Is There a Gender Gap in Health among Migrants in Russia?

Abstract This study investigates whether there is a gender gap in health among migrants. Focusing on migrants from Uzbekistan and Tajikistan in Russia, where there are high levels of both immigration and gender inequality, this paper exploits unique data collected by the authors recording detailed information on health and migration trajectories. We find that migrant women are on average in poorer health than migrant men. This gender gap is only partly explained by gender differences in observed socioeconomic, demographic, living and working characteristics and differences in pre-migration health. We show that migrant women’s health is more likely than men’s to deteriorate during migration. This women’s health disadvantage is sensitive to the migration profile, as it only appears after a certain time spent migrating and for migrants with a vulnerable legal status. These results call for targeted public health policies to address this gender health gap.


Introduction
In a world where approximately one in seven people is an internal or international migrant (Schenker, Castañeda, & Rodriguez-Lainz, 2014;Skeldon, 2017), migrants' health is crucial.First for their own welfare, but also for the economy of the countries of origin and destination.Unhealthy migrants represent a less productive workforce in the destination country and send fewer remittances home.It also challenges health systems at the destination, which must adapt to the health needs of these new mobile populations (Hern andez-Quevedo & Jim enez-Rubio, 2009), and at the origin country, if the healthiest leave and the sickest return (Abraido-Lanza, Dohrenwend, Ng-Mak, & Turner, 1999).
The past literature has shown that, on average, migrants are healthier than natives (Beiser, 2005;Lu & Qin, 2014) and has widely discussed the causes behind this so-called Healthy Immigrant Effect (HIE): the positive selection induced by individual or household choices (the healthiest, most productive migrate) and policies (such as medical screening); the negative selection out, i.e. the return of the oldest and sickest ('salmon bias') (Guillot, Khlat, Solignac, Gansey, & Elo, 2018;Wallace & Kulu, 2014).Another branch of the literature has nuanced the HIE by showing that this initial positive advantage diminishes over time, or even reverses, due to the many potential adverse factors that migrants face, such as lower socioeconomic position, stress, unhealthy habits, lack of social capital and lack of access to healthcare (Jusot, Silva, Dourgnon, & Sermet, 2009;Moullan & Jusot, 2014;Namer & Razum, 2018;Subedi & Rosenberg, 2014;Ukrayinchuk & Drapier, 2021;Wallace, Khlat, & Guillot, 2019).In addition, as health is partly the result of individual choices (Becker, 2007;Grossman, 1972) and partly socially constructed (Brabete, 2017;Case & Deaton, 2003), several personal and contextual factors are at play, interacting and producing different levels of health.As some migrants accumulate more vulnerabilities, there is considerable health heterogeneity among migrants, depending on the country of origin, age, generation or wave of migration, human capital, characteristics of the migration stay (working and living conditions, length of stay and legal status) (Agadjanian, Oh, & Menj ıvar, 2022;Gkiouleka, Huijts, Beckfield, & Bambra, 2018;Ll acer, Zunzunegui, Del Amo, Mazarrasa, & Bolumar, 2007).
A major source of heterogeneity among migrants is gender, which is all the more important as women account for almost half of the world's international migrants (UN DESA, 2020).As women in the general population report on average poorer health than men (Macintyre, Hunt, & Sweeting, 1996), there is reason to expect that the health of migrant women and men differs.Indeed, gender interacts with factors of health prior to migration and in migration.Before setting foot in the host country, the health of future men and women migrants may differ due to gender norms and inequalities in the country of origin, affecting their respective probability of migrating.Selection within migration may be gendered, with women not responding to traditional push factors in the same way as men (Docquier, Marfouk, Salomone, & Sekkat, 2012) and having gender-specific reasons for migrating (Antman, 2018;Ferrant & Tuccio, 2015;Ll acer et al., 2007;Ruyssen & Salomone, 2018).In line with the tied migration theory (Mincer, 1978), if women are tied-movers who migrate for marriage or family reunification, they may be selected less on the basis of their health and potential productivity (Atella, Deb, & Kopinska, 2019).Conversely, if women are increasingly independent migrants rather than the wives of migrants, as suggested by recent studies, the selectivity should converge (De Haas et al., 2019;Dumont, Martin, & Spielvogel, 2007).
However, in countries where women's independent migration remains stigmatized, only the best-educated and healthiest women will be able to overcome the social and cultural barriers to migration, leading to a higher positive selection bias among women than among men in terms of human capital (Hofmann, 2017).
After a gendered selection process, the migration journey itself may be harmful, perhaps more so for women in that they are more likely to be victims of violence (Calder on-Jaramillo, Parra-Romero, Forero-Mart ınez, Royo, & Rivillas-Garc ıa, 2020; Fleury, 2016;Ll acer et al., 2007).Once in the host country, gender shapes migrants' lives in several ways.Gender norms of the destination country will affect women's experience in migration and health.Racial and gender segregation is quite common in labor markets of host countries, with the result that migrant women occupy positions based on a gendered construction of their skills and roles, such as domestic and care services, while men are employed in construction and industry (Fleury, 2016;Hofmann, 2017).According to Cooper (2002) and Read & Gorman (2006), even after controlling for socioeconomic status, female migrants face a double disadvantage due to the potential discrimination associated with the two minority groups, which may affect their employment situation or access to care.Finally, depending on family status, women migrants may also be at a higher risk either of social isolation or be burdened with childcare and domestic duties (double day) (Ll acer et al., 2007).
Drawing on this literature, this article investigates whether a gender health gap exists among Central Asian migrants living in the Russian Federation and studies the reasons for this gap.Does it exist before migration, does it stem from a gendered selection in migration, or is it created or reinforced by the experience in migration (gender-specific barriers, occupational segregation, etc.)?
This paper makes a three-fold contribution to the literature.First, it contributes to the emerging yet scarce quantitative studies of the heterogeneous effects of migration experience on health especially associated with gender.If the gender perspective is deeply enshrined in qualitative studies that investigate gender as a social feature interacting with other social determinants, such as ethnicity, migration process or occupational segregation, to produce health inequality (Gkiouleka et al., 2018;King & Dudina, 2019;King, Dudina, & Dubrovskaya, 2020;Ll acer et al., 2007), this dimension is not the focus of most economic and epidemiological studies, where sex is rather seen as a control variable or one potential negative factor among others (Khlat & Guillot, 2017;Setia, Quesnel-Vallee, Abrahamowicz, Tousignant, & Lynch, 2012).Relying on statistical methods, this paper aims at making things comparable using a ceteris paribus reasoning, allowing us to test whether gender stays a cumulative disadvantage among migrants, even after adjusting for differences in socioeconomic characteristics (Gkiouleka & Huijts, 2020;Iglesias, Robertson, Johansson, Engfeldt, & Sundquist, 2003).It also enables us to check whether the potential gender gap varies according to migrants' trajectories.Our study makes a second contribution to the literature by using a retrospective health question to ascertain whether or not women migrants are more likely than men to experience a deterioration in health.This question allows us to control for time-invariant unobserved characteristics.Lastly, we focus on Central Asian migrants living in Russia, a large but understudied population, using a rich and original survey of 1,213 labor migrants from Uzbekistan and Tajikistan, that we compiled ourselves in 2019.This population is usually missing in general population surveys.
The results of our analyses reveal a significant gender gap in health among migrants, to the detriment of women.Differences in observed characteristics and pre-migration health only partly explain this gender disparity, which results from the migration experience and is reinforced by migration-related features such as the time spent in migration and the degree of institutional integration.These results call for targeted public health policies to address this gender health gap.
The remainder of the paper is structured as follows.Section 2 presents the context of migration and healthcare in Russia.Section 3 describes the data and the empirical framework.Section 4 provides the estimation results.Section 5 concludes and discusses avenues for future research.

International migration in Russia
Since 1991, migration in the post-Soviet area has grown as a result of economic crises and civil wars.Until 2015, Russia was ranked the second largest destination for migrants, and today it ranks fourth with 11.7 million migrants.Central Asians account for half of these migrants.This paper focuses on migrants from Tajikistan and Uzbekistan, who account for more than 47% of foreigners engaged in labor activities in Russia according to the Federal Migration Service 2011 estimates, mostly in unskilled jobs (Mukomel, 2013b).Remittances sent from Russia represent an important source of revenue: 12% of Uzbekistan's GDP and 27% of Tajikistan's GDP in 2020, maintaining it as one of the most dependent economies.
Despite the scale of this phenomenon, little is known about the health and living conditions of migrants in Russia, due to a lack of available data.Migrants are a highly mobile population that is generally not captured by national surveys such as the Russia Longitudinal Monitoring Survey (RLMS), sampled in residents' data (Kim, Matytsin, & Freije, 2019).Indeed, migrations Is there a gender gap in health among migrants in Russia?1929 to Russia are mostly temporary labor migrations.They are structured by policies aiming to make up the labor force shortfall, caused by high emigration and low fertility.Short-term stays are more encouraged than long-term ones, politically and legally.The duration depends on the legal documents obtained.Permanent residence permit is subject to restrictive quotas and difficult to obtain.Valid for five years, it can be renewed without any limitations.Temporary stay permit is subject to quotas as well and allows one to live and work for three years in the region where the permit was issued.Instead, many migrants pay for a patent that allows them to work in Russia for only one year, then they have to leave the country to renew their work permit.Even if they do not get the patent, citizens from the Commonwealth of Independent States (CIS) may enter Russia without a visa and stay for three months (that is what we call below a very vulnerable legal status, as explained in Section D of the online Supplementary Materials).Consequently, migration consists mainly of circular migrants who travel back and forth between Russia and their country of origin often on an annual basis or less (Danzer & Dietz, 2018;Zimmermann, 2014).
Although labor migrants play an important role in the Russian economy, attitudes towards migration are more negative than in other countries (Table S.M1 in the online Supplementary Materials).According to the independent Levada Center surveys, attitudes towards Central Asian migrants are increasingly negative: 38% of respondents in 2017 declared that they had a negative attitude towards labor migrants from Central Asia, compared to 31% in 2007, and 19% for migrants from Ukraine.Uzbekistanis and Tajikistanis are likely to face xenophobia, discrimination and violence.
Women made up 18% of citizens from Uzbekistan in Russia in 2016 and 16% from Tajikistan (Rocheva & Varshaver, 2018).Gender inequality is high in Russia, ranking 45 th out of 62 developed countries in the 2019 Gender Inequality Index.Gender norms are more traditional than in other destination countries (Table S.M1 in the online Supplementary Materials).It may affect women differently from men, resulting in harsher living or working conditions for women migrants that accumulate social disadvantages.However, gender norms in Tajikistan and Uzbekistan reflect a patriarchal society with a more gendered assignation of roles.Women are viewed as primarily responsible for the home and children (Table S.M2 in the online Supplementary Materials).These norms tend to translate into practices that limit women's access to basic services, potentially resulting in health disadvantage before migration (Ll acer et al., 2007), or in health advantage due to higher human capital selection (Hofmann, 2017).

Health and healthcare in Russia: what about migrants?
The Russian healthcare system is partly inherited from the so-called Semashko system of the USSR, which aimed at providing free universal care to all citizens.Since the early 1990s, the system is under-financed and various reforms were implemented to promote accessibility, quality and efficiency, such as mandatory health insurance (MHI) financed by social contributions and local budgets (Popovich et al., 2011).However, private expenditure still represents 35 to 50% of total health expenditure.It includes mostly out-of-pocket expenditures (private fees, public informal payments, medicines) and voluntary health policies (VHI) (Aistov, Aleksandrova, & Gerry, 2021;Danishevski, Balabanova, Mckee, & Atkinson, 2006;Mavisakalyan, Otrachshenko, & Popova, 2021).Several decades after free universal care was enshrined in the constitution, there are still strong disparities in the accessibility and quality of services among regions and income groups (Aistov et al., 2021;Manning & Tikhonova, 2012;Popovich et al., 2011).The majority of Russian citizens say that they have forgone care due to lack of time, lack of money or lack of connections.As a result, the health outcomes in the general population are low.
Migrants' access to health care depends on their legal status.For example, migrants with a residence permit are entitled to receive free of charge all the benefits guaranteed by MHI.Migrants without a residence permit benefit from free medical assistance only in case of life-threatening emergencies or childbirth (Mukomel, 2013a).When buying a patent, migrants should theoretically also take out VHI, which covers few services.However, migrants in Russia, regardless of their legal status, face the same financial, social and geographic barriers as natives, as well as migrant-specific barriers such as language and cultural gaps, potential discrimination, lack of information.One of the main problems cited by migrants is the uncertainty about the final amount of medical fees.It limits their access to care and ultimately affects their health (Demintseva & Kashnitsky, 2016;Kashnitsky & Demintseva, 2018;King et al., 2020;Rocheva & Varshaver, 2018).
Regarding the specific situation of women migrants from Central Asia in Russia, recent surveys highlighted the many health problems they face: deep fatigue and psycho-social risks, loneliness, lack of social support, and concerns about pregnancy and reproductive health (Agadjanian et al., 2022;Agadjanian & Zotova, 2012, 2019;King & Dudina, 2019;Zotova, Agadjanian, Isaeva, & Kalandarov, 2021). 1 Studies focusing on women highlighted intersecting factors of distress: gendered vulnerabilities, poorer working conditions, double burden (domestic work), love separation, higher family concern, stress of not achieving goal 2 (Zotova et al., 2021).Migration variables such as legal insecurity are detrimental to migrants' psychosocial well-being, but the effect seems more ambiguous for physical health (Agadjanian et al., 2022;Agadjanian & Zotova, 2019).

Data description
We conducted an original survey of 1,213 migrants from Tajikistan and Uzbekistan, in the Moscow region, between July and August 2019. 3We decided to focus on Central Asian citizens from Uzbekistan and Tajikistan as they represent the major part of labor migration from outside the Eurasian Economic Union.The other Central Asian country providing many migrants -Kyrgyzstanis part of an economic agreement (Eurasian Economic Union).Its citizens, therefore, do not need work permits, which greatly facilitates their integration resulting in a different type of migration stay.It is therefore a different context (with migrants less vulnerable, at least in terms of legal status) and a more feminine migration (38% of migrants from Kyrgyzstan are women (Rocheva & Varshaver, 2018), compared to only 16% and 18% from Tajikistan and Uzbekistan).Also, focusing on Moscow was a good compromise since it was not possible to sample all the regions but the capital remains the most attractive for migrantsalmost half of the Tajikistani migrants in Russia are said to be located there (Agadjanian & Zotova, 2012).
Since there is no sampling frame for temporary migrants, we could not make a probabilistic sampling based on residence.Also, as Agadjanian and Zotova (2012) noted this method is not accurate for this mobile population spending long hours at work, as well as the method of respondent-driven sampling used for rare and sensitive groups.Thus, we chose to recruit in different areas where migrants are living and working (residential zones, markets, tea places, restaurants, construction sites, etc.) asking the interviewers to diversify the locations as much as possible.The sample is a quasi random sampling with quotas for citizenship and gender.
As the interviews were uploaded to our cloud each day and geolocated, it enabled us to monitor the collection daily and check that the instructions, in particular the location criteria and quotas, were being respected.
Our criteria for inclusion in the survey were: being an adult Tajikistani or Uzbekistani migrant, i.e. at least 18 years old, having citizenship of Tajikistan or Uzbekistan and currently living in Russia.The sample excluded touristic stays: 97% of the migrants surveyed had been in Russia for more than a month; only 0.2% were there for the first time and less than a month; 80% of them were working, while the remaining 20% were looking for a job.For ethical reasons, we informed the respondents about the study and asked for their official consent to be involved in the study.
Is there a gender gap in health among migrants in Russia?1931 To make our sample more representative of the Tajikistani and Uzbekistani migrant population, we followed the citizenship and gender distribution in and around Moscow, with one-third Tajikistanis, two-thirds Uzbekistanis and one-fifth women, based on estimates given by the Federal Migration Service and previous studies (King & Dudina, 2019;Mukomel, 2014;Rocheva & Varshaver, 2018;Zotova et al., 2021).Among 1,213 migrants, women account for 249. 4 Our distribution of education and age is comparable to a previous survey of migrants across the CIS (Mukomel, 2017) (Section B in the online Supplementary Materials describes data collection).
The descriptive statistics (Appendix) on our sample verify the intuitions described in the introduction and converge with the results of other previous studies on this particular population (Hofmann, 2017;Mukomel, 2017;Rocheva & Varshaver, 2018).Most migrants interviewed are typical circular labor migrants who have been alternating temporary stays for years.Women are in poorer health than men.Women are on average better educated, older, more often divorced, and their place in the production system is less advantageous than that of men.These differences reflecting either gendered selection or migration experiences could partly explain the gap in health status, calling for ceteris paribus reasoning.
We use two main empirical models in order to estimate (1) the gender gap in migrants' current health status (baseline model) and the heterogeneous effects of migrant profile on this gender gap; (2) how gender influences health evolution (pseudo-dynamic model).

Baseline model
First, we examine whether the health status of migrants is likely to differ between women and men and whether this gap is due to differences in observed characteristics.
The health status is captured through four indicators which are commonly used and recognized as good predictors of mortality (Atella et al., 2019;DeSalvo, Bloser, Reynolds, He, & Muntner, 2006;Jusot et al., 2009;Lee, 2000;Moullan & Jusot, 2014).These are current selfassessed health, mental health, chronic diseases and illness/injury in the past six months; all are binary (see section D in Supplementary Materials for more details on the construction of the variables of interest).
The latent health status of the migrant i is denoted H Ã i , which represents health status on a continuous basis in equation (1).It is assumed to be determined by characteristics such as: sociodemographic characteristics (X i ) including gender our main variable of interest; migration trajectory features (M i ) such as cumulative time spent in Russia; living conditions (L i ) especially migrants' legal status 5 ; working conditions (W i ) measured among others by an index created using a multiple correspondence analysis; 6 and also other unobserved factors included in the error term ( i ).
The estimated model is therefore as follows: Migrants declare good (or very good) health (H 1 ¼ 1) and not suffering from depression/sadness (H 2 ¼ 1) if their latent health status is above a certain threshold, which may be different for each health outcome (h Ã 1 and h Ã 2 , respectively).Therefore, for migrant i, health outcome j, H ij (j ¼ ½1, 2), is determined as follows: Conversely, they will report having a chronic disease (H 3 ¼ 1) and having been ill in the past six months (H 4 ¼ 1) if their latent health status is below a certain threshold (h Ã 3 1932 S. Pellet and M. de Talanc e and h Ã 4 , respectively): These dichotomous models are estimated by probit regression, as commonly used in the literature (Moullan & Jusot, 2014;Subedi & Rosenberg, 2014).Average marginal effects (AME) are presented (Results in Section 4.1).
To investigate whether the gender gap in health status is constructed by migration and varies across migration profiles, we slightly modify the baseline model by incorporating interactions between gender and potential factors.We focus on the relationship between the gender gap and two migration-related factors: (1) duration of migration and (2) legal status.By interacting gender with the cumulative time spent in Russia (all migration stays), we first identify whether the gender gap widens with a longer length of stay or, on the contrary, whether it narrows as a result of the potentially greater integration of women.AME associated with gender are computed for different lengths of migration.We then explore how greater institutional integration through a more protective legal migration status can potentially reduce the health gap between migrant men and women.Gender is interacted with the degree of vulnerability of the legal status, allowing us to compute AME associated with gender for each legal status (Results in Section 4.2).

Pseudo-dynamic model
While we try to control for a maximum of factors, it is important to bear in mind that the present analysis does not claim to identify causal relationships, as several biases may arise from unobserved heterogeneity or reverse causality.One crucial issue when observing current migrants' health status is that we cannot ensure that the observed gender gap does not simply reflect a gender difference in health status prior to migration, which could be due to a gender health gap in the country of origin or to a gendered selection in migration as mentioned above (Atella et al., 2019;Brabete, 2017;Fleury, 2016).
To overcome this problem, we rely on self-assessed health pertaining to two stages of the life course: pre-migration (retrospective) health and in-migration (current) health.We start by including pre-migration self-reported health as a predictor of health in migration.Similar strategies including previous health status as a control have been used to investigate health status in migration in Indonesia and in Mexico (Lu, 2010;Ullmann, Goldman, & Massey, 2011).
We then take the analysis a step further by exploring the dynamics of self-reported health between the pre-migration and the current situation.As before, the latent health of migrant i at time t (H Ã it ) is assumed to depend on socio-demographic factors, migration trajectories (if applicable), current working and living conditions, as well as unobserved constant and timevarying individual characteristics (u i and it ).Each migrant was asked to report their health for two periods: before they first migrated (t ¼ O) and at the time of the survey when in migration (t ¼ T).The specification can be expressed as follows: 7 Theoretically, if the latent health before migration is higher than the latent health at the time of the survey, it implies that the individual has experienced a deterioration in health: Is there a gender gap in health among migrants in Russia?1933 This equation can be estimated using a probit model where the dependent variable (Deterio) is a dummy variable indicating whether the migrant's pre-migration self-reported health is better than his/her current self-reported health.It measures a deterioration in self-assessed health over time and migration.Focusing on health deterioration is justified by our data.For most migrants, self-rated health either deteriorated or remained the same, while very few experienced an improvement (Table S.M11 in the online Supplementary Materials).
By differentiating between pre-and post-migration health status, this 'dynamic' model allows us to remove time-invariant unobserved characteristics that affect pre-migration and current health status in the same way (u i ).For instance, if there is a gender declarative bias or fixed unobserved factors (such as individual risk aversion, genetic components of health, preferences, etc.) that impact pre-and post-migration health to the same extent, this 'dynamic' model will remove these biases. 8 We include labor market trajectories, as previous research has shown that adverse employment changes affect health (Burgard, Brand, & House, 2007;Dooley, Prause, & Ham-Rowbottom, 2000;Mossakowski, 2009;Ro, 2014;Ro & Goldberg, 2017).Based on retrospective questions about employment before the first migration, we distinguish five employment trajectories: (1) labor market entry, (2) exit from the labor market, (3) upward mobility, (4) downward mobility and (5) no change (Table A1 in the Appendix).In the model, we synthesize them as two dummies indicating whether the migrant has entered the labor market since first migrating and whether he/she has experienced downward employment mobility (Results in Section 4.3).
We finally re-estimate this pseudo-dynamic model by adding interaction variables to investigate whether the relation between gender and health dynamics varies according to the duration of migration and legal status.

Gender gap in current health status: baseline model
Table 1 reports the estimated gender gap for self-assessed and mental health while Table 2 presents the results for chronic disease and recent illness.We conduct this analysis in five steps.The first step includes only gender as an independent variable.We then successively add four sets of explanatory variables to assess whether the initial gender gap is due to differences in the distribution of observed characteristics between men and women.In the second step, we add sociodemographic controls.In the third step, we add migration-related characteristics and living conditions.In the fourth step, current working conditions in Russia are added. 9In the fifth and final step, we add migrants' hourly wages. 10 In line with the literature showing a female disadvantage in health for the non-migrant population (Almeida-Filho et al., 2004;Case & Deaton, 2005;Kobayashi & Prus, 2012;Malmusi, Borrell, & Benach, 2010;Subedi & Rosenberg, 2014), migrant women are also found to be in poorer health.Controlling for sociodemographic, living and working characteristics, the gender gap is smaller but is still high and significant for three out of the four indicators.Sudden illness or accidental injury is the only health outcome that is no longer significantly associated with being a woman (column (10), Table 2).Without any controls, women are 20 percentage points less likely to report being in good health (column (1), Table 1), compared to a difference of 10 percentage points when controlling for all observed characteristics (column (5), Table 1).Migrant women are also 8 percentage points more likely to experience sadness or depression (column (10), Table 1) or to suffer from a chronic disease (column (5), Table 2).These results suggest that the disadvantage of women migrants in health observed in Table A1 is not entirely due to differences in age or marital status or to worse living and working conditions.In order to quantify the proportion of the raw gender gap explained by observed characteristics, a Fairlie decomposition is estimated (Table S.M14 in the online Supplementary Materials).It shows that only socio-demographic differencesespecially marital status and agesignificantly explain the gender gap in self-assessed and mental health.Regardless of the measure of health considered, a large part of the gender gap is unexplained (36-87%).This residual gender gap may result from unobserved characteristics, an issue that is addressed in more detail by the pseudo-dynamic model.It could also be due to a differential effect of observed factors for men and women, a possibility that we investigate using interaction terms.

Heterogeneous effects: interaction models
To determine whether the gender gap is partly attributable to migration trajectories and conditions, we test the sensitivity of the gender differential to different profiles.In Tables 3 and 4, we estimate the AME of gender on health for different lengths of stay and different legal statuses, through probit models interacting gender with these migration profiles. 11 The gender gap is only significant for migrants who have spent more than a year in Russia (Table 3).These results suggest that health status is not different for male and female recent migrants.However, after a certain length of time spent in Russia, women begin to suffer a health disadvantage.Thus, women's health may deteriorate more rapidly during migration.
The relationship between gender and health also varies depending on migrants' legal status (Table 4).Migrant women are significantly less likely to be in good health only among legally vulnerable subgroups, i.e. those who only have a patent (significant at 1%) and, to a lesser extent, those who have no legal documents (at 10%).A more protective legal status reduces the health differential between men and women.In line with a previous study analyzing health within an intersectional perspective in other countries and showing a cumulative negative effect of being a woman from a poorer background (Gkiouleka & Huijts, 2020), we find that gender inequality is higher for the most vulnerable profiles.

Gender gap in health deterioration: pseudo-dynamic model
Controlling for self-assessed health before migration, women are still more likely to be in poorer health than men, even though the coefficients slightly decrease (Table 5).The gender Is there a gender gap in health among migrants in Russia?1937 difference in health status does not seem to reflect only a pre-migration gap.Table 6 reports the results of the pseudo-dynamic model, in which we compare self-assessed health before and during migration.Being a woman is strongly positively associated with the probability of undergoing a deterioration in health while in Russia, even after observed living and working characteristics are controlled for.Migrant women are on average 11 percentage points more likely to experience a deterioration in health than men.This strong relationship between gender and health dynamics confirms that a substantial part of the gender gap in health among migrants arises during migration.As in the baseline model, results from interacted models suggest that the gender gap we observe for health 'dynamics' appears after some time spent in Russia and only for migrants who have a less protective legal status (Table S.M16 in the Supplementary Materials).
The residual gender gap we found may result from several unobserved differences between women and men.Migrant women may experience threefold discrimination (as migrants, as a  minority and as women) (Ll acer et al., 2007) or fear being discriminated against by Russian doctors ( King & Dudina, 2019).Women may face specific barriers to obtaining healthcare, such as lack of time (women's double day) or lack of information about their rights, leading to poorer health.This suggestion is corroborated by some descriptive statistics in our sample.In line with King and Dudina (2019), in our sample women have less medical insurance, despite their greater needs, and declare more often than men that they have delayed or forgone care.

The role of the type of migration
We may expect differences in health outcomes for circular, temporary, and long-term migrants.If the type of migration affects health and is correlated with gender, the estimated gender gap may be biased.Although we control for the duration of migration, it is also important to differentiate between circular and other migrants.To this end, we distinguish new migrants and migrants with multiple stays in Russia.For the latter, we know the number of times migrants returned to their country of origin and the number of months they spent in Russia in the past year, which provides us with proxies for circular migration.As expected, migrants with several short-term stays have specific profiles, as they are for instance less likely to have relatives in Russia and are less institutionally integrated, as measured by legal documents and medical insurance (Table S.M17 in the online Supplementary Materials).When we control for the type of migration, whatever the proxy used, the gender gap in self-assessed health, chronic disease and mental health remains positive and significant (Table S.M18 in the online Supplementary Materials).

Potential recall bias
Second, a problem with the pseudo-dynamic model is the potential recall bias, as we use retrospective questions to capture health trajectories.If migrant women are more subject to recall bias than men when answering retrospective questions, especially when they migrated long time ago, gender gap estimates may be biased.
To mitigate this potential bias, we estimated the pseudo-dynamic model while retaining only those migrants who first migrated more recently, as the recall bias is likely to be lower for them.Regardless of the sample considered, women remain more likely to experience a deterioration in health during migration (Table S.M19 in the online Supplementary Materials).

The cohort effect
Third, the variable recording the cumulative time spent in Russia may potentially capture not only the total stay duration but also some differences between cohorts of migrants.Descriptive statistics computed separately for older and more recent cohorts show some disparities, with recent migrants having migrated at an older age, being less educated and less likely to have spoken Russian as a child.If selection mechanisms have changed over time for one gender and not the other, it could explain a potential gender gap for older cohorts.To check whether this affects our results, we run the same estimations, first with a continuous variable indicating the year of arrival, then with a binary variable indicating arrival before or after the 2014 economic downturn, and finally separately for cohorts that arrived before and after 2014. 12Our main results remain valid when controlling for the date of first arrival in Russia: the gender gap is always significant (Table S.M20 in the online Supplementary Materials).Nevertheless, when looking separately at migrants who first migrated before or after 2014, the gender gap in selfassessed current health is significant only for older cohorts of migrants (columns (3) and ( 4)).One explanation could be that selection in migration has changed over time with a more pronounced gender disadvantage in pre-migration health for older cohorts.However, this explanation is not supported by our results: women in older cohorts are not in poorer health than men before migration (columns ( 19) and ( 20)).This suggests that the gender gap in current selfassessed health among the older cohorts is likely to be due, once again, to migration experience, which has lasted longer for them than for recent cohorts.This is corroborated by the results regarding health 'dynamics ' (column (23)).For older cohorts, women are more likely than men to have experienced a deterioration in health since they migrated.

The role of family trajectories
Family situation may have evolved differently for men and women, which could explain the greater deterioration in women's health.Here we reach the limits of our dataset, which is not a panel and for which biographical data, date of marriage, divorce, childbirth, etc., are not available.It is also possible that the same family situation has a different influence on health according to gender.For example, leaving behind a child or elderly parents can be particularly detrimental to migrant women's health.Differentiated analyses for a variety of family contexts (with or without children/parents/partner in Russia, etc.) were carried out.First, when controlling for the type of relatives in Russia and stayed in the home country, the estimated gender gaps remain significant for the same outcomes, but are slightly reduced, suggesting that the family situation may play a role in explaining migrant women's health disadvantages.Second, the negative effect of gender on health seems to be driven rather by women being burdened with domestic duties (double day effect) rather than suffering from isolation (see the discussion in the online Supplementary Materials section G.).

Conclusion
This article contributes to the burgeoning literature on gender and health in migration, by investigating whether there is a gender gap in health among Central Asian migrants in Russian Federation, using an original database collected by the authors.It then proceeds to investigate whether this gap is produced (or reinforced) by the migration experience (gender-specific barriers, segregation in the host country, etc.).
We find a gender gap, with women being on average in poorer health than men.This gender gap appears with time spent in migration and concerns only migrants with a vulnerable legal status.We also show that, even though migrant women report being in poorer health than men before migration, their health is also more likely to deteriorate over time during migration.
Our findings suggest that the observed gender disparity is not entirely caused by a pre-migration health gender gap nor by observed gender differences in sociodemographic, migration, living or working characteristics.Gender disparity is reinforced by how migration is experienced, as men and women are affected differently by the factors identified.This situation could have serious consequences for the economies of both the host and origin countries since women experience a more rapid deterioration of their health and a decrease in their welfare, their productivity and sent remittances.It is particularly alarming if our results can be extrapolated to migrants from other countries, as women account for nearly half of the migrants worldwide.The situation calls for public policies and prevention measures targeting in particular women in more vulnerable circumstances and those forced to repeat their stay every year.
The residual gender gap we find may result from several unobserved differences between women and men.Migrant women may experience threefold discrimination (as migrants, as an ethnic minority and as women) (Ll acer et al., 2007) or fear being discriminated against by Russian doctors (King & Dudina, 2019).Women may face specific barriers to obtaining healthcare, such as lack of time (women's double day) or lack of information about their rights, leading to poorer health.This suggestion is corroborated by King and Dudina (2019), who focus on Central Asian women and show that they often delay their visits to doctors, and by our descriptive statistics.Women more often than men declare that they have forgone care and have less medical insurance, despite their greater health needs.
It is also possible that the gender gap varies depending on gender norms.Indeed, gender norms and attitudes towards migration may be a potential mechanism leading to women's health disadvantage.Although the nature of the data does not allow us to explore this, future research could investigate this issue and attempt to directly account for variations in gender norms.
Finally, as shown by the previous literature, social capital plays a large role in health (Rose, 2000).Migrant women may be more socially isolated, especially those who have followed their husbands and immigrant women working in the domestic sector, who have little chance of establishing social relationships (Ll acer et al., 2007).This greater social isolation can be detrimental to their health (Litwin, 2006;Wong, Yoo, & Stewart, 2007).Immigrant women lacking institutional integration may have fewer alternative resources and be less included in informal solidarity networks developed to face health shocks, a research track we follow in another ongoing study.We find that fewer women participate in informal solidarity networks that Is there a gender gap in health among migrants in Russia?1941 collect money to help migrants in difficult circumstances (financing legal documentation, emergency medical care, funerals, etc.).
Further research on the role of access to care and the role of alternative social resources regarding gendered health differences is already underway.If the disadvantage of women with vulnerable status is confirmed, it will reinforce the policy recommendation to improve migrants' integration and preventive medical care and to target in particular the most vulnerable women migrants.

Notes
1.These rare and rich studies present however some limitations about our research question.They focus on Central Asian women without comparing them with men.Also, they are generally based on ethnocultural perspective and use ethnicity as an identification variable.This paper investigates the migration experience of foreigners born abroad, using citizenship criteria.2. Job sectors such as care are less remunerating than male-oriented sectors such as construction.3.This survey was part of a larger project, the REFPoM project (Rituels et Economie Fun eraires Postsocialistes en contexte Migratoire), funded by the French National Research Agency (ANR).More details can be found on REFPoM site https://refpom.hypotheses.org/.4. Due to sample size, we could suspect statistical power limitation, thus we computed the minimum detectable effect of gender for each outcome: the estimated differences are high enough to let us conclude.See section C in the online Supplementary Materials.5. We distinguish four degrees of legal status: very vulnerable, vulnerable, more secure and very secure.See section D in Supplementary Materials for more details.6. See Tables S.M7 and S.M8 in the online Supplementary Materials for more details on the construction of working condition index.7. The coefficients associated with the different factors are indexed by the period (O or T) to allow characteristics to have a different impact on health depending on whether the individual is in the country of origin or Russia.For instance, gender does not necessarily have the same impact since norms and culture between the two countries are likely to differ.8. Ideally, we would like to observe all time-varying independent variables before and during migration.However, as we rely on cross-sectional rather than longitudinal data, only some characteristics before migration are reported using retrospective questions.9.The sample is reduced to migrants who were working at the time of the survey (93% of the sample) or who were not currently working but had worked in Russia in the past (6%, among them 75% who had been unemployed for four months or less).10.At this stage, we loose 79 observations due to difficulties in declaring earnings.11.We also estimate a model where gender and working conditions are interacted and find that the health gender gap is not sensitive to working conditions.Whatever the degree of difficulty, women have a disadvantage for the three first outcomes.However, for illness within the past six months, only women working in worse conditions are more likely to get sick (Table S.M15 in the online Supplementary Materials).12. Taking the date of an economic downturn is an easy way to divide into two different groups those who had potentially different conditions of departure and stay.We chose the 2014 downturn instead of the 2008, first to have more individuals in the sub-sample -they were fewer who arrived before 2008 -but also because the 2014-2015 crisis was very dramatic in Russia and led to a large decline in remittances (Figure S.M3 in the online Supplementary Materials).13.Pellet, S., and de Talanc e, M. (2021)."Is there a gender gap in health among migrants in Russia?".Erudite laboratoire d' economie Paris-Est.https://ideas.repec.org/s/eru/erudwp.htmlLEGOS seminar), to M. Leturq, E. Cambois, M. Seg u, G. Duth e Sophie Le Coeur (INED), to I. Chort (TREE seminar), to I. Ohayon and J. Thorez (EHESS seminar on Central Asia) and all other participants.We also thank the participants of the 2021 International Conference in Development Economics, the 37 th Applied Microeconomics Days, the 69 th Congress of the French Economic Association, the Population Association of America 2021 Annual Meeting and the Women on the Move workshop for their useful feedback.We thank the French Collaborative Institute on Migration for the fellowship and the members of the DYNAMICS department for the inspirational research discussions.Finally, we are very grateful to the referees for their precious comments that helped to significantly improve the quality and clarity of the article.An older version of this manuscript were deposited on an academic website as a preprint for non-commercial purposes. 13

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
We acknowledge the French National Agency of Research for financing REFPoM project (Rituels et Economie Fun eraires Postsocialistes en contexte Migratoire, ANR-17-CE41-0003) and our respective laboratories and research units (Nanterre University, Universit e Gustave Eiffel, Universit e Paris-Est Cr eteil and INED) for their substantial material, financial and logistic help.

Table 1 .
Estimates of the role of gender on self-assessed and mental health

Table 2 .
Estimates of the role of gender on chronic disease and recent illness

Table 3 .
Sensitivity of the gender health gap to time spent in Russia We control for all observable characteristics, such as in baseline model, reported in Table S.M12.Robust standard errors in parentheses: ÃÃÃ , ÃÃ and Ã denote a significance at respectively 1%, 5% and 10%.

Table 4 .
Sensitivity of the gender health gap to legal vulnerability We control for all observable characteristics, such as in baseline model, reported in TableS.M12.Robust standard errors in parentheses: ÃÃÃ , ÃÃ and Ã denote a significance at respectively 1%, 5% and 10%.

Table 5 .
Estimates of the gender health gap controlling for pre-migration self-assessed health ÃÃÃ and ÃÃ denote significance at respectively 1% and 5%.Other controls not reported are those included in TableS.M12.

Table 6 .
Estimates using a pseudo-dynamic model