Origin-country gender norms, individual work experience, and employment among immigrant women in Sweden

Many Western countries are built on a dual-earner model and have high levels of female labor force participation. Increasing the labor market activity of immigrant women is therefore seen as a key part of immigrant integration. However, female labor force participation (LFP) differs substantially between countries, reflecting differences in work-related gender norms that can continue to influence preferences and behaviors after migration. In this study, we investigate how origin-country gender norms and migrant selection interact to produce post-migration outcomes. Our data shows that immigrant women in Sweden have a higher level of pre-migration work experience than expected based on origin-country female LFP, indicating positive selection. Furthermore, the association between origin-country LFP and post-migration employment varied with work experience. For women without origin-country work experience, origin-country LFP was positively associated with employment in Sweden. For women with origin-country work experience, origin-country LFP however was not associated with higher likelihood of employment in Sweden. Though our focus is on immigrant women, we also include immigrant men in our analysis to test our prediction more thoroughly. For men without origin-country work experience, origin-country LFP was negatively associated with employment in Sweden, while we found no association for men with origin-country work experience. Our results show that migrant selection is a crucial factor in understanding the relationship between origin-country LFP and post-migration labor market outcomes, and that these patterns vary with gender. Policy interventions targeting immigrant women from countries with low female LFP should therefore not assume that women arrive socialized with gender-norms that hinder labor market activity.


. Introduction
In many Western countries, immigrants' labor market integration has become a key policy issue, particularly regarding immigrant women (Towns, 2002;Larsson, 2015;European Migration Network, 2022). In general, immigrants tend to have worse labor market outcomes than non-immigrants (Bevelander and Pendakur, 2014;Gorodzeisky and Semyonov, 2017), and women in general tend to have lower labor market participation and employment rates than men (Albrecht et al., 2003). Immigrant women are in turn less likely to be active in the labor market than non-immigrant women, less likely to be employed conditional on being active, and more often in unskilled jobs conditional on being employed (Ballarino and Panichella, 2018).
/fhumd. . Societal gender norms often prescribe different behaviors for women and men, which shape gendered practices in the division of labor. These norms and practices are cultural, in the sense that they relate to values, beliefs, or preferences that are likely to be shared among people experiencing similar socialization processes (Polavieja, 2015). And while a person's values, beliefs, and preferences change over their life course-due to new experiences or circumstances, institutional pressure, or exchange with other people-change is often gradual. As a result, individual practices often continue to be influenced by norms encountered during early socialization to some degree (Perales et al., 2019). Thus, the labor market outcomes of immigrant women can be affected by both the work-related gender norms from their countries of origin, and the work-related gender norms in their countries of destination (Van Tubergen et al., 2004;Blau, 2015;Uunk et al., 2016;Kanas and Steinmetz, 2021).
To explore the relationship between origin-country gender norms and post-migration outcomes, previous research has used the rates of female labor force participation (henceforth LFP) in the country of origin as an indicator of the work-related gender norms that were dominant in the pre-migration socialization process. Assuming that higher female LFP is an indication of societal work-related norms that encourage female labor market participation, this indicator has then been used to predict post-migration labor market outcomes for immigrant women, finding consistent associations between higher origin-country female LFP and better labor market outcomes in several destination countries (Antecol, 2000;Van Tubergen et al., 2004;Blau et al., 2011).
However, a problem in this line of research is the failure to account for migrant selection-how individual experience relate to societal norms and what consequences that this selective experience may lead to. Immigrant women with work experience from a country with low a female LFP are positively selected in that they managed to overcome the normative or structural barriers to participate in the labor market and find a job. This implies strong determination, high cognitive or interpersonal skills, or the presence of other facilitating factors. Such characteristics and factors should in turn also be beneficial for post-migration labor market outcomes (for similar arguments see: Feliciano and Lanuza, 2017;Engzell and Ichou, 2020;Schmidt et al., 2022). Importantly, the lower origin-country female LFP, the more positively selected will women with work experience be, leading to the opposite association with post-migration outcomes from what previous research has postulated.
The aim of this study is to investigate how origin-country gender norms and migrant selection interact in influencing post-migration outcomes. Using immigrants in Sweden as our case, we model the probability of being employed in the destination country on genderspecific origin-country LFP, individual work experience, and an interaction between the two as a measure of selection. Our prediction is that the gender-specific LFP in the country of origin has different effects on employment after immigration depending on individual pre-migration work experience. Though our focus is on immigrant women, the effect of selection should exist regardless of gender. We therefore also include immigrant men in our analysis to test our prediction more thoroughly. By investigating the interaction between origin-country gender norms and migrant selection, we contribute to both the general field of research on female labor market integration and the growing body of literature emphasizing migrant selection.
From a policy perspective, our design makes an additional contribution. Low female LFP in a country of origin means that immigrant women from that country are both less likely to have been exposed to work-oriented norms during early socialization, and less likely to have work experience. It is therefore impossible to separate the effects of context-level and individual-level factors on employment probability without controlling for pre-migration work experience. This in turn makes it difficult for policy makers to gauge whether initiatives against unemployment should target gender norms relating to work, or the lack of individual work experiences. By considering both aspects simultaneously, our results can better inform policy on which interventions are likely to be effective.
The article is structured as follows. In the next section, we present the Swedish context. In international comparison, the Swedish welfare system is strongly geared toward a dual-earner labor market, and female LFP is high. However, the country's welfare and migration policies have also been suggested to cause different migrant selection to Sweden than in continental Europe or the USA (Polavieja et al., 2018). This makes Sweden an interesting case for studying labor market outcomes for immigrant women. After this, we describe the theoretical underpinnings of our study. We also survey previous research on origin-country LFP and post-migration labor market outcomes. There is a wealth of research in this area. However, to the best of our knowledge there is currently only one study, from the USA, that also includes individual work experience in the country of origin (Blau and Kahn, 2015). There is consequently a knowledge gap that we address in this article. Having presented our hypotheses, we describe our data, and detail our variables and analytical strategy. Finally, we present our results, and conclude by discussing their wider implications for integration policies.

. Context
Labor market incorporation of immigrants in Europe is typically understood as a trade-off between employment and job quality, with policies in continental and northern Europe prioritizing job quality over swift employment (Kogan, 2006;Reyneri and Fullin, 2011). This is also the case for Sweden. Still, the Swedish labor market is relatively open to immigrants (Solano and Huddleston, 2020) and integration programs focus on socio-economic incorporation (Borevi, 2014). Sweden also has a welfare system based on a dualearner households (Oláh and Bernhardt, 2008), meaning that labor market policies and welfare institutions are designed to encourage the employment of immigrant women (Stier et al., 2001;Bevelander and Pendakur, 2014;Boeckmann et al., 2015). However, gender and immigrant background often intersect in non-additive ways (Browne and Misra, 2003;Cho et al., 2013). Immigrant women consequently face gender-specific obstacles that can lead to further disadvantages that are not easily described in terms of trade-offs (Cantalini et al., 2022).
To a large extent, immigrants to Europe come from countries with more traditional gender norms (Röder and Mühlau, 2014). Cross-sectional studies have found that traditional, male singleearner, attitudes are negatively related to immigrant women's labor force participation and work hours (Kanas and Müller, 2021). However, since culture is malleable, destination-country gender norms are also a key factor in the labor market participation of immigrant women (Kanas and Müller, 2021;Kanas and Steinmetz, 2021). In international comparison, Sweden scores high in genderegalitarianism (Inglehart et al., 2017). The gender gap in labor force participation in the Swedish population aged 15 and over is likewise low in international comparison, at 8 percentage points in 2011 (International Labor Organization, 2022). The strong societal norm of female LFP also means that immigrant women have high LFP in Sweden compared to international standards (Kanas and Müller, 2021). Still, the gender gap in LFP among immigrants is more than 12 percentage points-substantially larger than among nonimmigrants (Statistics Sweden, 2021). Because of this, the integration of immigrant women is often framed as a particularly problematic issue in the political discourse (Towns, 2002;de los Reyes, 2021).
Since this study focuses on a single country, we are not able to explicitly test how the Swedish context influences the results. Yet, Sweden's open labor market, labor market focus in integration policy, dual-earner model, high female LFP, and prevalence of gender egalitarian values, all suggest a structural push toward higher LFP for all immigrant women. This in turn leaves less room for variation based on origin-country factors. The Swedish context is therefore likely to have a suppressing effect on the association between origincountry LFP and employment outcomes.

. Theory
The labor market outcomes of immigrant women can be attributed to several factors. As for men, legal and structural barriers, lack of relevant human and social capital, and discrimination may all cause unfavorable results (Tibajev, 2022). Compared to immigrant men, immigrant women may face a double disadvantage in the labor market of being both migrants and women (Raijman and Semyonov, 1997). Research has suggested that gender norms play a role in this. Gender norms almost universally stipulate that men should participate in the labor force but norms vary in the degree to which women are also expected to do so. Importantly, there is often an expectation on women to do a larger share of unpaid household work, which hamstrings their possibilities to participate in the labor force and results in women's work not being recognized (Morokvasic, 1984). Differences in such gender norms have been suggested to explain the differing labor market outcomes of immigrant women from different countries, since immigrants bring cultural norms and social practices from their origin countries when they migrate (Read and Oselin, 2008).
The so-called cultural explanation was proposed by Reimers (1985), who noted that the higher degree of labor force participation of married black, compared to white, women was rooted in a particular historical experience of blacks in America, and not reducible to current conditions alone. On a similar note, she argued that different immigrant groups to the USA might have different views about "male and female roles in the family and about wives and mothers working outside the home, as well as [. . . ] the value placed on children, family size, household composition, and the education of women" (Reimers, 1985, p. 251). Such cultural differences might in turn result in systematic differences in behavior when faced with similar constraints and opportunities. This explanation was further developed by Antecol (2000), who proposed using gender differences in LFP in the countries of origin as an indicator of historical experiences that influence current behavior in the country of destination. Concretely, she suggested that being socialized into a context with a higher degree of female LFP acted as a "portable factor" (Antecol, 2000, p. 419) leading to a higher likelihood of female labor market activity in the country of destination, net of other factors that influence participation.
. . Origin country work-related gender norms and post-migration outcomes Building on Antecol, several studies have used female LFP in a country as an indicator of gender norms relating to work. Origin-country female LFP at the time of emigration is thus understood as a proxy for the societal norms regarding gender and work that dominated during a person's pre-migration socialization process. Studies in this tradition have found that immigrant women from countries with low female LFP are less likely to work after immigration than are immigrant women from countries with higher female LFP (Antecol, 2000;Blau et al., 2011;Polavieja, 2015;Gay et al., 2018;He and Gerber, 2020;Muchomba et al., 2020). Origincountry work-related gender norms have also been found to influence the labor market outcomes of children of immigrants (Eylem Gevrek et al., 2013;Finseraas and Kotsadam, 2017;McManus and Apgar, 2019), highlighting the roles of parents as socialization agents for transmitting work-related gender norms across generations (Fernández and Fogli, 2009).
Associations between origin-country female LFP and postmigration outcomes have also been reported from European countries, including Germany (Krieger, 2020), Italy (Scoppa and Stranges, 2019), and the Netherlands (Kok et al., 2011). There are additionally a number of comprehensive studies, assessing the association between origin-country female LFP and labor market outcomes in more than one country. Van Tubergen et al. (2004) included the USA, Australia, and 16 European countries, and found a positive association between origin-country female LFP and postmigration likelihood of labor force activity for immigrant women. Similarly, Kesler (2018) found a positive association between female LFP in the country of origin and the post-migration labor force participation for immigrant women in 16 European countries. Bredtmann and Otten (2015) used survey data on 26 European countries from the European Social Survey 2002-2011 and found a positive correlation between origin and destination country female LFP. In contrast, a recent study using data from European Social Survey, found that the association between female LFP in the country of origin and employment in the country of destination was only present for Muslim-majority countries of origin (Blekesaune, 2021).
Studies using Swedish registry data have found a substantial positive association between origin-country female LFP and the likelihood of establishment on the Swedish labor market for immigrant women (Neuman, 2018;Grönlund and Fairbrother, 2022). However, when controlling for individual fixed effects, Neuman (2018) found that immigrant women from countries with low levels of female LFP in fact worked to a higher degree than those from countries with high levels of female LFP. This suggests that factors unobserved in destination-countries' registry data play an important role in shaping the post-migration outcomes of immigrant women, highlighting the need to account . /fhumd. .
for origin-country work experience and its relation to the origincountry LFP.

. . Selection
As outlined above, there is wealth of empirical evidence indicating that work-related gender norms from the country of origin influence post-migration labor market outcomes for immigrant women. A recurring problem in this literature, however, has been the inability to separate societal work-related gender norms from individual work experience, and to explicitly model the joint effect of the two on post-migration outcome. That is, to account for migrant selection, the fact that migration processes are not random, and that there are systematic differences between migrating and sedentary parts of a population that can impact post-migration outcomes (Feliciano, 2020).
Cultures are never monolithic, and individual practices may differ substantially from those prescribed by societal norms. Nonconformity with dominant gender norms can be an important factor in the decision to migrate (Hofmann, 2014). Women seeking more egalitarian gender norms, women with a strong motivation to work, and women forced to work to provide for their family or compelled by other circumstances are thus subsets of immigrant women where the association between origin-country gender norms and postmigration outcomes might diverge from the general pattern (He and Gerber, 2020). This is because finding a job is facilitated by skills and education, but also hard-to-observe factors such as motivation, effort, cognitive skills, and non-meritocratic advantages.
When work-related gender norms are biased against women, finding a job as a woman takes more effort and determination given the same levels of skills and education. Immigrant women with work experience from a country with low female LFP are thus likely to be positively selected on unobserved factors relating to their capacity to find employment. This positive selection can, in turn, be helpful for entering the post-migration labor market (see Schmidt et al., 2022). An attempt at assessing the impact of selection on the association between origin-country female LFP and post-migration outcomes was made by He and Gerber (2020), who used the respective timing of migration by husbands and wives as a proxy for the relative work-orientation of individual immigrant women. Their assumption was that wives who migrate before their husbands are likely to be less traditional in their gender roles, than wives who migrate after their husbands. Focusing on immigrants to the USA, they found that origin-country female LFP influenced the likelihood of post-migration employment more for follower migrants than for unmarried, lead, and concurrent migrants. This suggests that selection can have an impact on the relation between origin-country LFP and labor market outcomes in the destination country.
Still, to the best of our knowledge, only one study has so far used data on pre-migration work experience to separate between country-level work-related gender norms and individual-level workrelated practices. Analyzing the outcome of annual working hours in the USA, Blau and Kahn (2015) used the New Immigrant Survey to show a negative interaction between origin-country female LFP and individual experience. The positive impact of origin-country LFP on hours worked was stronger for women who did not themselves work before migrating, indicating that both pre-migration norms and work experience are important factors for post-migration labor market outcomes. The authors interpreted the negative interaction as a substitutionary effect between culture and human capital, i.e., that origin-country gender norms have an effect when individual job-related human capital is absent.
Based on the discussion above, we hypothesize that the association between origin-country female LFP and immigrant women's post-migration labor market outcomes will depend on the immigrant woman's individual work experience. For women without work experience, higher origin-country female LFP should have a positive effect on post-migration employment, as noted by previous research. Women from countries with high female LFP are socialized into work-related gender-norms promoting female labor market participation, yielding a higher probability of post-migration employment. This results in our first hypothesis.
H1: For women without own origin-country work experience, there will be a positive correlation between origin-country female LFP and probability of employment in Sweden.
For women with pre-migration work experience, female LFP in the country of origin should not have the same effect. Instead, we hypothesize that the association should be negative due to selection. Being a woman from a country with low female LFP and having worked nevertheless implies a stronger-than-average propensity for labor market participation or the presence of characteristics or factors that facilitate entry into employment even if employment for women is rare. These women should therefore be positively selected and consequently have the highest probability of post-migration employment. This results in a second hypothesis.
H2: For women with own origin-country work experience, there will be a negative correlation between origin-country female LFP and probability of employment in Sweden.

. . Selection and post-migration outcomes for men
Research on the association between origin-country gender norms and post-migration outcomes has typically focused exclusively on immigrant women. The reason for this is that the work-related gender norm for men is largely invariant. Men are generally expected to participate in the labor market, regardless of country and historical moment-at least during the last centuries. The male work-related norm is also largely unaffected by female work-related norms, since both single-earner and dual-earner norms assume that men participate in the labor market.
Even so, some of the aforementioned studies have included men as a falsification test. Their findings have generally been that the effect of origin-country LFP is negligible for men (Blau et al., 2011;Blau and Kahn, 2015;Neuman, 2018), even if some studies reported more substantial effects (Finseraas and Kotsadam, 2017;Grönlund and Fairbrother, 2022) or mixed findings (Neuman, 2018). In contrast to previous research, we however argue that including men allows for a more thorough investigation of how selection impacts the relation between origin country gender norms, individual work experience, and post-migration labor market outcomes. We therefore include men not as a falsification test, but to test our predictions about the effects of selection more thoroughly.
The universal norm that men should be participate in the labor market implies that variation in male LFP is not caused by workrelated gender norms. Variations across time and space can instead be understood as a function of structural differences in labor markets and education systems. However, individual experiences might still differ from the societal norm, meaning that selection can be a factor also for men. Concretely, men without origin-country work experience should be negatively selected in relation to their genderspecific LFP. However, in the case of men, it is not because of differences in norms, but because men coming from countries with high male LFP had more ample opportunities to work, and the fact that they did not implies the presence of unobserved characteristics or factors that reduce the likelihood of finding employment in the future-such as disabilities or stigmatization. These unobserved characteristics or factors can in turn be assumed to also have a negative influence on the likelihood of post-migration employment. Following the reasoning of a universal male work norm, men with origin-country work experience should on the other hand be unaffected by origin-country male LFP. This gives us our two final hypotheses.
H3: For men without own origin-country work experience, there will be a negative correlation between origin-country male LFP and probability of employment in Sweden.
H4: For men with own origin-country work experience, there will be no correlation between origin-country male LFP and probability of employment in Sweden.
. Data and methods

. . Data and sample selection
To study how selection affects the association between origincountry LFP and post-migration employment, we use individual level data on immigrants from the Level-of-Living Survey for Foreign Born and Their Children (Migrant-LNU). Migrant-LNU is a comprehensive high-quality data set, collected by Statistics Sweden in 2010-2012 through retrospective interviews with a representative sample of the immigrant population in Sweden. The inclusion criterion was that the person had first immigrated to Sweden at least five years prior to the start of data collection. Data consists of a random sample, stratified on seven regions of origin and three age categories, allowing both comparisons between the strata and representativeness of the whole population. The response rate was 49.9 percent, yielding 3,448 respondents. The comprehensiveness of the data, including detailed information about the respondents lives both before and after migration, makes the data set particularly wellsuited to study the connection between pre-migration experiences and post-migration outcomes. More information about the Levelof-Living Surveys can be found at the Swedish Institute for Social Research (2017).
Given the focus on labor market experience before and after migration, we restricted our analytical sample to respondents who were at least 15 years old at the time of migration and were no older than 64 at the time of the survey. The lower age limit was set to ensure that respondents were old enough at the time of migration to have been exposed to origin-country gender norms during their socialization and had some opportunity to enter the labor market before migrating. The upper age limit was in turn set to ensure that respondents were still of working age when surveyed about their current labor market situation in Sweden. This excluded 1,578 respondents, mostly because many immigrants in the survey had come as children. Additional list-wise deletions on the included variables excluded 354 respondents, leaving an analytical sample of 1,516 (813 women and 703 men).
All results are weighted with post-stratification weights that account for the stratified sampling technique, as well as the differing non-response rates across the 21 strata and key demographic variables.

. . Variables
Our outcome variable is current labor market status, defined dichotomously as unemployed or employed (0/1). Based on the Migrant-LNU questions of labor market activity in the week preceding the survey, we define being employed as having gainful employment, full-time or part-time, or being self-employed. Respondents without employment/self-employment were instead coded as unemployed. From the latter group, we excluded respondents who were studying (and not simultaneously having or seeking employment) and the long-term ill or early retired. The reason for this being that these categories are not easily definable in terms of work-related gender norms or relation to the labor market since individuals in these categories can either be early retired after a lifetime of work or long-term ill for unrelated causes.
Gender and origin-country work experience are dichotomous variables based on survey responses in the Migrant-LNU. An alternative to distinguishing dichotomously between having and not having had employment before migration would be to use years in employment before immigration to Sweden as a continuous variable. However, this specification is less suitable for measuring selection since an individual is not twice as selected if she worked two years instead of one. Consequently, using a continuous variable specification added no additional explanatory power to the models. Therefore, we opted for the dichotomous distinction between having been employed or not.
Operationalizing origin-country work-related gender norms, previous studies differ in their choices of LFP measures. Some studies have used the LFP rate, measuring the share of women that are active in the labor market (Van Tubergen et al., 2004;Fernández and Fogli, 2009;Kesler, 2018;McManus and Apgar, 2019), while other have used the ratio of female to male LFP (Antecol, 2000;Blau et al., 2011;Blau and Kahn, 2015;He and Gerber, 2020). The latter choice has been informed by a desire to separate workrelated gender norms from other factors that affect labor market participation, specifically labor market structure and educational expansion. Countries can differ systematically in economic situation, the timing and degree of structural transformations, and in the degree of participation in education. These factors also affect LFP, for example since subsistence farmers and students are not participating in the labor force.
Despite these concerns, we have opted to use the more straightforward LFP rate. The main reason is that it relates to our measure of selective experience more directly than any alternative. How common or uncommon an individual's pre-migration work experience is depends on the proportion of other people in work (the LFP rate). Similar to previous research, we understand observed LFP for a specific country, year, and gender as the outcome of two processes: gender-specific norms regarding work and economic structure. The indicator for women will emphasize work-related gender norms, such as an ideal that women should focus on raising a family rather than working. For men, by contrast, the indicator emphasizes the structural component based on the theoretical assumption that men everywhere are expected to participate in the labor market. In other words, to the extent that male LFP rates differ from 100%, it is not because work-related gender norms stipulate that men should not work but because of higher participation in education, larger share of retirees, or similar structural factors. The data on origin-country LFP comes from the ILO Yearbooks for Labor Statistics (International Labor Organization, 1971-2000 and the ILOSTAT Explorer database (International Labor Organization, 2022). Since work-related gender norms shift over time, female LFP have changed substantially over the 20th century in many countries. To capture this, we have used multi-year records and calculated LFP at the year of migration for each observation in the data following a similar approach by Tibajev (2019). In this way, we can appropriately approximate both gender norms and selectivity of the individuals' experiences given the time and place of the premigration socialization. Respondents in the sample come from 171 countries of origin with 974 unique origin-country years. Quality of data sources and definitions vary inescapably across time and space. In our coding, we have accepted numbers as presented and to our best ability calculated the gender-specific LFP for the same age groups across all country-years with the age restriction of 15 years and up as the anchor (following ILO definitions).
Control variables are migration age and years since migration (YSM), both squared, to account for the chances of accumulating work experience in either context and the degree of socialization into pre-migration and post-migration gender norms. We also include education level and place of highest education, to control for the individual's human capital. Region of origin , based on the country the respondent lived in before the age 16, is included to account for origin-specific effects apart from work-related gender norms, such as discrimination in the labor market. Lastly, we included reason for migration to account for the effects of the overall integration trajectory in the destination country connected to different motives for migration. We additionally tested the variables family status and number of children but discarded them for parsimoniousness as they did not have any influence over the main effects.

. . Analytical model
Our theoretical discussion and hypotheses concern the degree to which the effect of origin-country LFP will vary depending on gender and individual work experience. We therefore test the four hypotheses using logistic regression, regressing a threeway interaction between the main variables gender-specific origincountry LFP, gender, and individual origin-country work experience on the probability of current employment. The interaction model allows the effect of origin LFP to vary across the four combinations of gender and work experience, while holding constant for all the included control variables.
The regression results are presented as gender-specific average marginal effects (AME) on the probability of being employed per one percentage point difference in origin LFP, separated by gender and origin-country work experience. Presenting the results like this makes the analysis directly conform to the hypotheses, as each AME will correspond one prediction. They are also a useful way of displaying results when the effect size of a variable is contingent on the status of some other variable, i.e., in an interaction (Mize, 2019). Full results are available in the Supplementary material.

. . Alternative specifications
The design and methodological choices can have a substantial impact on the results of a quantitative analysis of the same dataset and research questions (Breznau et al., 2022). This study includes five particularly important choices. For transparency, we discuss them below and offer alternative specifications for each choice.
First, we have opted for employment in Sweden as the outcome variable. An alternative would be to instead use participation, i.e., also including the unemployed as a positive case and leaving individuals doing housework or other activities as non-participating. Differences between the two outcomes is that employment has a stronger emphasis on human capital as an explanation, while participation relies more on gender-specific norms (for a similar argument see Blau and Kahn, 2015). As an alternative specification, we therefore use participation as the outcome.
Second, we have coded students and long-term ill/early retired as missing values in the outcome variable, effectively excluding them (258 respondents) from the sample. As an alternative specification, we include these categories (and the additional 258 respondents), both as non-employed and non-participating in the labor market.
Third, as discussed above, our choice was to use the LFP rate rather than ratio, when approximating work-related gender norms and the selectivity of women's work experience. Previous studies have however argued that the female to male LFP ratio is a more accurate approximation of the work-related gender norm for women, net of structural constraints that affects both genders (approximated . /fhumd. . by male LFP). To test whether this affects our results, we used the female LFP ratio as an alternative specification. For men, this would create a variable that is a constant of one across all observations, and we therefore keep the male LFP rate as the measure also is this alternative specification. Fourth, we have used values in the ILO origin-country data as they are presented in the yearbooks, using the gender-specific employment rate for 15 years and up (or closest possible version).
Since not all countries are covered for all years, respondents who migrated before we have any datapoints from a specific country was assigned the earliest available LFP rate for that country. For other missing years, we have estimated a linear trend between the two closest years with the available data. The ILO data, especially regarding earlier years, comes from different data sources, with different definitions over time and space, and is of varying quality. An alternative approach would be to model an underlying LFP rate as a function of the recorded values, thereby treating drastic changes and extreme datapoints as errors rather than real fluctuations. As an alternative specification we therefore predicted LFP values for each country, year, and gender using an OLS regression with the LFP rate as the dependent variable and a cubic polynomial of years as the independent variable.
Fifth, we restricted our sample to only include persons who were older than 15 at the time of migration, and 64 or younger at the time of survey. As outlined above, the reason for the lower bound was to ensure that persons in the sample both had been exposed to origin-country gender norms during their socialization process and had had some opportunity to enter the labor force in the country of origin. The lower age limit follows the definition of working age used by the ILO and thus the rates of origin-country LFP in this study. However, it can be argued that a lower age limit of 15 years old is too low, and that persons at the lower ages have not yet had enough time to enter the labor market in the country of origin. As an alternative specification, we instead limited the sample to only persons with a migration age of 20 or over, excluding 255 additional respondents .
/fhumd. . who were not in prime working age in both the country of origin and destination from the sample. Beside our main model, we include the results from all these alternative specifications, a total of 32 models, in the Supplementary material and briefly discuss them as a sensitivity analysis at the end of the results section.
. Results . . Descriptive statistics Table 1 displays the descriptive statistics for all included variables, separate for women and men. The measured labor market outcomes in Sweden-participation and employment-are very high. It is important to remember that students and the long-term ill/early retired have been exclude from the sample based on our definitions. When included, as is visible in the alternative specification of these variables, the proportion of employed and participating is reduced. In comparison, the overall LFP rates in the survey year of 2011 in Sweden, however for the slightly broader age range of 15-64, were 77 percent for women and 84 percent for men (International Labor Organization, 2022).
Immigrant women have a lower rate of both participation and employment compared to men. The differences are small, highlighting the strong dual breadwinner norm in Sweden. By comparison, the gendered pattern is much more pronounced in the origin-country variables, with women both coming from countries with considerably lower gender-specific origin-country LFP and having themselves been to a lesser extent employed than men. It is also worth noting that the variation in origin-country LFP is much larger among women than among men. Our theoretical assumption that male labor market participation is globally the norm and has little variation is thus confirmed in the data. Table 1 also shows the alternative measure female LFP ratio, i.e., female LFP divided by male LFP for the same country-year. The correlation between the LFP rate and ratio for women is 0.96. For men, the ratio is undefined. A gender difference evident in the table is the relationship between origin-country LFP rate and work experience. Women have a higher probability of origin-country work experience than their average LFP rate would suggest, while the relationship is the opposite for men. Table 1 also displays the distribution of the control variables.

. . Origin-country LFP
To better illustrate the degree of selection, Figure 1 displays the predicted probability of origin-country employment over origincountry LFP rate for women and men, respectively. The figure confirms two aspects already evident in Table 1: that fewer women have on average been employed in their origin country, and that the gender-specific origin-country LFP rate is both lower and varies considerably more for women than for men.
The figure also shows that the connection between origin-country LFP rate and the individual immigrant's experience are quite different depending on gender. For women, there is a mostly a positive association, with women coming from high-LFP countries also being more likely to have had own work experience in the origin country FIGURE Probability of origin-country employment based on a logistic regression with gender-specific origin-country LFP (squared polynomial). Prediction ranges from th to th percentile of origin-country LFP.

FIGURE
Main results. AME (with % confidence intervals from robust standard errors) on employment for gender-specific origin-country LFP rate, separated by gender and origin-country work experience. Full results from logistic model in Supplementary Table . compared to women coming from low-LFP countries. For men, there seem to be a negative association, though it is very small across the narrow range of male origin-country LFP.

. . Predicting labor market activity in Sweden
Immigrant women and men in Sweden thus have origin-country work experiences to a degree that deviates from the countrylevel averages. To test to what extent this impacts post-migration employment outcomes, we fitted a logistic regression model with a three-way interaction between gender of the respondent, gender-specific origin-country LFP, and individual origin-country employment experience. The results are presented below. Based on our four hypotheses, Figure 2 displays the AME of one percentage .
/fhumd. .  Table 1. H1 predicted that there would be a positive correlation between origin-country LFP and employment probability in Sweden for women without own origin-country work experience. Our result is that, on average, an increase of one percentage point in female origincountry LFP rate is associated with 0.30 percentage points higher probability of employment in Sweden. To illustrate, comparing two women without origin-country experience of employment and 22 and 60 percent female LFP rate-one standard deviation below and above the mean, respectively-the predicted difference in employment probability in Sweden is about 10% points. H1 is therefore confirmed.
H2 predicted that for women with own origin-country employment experience, the correlation between origin-country LFP rate and employment probability in Sweden should be negative. As seen in Figure 2, results indicate that there is a zero association between the origin-country female LFP rate and employment probability in Sweden for women with own experience. Since the AME is very small and nowhere near statistical significance at traditional levels, H2 is rejected.
For men, H3 predicted that there would be a negative association between the male origin-country LFP rate and the employment probability for men without origin-country experience, and H4 predicted a zero association for men with origin-country experience. Both hypotheses are confirmed. For men without origin-country employment, a difference in one percentage point higher origincountry LFP rate is associated with on average 0.48 percentage points lower probability of employment in Sweden, or thirteen percentage points lower probability of employment comparing two men coming from origins with 66 and 82 percent LFP rate, one standard deviation below and above the mean, respectively.

. . Sensitivity analysis
As discussed in the section on methods and data, there is a possibility to make other choices in designing the study. Specifically, to (A) use participation as the outcome variable; (B) include students and the long-term ill/early retired as neither participating nor having employment; (C) use the female origin-country LFP ratio between males and females rather than rate; (D) use predicted LFP rate; (E) only include immigrants who have been in prime working-age in both the origin country and Sweden. Figure 3 displays the results using all 32 combinations of design choices for the results concerning H1, i.e., the probability of employment in Sweden for immigrant women without origincountry work experience. The main result (reproduced on the far left in the figure) is replicated regardless of model specifications. We therefore conclude that there is a positive association between origincountry LFP and employment probability in Sweden for women without origin-country work experience, in accordance with H1 and regardless of design choices.
Results for H2-4 are available in the Supplementary material. In short, women with origin-country work experience had zero correlation between origin-country LFP and employment or participation probabilities in Sweden, regardless of specification, rejecting H2 as in the main results. For men without origin-country work experience, there was a negative correlation between LFP and employment but no correlation between LFP and participation, partially confirming H3. For men with origin-country work

. Discussion
Immigrant women typically face multiple disadvantages in destination-country labor markets, including a lower likelihood of being employed. Some of these disadvantages are shared with immigrant men, but gender and immigrant background intersect in multiple ways (Browne and Misra, 2003;Cho et al., 2013). Because of this, some factors affect immigrant men more, while others pertain particularly to immigrant women. One such factor is the gender norms surrounding the division of labor, that often prescribe for women to do more unpaid house labor while prescribing that men should be more active in the labor market. The Swedish welfare system assumes and facilitates for a dualearner household, and female LFP is consequently high. Many other countries-including many countries of origin for immigrantsinstead assumes male single-earner households. Migrants have therefore often been exposed to corresponding gender norms during their pre-migration socialization processes, which can continue to shape preferences and behaviors also after immigration. Over the last 20 years, research has found consistent associations between origin-country LFP and post-migration labor market outcomes for immigrant women. This suggest that the gender norms a person is exposed to during their socialization process has a lingering effect, even when the context-and surrounding societal norms-change.
However, as we know from the literature on migrant selection, immigrants can be expected to have some characteristics, experiences or resources that sets them apart from the sedentary population (Feliciano, 2020). These characteristics, experiences, or factors could in turn have consequences for post-migration outcomes, including for the relation between origin-country gender norms and postmigration labor market integration. Despite this, most of the previous research has been unable to account for this selection, and its consequences for labor market outcomes are therefore unknown.
In this study, we have contributed to the research field by explicitly focusing on the question of selection by modeling the probability of being employed in Sweden on origin-country LFP, individual work experience, and an interaction between the two as a measure of selection. We tested four hypotheses, one for each combination of gender and individual origin-country work experience. For women without origin-country work experience, we hypothesized that origin-country LFP should be positively associated with employment in Sweden (H1). For women with origin-country work experience, we instead hypothesized that origin-country LFP should have a negative association with employment probability in Sweden, since women with work experience from countries with low female LFP should be more positively selected (H2). For men, we hypothesized that origin-country LFP-here measuring economic structure rather than the universal norm that men should be active in the labor market-would have a negative association with employment in Sweden for men without origin-country work experience, since men with no work experience should be more negatively selected the more common labor force participation was in their country of origin (H3). Finally, we hypothesized that there should be no correlation between gender-specific origin-country LPF and employment probability in Sweden for men with origincountry work experience, since the labor market structure of the origin country should have no independent effect on post-migration employment probabilities (H4). H1, H3, and H4 were confirmed by our analysis, but H2 was rejected.
Confined to a single destination country, the study is limited in what it can say about the combination of origin and destination country gender norms and their effect on different categories of immigrants. As mentioned in the section on the Swedish context, we expected that the association between origin-country LFP and employment in Sweden would be suppressed by the Swedish context because of a generally high LFP for both immigrant women and men. This may have been the reason for why H2 could not be confirmed. There is not much room for variation in employment probabilities for women who both worked in the origin country and arrived in a country where labor market participation is the norm for women, which might have created a ceiling effect. If so, there might still be an effect of positive selection on other labor market outcomes, such as job quality or income. Future studies could therefore extend the scope of inquiry by including such outcomes in the analysis.
The sensitivity analysis showed that the results were stable across different specifications and design choices. Still, our study has limitations. Migrant-LNU had a response rate of 49.9 percent. While all results are weighted to account for uneven response rate across sample strata and key demographic variables, it is possible that there is selection on unobservable characteristics. Migrant-LNU only include immigrants who first immigrated to Sweden at least 5 years before the start of data collection. This limits the population that our results can be generalized to. As with all surveys of immigrant populations, there is also a risk that the results are biased by selective outmigration if individuals who were unsuccessful at finding employment move out of the country at a higher rate than individuals who were successful. Another limitation is that origin-country LFP data might not always be reliable, and that inconsistent definitions between countries or across time can impact data quality.
Despite these limitations, our results indicate that there are gender-specific selection patterns among immigrants to Sweden regarding pre-migration work experience, and that this migrant selection impacts the relation between origin-country gender norms and post-migration labor market outcomes. From a policy perspective, our study shows that women coming from low female LFP countries are positively selected on work experience (Figure 1), and that the negative association between female LFP and employment probability does not exist for these positively selected women (Figure 2). While Swedish integration policy is focused on labor market entry-for both immigrant women and men-it has been criticized for a gap between intent and outcomes (Wiesbrock, 2011), and for reproducing ethnic differences through policy design (Kamali, 2006). Our results indicate that policy interventions targeting immigrant women from countries with low female LFP should not assume that all these women arrive socialized with gender-norms that hinder labor market activity, and instead tailor policy and interventions to individual experiences.
In this study, we have highlighted the importance of both factoring in and contextualizing pre-migration experiences when .
/fhumd. . seeking to understand post-migration outcomes. We know from previous research that not only experience, but complex webs of individual characteristics and social positions affect immigrant women's labor market behavior and opportunities (Webster and Haandrikman, 2022). A possible venue for future research is therefore to explore to what extent education or class-based selection also influences the effect of origin-country gender norms on post-migration labor market outcomes, to better understand the mechanisms of these selection patterns. Future research should also combine pre-migration variables with a multi-destination design, to better understand how the continued socialization and change in opportunity structure in different countries affect labor market results.

Data availability statement
The data analyzed in this study is subject to the following licenses/restrictions: The Level of Living Surveys (Levnadsnivåundersökningarna, LNU) include sensitive information on individuals, as defined by the Swedish Personal Data Act (Personuppgiftslagen, PUL, SFS 1998:204;§ 13). This kind of information may be used in research if the project is compatible with the Act Concerning Ethical Review of Research Involving Humans (Etikprövningslagen, EPL, SFS 2003:460). Such compatibility is examined by an Ethical Review Board (Etikprövningsnämnden, EPN, www.epn.se/en). Therefore, an approval by EPN for the project in which LNU data is to be used is necessary for access to LNU data. Requests to access these datasets should be directed to https://www. sofi.su.se/english/2.17851/research/three-research-units/lnu-levelof-living/apply-for-lnu-data.