Migration, Household Tasks, and Gender: Evidence from the Republic of Georgia

Abstract

We examine whether migration affects the gender division of household tasks and participation in leisure within origin-country households using survey data from the Republic of Georgia. Our theoretical framework identifies two sets of mechanisms whereby migration might influence gender differences in home activities: migrant experience effects and migrant absence effects. We test for both types of effects on the probability that men and women perform gender atypical household tasks and engage in leisure activities by comparing households with and without currently absent and return migrants using probit regressions. We find evidence for both migration absence and migration experience effects on gender differences in housework and leisure. However, these effects are complex and contradictory: generally, male migration tends to exacerbate gender differences in the sending household while female migration tends to ameliorate them.

INTRODUCTION

International migration can re-shape gender norms and relations in a variety of ways. Most obviously, migrants themselves often encounter different ideas and expectations about gender roles and engage in different gender-specific behaviors in host countries compared to their origin countries (e.g. Hondagneu-Sotelo 1992; Menjívar 1999; Parrado and Flippen 2005). Recent scholarship suggests that international migrant flows can also influence gender relations in migrant-sending countries (Menjívar and Agadjanian 2007; Hoang and Yeoh 2011). We analyze how migration affects gender differences in an important arena – household activities – in the Republic of Georgia. Specifically, we investigate whether men and women who return from migration and those who are left behind are more or less likely to engage in gender-atypical household tasks and to participate in leisure activities. By doing so, we bring a developing country and the process of migration into the literature on gender and housework.

We propose two sets of mechanisms whereby international migration can influence gender differences in household tasks and leisure time in migrant-sending countries: migrant absence effects refer to reallocations of household tasks among those left behind necessitated by the departure of one member of the household; migrant experience effects are reallocations that result when migrants return to the household with different gender norms, expectations, or bargaining power due to the experience of living and working abroad. These concepts highlight distinct processes and temporal dimensions, but they are not mutually exclusive. Migrant absence effects stem from the functional imperative that someone else perform tasks the migrant used to do in the household. They might be strictly temporary, lasting only so long as the migrant remains abroad, unless the left-behind household members become accustomed to doing the new tasks. Migrant experience effects, by contrast, are probably more enduring changes in family relations. If migrants adopt different gender norms and/or acquire new economic resources while abroad, they may have both the will and the leverage to renegotiate time allocation for the longer term when they return to their left-behind households.

The Republic of Georgia is an especially suitable site for a study of migrant absence and migrant experience effects due to its classically patriarchal gender relations (Kandiyoti 1988; Hofmann forthcoming) and high levels of labor out-migration (Badurashvili 2004; Hofmann and Buckley 2012; Gerber and Torosyan 2013). We analyze data from a 2008 household survey, Georgia on the Move (GOTM), especially designed to assess the impact of migration on Georgian households along various dimensions. Our analyses indicate that international migration does alter gender differences in household tasks and leisure time. We find robust evidence for both migrant absence and migrant experience effects. However, the pattern of effects is complex and differs for male and female migrants and household members.

PRIOR LITERATURE ON GENDER, MIGRATION, AND HOUSEWORK

Two major strands of research pertain to the issue of whether international migration affects gender differences in housework and leisure in migrant-sending communities: studies of broader linkages between gender and migration and the literature on gender differences in time allocated to housework and (often implicitly) leisure.

Gender and Migration

Many studies have examine how gender relations in sending countries affect migration decisions (Massey et al. 2006; Stecklov et al. 2010; Nobles and McKelvey 2012). Here we focus on the reverse: potential effects of migration on gender relations. Some have argued that migration can help emancipate women from patriarchal attitudes and practices that often prevail in migrant-sending countries (Boserup 1970; Guendelman and Perez-Itriago 1987; Grasmuck and Pessar 1991). They note that migrant-receiving countries provide more access to jobs and income for migrant women than they would have in their origin societies. Access to these resources reduces their dependency on men and the family, enhancing their bargaining leverage vis-à-vis husbands – for example, by making divorce threats more credible and effective (Foner 1987).

Others conclude that the relationship between international migration and gender relations is uneven and shifting (Kibria 1990; Pessar and Mahler 2003; Parrado and Flippen 2005; Parrado, Flippen, and McQuiston 2005). Women’s employment does not necessarily lead to independence, nor does it automatically undermine patriarchal gender relations at home (Ferree 1979; Castro 1986; Kibria 1990; Menjívar 1999; George 2005). The type of labor market incorporation, contexts of reception, and assimilation patterns migrants experience influence whether and how international migration affects gender relations in their households (Glenn 1986; Espiritu 1997; Pessar 1999; Parrado and Flippen 2005; Derby and Schmalzbauer 2013). In sum, the transformation of gender relations among international migrants is complex and multi-dimensional: migration might promote gender egalitarianism in some cases, but not others.

These studies examine migrants while they are living abroad, leaving open the question of whether migrants’ experiences continue to shape their gender norms and relations when they return home, as many migrants do. Moreover, in migrant-sending countries the impact of international migration depends not only on how migrants’ experiences carry over upon their return, but also on how household members adapt to a migrant’s absence. Norms are difficult to measure on surveys. Here, we focus on housework activities and leisure, concrete practices that yield insight into gender relations. A large body of social science scholarship, albeit one focused mainly on developed countries, examines gender differences in housework and leisure.

Gender, Housework, and Leisure

Sex differences in the allocation of time to housework and leisure are a central aspect of patriarchal gender regimes, which rely on the separation between private and public spheres and allocate money-making activities to men and household activities to women (Connell 1987). In many countries women, even when employed, spend much more time doing housework than men (England and Farkas 1986; Evertsson and Nermo 2004; Heisig 2011). Different tasks in the home carry distinctively gendered meanings: women do much of the routine and care work while men do time-flexible activities associated with material provision and maintenance of the house (Twiggs, McQuillian and Ferree, 1999). The two main theoretical frameworks that explain variation, reproduction, and change in housework patterns focus, respectively, on economic bargaining and gender norms and performance.

Drawing on rational choice models of household relations (England and Farkas 1986), the economic perspective assumes housework is universally undesirable and individuals seek to minimize time spent on household tasks and maximize leisure time using their economic power as bargaining leverage (Bittman et al. 2003). Because earnings provide bargaining power, individuals with higher shares of earnings can shift housework to their partners. Empirically, individuals who earn the most do less housework (Brines 1994; Bittman et al. 2003; Evertsson and Nermo 2004; Schneider 2011). Therefore, if international migration changes the distribution of bargaining power within the household – for example, by providing migrant women with more earning power or income – it could have consequences for the allocation of household tasks and leisure time between women and men in the household.

Gender theory proposes that gender norms and expectations, not economic leverage, are the key source of differences between men and women in time allocated to household tasks and leisure. Daily activities such as ironing, cleaning dishes, or cooking constitute performances that accomplish and reinforce socially constructed gender identities and relations (West and Zimmerman 1987). Household tasks are not universally undesirable, but they are differently associated with ideals of masculinity and femininity. Women do more housework than men even after controlling for relative earnings, and the allocation of housework and leisure time between partners is less sensitive to the relative earnings of male and female partners than the bargaining theory implies (Brines 1994; Bittman et al. 2003; Evertsson and Nermo 2004; Schneider 2011). The patterning of norms and expectations linking gender to specific household tasks and leisure activities varies across nations based in part on economic, political, and cultural factors (Davis and Greenstein 2004; Fuwa 2004; Hook 2010; Treas and Tai 2012). By implication, international migration can affect gender differences in household tasks and leisure time by altering the gender norms and identities of migrants.

TWO MECHANISMS: MIGRANT ABSENCE AND MIGRANT EXPERIENCE EFFECTS

We distinguish two main types of effects that international migration can have on gender differences in housework and access to leisure within sending-countries. Migrant absence effects refer to changes in the division of household labor that occur in response to the absence of a member of the family. These rearrangements might depart from prevailing gender norms by necessity. Male migration can compel women to take on male tasks, which might increase their authority in the household (Parrado, Flippen, and McQuiston 2005; Menjívar and Agadjanian 2007). The absence of a female migrant might require men to conduct female-type activities: some qualitative studies describe men increasing their housework in women’s absence (Gramburd 2000; Hoang and Yeoh 2011), even if they resent it (Asis, Huang, and Yeoh 2004). However, responsibilities for housework, childcare, and eldercare may be difficult to transfer to men when women migrate (De Jong 2000; Curran, Garip, and Chung 2005). The presence of other female relatives in the household to take on the tasks of absent migrant women can attenuate men’s involvement in female-typical housework (Parreñas 2005; Derby 2010). Men might even reduce their housework when their partners migrate, as if their migration signaled a gender deviant arrangement that must be compensated for (Ashwin and Lytkina 2004; Parreñas 2005).

Most of these findings rely on qualitative studies, so it is unclear precisely how common migrant absence effects are and what factors predict whether they occur or not. Moreover, none indicate whether ostensible migrant absence effects persist over time or change upon migrants’ return: left-behind household members could become accustomed to doing the tasks previously performed by the migrant and persist in doing them even after the migrant returns. Or the return of the migrant could result in a restoration of the status quo ante, rendering migrant absence effects temporary.

Migrant experience effects are changes that happen when the migrant returns to the origin country with different bargaining power and/or gender norms as a result of their migration experience. For these effects to occur, migrants’ experiences have to either transform gender norms and expectations or provide them with more bargaining power. Empirical studies provide evidence that gender norms and relative bargaining power can shift in ways favoring more gender egalitarian arrangements while migrants are living abroad (Guendelman and Perez-Itriago 1987; Hondagneu-Sotelo 1992; Parrado and Flippen 2005; Parrado, Flippen, and McQuiston 2005). But the extent of change depends on the types of jobs migrant women have and the status of migrant men (Glenn 1986; Menjívar 1999).

Whatever happens while migrants are abroad, migrant experience effects also require that these experiences travel back home along with the migrant. Little is known about this additional step, and whether and how often it happens is impossible to say a priori. For example, returned migrant women might have greater leverage to renegotiate housework arrangements so that they perform fewer female-typed household tasks, which would constitute a migrant experience effect on the returned migrant, and their male partners might be compelled to do more female-typed housework, a migrant experience effect on non-migrant household members. But qualitative research provides cases where gender norms in the sending country impede such changes (Guendelman and Perez-Itriago 1987), particularly where female migration is stigmatized (Hoffman and Buckley 2012).

Altogether, plausible theoretical cases can be made for both migrant absence and migrant experience effects on gender differences in housework and access to leisure within households in migrant-sending countries. But it is unclear just how typical and enduring these effects are. These issues constitute an important and continuing gap in our understanding of the potential impact of migration on gender relations in sending countries.

GENDER AND MIGRATION IN THE REPUBLIC OF GEORGIA

The Republic of Georgia is a particularly suitable site for analyzing the link between international migration and gender relations due to its tradition of patriarchal gender relations and its high levels of both male and female labor migration. Historically, Georgia’s gender regime has been one of “classic patriarchy” characterized by patrilineal extended households where senior men have most authority and senior women have variable sway limited to domestic affairs (Kandiyoti 1988; Hofmann forthcoming). Mars and Altman (1983) provide an account of traditional Georgian norms. Males typically aspire to the status of strong and honorable men, usually achieved by asserting dominance within the household and demonstrating “manliness” through displays of consumer goods, dressing up, organizing feasts, and bouts of excessive and competitive drinking. Women are expected to be submissive to the men in their household and to maintain family honor by displaying modesty (including sexual) outside the household. Masculine and feminine positions within the household are clearly distinguished, with little competition or overlap.

Some aspects of the gender regime changed during the years of Soviet rule, while others remained the same (Gal and Kligman 2000). Women gained equality in the public sphere, with equal access to education and employment, but not in the private sphere, where housework and care for children and the elderly persisted as women’ responsibilities (Zurabishvili et al. 2009; Zurabishvili and Zurabishvili 2010). The transition to a market economy reduced economic and employment opportunities for everybody, but more for women than men (Jashi 2005). As in other post-Soviet countries, nationalist and religious revivals embraced conservative gender ideals that reinforce women’s duty as mothers and depict men as ideal breadwinners (Gal and Kligman 2000; Ishkanian 2002; Jashi 2005; Heyat 2006). Public opinion barometers indicate that Georgians hold fairly traditional attitudes about gender relations, including preferences for male authority in the family (Naskidashvili 2011; United Nations 2013). Nevertheless, economic necessity compels most women to take part in economic activities, including small-scale businesses in the informal economy, perpetuating women’s dual responsibility as housewives and breadwinners (Jashi 2005).

The violent separatist conflicts, other political turmoil, and severe economic crisis that have befallen Georgia since the collapse of the Soviet Union spurred waves of massive international migration, much of it temporary labor migration (Badurashvili 2004; Hofmann and Buckley 2012; Gerber and Torosyan 2013). According to one estimate, more than a fifth of the country’s population was living abroad in 2005 (Chindea et al. 2008). Lone female labor migration from Georgia is much more common than from other countries, including Mexico (Hofmann and Buckley 2013), probably due to the socioeconomic position of Georgian women and the growing demand for female migrant labor in the domestic and care work sectors of Western countries (Badurashvili 2004; Chindea et al. 2008; Zurabishvili and Zurabishvili 2010). Yet this demographic trend has evoked a cultural backlash that stigmatizes migrant women, particularly mothers (Hofmann and Buckley 2012).

Our empirical analysis uses quantitative data to examine whether migrant absence and migrant experience disrupt and/or exacerbate the traditional patriarchal gender differences in household tasks that prevail in many Georgian households. Because both male and female labor migration is so common in Georgia and our data come from a survey specifically designed to study migration issues, we can test for effects on the household activities of both men and women who are left behind. Our theoretical framework proposes two possible types of effects. Migrant absence effects occur while the migrant is still abroad; therefore, we test for them by comparing the probability that men and women engage in gender-atypical tasks and in leisure activities across households with and without currently absent migrants. Migrant experience effects occur after the migrant has returned home, and these effects may pertain to return migrants themselves, to their partners, and to the other members of the household who did not migrate. We assess their presence by comparing return migrants, the partners of return migrants, and other members of return migrant households to similar members of households with no migration experience. The next section describes our data and measures; then our analytical approach is explained in more detail.

DATA AND MEASURES

The “Georgia on the Move” (GOTM) survey was part of a six-country study of the relationship between migration and development funded by the Global Development Network (GDN) and the Institute for Public Policy Research (IPPR).² The GOTM survey was designed and implemented (using face-to-face interviews) by the Caucasus Research Resource Centers (CRRC) and International School of Economics at Tbilisi State University (ISET), with the help of external advisors and the GDN’s Project Management Team. Target sample volume was allocated equally across three strata: absent migrant households (at least one member currently living abroad), return migrant households (at least one member who previously lived abroad for at least three months), and non-migrant households (with neither current nor return migrants).

Primary sampling units (PSUs) were voter precincts randomly sampled within Tbilisi (the capital), other cities, and rural villages, with the number of PSUs in each proportionate to population size. The researchers conducted block enumerations of households by migration status within each selected PSU, which were used to randomly sample households within each migration-status stratum. Due to some errors in the enumeration and variation in response rates (overall, 70%) by strata, the final sample of 1482 households included 464 absent migrant households (31.3%), 345 return migrant households (23.8%), and 673 non-migrant households (45.4%). Overall, the survey collected data for over 5800 individuals. The interviews were conducted in November–December 2008, after the August invasion by Russian troops. For the analysis we chose to work with the subsample of adult (18 years and older) household members, which is 4700 individuals – 2260 men and 2440 women.³

Dependent Variables

The survey included a battery of questions ascertaining which of a list of 12 activities respondents engage in most, second most, and third most frequently at home. We collapse the three “frequency” levels into dummy variables indicating whether the respondent engages in each activity as one of the three most frequent activities in his/her home life (Table 1). About one–fifth of surveyed individuals did not report any activities, resulting in missing values.

TABLE 1.

Percentage of Respondents Who List Each Activity as One of Their Three Main Activities

Main household activities	Men %	Women %	Type of activity
1. Cooking	5.08	78.86	Female
2. Doing dishes/laundry/ironing/cleaning	4.26	78.61	Female
3. Repairing your home	10.28	1.75	Male
4. Collecting water	11.45	6.40	Male
5. Collecting firewood	31.89	1.75	Male
6. Growing/collecting food, looking after animals	25.99	14.45	Male
7. Shopping for food and household items	24.65	15.71	Male
8. Caring for children	3.50	21.05	Female
9. Caring for the sick/old	1.81	3.35	Female
10. Resting, recreation (e.g. chatting, watching TV, doing sports)	67.00	35.40	Leisure
11. Social occasions/visiting family and friends	35.86	10.81	Leisure
12. Community work	4.56	2.57	Leisure
Total individuals	1712	2062	3774

Note: Data are from the Georgia on the Move Survey.

For our analyses we separate the first nine types of activities, which are various household tasks from the last three, which we construe as leisure pursuits. As expected from our discussion of Georgian gender norms, household tasks are clearly divided between men and women. We characterize tasks as either “male” or “female” tasks depending on whether men or women are proportionately more likely to claim them as one of their three most frequent household activities. Male tasks include home repairs, collecting firewood or water, growing/collecting food, and shopping for food and household items. Shopping could be considered a gender-neutral task, given that the share of women doing it is also quite high and the Georgian “Gender and Generations Survey” finds women are slightly more likely to engage in shopping for food than are men.⁴ But we treat it is a male task in order to apply our definition consistently. As we discuss below, we check the robustness of our results by re-doing our analyses after treating shopping as a neutral task. Female tasks are cooking, cleaning, caring for children, and caring for sick or old household members.

Given that the final three activities represent forms of recreation, socializing, and elective pursuits outside of the household – not housework – we combine them into a “leisure” category that we analyze separately as an indirect measure of freedom from the burdens of housework. “Community work” may not be a leisure activity in the strict sense, but it does often have a social component and it represents a voluntary use of one’s time. Intuitively, engaging in leisure activities as one of three most common activities implies that one has enough spare time to indulge one’s tastes for entertainment, socializing, or voluntary work. Thus, it reflects the quantity of disposable time available after fulfilling one’s work obligations inside and outside the home. If, for example, the absence of migrants adds a burden to their partners and other household members left at home, it will reduce leisure time and, on average, the probability that a leisure activity is one of the three most common pursuits at home. Men’s greater access to leisure is the flip side of women’s greater responsibilities for household tasks. As Table 1 indicates, Georgian men are far more likely to list “leisure” pursuits among their three main activities, which is in line with our earlier discussion of traditional behavior of Georgian males.

Based on the types of household tasks 1–9 performed by each surveyed individual, we create a dummy variable indicating whether the respondent performs at least one gender atypical activity as part of his/her three most common activities. We also construct a dummy variable indicating that the respondent engages in at least one of the leisure activities 10–12. These are the two dependent variables in our analyses. Table 2 reports tabulations of these two variables for men and women from different types of households. In addition to separating absent and return migrant households, we also distinguish members of return migrant households who stayed in Georgia (“locals”) from the return migrants themselves.

TABLE 2.

Means on Dependent Variables, by Household Types

Type of household (HH)	Total		% Doing gender atypical tasks		% Resting and socializing
	Men	Women	Men	Women	Men	Women
Absent-migrant HHs	447	629	12.75	42.77***	80.31**	48.17***
Return-migrant HHs, local	218	377	7.80*	33.95	84.86	48.28***
Return-migrants	254	135	6.30**	34.81	84.25	53.33
Non-migrant HHs	793	921	11.10	36.37	84.87	55.59

Note: Data are from the Georgia on the Move survey.

^*p < 0.10,

^**p < 0.05,

^***p < 0.01.

For both men and women the probability of doing gender-atypical activities is higher when there is a migrant absent from the household compared to men and women in non-migrant households (though the difference is not statistically significant for men). Males in households with return migrants are less likely to do gender-atypical activities. The probability of engaging in leisure activities is lower for both genders in absent migrant households: apparently, having somebody in migration decreases the time available for leisure, presumably due to added housework or labor market responsibilities. We also observe a significant drop in leisure activities for local women in return-migrant households.

In the empirical analysis, we further explore differences across household types in the dependent variables and examine other correlates of doing gender-atypical tasks and leisure activities by estimating probit models for gender-atypical household tasks and for leisure activities for both men and women. In particular, we distinguish the effects of an absent or return migrant on the migrant’s partner from the effects on other household members. We also introduce covariates to isolate the effects of migration from the effects of other variables related to migration.

Covariates

In addition to measures of migration status, the data contain a variety of demographic and socio-economic characteristics at individual and household levels. We group our covariates of interest into migration status, other partner and individual characteristics, and household structure, equipment, and locality.

Migration Status

We test for absent migrant effects using dummy variables indicating, respectively, that the respondent is the partner of an absent migrant and that the respondent is a member of an absent-migrant household. If these variables have statistically significant effects on the probability of engaging in gender-atypical tasks or leisure, it is evidence of migrant absence effects at the individual (partner) and/or household level. Unlike most studies of gender differences in household tasks and leisure, we do not restrict our attention to spouses, because extended family arrangements are more typical in Georgia. The household activities of non-partner household members could also be affected by the absence of a migrant.

Using the same logic, we test for migrant experience effects using dummy variables denoting return migrants themselves and also denoting members of return-migrant households. We further distinguish return migrants from different destination countries. A sizable share returned from Russia, Georgia’s largest migration destination at the time of the survey. The Georgian diaspora in Russia is very large, and many migrants join a familiar environment upon their arrival there, minimizing exposure to foreign norms. Also, although Russia has had a high rate of female labor force participation both during and after the Soviet period (Gerber and Mayorova 2006), Russia’s gender norms within the household are far from egalitarian. For these reasons, we separate migrants returning from Russia from other return migrants when testing if the migration experience affects the probability of engaging in household tasks or leisure.

Other Partner and Individual Characteristics

In addition to partner migration status, we also control for partner’s presence (some are away from home but not abroad – most likely they are internal migrants), employment, and self-employment, as each of these may affect household activities. At the individual level, we control for personal characteristics likely both to affect housework and to be related to individual, household, and partner migration status: age, self-reported health, completion of higher education, and individual economic status. Hired employment is the baseline category for economic status, which is specified using dummy variables for the other categories: in education, self-employed, unpaid family worker, unemployed, retired, other (which includes military, and some other special cases). These categories are not mutually exclusive: respondents could report several statuses (for example, being retired and self-employed).

Household Characteristics

The composition of a household by age and gender affects both the quantity and profile of household tasks that need to be performed and the availability of household members to perform those tasks in a “gender typical” fashion. For example, other adult females in the household could well mitigate the pressure on males whose partners migrate abroad to take on female-typical activities. Therefore, we control for household structure with a series of count variables measuring how many family members of different ages of each gender, including children, are in the respondent’s household. We also control for the locality of the household (rural versus urban) and we include dummy variables for possession of several household appliances. Both locality and equipment could influence the profile of household tasks and attitudes about who should do them; for instance, families in rural areas might lack close access to water, making it less likely for women to carry water (given the distance required). More traditional attitudes to gender roles in rural areas might make the gender separation of tasks even stronger. Appliances affect the difficulty and time required for some tasks: a washing machine makes doing laundry more efficient and a refrigerator mitigates the need to shop for food daily. A TV or a computer could serve as a source of information and education, promoting a less traditional approach to housework. But they could also draw people into leisure instead of doing household tasks. Table 3 reports summary statistics for covariates, restricting the sample to the observations used in the analysis below (which have no missing values for any of the covariates).

TABLE 3.

Summary Statistics for Independent Variables

Variables	Men (N = 1685)				Women (N = 2027)
	Mean	SD	Min	Max	Mean	SD	Min	Max
Migration status
Partner is absent migrant (AM)	0.034	0.181	0	1	0.055	0.228	0	1
Local member of AM household (AM HH)	0.227	0.419	0	1	0.252	0.434	0	1
Return migrant (RM) from Russia	0.066	0.248	0	1	0.028	0.165	0	1
RM from other destinations	0.084	0.277	0	1	0.037	0.190	0	1
Partner is RM	0.037	0.188	0	1	0.084	0.278	0	1
Local member of RM household (RM HH)	0.128	0.334	0	1	0.183	0.387	0	1
Other partner and personal characteristics
Partner away, not in migration	0.068	0.251	0	1	0.230	0.421	0	1
Partner is employed	0.133	0.340	0	1	0.182	0.386	0	1
Partner is self-employed	0.094	0.292	0	1	0.145	0.352	0	1
Age	43.97	17.46	18	97	45.62	17.42	18	96
Has higher education	0.261	0.439	0	1	0.276	0.447	0	1
In good/excellent health	0.466	0.499	0	1	0.395	0.489	0	1
In education	0.068	0.251	0	1	0.057	0.231	0	1
Unpaid family worker	0.367	0.482	0	1	0.550	0.498	0	1
Self-employed	0.217	0.412	0	1	0.125	0.331	0	1
Unemployed	0.268	0.443	0	1	0.267	0.443	0	1
Retired	0.173	0.378	0	1	0.212	0.409	0	1
Other economic status	0.064	0.245	0	1	0.060	0.238	0	1
Household structure, equipment, location, and other variables
# Children (0—6)	0.275	0.558	0	3	0.297	0.579	0	3
# Children (7—17)	0.485	0.764	0	6	0.508	0.781	0	6
# Men (18–24)	0.242	0.499	0	3	0.272	0.527	0	3
# Men (25–45)	0.421	0.605	0	4	0.669	0.648	0	4
# Men (46–55)	0.164	0.377	0	2	0.304	0.466	0	2
# Men (56–65)	0.092	0.297	0	2	0.163	0.372	0	2
# Men (66+)	0.110	0.315	0	2	0.183	0.388	0	2
# Women (18–24)	0.254	0.499	0	3	0.220	0.481	0	3
# Women (25–45)	0.662	0.597	0	3	0.380	0.558	0	3
# Women (46–55)	0.356	0.485	0	2	0.176	0.390	0	2
# Women (56–65)	0.209	0.409	0	2	0.092	0.292	0	2
# Women (66+)	0.218	0.416	0	2	0.126	0.337	0	2
Rural household	0.430	0.495	0	1	0.389	0.488	0	1
Household has TV	0.840	0.367	0	1	0.845	0.362	0	1
Household has PC	0.180	0.385	0	1	0.184	0.388	0	1
Household has refrigerator	0.747	0.435	0	1	0.743	0.437	0	1
Household has washing machine	0.328	0.470	0	1	0.336	0.472	0	1

RESULTS

Table 4 presents the results of our simple probit regressions for performing gender atypical tasks and for citing leisure as one of the three most common activities. These estimates do not account for the non-independence of observations within households, nor do they address the potential endogeneity of migration status and housework (for example, more “Western-oriented” households may be simultaneously more likely to send female members abroad to work and to have males engaging in gender-atypical household tasks). Below we report the results of sensitivity analyses we conducted to deal with these potential problems in our estimations.

TABLE 4.

Probit Models for Gender-Atypical Tasks and Leisure, by Gender

Variables	Men (N = 1685)				Women (N = 2027)
	Gender atypical tasks		Leisure		Gender atypical tasks		Leisure
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Migration status
Partner is AM	0.741	0.206	−0.937	0.222	0.543	0.142	−0.424	0.143
Local member of AM HH	0.197	0.122	−0.269	0.127	0.221	0.081	−0.286	0.082
RM from Russia	−0.692	0.264	0.028	0.187	0.065	0.196	0.009	0.194
RM from other destinations	−0.278	0.187	−0.023	0.183	0.153	0.171	−0.031	0.169
Partner is RM	0.453	0.254	−0.065	0.285	0.020	0.136	−0.355	0.137
Local member of RM HH	−0.053	0.157	−0.034	0.160	−0.054	0.100	−0.039	0.099
Partner and personal characteristics
Partner away, not migration	0.579	0.161	−0.133	0.186	0.072	0.099	−0.036	0.098
Partner is employed	0.146	0.146	−0.245	0.139	0.139	0.090	−0.391	0.089
Partner is self-employed	−0.048	0.184	−0.144	0.140	0.018	0.098	−0.108	0.098
Age	0.001	0.005	−0.002	0.005	0.013	0.003	0.000	0.003
Has higher education	0.212	0.109	0.006	0.124	−0.309	0.076	0.228	0.074
In good/excellent health	0.063	0.106	−0.128	0.107	−0.011	0.075	0.029	0.075
In education	0.290	0.211	1.257	0.476	−0.412	0.175	1.008	0.195
Unpaid family worker	0.135	0.105	−0.709	0.100	0.109	0.067	−0.322	0.065
Self-employed	−0.042	0.126	−0.080	0.107	0.118	0.096	0.027	0.098
Unemployed	0.231	0.107	0.013	0.113	−0.034	0.076	−0.206	0.075
Retired	0.059	0.171	0.223	0.164	−0.476	0.115	0.277	0.114
Other economic status	−0.320	0.235	0.212	0.163	−0.142	0.135	0.220	0.138
Household structure, equipment, and type of residence
# Children (0–6)	0.265	0.082	−0.089	0.091	−0.157	0.060	−0.241	0.058
# Children (7–17)	−0.050	0.065	−0.129	0.065	−0.087	0.042	−0.135	0.042
# Men (18–24)	−0.162	0.101	0.091	0.100	0.145	0.060	−0.129	0.060
# Men (25–45)	−0.076	0.082	−0.009	0.082	0.009	0.053	−0.094	0.053
# Men (46–55)	−0.095	0.142	0.195	0.154	−0.127	0.081	0.115	0.080
# Men (56–65)	0.189	0.151	0.467	0.191	−0.055	0.097	−0.079	0.097
# Men (66+)	0.104	0.138	0.112	0.148	−0.082	0.094	−0.005	0.093
# Women (18–24)	−0.127	0.102	−0.019	0.103	0.026	0.065	0.077	0.065
# Women (25–45)	−0.177	0.095	0.208	0.100	−0.020	0.061	0.267	0.061
# Women (46–55)	−0.531	0.119	0.156	0.124	−0.251	0.094	0.337	0.094
# Women (56–65)	−0.357	0.129	0.018	0.131	−0.246	0.117	0.220	0.116
# Women (66+)	−0.115	0.116	−0.147	0.119	−0.159	0.095	−0.028	0.094
Rural household	−0.207	0.114	−1.089	0.118	0.506	0.073	−0.395	0.074
Household has TV	−0.152	0.127	−0.083	0.127	0.048	0.095	0.014	0.095
Household has PC	−0.154	0.139	0.703	0.232	−0.106	0.091	0.140	0.089
Household has RF	−0.155	0.113	−0.175	0.115	−0.090	0.080	−0.117	0.080
Household has WM	−0.098	0.113	0.114	0.119	0.023	0.075	0.118	0.075
Constant	−0.872	0.269	2.217	0.296	−0.872	0.203	0.526	0.201
Pseudo-R2	0.110		0.298		0.111		0.148

Note: AM HH: absent migrant households; RM HH: return migrant households. Estimates are based on the analysis sample from the Georgia on the Move Survey.

Bold coefficients are statistically significant at p < 0.05, two-tailed; bold italicized coefficients at p < 0.05, one tailed.

We find strong and consistent evidence of absent migrant effects: for both men and women, the absence of a partner who is migrating abroad is associated with elevated probability of doing gender-atypical household tasks and lower probability of engaging in leisure as a main activity. These effects are large in magnitude: the marginal effects of having an absent migrant partner on the probability of doing gender atypical tasks are 0.17 for men and 0.21 for women, and the corresponding (negative) marginal effects on the probability of taking part in leisure are 0.22 and 0.16 respectively.⁵ Moreover, migrant absence also affects adult household members other than the migrant’s partner, producing lower probability of leisure activity for all household members and higher probability of doing male tasks among female household members. However, for men the effect of migrant absence falls mainly on the migrant’s partner: other men in absent migrant households exhibit no significant change in their tendency to perform gender atypical tasks, while their probability of engaging in leisure activities is reduced by a mere 0.041. The results provide clear and compelling evidence that in Georgian households the absence of a migrant from the household reduces access to leisure time of the householders left behind and also tends to increase the probability that they take on gender typical housework, with the strongest effects experienced by the migrant’s partner.

The evidence for migrant experience effects is more uneven. On one hand, the predicted probability of doing gender atypical tasks is 0.035 lower for male return migrants from countries other than Russia and 0.067 lower for men coming back from Russia. It would seem that upon their return from migration men have more bargaining power, enabling them to avoid female-type household tasks, and this effect is particularly strong for men who work in Russia. Female members of return migrant households are neither more nor less likely to engage in gender atypical tasks, except for the partners of return migrants, who are about as less likely to take part in leisure activities as are women whose partners are currently living abroad. Thus, male migration only adds to the housework of their female left-behind partners and reinforces gender inequality in household activities even upon their return. There is mixed evidence of migrant experience effects for female migrants: return migrant women do not have more access to leisure time, nor are they more likely to engage in gender-atypical tasks. But their male partners are more likely to engage in female-typical housework, which could be the result of greater bargaining power or more egalitarian gender norms on the part of the return migrant women.⁶ It could also result from left-behind males growing accustomed to performing gender atypical tasks while their partners are living abroad, or it could be a combination of female migrant experience effects and persisting male migrant absence effects.

The control variables by and large have intuitive effects on both gender atypical tasks and leisure activities. Males whose partners are absent but not abroad (presumably they are internal migrants) are more likely to engage in female-typical tasks. Having an employed partner reduces the probability of engaging in leisure activities for women. It is noteworthy that higher education is associated with elevated probability that men do female tasks (consistent with the expectation that higher education helps spread feminist norms) but diminished probability that women do male tasks. More educated men seem to have more egalitarian (less traditional) views on division of household tasks, while educated women have the option to avoid heavy male tasks. Both male and female students are substantially more likely than their working counterparts to engage in leisure activities. Unemployed males are more likely than working males to take part in female tasks, which is consistent with exchange theory and contrary to the “gender performance” explanation of gender differences in housework. The presence of young children has opposite effects on engagement in gender atypical tasks for men and women: men start helping more with taking care of children and with other female tasks, while women withdraw their time from male tasks to perform the increased load of female duties. Also, children make for more work and less leisure: young children reduce participation of women in leisure activities, while older children reduce leisure activity for both sexes.

The number of adult women in a Georgian household is inversely related to the probability that men engage in gender atypical tasks, suggesting that when more women are available in the household to do “women’s work” men are less likely to do it than when there are fewer women available. This result is consistent with previous studies indicating that extended family declines men’s involvement in housework (Parreñas 2005; Derby 2010). At the same time, the presence of more adult women in the household is linked to more leisure time for women (presumably because there are more hands available to help with the tasks that need doing) and less engagement in male tasks. The latter effect is curious, but it could reflect deeper patriarchal norms in households with more women.

Men in rural households are less inclined to take part in female tasks, which is an expected observation for more traditional rural families. However, rural women are more likely than urban women to do male tasks, most likely because they include such activities as chopping wood and fetching water that are less common in urban households. Both men and women in the countryside typically have substantially less time available for leisure activities. The household appliances we examined tend not to be associated with household tasks, aside from a greater tendency for males in households with computers to report taking part in leisure activities.

Sensitivity Analyses

As noted, our simple probit estimations ignore the likely correlation of residuals within households and the potential endogeneity of household activities and migration status. We performed five sensitivity analyses to address these concerns. We report the results in Table 5, limiting our focus to the key variables of interest (migration status and partner characteristics). The first columns in Table 5 (model 1) reproduce the basic probit estimates, as a benchmark.

TABLE 5.

Sensitivity Analyses, Migration Effects

Variables	1. Simple Probit, all data		2. RE, all data		3. RE, sub sample		4. RE, corrected		5. RE extended, subs sample		6. RE extended, corrected
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
No of observations	1685		1685		1208		1208		1195		1195
No of panels	NA		1131		649		649		641		641
A. Men doing gender atypical household tasks
Partner is AM	0.74	0.21	0.81	0.24	0.55	0.34	−0.36	0.50	0.63	0.33	−0.37	0.49
Local member, AM HH	0.20	0.12	0.23	0.14	0.00	0.17	−0.15	0.24	0.12	0.17	−0.03	0.23
RM from Russia	−0.69	0.26	−0.80	0.31	−1.37	0.63	−1.23	1.01	−1.35	0.63	−1.00	0.99
RM other destinations	−0.28	0.19	−0.30	0.22	−0.60	0.34	−0.41	1.11	−0.46	0.33	−0.17	1.09
Partner is RM	0.45	0.25	0.55	0.30	0.81	0.46	3.09	1.59	0.74	0.45	3.28	1.51
Local member, RM HH	−0.05	0.16	−0.04	0.18	−0.09	0.21	0.09	1.05	0.05	0.21	0.30	1.03
Rho	NA		0.24	0.13	0.38	0.11	0.39	0.11	0.28	0.13	0.26	0.13
B. Men doing leisure activities
Partner is AM	−0.94	0.22	−1.12	0.28	−0.70	0.33	−0.92	0.52	−0.85	0.36	−0.84	0.54
Local member, AM HH	−0.27	0.13	−0.33	0.16	−0.20	0.16	−0.37	0.22	−0.29	0.17	−0.29	0.23
RM from Russia	0.03	0.19	0.05	0.22	0.29	0.31	−0.62	0.81	0.32	0.33	−0.35	0.82
RM other destinations	−0.02	0.18	−0.06	0.22	0.33	0.30	−1.16	0.99	0.40	0.32	−0.75	1.00
Partner is RM	−0.06	0.29	−0.07	0.34	−0.08	0.41	−0.44	0.63	−0.44	0.44	−0.44	0.66
Local member, RM HH	−0.03	0.16	−0.04	0.19	0.01	0.20	−1.26	0.92	0.02	0.22	−0.78	0.94
Rho	NA		0.29	0.12	0.35	0.10	0.38	0.11	0.26	0.12	0.21	0.13
No of observations	2027		2027		1420		1420		1402		1402
No of panels	NA		1352		736		736		727		727
C. Women doing gender atypical household tasks
Partner is AM	0.54	0.14	0.57	0.15	0.41	0.19	0.43	0.27	0.50	0.19	0.45	0.27
Local member, AM HH	0.22	0.08	0.23	0.09	0.08	0.09	0.15	0.12	0.16	0.10	0.20	0.12
RM from Russia	0.06	0.20	0.07	0.21	−0.23	0.26	0.45	0.55	−0.19	0.26	0.61	0.54
RM other destinations	0.15	0.17	0.17	0.18	−0.07	0.22	1.01	0.75	0.05	0.22	1.33	0.75
Partner is RM	0.02	0.14	0.02	0.14	0.19	0.18	0.16	0.23	0.20	0.19	0.20	0.24
Local member, RM HH	−0.05	0.10	−0.05	0.11	−0.22	0.13	0.72	0.70	−0.22	0.12	1.07	0.69
Rho	NA		0.09	0.07	0.13	0.07	0.14	0.07	0.07	0.07	0.06	0.07
D. Women doing leisure activities
Partner is AM	−0.42	0.14	−0.45	0.15	−0.91	0.21	−1.07	0.29	−0.92	0.21	−1.06	0.29
Local member, AM HH	−0.29	0.08	−0.30	0.09	−0.30	0.10	−0.45	0.13	−0.39	0.11	−0.47	0.13
RM from Russia	0.01	0.19	0.02	0.21	0.32	0.28	−0.27	0.59	0.27	0.28	−0.63	0.58
RM other destinations	−0.03	0.17	−0.02	0.18	0.17	0.24	−0.49	0.79	0.10	0.23	−0.94	0.78
Partner is RM	−0.36	0.14	−0.39	0.15	−0.80	0.21	−0.91	0.25	−0.73	0.21	−0.90	0.25
Local member, RM HH	−0.04	0.10	−0.04	0.11	−0.05	0.13	−0.72	0.74	−0.10	0.13	−1.27	0.73
Rho	NA		0.12	0.07	0.23	0.07	0.22	0.07	0.14	0.07	0.13	0.07

Note: AM HH, absent migrant households; RM HH, return migrant households; RE, random effects. Estimates are based on the analysis sample from the Georgia on the Move Survey.

Bold coefficients are statistically significant at p < 0.05, two tailed; bold italicized coefficients at p < 0.05, one tailed.

We perform random effects estimation as our first robustness check (model 2). There is significant unobserved household-level heterogeneity: the correlation of observations within the same households, Rho, differs statistically from zero, indicating the need for random effects estimation to control for household heterogeneity.

However, if household level unobservables are correlated with the covariates, simple random effects estimation produces inconsistent estimates. To address this issue, we apply a correction to random effects originally proposed by Mundlak (1978) and further developed by Wooldridge (2002), which we describe in the Appendix (model 4, Table 5). The “Mundlak” correction can only be applied to households with two or more men (women), where we can compute group averages for the covariates, hence we have fewer observations for these estimates. To facilitate comparison across different estimation techniques, we re-estimate simple random effects models for this reduced subsample of households, (model 3, Table 5). Finally, the standard Mundlak correction uses only covariates that vary within households, so models 3 and 4 exclude household-level covariates. However, because household-level variables are important predictors of behavior in our model (as seen in Table 4) we add them back in models 5 and 6. We lose a small number of additional households in models 5 and 6 relative to models 3 and 4, due to missing data on household-level covariates.

Given that the simple random effect model might suffer from heterogeneity bias, while estimation with Mundlak correction uses only a subsample of data (which basically omits households with one adult man and/or one adult woman, leading to loss of observations and selectivity), we consider all the estimates together. Starting with the relationship between migration and males’ engagement in female tasks, the sensitivity analyses cast some doubt on the findings that the absence of female migrants increases the probability that their male partners are more likely to engage in female tasks: notably, after the Mundlak correction the sign of the effect changes. The robustness checks also call for caution in interpreting the negative effect of return migrant status for males who had migrated to Russia. However, it could be that the loss in statistical power resulting from our differentiation of migration to Russia and migration to other destinations is to blame. The correction confirms our inference (based on a one-tailed test) that male partners of female return migrants are more likely to perform female-typical tasks after their partners return.

The robustness checks generally confirm our findings regarding leisure activities: when a household member is working abroad, Georgian men and women are less likely to be engaged in leisure activity, and this is an especially strong effect for men. But this is purely a migrant absence effect: there is no evidence that the pattern holds once the migrants return.

While their partners are working abroad in migration, women are more often performing some male-typical tasks. This effect is relatively large and robust across different specifications. There seems to be also some spillover effect on the rest of females in absent migrant households, but the effect is much smaller. This, too, is exclusively a migrant-absence effect: when husbands return from migration, women go back to their pre-migration distribution of tasks.

Finally, when it comes to women resting and socializing, the impact of migration absence is robust and consistent. Women with household members working abroad are much less likely to rest or socialize, compared to women with similar characteristics from non-migrant families, and this effect persists when their male partners return home. Thus, our evidence is consistent that the labor migration of a male partner results in significantly lower probability that Georgian women engage in leisure activities, both during the partner’s migration and after their return.

In an additional sensitivity check we dropped “shopping for food and household items” from male tasks and re-estimated the model for women after removing shopping from the list of male tasks. The only inference that changes is that the effect of living in household with absent migrants becomes insignificant. However the effect on the partner of the absent migrant stays almost unchanged. The other three regressions are not affected by re-definition of male tasks.⁷

Finally, we considered whether the effects of the migration status variables vary by household composition. In particular, it could be that migrant absence only increases the probability of performing gender-atypical tasks by left behind males if there are no other female adults in the household to pick up the duties of absent female migrants. We entered the appropriate multiplicative terms in our models to test for such interactions, and in no case were their effects significant. Thus, our additive controls for the presence of male and female adults in different age categories adequately take into account their potential confounding role.

DISCUSSION

Qualitative studies have suggested that migrant absence and migrant experiences both might affect gender differences in the type and amount of housework within households in migrant-sending countries. Our study of Georgia provides initial quantitative evidence on whether these effects are widespread enough to be statistically discernable. Because labor migration is an important fact of life in many low income countries, this is a key step toward incorporating them into the literature on gender and housework.

We tested empirically for migrant absence and migrant experience effects on the probability that Georgian men and women engage in gender-atypical housework tasks and that they partake in leisure activities. The absence of migrants leads to more involvement of males in female tasks, particularly for partners of absent migrants. However, this apparent effect is not robust to controls for unobserved household-level variables and thus must be viewed with caution: it could represent unobserved heterogeneity rather than a causal effect. But the Mundlak correction may be overly conservative because it essentially limits the sample to extended family households with multiple left-behind adults of the opposite gender to the respondent. We find consistent evidence that the migration of household members and (especially) female partners leads to less engagement in leisure activities by Georgian males during the migration period, an effect that persists for male partners of female migrants when they return home. This could result from greater bargaining power of women vis-à-vis their partners after migration, their greater unwillingness to take on gender typical housework tasks due to their migration experiences, males growing accustomed to doing female tasks while their partners are absent, or some combination of these. At the same time, we do not find that female return migrants are more likely to engage in either leisure or gender atypical household tasks. Male return migrants, though, are less likely to engage in female tasks.

The absence of migrants, especially male partners, leads to an increase in Georgian women’s engagement in gender-atypical household tasks and a substantial reduction in their leisure activities. The former effect is limited to the period of the migrant’s absence, while the latter persists after the migrant returns. Thus, male migration tends to exacerbate gender inequality in housework. Migration can reinforce patriarchal gender norms, as expressed in daily activities within the household.

In sum, migration behavior does affect gender differences in housework and leisure for those left behind in the sending country, but in complex and contradictory ways. Both males and females are less likely to engage in leisure activities when their partners (and, to some degree, other adult household members) are working abroad and more likely to engage in gender atypical household tasks (though for left-behind men this could be due to unobserved heterogeneity.) Possible migrant experience effects cut both ways in terms of gender egalitarianism in housework: male return migrants are less likely to perform female household tasks than otherwise similar males who have not migrated, but the male partners of female migrants are more likely to engage in “female” housework after their partners return (though this effect is marginally significant in most of the estimations). Altogether, it appears that the migration of Georgian males exacerbates gender differences in the type and amount of housework performed with Georgian households, while the migration of Georgian women tends to ameliorate those differences.

International migration has great potential to reshape gender relations in sending countries. An emerging tradition in international migration research examines how the migration experience affects women’s opportunities to gain on men in the labor market. But the labor market is only one sphere of gender inequality: home activities (including both household tasks and access to leisure) are a key site for understanding gender relations more broadly in a society. Moreover, the social consequences of migration are hardly limited to the experiences of migrants during the course of migration: they can be equally significant in migrant-sending communities and within the households of those left behind. Our study uses data from a survey explicitly designed to assess the impact of migration to test for possible effects of both migrant absence and migrant experience on participation in gender-atypical tasks and in leisure activities in households residing in a migrant-sending country. We found mixed evidence for these types of effects, which suggests that future studies of migration and gender, as well as analyses of gender and housework in lower income countries, should attend to the role migration can play in shaping gender relations within the home in sending communities. It is, of course, possible that the effects of migration in this sphere will vary across different sending countries based on their prevailing gender norms, the nature of international migration that characterizes them, and their most common destination countries. Our theoretical framework and research design should prove useful to scholars who undertake studies of the nexus of migration, gender, and housework in other contexts. As similar studies of other migrant-sending countries emerge, it may become possible to identify systematic country-level sources of variation in how migration shapes male and female activities in the home within sending countries. More broadly, our study suggests that scholars interested in the impact of migration on gender relations should continue to expand their lens outside the labor market and identify other areas where gender differences in sending countries are either reinforced or challenged by migration. Civic engagement, political participation, and educational attainment are three such areas that seem particularly promising territory for future studies.

We conclude with some brief methodological observations. Although survey-based studies are the most promising for measuring the extent and nature of migrant absence and migrant experience effects on gender relations within the home in migrant sending countries, qualitative studies are better suited for exploring the mechanisms that produce these quantitative patterns. Ideally, the two different methodological approaches will offer complementary insights as research on how migration affects gender norms and behaviors grows. In addition, cross-sectional survey data suffer from inherent limitations for the purpose of identifying causal effects. We have done our best to correct our estimates for potential biases, but it would be preferable to have panel data available. Longitudinal data sets are difficult to come by in migrant-sending countries, and it is especially difficult to follow households whose members migrate frequently. However, our results call attention to the need to collect such data to better measure migrant absence and migrant experience effects as part of a broader effort to understand how large-scale labor migration affects gender relations – in Georgia and in other sending countries.

APPENDIX

We can write the regression model in a manner that decomposes the residual into household-level and individual-level components:

$$y_{\mathit{\text{ij}}}^{\ast }=\hat{\mathbf{a}}\prime {\mathbf{X}}_{\mathit{\text{ij}}}+{\mathrm{\alpha }}_j+{\mathrm{\epsilon }}_{\mathit{\text{ij}}},$$

where ij denotes person i in household j, y* is the latent variable, X_ij is the set of observed covariates, and the error term is expressed as the sum of a household-specific component, α_j and an individual component, ε_ij.

The Mundlak correction method is built on projecting the fixed household effects on the within-group means of the covariates:

$${\mathrm{\alpha }}_j=\stackrel{\sim }{\mathbf{a}}\prime {\overline{\mathbf{X}}}_j+{\omega }_j.$$

Here X̄_j is the set of unit-specific (i.e. household-level) mean values for each of the model covariates; it is assumed to contain any information in α_j that is correlated with the covariates, leaving ω_j uncorrelated with any of the original covariates as well as their group means. By substituting this projection of fixed effects into the original model we obtain the following equation:

$$y_{\mathit{\text{ij}}}^{\ast }=\hat{\mathbf{a}}\prime {\mathbf{X}}_{\mathit{\text{ij}}}+\stackrel{\sim }{\mathbf{a}}\prime {\mathbf{X}}_{\mathit{\text{ij}}}+{\mathrm{\omega }}_j+{\mathrm{\epsilon }}_{\mathit{\text{ij}}},$$

In this specification the unit-specific component is not correlated with covariates, so random effects probit estimation can be used to estimate this “corrected” version of the model.