and statistically insignificant coefficients of the sex ratio by education brackets and of its interactions. No additional impact is found for SC and CC. Furthermore, all their coefficients are different from the couples' sample, which emphasizes the bargaining power effect on husbands and wives. Only the coefficient reflecting the impact on HS graduates has a large magnitude for single men. However, the coefficients are never significant and the single men's very small sample size, especially in 1980, may explain the imprecise estimate. This lack of impact on singles is consistent with the findings of Chiappori, Fortin, and Lacroix (2002). We also estimate these effects on an expanded sample of unmarried, nonseparated individuals. Results are reported in Table 4 and support our findings and interpretation for singles.

Our empirical results emphasize that both the local dimension of spouse availability and the economic attractiveness of mates affect spouses' bargaining power and labor supply. This evidence represents the first empirical support for the bargaining power effect of a quality sex ratio by education and for its stronger impact at higher levels of educational attainment. Further evidence presented below, together with the discussion of various alternative explanations, should help making this claim convincing.

C. Impact on Older Couples

The bargaining power effect of our sex ratio by education is also estimated on subsamples of couples older than 30 yr. We report the results in Table 5. Couples in their mid-30s and above exhibit a stronger impact of the education sex ratios than the entire sample, especially for the CC category. With a 10 percentage point increase in the education sex ratio, the associated decline in wives' labor supply is 34.2 annual hours for CC in 2000 and 23.6 annual hours in 1990 (columns 1 and 3), while for husbands, the increase is of 15.7 and 15.9 annual hours, respectively (columns 2 and 4). The impact for SC is about 10 h for wives and 7 h for husbands in both decades as seen in columns 2 and 4. The role of HS graduates' sex ratio is still negligible. We believe that these results reflect different informational values about the quality of potential mates that educational attainment conveys at different stages of life. At older ages, education is a better predictor of mate quality and economic prosperity because enough time elapsed to establish social status and wealth. Especially if one has a high educational attainment, the signal given by the education sex ratio is very quality informative, so that the effect of such outside marriage market opportunities on bargaining power is very strong. Education matters more in marriage choices when prosperity is directly at stake: this is the case for "older" couples looking at their marriage prospects since the returns to education are already realized. Evidence from the literature actually suggests that later age at union promotes stronger educational homogamy. In particular, men and women aged 30 yr or above are less likely to be with partners with a different level of educational attainment than are persons in their 20s (Qian 1998).

D. Race

Running our main labor supply specification on the subsample of white couples yields the same results as the full-sample regressions (Table 6). The education sex ratio (14) has a negative effect on wives' labor supply and positive effects on husbands', with a significantly stronger impact for the CC than for SC wives. The coefficient of HS graduates is not significant. As reported in columns 1 and 2, in 2000, with a 10 percentage point increase in the education sex ratio, SC wives experience a reduction in their annual hours of 8.3 (p = .02), while their spouses increase theirs by 3.8 (p = .1). Moreover, wives in the highest education category reduce their annual hours worked by 25.6 h (p = .001), and their spouses experience an increase of 12 annual hours (p = .03). This impact is comparable to the one in Grossbard-Shechtman and Neideffer (1997), who find a similar effect of city-level quantity sex ratios on the labor supply of white wives. Because of the very small black population present in the CPS, we could not estimate meaningful regressions for only black couples. However, comparing the findings for white couples to those for the entire sample, it seems that blacks and whites respond to bargaining power effects of the education sex ratio in a similar manner. In any case, in our full sample, we run a similar regression to check whether the bargaining power effect of our within-race quality sex ratio varies across races. Each of the three variables concerning the education sex ratio is interacted with a dummy variable for race, in order to capture a possible differential effect. No evidence of a different impact across races was detected.

E. Labor Force Participation

The bargaining power impact of the education sex ratios is also estimated on the labor participation decision of couples and singles. The effects on the extensive margin of the labor supply decision are reported in Table 7. We find a somewhat significant impact on SC and CC spouses, for whom the direction of the education sex ratio effects is in accord with our bargaining power interpretation. Nonetheless, the pattern of significance is consistent for both spouses only in 1980. Singles do not exhibit any significant impact. However, these findings on the extensive margin are subject to theoretical caveats. Our theoretical background of collective labor supply models (Chiappori, Fortin, and Lacroix 2002) relies on, and yields predictions for, labor supply behavior rather than participation decisions of spouses. Bargaining power effects of the sex ratio or other "distribution factors" are predicted as continuous changes in spouses' labor supplies.

F. HS Graduates

Bargaining power in HS graduate households does not seem to be affected by the relative number of men and women who are HS graduates in a metropolitan area. We suggest that the lack of an effect at this relatively low level of education could be due to individuals not deriving significant marital gains in terms of their educational attainment, so that they are not affected by the mate quality dimension "education." Another force driving this result could be that HS graduates, the black ones in particular, have a lower remarriage rate than higher educated individuals (Smock 1990). HS graduate couples may be less sensitive to variations in remarriage market opportunities such as fluctuations in the sex ratio. Moreover, HS graduate couples exhibit more rigid labor supplies than higher educated spouses. Wives especially work more hours and are less flexible in specializing in nonmarket activities than college graduates (Pencavel 1998a). Low-educated couples are likely to hold jobs with fixed amount of hours to supply and they may not be able to respond to bargaining power shifts in terms of labor supply.

We also consider the hypothesis that these individuals do not exhibit strong assortative mating behavior by education because the bracket is too narrow and they may also look for mates "above," in the SC pool. We thus construct a modified quality sex ratio, in which couples with a HS graduate wife are associated with the sex ratio of HS graduates plus SC men and women. There is no evidence to support the hypothesis. The bargaining power effect for them is not significant for husband or wife, while for SC and CC couples, it remains significant, and with an increasing impact across educational brackets.

V. ALTERNATIVE EXPLANATIONS

A. Sex Ratios As' Proxy for Local Labor Market Conditions

It may be possible that the labor supply of married women falls not as a result of the bargaining power effect of mate availability by education brackets but due to poor local economic opportunities for women. High values of our quality sex ratio by metropolitan area may suggest male workers outnumbering female workers because of a local labor market with gloomy perspectives for women. Similarly, it could be that more educated women, whose labor supply is high, live in metropolitan areas where there are better job opportunities for them, so that the negative coefficient of our education sex ratio represents labor market instead of bargaining power fluctuations. There are at least five reasons to believe that the local economy hypothesis does not provide a plausible alternative explanation for our findings. First, our labor supply regressions include individuals' wages and experience, state unemployment rate, and total and female labor force participation rate, which account for the effects of variation in labor market opportunities, specifically for women. Specifications without local labor market controls yield, however, the same pattern of education sex ratio effects. Second, our findings are robust to adding individual controls for industry and occupation categories in our labor supply regressions. Third, it is difficult to understand why the labor supply of men married to these women, but not other men, should be higher in those metropolitan areas if it were just a labor market fluctuation. Fourth, single or nonseparated, unmarried women with similar demographic and labor market characteristics did not experience the same impact of the education sex ratios as married women. Fifth, imbalances in the number of men and women should affect marriage market conditions more than labor market conditions (Grossbard-Shechtman and Neideffer 1997). As Grossbard-Shechtman and Neideffer contend, men and women are poor substitutes in marriage, but they can be very good substitutes in the workplace, wherever gender discrimination is not practiced. Thus, sex ratios affect bargaining power more than they affect labor market outcomes such as wages.

We also address the related alternative explanation of potential endogeneity of sex ratios driven by endogenous mobility. Controlling for labor market opportunities, comparing couples' to singles' labor supply effects, and especially the cross-sectional nature of our analysis should rule out this concern. We follow the approach by Chiappori, Fortin, and Lacroix (2002) and Grossbard-Shechtman and Neideffer (1997), who run cross-sectional analyses with sex ratios as exogenous variables.

B. Sex Ratio Including Married and Same-Sex Partners

It may seem that our education sex ratio does not capture the actual availability of mates in a local marriage market because both married individuals and same-sex partners are included in the computation of our variable. Its lack of significance in our unmarried samples may be attributed to large percentages of unmarried men or women having same-sex partners. We believe that our ratio of the total number of men and women present in a metropolitan area does represent a reliable sex ratio for three main reasons. First, there is considerable evidence in the literature that relatively little benefit is realized from refinements such as computing sex ratios separately by marital status (Fossett and Kiecolt 1991; Freiden 1974). Second, we control for the prevalence of same-sex unmarried households in 2000 and 1990, the only two decades when the Census provides this information, constructing two ratios: the number of homosexual relationships out of the total number of households and the number of lesbian relationships out of the total number of households. These metropolitan level controls ensure that our education sex ratio is an index of the tightness of the heterosexual marriage markets. Third, to the extent that the sizes of the male and female homosexual populations vary together, their impact on the validity of the sex ratio would be reduced (Fossett and Kiecolt 1991).

C. Marital Gains from Specialization

It is known that if the education of the husband is higher than the wife's, there are gains from the wife specializing in household production and thus working less in the labor market (Becker 1991). Our quality sex ratio may capture the presence of these gains, showing that when the education gap of married couples increases (i.e., the number of highly educated men increases), married women's labor supply decreases and their husbands' increases. However, this link cannot represent an alternative explanation to our bargaining power interpretation for three reasons. First, our sample consists of already married couples, while the specialization effect should be present only for couples formed after any sex ratio change. When we restrict our sample to older couples who are likely to have gotten married many years prior to the decade under analysis, our bargaining power interpretation still holds. Second, when we focus on a subsample of couples who did indeed perfectly sort by education brackets (i.e., no peculiar gain from specialization should be present for them), our results still hold. Third, we consider positive assortative mating within education brackets, so that men and women are affected by fluctuations in the sex ratio only in their own education group. In this case, the education gap of potential spouses and the corresponding gains from specialization are small.

D. Welfare Programs for Women

Welfare programs favorable to women may discourage female labor supply or increase the bargaining power of married women by enhancing the value of single motherhood. However, by definition, welfare programs benefit only low-income households, while our results hold for all levels of income. Additionally, there is no reason why the pattern of the main welfare benefits such as Aid to Families with Dependent Children (AFDC) (Temporary Assistance for Needy Families [TANF]), Earned Income Tax Credit (EITC), and mandated benefits should vary across metropolitan areas to be more favorable to women where women are relatively scarce.

VI. CONCLUSIONS

This paper further explores the role of sex ratios on intrahousehold bargaining power and spouses' labor supplies by constructing a quality sex ratio by metropolitan area and education brackets. We test whether this education sex ratio affects the balance of power of couples in the corresponding education brackets, within the framework of a collective labor supply household model. We also test that the bargaining power effect of our education sex ratio is greater as the assortative mating order by education increases. We use CPS and Census data for 2000, 1990, and 1980 at the metropolitan level and find that married women significantly reduce their supply of market labor, while their husbands increase theirs as the corresponding education sex ratio becomes more favorable to women. Couples with CC wives exhibit a stronger impact of the quality sex ratio on their bargaining power than those with SC wives, whose estimated quality sex ratio coefficient is in turn larger than for HS graduates. Our bargaining power interpretation is strengthened by the fact that unmarried men and women do not exhibit any significant impact of the education sex ratio on their labor supply. Alternative explanations such as local labor market opportunities, marital gains from specialization, welfare programs, and inclusion of married and same-sex partners in the sex ratio are rejected.

The findings presented here are consistent with theories where favorable sex ratios increase female bargaining power in the marriage market. Furthermore, our results indicate that both the local and the quality dimensions of sex ratios are relevant to explaining household behavior. Our evidence represents the first empirical support for the bargaining power effect of a quality sex ratio by education and for its stronger impact as higher levels of educational attainment are considered.

ABBREVIATIONS

CC: College-College Plus

CPS: Current Population Survey

HS: High School

MLE: Maximum Likelihood Estimator

SC: Some College

SF4: Summary File 4

doi: 10.1111/j.1465-7295.2008.00132.x

REFERENCES

Angrist, J. "How do Sex Ratios Affect Marriage and Labor Markets? Evidence from America's Second Generation." Quarterly Journal of Economics, 117, 2002, 997-1038.

Becker, G. A Treatise on the Family. Cambridge, MA: Harvard University Press, 1991.

Benham, L. "Benefits of Women's Education within Marriage." Journal of Political Economy, 82, 1974, S57-S71.

Bertrand, M., E. Duflo, and S. Mullainathan. "How Much Should We Trust Differences-in-Differences Estimates?" Quarterly Journal of Economics, 119, 2004, 249-75.

Blundell, R., P.-A. Chiappori, T. Magnac, and C. Meghir. "Collective Labor Supply: Heterogeneity and Nonparticipation." Mimeo, UCL, 2002.

Chiappori, P.-A, B. Fortin, and G. Lacroix. "Household Labor Supply, Sharing Rule and the Marriage Market." Journal of Political Economy, 110, 2002, 37-72.

Chiappori, P.-A., M. Iyigun, and Y. Weiss. "Gender Inequality, Spousal Careers and Divorce." Mimeo, Columbia University, 2005.

--. "Investment in Schooling and the Marriage Market." Mimeo, Columbia University, 2007.

Donni, O. "Collective Female Labor Supply: Theory and Application." The Economic Journal, 117, 2007, 94-119.

Eckstein, Z., and K. Wolpin. "Why Youths Drop Out of High School: The Impact of Preferences, Opportunities, and Abilities." Econometrica, 67, 1999, 1295-339.

Fossett, M., and J. Kiecolt. "A Methodological Review of the Sex Ratio: Alternatives for Comparative research." Journal of Marriage and the Family, 53, 1991, 941-57.

--. "Mate Availability and Family Structure among African Americans in U.S. Metropolitan Areas." Journal of Marriage and the Family, 55, 1993, 288-302.

Freiden, A. "The US Marriage Market," in Economics of the Family, edited by T. Schultz. Chicago: University of Chicago Press, 1974, 352-71.

Grossbard, S., and C. Amuedo-Dorantes. "Cohort-Level Sex Ratio Effects on Women's Labor Force Participation." Review of Economics of the Household, 5, 2007, 249-78.

Grossbard-Shechtman, A. "A Theory of Allocation of Time in Markets for Labour and Marriage." Economic Journal, 94, 1984, 863 82.

Grossbard-Shechtman, S. On the Economics of Marriage. A Theory of Marriage, Labor and Divorce. Boulder, CO: Westview Press, 1993.

Grossbard-Shechtman, S., and M. Neideffer. "Women's Hours of Work and Marriage Market Imbalances," in Economics of the Family and Family Policies, edited by I. Persson and C. Jonung. New York: Routledge, 1997, 100-16.

Iyigun, M., and R. Walsh. "Building the Family Nest: PreMarital Investments, Marriage Markets and Spousal Allocations." Review of Economic Studies, 74, 2007, 507-35.

Lichter, D., F. LeClere, and D. McLaughlin. "Local Marriage Markets and the Marital Behavior of Black and White Women." American Journal of Sociology, 96, 1991, 843-67.

Lundberg, S., and R. Pollak. "Bargaining and Distribution in Marriage." Journal of Economic Perspectives, 10, 1996, 139-58.

Mare, R., and C. Schwartz. "Trends in Educational Assortative Marriage from 1940 to 2003." California Center for Population Research, 017-05, 2005.

Pencavel, J. "Assortative Mating by Schooling and the Work Behavior of Wives and Husbands." The American Economic Review, 88, 1998a, 326-9.

--. "The Market Work Behavior and Wages of Women: 1975-94." The Journal of Human Resources, 33, 1998b, 771-804.

Qian, Z. "Changes in Assortative Mating: The Impact of Age and Education." Demography, 35, 1998, 279-92.

Rumberger, R.W. "Dropping Out of High School: The Influence of Race, Sex, and Family Background." American Educational Research Journal, 20, 1983, 199-220.

Smock, P.J. "Remarriage Patterns of Black and White women: Reassessing the Role of Educational Attainment." Demography, 27, 1990, 457-73.

Tiefenthaler, J. "The Productivity Gains of Marriage: Effects of Spousal Education on Own Productivity across Market Sectors in Brazil." Economic Development and Cultural Change, 45, 1997, 633-50.

Weiss, Y., and R. Willis. "Match Quality, New Information, and Marital Dissolution." Journal of Labor Economics, 15, 1997, S293-329.

BRIGHITA NEGRUSA and SONIA OREFFICE *

* We thank Pierre-Andre Chiappori, Murat Iyigun, Jeanne Lafortune, Curtis Simon, Robert Tollison, John Warner, the editor, and the referees for helpful comments and suggestions. Errors are ours.

Negrusa: Consultant, NERA Economic Consulting, 1166 Avenue of the Americas, New York, NY 10036. Phone 212 345 3388, Fax 212 345 4650, E-mail brighita. negrusa@nera.com

Oreffice: Assistant Professor, City College of New York-CUNY, Department of Economics, 160 Convent Avenue, New York, NY 10031. Phone 212 650 6205, Fax 212 650 6341, E-mail oreffic@gmail.com

(1.) As a robustness check, we expand our control group to include all unmarried household heads who are not separated ("separated" individuals might not represent a one--decision maker household). While the sample size increases, the education sex ratio effects remain insignificant.

(2.) Research shows that people consider mates drawn from relatively broad age ranges. While mean age differences between husbands and wives are relatively small, there is considerable variation around this central tendency as many marriages involve larger age differences. Competition and substitution across age categories are considerable (Fossen and Kiecolt 1993). Sex ratios accounting for wives being younger than husbands are reported to have the same impact (Chiappori, Fortin, and Lacroix 2002). We also compute the sex ratio for the age bracket 1844 and obtain similar results.

(3.) We refer the reader to "Sex Ratios As Proxy for Local Labor Market Conditions" subsection for a comprehensive discussion of alternative explanations.

(4.) We only exclude household observations where neither spouse works, given that this analysis measures bargaining power changes through labor supply. The inclusion of nonworking men is relevant, given the decrease in male labor force participation rates in the past 30 yr.

(5.) The latter are presence of young children or number of family members only affecting the participation decision but not labor supply. Tables report estimation with identification from statistical distribution assumptions.

(6.) The participation decision depends on the number of children, dummies for age brackets, education, race, and measures of local economy.

(7.) Tables report estimation with the predicted spouse's and own wages.

(8.) In CPS 2000, education is recorded as degrees attained rather than years of schooling completed as in 1990 and 1980. We thus assigned number of years of schooling to the corresponding degrees.

(9.) In 1980, the available cross-tabulations only provide the age bracket "25 and older." Also, the education brackets in Census are focused on years of schooling rather than degrees obtained as is the case of the following decades. However, the broad education categories and age ranges used to construct our education sex ratio are not affected.

(10.) The available number in 1980 is so small due to the fact that the CPS identifies only 44 metropolitan areas in 1980.

(11.) A 10 percentage point increase in one of our education sex ratios corresponds to at least two-thirds of a standard deviation of the ratio with respect to its mean.

(12.) The absence of such a bargaining power effect may be due to strong rigidities in the labor supply schedules of such low-educated couples, or to the lack of sorting behavior by this demographic group. See "Labor Force Participation" subsection for a more detailed discussion.

(13.) We also estimate the impact of our quality sex ratio on a subsample of couples who actually sort in marriage by education bracket, that is, on couples where wives' education belongs to the same education bracket as their husbands'. We find a similar pattern of results as in our main specification.

(14.) For the white subsample, EdR is computed using data only for white men and women.
TABLE 1
Summary Statistics

 2000

 Mean SD Mean SD

 White Black
Variable
Education ratio HS
 graduates 0.98 0.06 1.22 0.63
Education ratio SC 0.89 0.05 0.94 0.43
Education ratio 1.02 0.07 0.95 0.48
 college and above
Number of 173 173
 observations (a)

 Wives Husbands

Hours worked (b) 1,745.53 713.46 2,309.25 547.44
Log of wage (b) 2.59 0.72 3.01 0.64
Age 39.18 7.88 41.2 8
Education 14.19 2.16 14.44 2.37
Household nonlabor income 6,153 14,748 6,153 14,748
Number of children 0.415 0.7 0.415 0.7
 younger than 6 yr
Number of family members 3.5 1.19 3.5 1.19
Dummy for black 0.088 0.28 0.088 0.28
Number of observations 9,235 9,235

 Single Men Single Women

Hours worked (b) 1,759.24 652.6 2,098.87 583.45
Log of wage (b) 2.24 0.77 2.66 0.58
Age 32.81 8.88 37.92 8.44
Education 13.17 1.68 13.68 2.15
Household nonlabor income 3,601 5,971 6,885 13,938
Number of children 0.53 0.71 0.094 0.36
 younger than 6 yr
Number of family members 2.75 1.02 2.36 0.8
Dummy for black 0.6 0.48 0.3 0.46
Number of observations 696 153

 1990

 Mean SD Mean SD

 White Black
Variable
Education ratio HS
 graduates 0.86 0.07 1.03 0.41
Education ratio SC 0.92 0.06 0.95 0.47
Education ratio 1.21 0.09 1.06 0.53
 college and above
Number of 181 181
 observations (a)

 Wives Husbands

Hours worked (b) 1,638.36 722.51 2,223.85 568.64
Log of wage (b) 2.19 0.67 2.69 0.61
Age 40.49 9.75 42.88 10.13
Education 13.8 2.03 14.33 2.22
Household nonlabor income 5,154 12,569 5,154 12,569
Number of children 0.37 0.68 0.37 0.68
 younger than 6 yr
Number of family members 3.4 1.19 3.4 1.19
Dummy for black 0.07 0.25 0.07 0.25
Number of observations 11,894 11,894

 Single Men Single Women

Hours worked (b) 1,775.76 632.01 2,057.73 525.43
Log of wage (b) 1.98 0.63 2.34 0.65
Age 32.21 8.08 34.59 10.3
Education 12.96 1.68 13.71 2.04
Household nonlabor income 3,634 5,433 5,352 9,264
Number of children 0.58 0.77 0.11 0.43
 younger than 6 yr
Number of family members 2.76 1.05 2.46 0.8
Dummy for black 0.62 0.48 0.19 0.39
Number of observations 562 185

 1980

 Mean SD Mean SD

 White Black
Variable
Education ratio HS
 graduates 0.7 0.04 0.78 0.09
Education ratio SC 0.96 0.08 0.9 0.14
Education ratio 1.5 0.08 0.95 0.25
 college and above
Number of 34 34
 observations (a)

 Wives Husbands

Hours worked (b) 1,464.74 743.14 2,211.26 492
Log of wage (b) 1.57 0.75 2.21 0.59
Age 36.33 8.87 39.17 9.42
Education 13.44 1.9 13.44 1.9
Household nonlabor income 1,810 5,572 1,810 5,572
Number of children 0.42 0.69 0.42 0.69
 younger than 6 yr
Number of family members 3.66 1.31 3.66 1.31
Dummy for black 0.075 0.29 0.075 0.29
Number of observations 4,597 4,597

 Single Women Single Men

Hours worked (b) 1,670.07 629.77 2,001.37 528.29
Log of wage (b) 1.55 0.62 2.02 0.54
Age 34.98 9.88 35.61 10.43
Education 12.9 1.57 13.87 1.92
Household nonlabor income 3,076 4,907 2,923 4,657
Number of children 0.45 0.68 0.072 0.26
 younger than 6 yr
Number of family members 2.97 1.13 2.62 0.94
Dummy for black 0.47 0.49 0.19 0.39
Number of observations 548 107

Note: The sample contains data from the CPS March supplement and
U.S. Census years 2000, 1990, and 1980.

(a) Number of Census metropolitan areas present in the CPS sample.

(b) For women and men with positive hours of work.

TABLE 2
Estimation of the Labor Supply Regressions of Wives and Husbands

 2000

Variable Wives (1) Husbands (2)

Education ratio -59.47 (102.25) -31.77 (92.61)
Education ratio x -100.7 *** (31.38) 40.27 * (22.18)
 dummy SC
Education ratio x -263.44 *** (71.65) 96.74 ** (50.88)
 dummy CC
Log of wage of wife 694.59 * (375.9) 313.17 (336.22)
Log of wage of 738.83 (463.87) 126.76 (197.89)
 husband
Age of husband -17.43 (13.64) 5.11 (6.33)
Age of wife 11.83 (8.77) 5.65 (5.45)
Education of husband -99.52 *** (38.09) 9.34 (19.64)
Education of wife 36.1 (31.73) -39.1 * (24.03)
Household nonlabor -1.44 *** (0.59) 0.02 (0.58)
 income
Number of children -149.48 *** (16.43) 3.12 (10.99)
 younger than 6 yr
Number of household -106.95 (8.70) 20.84 *** (5.26)
 members
Dummy for black 244.52 ** (102.66) -133.97 ** (68.06)
Inverse Mill's ratio 23.33 (-12.08) 12.91 (-13.05)
Observations 9,235 9,235

 1990

Variable Wives (3) Husbands (4)

Education ratio 169.41 (118.95) -78.48 (85.21)
Education ratio x -102.59 ** (42.72) 56.38 * (29.67)
 dummy SC
Education ratio x -175.17 ** (73.23) 131.48 *** (53.28)
 dummy CC
Log of wage of wife 1,617.47 * (945.66) -223.51 (685.21)
Log of wage of 387.23 (371.11) 1,346.66 *** (339.68)
 husband
Age of husband -4.31 (4.1) 15.54 *** (5.31)
Age of wife 4.47 (6.73) -5.05 (4.59)
Education of husband -46.86 * (28.08) -217.06 *** (57.63)
Education of wife -130.71 (98) -8.42 (69.94)
Household nonlabor -3.59 *** (0.82) -3.1 *** (0.77)
 income
Number of children -170.84 *** (20.27) 5.29 (9.75)
 younger than 6 yr
Number of household -121.54 (9.8) 1.9 (5.33)
 members
Dummy for black 356.63 *** (62.4) -19.91 (53.27)
Inverse Mill's ratio 60.45 (-14.21) -0.12 (-9.47)
Observations 11,894 11,894

 1980

Variable Wives (5) Husbands (6)

Education ratio 336.75 (229.46) -187.09 (176.25)
Education ratio x -192.5 * (108.84) 118.33 * (69.88)
 dummy SC
Education ratio x -363.75 ** (188.52) 183.5 * (114.04)
 dummy CC
Log of wage of wife 311.13 (1,211.88) 385.57 (875.13)
Log of wage of 2,147.05 *** (817.68) -1,328.39 *** (514.13)
 husband
Age of husband -1.14 (5.05) 3.91 * (2.09)
Age of wife -7.62 (7.85) -0.66 (5.7)
Education of husband -155.44 *** (56.22) 113.57 *** (35.62)
Education of wife 56.91 (118.27) -45.91 (84.24)
Household nonlabor -1.9 (1.9) -4.1 *** (1.7)
 income
Number of children -270.4 *** (24.27) 27.95 ** (12.3)
 younger than 6 yr
Number of household -100.02 (12.33) 13.21 (3.80)
 members
Dummy for black 428.53 *** (69.18) -213.41 *** (66.56)
Inverse Mill's ratio 85.79 (-17.8) 101.49 (-18.59)
Observations 4,597 4,597

Notes: Data are from the CPS March supplement and U.S. Census
years 2000, 1990, and 1980. Estimated coefficients and standard
errors (in parentheses) are clustered by metropolitan area. All
tables report regressions with the same set of covariates
described in Section III. Regressions are corrected for sample
selection with Heckman MLE. Parameters for household nonlabor
income are multiplied by 1,000.

Significant coefficients are bolded.

* Significant at 10%; ** significant at 5%; *** significant at 1%.

TABLE 3
Effect of Education Ratio on Annual Hours Worked of Couples and
Singles

 2000

Variable Wives (1) Husbands (2)

Education ratio -59.47 (102.25) -31.77 (92.61)
Education ratio x dummy SC -100.70 *** (31.38)# 40.27 * (22.18)#
Education ratio x dummy CC -263.44 *** (71.65)# 96.74 ** (50.88)#
Number of observations 9,235 9,235

 Single Women (7) Single Men (8)

Education Ratio -93.49 (312.81) 793.40 (678.12)
Education ratio x dummy SC -95.46 (149.94) -343.37 (439.32)
Education ratio x dummy CC 32.02 (225.08) -535.88 (613.58)
Number of observations 696 153

 1990

Variable Wives (3) Husbands (4)

Education ratio 169.41 (118.95) -78.48 (85.21)
Education ratio x dummy SC -102.59 ** (42.72)# 56.38 * (29.67)#
Education ratio x dummy CC -175.17 ** (73.23)# 131.48 *** (53.28)#
Number of observations 11,894 11,894

 Single Women (9) Single Men (10)

Education Ratio 49.40 (256.50) 364.53 (590.63)
Education ratio x dummy SC -61.76 (184.88) -212.91 (229.04)
Education ratio x dummy CC 214.63 (297.43) -264.91 (408.08)
Number of observations 562 185

 1980

Variable Wives (5) Husbands (6)

Education ratio 336.75 (229.46) 187.09 (176.25)
Education ratio x dummy SC -192.50 * (108.84)# 118.33 * (69.88)#
Education ratio x dummy CC -363.75 ** (188.52)# 183.5 * (114.04)#
Number of observations 4,597 4,597

 Single Women (11) Single Men (12)

Education Ratio -130.14 (473.75) -1,005.82 (865.94)
Education ratio x dummy SC 65.98 (200.60) 1.90 (385.56)
Education ratio x dummy CC 208.15 (325.33) 417.66 (522.43)
Number of observations 548 107

Notes: Data are from the CPS March Supplement and U.S. Census
years 2000, 1990, and 1980. Estimated coefficients and standard
errors (in parentheses) are clustered by metropolitan area. All
tables report regressions with the set of covariates described in
Section III. Regressions are corrected for sample selection with
Heckman MLE. Single individuals report marital status "never
married." In 1980, nonseparated, unmarried individuals were
included due to scarcity of observations.

Significant coefficients are bolded.

* Significant at 10%; ** significant at 5%; *** significant at 1%.

Note: Significant coefficients indicated with #.

TABLE 4
Effect of Education Ratio on Annual Hours Worked of Unmarried,
Nonseparated Individuals

 2000

Variable Women Men

Education ratio -42.16 (232.22) 483.13 (427.22)
Education ratio x dummy SC 16.98 (112.7) -131.13 (203.74)
Education ratio x dummy CC 84.48 (167.76) -141.36 (321.88)
Number of observations 1,660 408

 1990

Variable Women Men

Education ratio -84.1 (155.4) -58.37 (379.72)
Education ratio x dummy SC -78.68 (77.02) 4.07 (131.6)
Education ratio x dummy CC 60.12 (138.19) 5.17 (248.37)
Number of observations 2,063 455

Notes: Data are from the CPS March Supplement and U.S. Census
years 2000 and 1990. Estimated coefficients and standard errors
(in parentheses) are clustered by metropolitan area. Regressions
run with the covariates described in Sec tion 111. Regressions
are corrected for sample selection with Heckman MLE. In 1980, the
estimated coefficients shown in Table 3 already refer to the
nonseparated sample, as reported in the footnote to that table.

TABLE 5
Effect of Education Ratio on Annual Hours Worked of Older Couples

 2000

Variable Wives (1) Husbands (2)

Education ratio 59.05 (97.94) -6.96 (103.62)
Education ratio x dummy SC -98.23 *** (38.70)# 71.1 *** (28.87)#
Education ratio x dummy CC -341.92 *** (95.23)# 157.02 ** (57.78)#
Number of observations 6,166 6,166

 1990

Variable Wives (3) Husbands (4)

Education ratio 278.6 (182.1) -160.57 (117.94)
Education ratio x dummy SC -126.15 ** (62.33)# 66.3 * (38.25)#
Education ratio x dummy CC -236.28 ** (101.01)# 159.1 ** (69.09)#
Number of observations 6,501 6,501

Notes: Data are from the CPS March Supplement and U.S. Census
years 2000 and 1990. Estimated coefficients and standard errors
(in parentheses) are clustered by metropolitan area. Regressions
with the same covariates are described in Section 111.
Regressions are corrected for sample selection with Heckman MLE.
Old couples are those with wives aged 34 65 yr and husbands aged
38-65 yr.

Significant coefficients are bolded.

* Significant at 10%; ** significant at 5%;
*** significant at 1%.

TABLE 6
Effect of Education Ratio on Annual Hours Worked of White Couples

 2000

Variable Wives Husbands

Education ratio 119.31 (183.82) -109.29 (133.66)
Education ratio x dummy SC -83.45 ** (37.81)# 38.27 * (24.11)#
Education ratio x dummy CC -256.03 *** (78.19)# 120.78 ** (57.51)#
Number of observations 8,546 8,546

 1990

Variable Wives Husbands

Education ratio 15.45 (157.73) -141.51 (100.67)
Education ratio x dummy SC -83.24 ** (42.96)# 53.75 * (32.65)#
Education ratio x dummy CC -139.40 * (75.83)# 136.38 ** (58.27)#
Number of observations 11,167 11,167

Notes: Data are from the CPS March Supplement and U.S. Census
years 2000 and 1990. Estimated coefficients and standard errors
(in parentheses) are clustered by metropolitan area. Regressions
run with the covariates described in Section III. Regressions
are corrected for sample selection with Heckman MLE.

Significant coefficients are bolded.

* Significant at 10%; ** significant at 5%; *** significant at 1%.

TABLE 7
Effect of Education Ratio on Labor Force Participation of Couples
and Singles

 2000

Variable Wives (1) Husbands (2)

Education ratio -.0044 (0.057) -.011 (0.013)
Education ratio x dummy SC .011 (0.014) .0067 * (0.004)#
Education ratio x dummy CC -.023 (0.03) .017 ** (0.008)#
Number of observations 9,235 9,235

 Single Women (7) Single Men (8)

Education ratio .284 (0.18) -.104 (0.123)
Education ratio x dummy SC 055 (0.041) -.026 (0.056)
Education ratio x dummy CC -.076 (0.11) 03 (0.13)
Number of observations 696 153

 1990

Variable Wives (3) Husbands (4)

Education ratio -.058 (0.074) -.004 (0.019)
Education ratio x dummy SC -.032 (0.02) 004 (0.006)
Education ratio x dummy CC -.062 * (0.035)# 009 (0.01)
Number of observations 11,894 11,894

 Single Women (9) Single Men (10)

Education ratio 058 (0.194) -.156 (0.161)
Education ratio x dummy SC -.087 (0.121) 111 (0.082)
Education ratio x dummy CC -.189 (0.192) 128 (0.201)
Number of observations 562 185

 1980

Variable Wives (5) Husbands (6)

Education ratio 281 (0.18) -.02 (0.022)
Education ratio x dummy SC -.179 ** (0.073)# 004 (0.006)
Education ratio x dummy CC -.325 *** (0.116)# .02 * (0.012)#
(0.012)
Number of observations 4,597 4,597

 Single Women (11) Single Men (12)

Education ratio 035 (0.255) 145 (0.213)
Education ratio x dummy SC -.027 (0.122) .01 (0.13)
Education ratio x dummy CC -.035 (0.241) -.041 (0.182)
Number of observations 548 107

Notes: Data are from the CPS March Supplement and U.S. Census
years 2000, 1990, and 1980. Estimated changes in probability and
standard errors (in parentheses) are clustered by metropolitan
area.

Significant coefficients are bolded.

* Significant at 10%; ** significant at 5; *** significant at 1%.
COPYRIGHT 2010 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Negrusa, Brighita; Oreffice, Sonia
Publication:Economic Inquiry
Geographic Code:1USA
Date:Jul 1, 2010
Words:11328
Previous Article:Does the river spill over? Estimating the economic returns to attending a racially diverse college.
Next Article:Piece-rate contracts for other-regarding workers.
Topics:

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters