2.1 Previous literature
Previous theories on the determinants of intermarriage also have implicit predictions about relative market labor supply by spouses. Intermarriages exhibit positive assortative mating among spouses—natives and immigrants with similar level characteristics tend to intermarry. This is particularly true of individuals with higher levels of education and proficiency in the English language (Chiswick and Houseworth 2011; Furtado 2012). Immigrants who arrive in the USA at an early age and have a similar experience of US life as natives have a higher propensity to intermarry. Nottmeyer (2014) also shows that intermarried immigrants score high on the personality traits of “openness” and “extraversion.” Immigrants and natives who intermarry may be less inclined to follow gender-based household division of labor.
Becker’s work on family division of labor (1981; 1985) posits that husbands and wives choose to specialize in market versus house work based on their productivities and abilities. Couples choose the degree of specialization based on the expected costs and benefits (Stratton 2005). Married households can allocate spousal hours more efficiently, compared to single people, to maximize home production and work income (Lundberg and Pollak 2007). The natural assumption is that a partnership decision like marriage will affect not only how individuals decide to supply labor but also how spouses jointly decide on their market labor supply.
A couple that is similarly matched on human capital variables like education or language proficiency is unlikely to have differing comparative advantages. Benefits from specialization are limited. Intermarriages, where partners have more equal endowments of human capital, may be characterized by more equal division of labor market work. However, it is not clear that this “equality” is true for both male and female intermarriages. The average levels of education can differ by sex of the native partner (Jasso et al. 2000). For example, in the USA, native husbands and their wives have substantially more schooling than native wives and their husbands. This can affect traditional gender-based division of market labor heterogeneously by sex of the immigrant. Also, couples may not always match on human capital traits—for example, if natives marry immigrant women from traditional societies to reemphasize home-building and child-rearing (Basu 2015), household specialization might be efficient in certain intermarried female families. Educated immigrants, particularly women, may exchange their labor market aspirations for higher societal status from marriage to a native (Grossbard-Shechtman 1993)—this is a special case of status exchange.
The productivity hypothesis of intermarriage states that immigrants marry natives to compensate for their lack of host-country-specific human capital.Footnote 5 Intermarriage can encourage an immigrant to acquire more human capital, via added incentives in learning the language and culture of the home country as well as attachment to the labor market (Meng and Gregory 2005).Footnote 6 English proficiency or extended stay in the USA is not merely determinants of intermarriage; rather, intermarriage affects these variables and via these channels can affect household specialization. We recognize that these immigrant characteristics can change over the course of the marriage. Given their importance in determining type of marriage and labor market hours, we retain them as controls in our subsequent estimations.
A native spouse can also reduce information costs surrounding local job markets and institutions and increase employment opportunities (Furtado and Theodoropoulos 2010). Factors that improve an immigrant’s labor market options can imply similar work hours as their native spouse and thereby lower household specialization.
In certain cases, acquisition of human capital over the course of a marriage might occur in intra-marriages. Labor market effort and human capital investment can be better coordinated among intra-married couples and not intermarried couples. Baker and Benjamin (1997) show for Canada, upon arrival, intra-married immigrant women work in low-paying, high-hour jobs to finance their husband’s human capital investment in a credit-constrained labor market. Eventually, both husbands and wives move to better jobs. This is the family investment hypothesis in intra-marriages (Eckstein and Weiss 2002). Intermarried wives do not have to perform this borrowing function for native husbands.Footnote 7 Intermarriages are also characterized by high family incomes (Pew Research Center 2012). The income effect of a high-earning native husband can reduce a wife’s labor supply (Basu 2015).
The above discussion indicates that own and relative (to spouse) human capital—like education and age—are important determinants of intermarriage and household market specialization. The discussion also indicates that the variables do not affect male and female immigrants similarly, and hence, effects of intermarriage on household labor supply should be considered separately for male and female immigrants.
Besides human capital, the distribution of bargaining power can differ between intermarriages and intra-marriages. Cross-racial and cross-nativity marriages are more likely to end in divorce (Milewski and Kulu 2014; Adserà and Ferrer 2014). If options outside the marriage are more attractive, and the threat of divorce is greater, gender-based division of labor is less optimal in the marriage (Becker 1985). Specialization is attractive when a marriage is expected to continue. More children increase the opportunity cost of market hours particularly for the primary-care giver, often the mother. Intermarried households have lower fertility rates and are less stable than immigrant households. Overall, intermarriages should be characterized by less specialization than intra-married households, though clearly the sex of the intermarried immigrant matters. Also based on this discussion, characteristics of the marriage like duration of the marriage, number of children, and presence of young children are important determinants of household specialization.
On the other hand, cohabiting with a native, with or without marriage, can tilt bargaining power away from an immigrant. Marriages are costly to dissolve, compared to cohabitations (Stratton 2005). Household bargaining models posit that the partner with the more attractive outside options can dictate household allocation of hours (Lundberg and Pollak 1996).Footnote 8 In any form of native-immigrant partnerships, this is presumably the native partner. In legal marriages particularly, the immigrant can depend on the native spouse for legal residence in the USA. This paper also examines unmarried cohabiting native-immigrant couples to identify the sources of bargaining power—the act of marriage over and above cohabitation with a native.
A higher sex ratio, defined as the proportion of immigrant women to men available in one’s marriage market, increases the rate of intra-marriage for men and lowers it for women. Regional sex ratios are also inversely related to married women’s labor force participation in the USA (Grossbard and Amuedo-Dorantes 2008). Clearly, these variables affect the household specialization of married couples, and the effects differ by sex of the immigrant. Our regressions control for state of residence and regional female-male sex ratio. Also included is the proportion of immigrants from one’s birthplace that lives in one’s metropolitan—this variable seeks to measure the segregation of immigrants.
An immigrant’s place of birth influences their labor market participation (Basu 2016).Footnote 9 Culture has an important impact (Gevrek et al. 2013), and immigrant women from countries with higher female labor participation exhibit the same in the host country. Source-country differences play an important role in immigrant labor supply and therefore household specialization.
Previous studies relating intermarriage to labor market outcomes of immigrants have concentrated on immigrant wages or employment. Intermarriage can affect how a couple allocates their labor market hours. Nottmeyer (2014) finds that intermarriages in Germany are characterized by lower labor market specialization for both men and women. Conducting a similar exercise for immigrants in the USA would present a clearer picture of immigrant income assimilation. A study of the sources of gender differences in household specialization can contribute to our understanding of the US gender wage gap, as more immigrants enter the labor force.
2.2 Variable construction and empirical specification
The variable of interest S
ih
, or the dependent variable, is the degree of specialization in labor market hours in household h. The index is constructed from the point of view of immigrant i. Following previous work (Stratton 2005; Bonke et al. 2008; Nottmeyer 2014), this is defined as:
$$ {S}_{i h}=\left(\frac{ \max \left\{{H}_{i, h};{H}_{j, h}\right\}}{H_{i, h}+{H}_{j, h}}-0.5\right)*2 $$
where H
i,h
and H
j,h
= usual weekly market hours supplied by immigrant i and spouse j, respectively. Spouse j can be a native or another immigrant. It bears clarification that while the index carries the subscript i, it is common for spouses i and j. S
ih
is a degree of specialization chosen by the household h based on the abilities of the spouses. Subscript i helps to differentiate between spouses, since individual and relative characteristics will be included in the estimation.Footnote 10
The value of this measure ranges from 0, where hours supplied by both spouses is the same. This is the case of no specialization in labor market work. The other extreme is complete specialization in labor market work by one spouse, while the partner stays out of the labor market entirely. Here, the index takes a value of 1. The index allows for a continuum of incomplete but increasing specialization between the values of 0 and 1.
This index is gender-neutral. In our sample, which we discuss in the next section, households generally follow the traditional model of male-breadwinner-female-homemaker. Thus, despite its gender-neutrality, if household h has a higher value of the specialization index than household h′, it is likely that household h follows more traditional gender roles. A concern with this index is that it uses aggregate weekly labor market hours worked by individuals—this is a facet of the data. A person who works more during the weekdays may trade off with their spouse over the weekends—thereby pointing at specialization—but in aggregate, this might not be visible.
By construction, the index is bounded between 0 and 1. It has positive mass at these limits. The two-limit Tobit estimation is used to model corner solutions. This is a special case of the censored regression model, where the dependent variable is simultaneously censored from above and below. The model supposes there is a latent unobserved variable S
*
ih
= β Inter
ih
+ γX
ih
+ ε
ih
, with ε
ih
as a normally distributed error term ~N(0, σ). The observed specialization S
ih
equals latent variable S
*
ih
when it lies between 0 and 1. Spouses solve their labor market hour allocation comparing the costs and benefits from different degrees of specialization. The result of this optimization exercise is S
ih
.
The latent variable linearly depends on the vector of observed explanatory variables via the coefficient parameter. The important explanatory variable is Inter
ih
which equals 0 if both spouses are immigrant and 1 if exactly one spouse is a native.Footnote 11 The coefficient of interest is β. Estimations are conduced separately for married male and female immigrants, and we obtain different values of β depending on gender. The coefficient β shows the average difference in observed specialization between intermarried and intra-married immigrants, for a given gender.
Based on our discussion of intermarriages and specialization in these marriages, other controls included in vector X
ih
are the following: education of spouse i, a square term in experience for i, more years of education and age for i compared to spouse j,Footnote 12 years spent in the USA by immigrant i, and their English language proficiency. Veteran status of both spouses are also included. In addition, marriage characteristics—like age of marriage of i, duration of marriage, family size, number of children, and age of the youngest child in the household—are also included. Additionally, we include a sex ratio variable estimating the proportion of women to men for i’s age group, country of birth, and metropolitan statistical area (MSA). A control for immigrant concentration is also included—which specifies the proportion of own-country immigrants living in one’s MSA.Footnote 13 Birthplace controls and state fixed effects are included in some specifications.
The coefficients from a Tobit estimation are interpreted similar to ordinary-least-square coefficients. The linear effects are on the uncensored latent variable, not the observed outcome. Marginal effects, assuming the specialization index is greater than 0 (excluding cases of no specialization), or less than 1 (excluding cases of complete specialization) or where the specialization index lies between 0 and 1 (incomplete specialization), are available upon request. While the magnitudes of these effects are different from the results presented in subsequent tables, qualitative results were the same.