Human capital, gender, and labor force incorporation: The case of immigrants from the Former Soviet Union

Abstract

Women immigrating to the United States from the Former Soviet Union (FSU) were expected to incorporate seamlessly into the US labor force because of their strong educational and professional backgrounds. Using 2000 Census data, we find that FSU women were less successful than both FSU men and other non-Hispanic white female immigrants. After controlling for other factors, FSU women were more likely to rely on public assistance and less likely to be employed. If employed, they worked in less prestigious occupations and earned much less. These findings draw attention to the particular difficulties of incorporation of this wave of relatively advantaged immigrants.

Introduction

In the 1990s the Former Soviet Union (FSU) became the largest single source of non-Hispanic white immigrants to the United States. While immigration from Western Europe and Canada was declining, the annual number of FSU immigrants soared from under 3000 in 1988 to nearly 60,000 in 1991, constituting more than a third of all immigrants from Europe and Canada by that time (Figure 1). This study evaluates the incorporation of this new immigrant stream into US society, with a particular focus on labor market outcomes for women.

Figure 1.

Number of immigrants to the US from selected regions of origin, 1971–1999^a

Source: INS (1980, 1983, 1994, and 1999 Statistical Yearbooks).

^aFormer Eastern Bloc numbers exclude immigrants from Eastern Germany. Immigrants entering between July and September 1976 were excluded from the figure.

Especially compared to the equally fast growing numbers of Latin American and Asian immigrants at this time, expectations were high for the quick and successful incorporation of these FSU immigrants. They ‘were viewed as high-skilled workers who would easily adjust economically and who, in any case, would be taken care of by an involved Jewish community’ (Chiswick, 1993: 266). The Soviet Union was known for very high rates of labor force participation for women and opportunities for women in occupations that are predominantly male in the US. That is not to say that the Soviet labor market was consistently welcoming to women. Soviet women were expected to work even while they held primary responsibility for domestic labor and care giving in the home. And unlike the US, where women are concentrated in relatively few occupations (Padavic and Reskin, 2002), Soviet women were represented in a wide variety of occupations, but routinely paid less and promoted more slowly than their male counterparts (Khotkina, 1994). Nevertheless, women from the FSU are likely to have arrived in the US with exceptional professional skills and experience, well placed for employment in their new home.

On the other hand, women’s choices were also influenced by household-level strategies (Baker and Benjamin, 1997; Duleep and Sanders, 1993). We show below that over 75 percent of FSU women in our study were married at the time of the 2000 census, and only 4 percent were never married. Hence many decisions by women were likely to be linked to the situation of a spouse or former spouse. One recent study suggests that FSU families often rely on women to find low-skilled jobs soon after emigration in order to financially support the household while men take longer to search for better jobs and pursue additional education in the US (Gold, 2003). When husbands reenter the labor force, their wives typically do not retrain for a better position and they may drop out of the labor force altogether. Based on in-depth interviews with immigrants living in Boston, Remennick (2007) found that college-educated Soviet women who previously worked in highly skilled and technical occupations frequently chose to work in lower-skilled jobs after moving to the US. For example, an engineer might decide to become a youth counselor in order to find a position more quickly, or she might opt out of the labor force altogether. For some younger women, she argues, immigration created an opportunity to leave the labor force.

In addition, as discussed below, a large share of these men and women had refugee status that entitled them, among other benefits, to public assistance programs. The option of welfare might delay their search for paid employment beyond what is typical of other immigrants.

To address these differing expectations this study is based on two main comparisons. First, women are compared with men among FSU immigrants, allowing us to draw conclusions about gender differences in labor force incorporation. Second, FSU women are compared with non-Hispanic white women immigrants from other countries to test whether there are specific effects of FSU origin. These comparisons require that we control properly for other predictors. One reason is simply to factor out the effects of compositional differences between men and women or between women of different origins, such as differences in education level or year of arrival. Another reason is to assess differences in the strength of these predictors by gender and national origin. We therefore begin by reviewing predictors of labor force outcomes for all immigrants, then turn specifically to how gender and FSU background may make a difference.

Human capital and immigrant integration

There are two main categories of explanations for variation in immigrants’ labor market outcomes, human capital and immigrant integration. Human capital, encompassing individual characteristics such as educational attainment, work experience, and firm-specific training, is naturally an important predictor of employment outcomes. Read and Cohen (2007) find that human capital characteristics – and educational attainment in particular – are the strongest predictors of women’s entry into paid employment. Immigrant integration is a broad term describing the opportunities that immigrants have to acquire knowledge, credentials and skills specific to the host country that are associated with greater labor force participation and higher wages (Chiswick, 1978). Logan et al. (2003; see also Zhou and Logan, 1989) include generational differences, time in the US, citizenship, and English-language ability in their models of labor force participation and earnings for both men and women.

Immigrants living in the host country for longer periods of time, for example, are generally assumed to have better facility with English, more firm- or occupation-specific training, educational credentials confirmed or acquired in the US, and better information about job opportunities and the norms of the US job market. Using the 1970 US Census, Chiswick (1978) found that after 10 to 15 years in the US, the earnings of foreign-born men equaled or exceeded those of native-born men. Using a pooled sample of 1970, 1980 and 1990 US Census data, Schoeni (1998a) found a similar effect of duration of residence in the US for female immigrant labor force participation, finding that over time immigrant women’s labor force participation came to mirror that of native-born women. Like Chiswick, he found that the largest increase in immigrant women’s labor force participation occurred in the first 10 years after arrival. Attaining citizenship can be seen as another indicator of immigrant integration, although one reason that citizenship might matter is that employers may prefer to hire citizens over other workers, holding other characteristics equal. However, in his study, Chiswick (1978) found that citizenship explained none of the difference between the earnings of foreign-born and native-born men. Surprisingly studies of women have rarely asked whether citizenship itself contributes to the differences between immigrant and native-born women.

Gender differences in immigrant incorporation

Writing in the early 1990s, Pedraza (1991) observed that, although a majority of immigrants to the US over the last half of the 20th century were women, most studies of immigrant incorporation still focused exclusively on men or otherwise failed to address how gender affects immigrant outcomes. To varying extents, subsequent scholarship has expanded to include gender perspectives. Especially relevant in this study are the effects of household constraints and labor market segmentation, which are likely to have greater impacts on women than on men.

A large literature on households, work and gender finds that children, especially preschool-aged children, have a large impact on women’s labor market entry. Having more children poses additional challenges to the number of hours mothers can work and the types of jobs they can accept or maintain (Cohen and Bianchi, 1999). These same constraints, however, do not generally apply to men (Kaufman and Uhlenberg, 2000). The presence of grandparents is another factor that is more likely to affect women than men. It can facilitate women’s labor market participation if grandparents are willing and able to provide childcare. On the other hand, ailing grandparents could pose an additional caregiving burden on women, thus dampening their labor force participation (Remennick, 1999). The total level of earned income from other household members may also affect whether women participate in the paid labor force, although the relationship between other earned income and women’s propensity to work may differ across immigrant groups. Greenlees and Saenz (1999) find a negative relationship between the level of husband’s earned income and the probability that their wives will work, whereas Read and Cohen’s (2007) study of multiple immigrant groups paints a somewhat mixed picture. They find no relationship between earned income from other household members and women’s likelihood of employment for Arab, Cuban and Chinese immigrant women; a negative relationship for Filipina, Asian Indian, Iranian, White, Mexican and Japanese women; and a positive relationship for Korean, Vietnamese and Puerto Rican women.

A number of studies also include measures of ethnic group concentration in the local labor market as one important predictor of labor force outcomes for immigrant workers (Logan et al., 2003). High concentrations of co-ethnics suggest the presence of ethnic networks that can be tapped for information about job opportunities and advice on the job seeking process, as well as the potential to be employed by co-ethnic small business owners. However, high concentrations of co-ethnics can also inhibit the acquisition of English and other skills valuable to employers outside of ethnic niches, effectively trapping workers within enclave employment for the long term.

Pedraza (1991) argues that studies of female immigrants must also examine the occupational concentration of women as a distinct feature of the labor market. Compared to men, she suggests, immigrant women are occupationally concentrated ‘along a much smaller spectrum of choices … most of them cluster in just a few occupations. They become domestic servants, work for the garment industry, donate their labor to family enterprises, or most recently, work in highly skilled service occupations, such as nursing’ (Pedraza, 1991: 314). As is also true for US-born women, concentration into a limited range of occupations contributes to the persistent wage gap between men and women workers. Female-dominated occupations tend to pay less and be less prestigious than other occupations (Padavic and Reskin, 2002; Petersen and Morgan, 1995). In addition, men who work in female-dominated occupations tend to be promoted faster and occupy more prestigious positions than their female coworkers (Wilson, 1995) whereas women working in male-dominated occupations are less likely to access career advancement opportunities than their male counterparts (Maume, 1999).

Effects of FSU origin

We turn finally to the possible differences between immigrants from the Former Soviet Union and other non-Hispanic white immigrants (largely from Western Europe and Canada). In this comparison, those with FSU origins may actually be disadvantaged. Chiswick (1993) found that male FSU immigrants experience as much as 33 percent lower earnings than other European immigrants to the US, controlling for other predictors.

A common assumption in the literature is that immigrants are likely to have been underemployed or employed as low-skilled workers in their home countries. As already noted, these background assumptions do not hold in the case of FSU immigrants. We will compare the education levels for both men and women from FSU and non-FSU origins, and evaluate their impact on labor force outcomes, expecting similar results.

FSU immigrants face challenges similar to other immigrants in applying foreign educational credentials in the US labor market. In particular it is believed that they have acute problems of language proficiency. The best evidence on this issue is from a study of FSU men using the 1980 census by Chiswick (1993). Chiswick found that, net of other factors, FSU men – especially the more recent arrivals – were much less likely to be fluent in English than other European immigrants. Furthermore, although more years of education improved English language ability for all immigrants, it had a smaller impact on FSU immigrants.

Another distinctive factor is refugee status, a characteristic that cannot be measured directly with our data but which likely affects outcomes for FSU immigrants. A substantial share of the immigrant stream from the former Soviet Union to the US has been made up of religious minorities, Jewish and evangelical Christians in particular. Armenians are also a substantial share of FSU immigrants to the US (Heitman, 1991). Pedraza (1991) suggests that refugee status is an advantage, implicitly referring to the well known distinction between Jewish immigrants and Italian ‘birds of passage’ a hundred years ago. She points out that refugees are likely to arrive as permanent migrants, forced to ‘make their future in the new land’ and bringing ‘families who intend to remake their lives and homes’ (Pedraza, 1991: 312).

The anticipation of permanent settlement may be an advantage, but research on other refugee streams suggests that refugees are disadvantaged. They are less likely to enter the US with skills specific to the US labor market, and are more likely than other immigrants to have been unemployed prior to entering the US. Refugees are also less likely to speak English or to make a fast transition to a new labor market (Fix et al., 2003). However, they are more motivated and more able to become citizens and to naturalize within a relatively short time after arrival (Passel and Clark, 1998). Schoeni (1998b) examines a variety of immigrant and native-born women’s labor market outcomes including the largely refugee immigrant women from Indochina and Vietnam. In 1990, he finds that these women work the largest number of hours per week and weeks per year, though their weekly earnings remain among the lowest among immigrant women (even after controlling for poor English-language proficiency).

In short there are reasons to expect refugee status to have both positive and negative effects, some of which may disappear after introduction of controls for family composition and human capital. There is another dimension to this issue that specifically affects FSU immigrants, and this is their distinctive juridical status. Soviet immigrants were granted presumptive refugee status by the US government through much of the 1980s, making it easier to enter the country, to qualify for public assistance, and to sponsor members of their immediate family, as well as shortening the waiting period to apply for citizenship. In 1990 nearly all FSU immigrants (92.3%) were given refugee status (US INS, 1991–2002). As the numbers of FSU immigrants grew during the 1990s, this share dipped to 80 percent in 1994 and fell to 43.9 percent by 1999. But by then about a quarter were admitted as family members of US citizens, which partly reflected the refugee status of earlier arrivals.

Formal refugee status may make it easier for persons with lower skills (including language skills) to enter the country and provide an economic support that might delay entry into the labor market. Therefore we will look directly at public assistance use as both a dependent variable, expecting this to be a more important factor for FSU immigrants. We note, however, that the circumstances faced by the most recent wave of FSU immigrants in the US may differ from those of previous arrivals. In the early 1980s, the coverage period of resettlement assistance for refugees was trimmed from three years to eight months, and a sharp decline in public benefit receipt among immigrants was observed after welfare reform in the mid-1990s (Passel and Clark, 1998).

Hypotheses

To summarize the current state of theory on these effects, this section lists the specific hypotheses that will be tested through models of receipt of public assistance and three measures of labor market incorporation (employment, occupational prestige and weekly earnings). For convenience, we will refer here simply to ‘better or worse’ labor force outcomes, where being employed, having higher occupational prestige and earnings, and not receiving public assistance are treated as ‘better’.

We expect that the probability of employment among recently arrived women is comparable to or higher than that of FSU men, but that the improvement in men’s labor force incorporation (especially as measured by occupational prestige and earnings) is much steeper than the improvement in women’s incorporation over time.

Presence of children in the household, particularly of preschool-aged children, should show a negative impact on women, but bear no relationship or a positive relationship (because of male breadwinner expectations) to men’s labor force outcomes. However, the presence of extended kin and specifically of grandparents in the household may facilitate women’s labor force participation, allowing women to enter the workforce and earn more by working longer hours, or hamper it by adding additional care giving obligations. No relationship should exist between grandparent co-residence and men’s labor force participation.

Another source of disadvantage for FSU women relative to their male co-ethnics is the gender structure of the US labor market. In particular, we expect that women working in more female-dominated occupations will earn less and have lower occupational prestige, but that the sex ratio of the occupation will not affect men in the same way. Men working in more female-dominated occupations might have greater opportunities for promotion, but their occupations may tend to be relatively devalued.

Past work on gender and the benefits of ethnic enclaves has found that the benefits realized by male immigrants of ethnic economies may not be shared by their female co-ethnics (Zhou and Logan, 1989). We will use residence in New York City as an indicator of the probability of access to an ethnic FSU labor market. Orleck (1999) estimates that approximately 50 percent of the Soviet émigrés who entered the US since the mid-1970s settled in New York City. In our sample 30 percent of all FSU immigrants arriving after 1989 still resided in the New York metro area in 2000. Another group-specific reason to expect a beneficial effect for FSU immigrants is the high concentration of Jewish resettlement agencies serving Soviet Jewish immigrants. The two largest Jewish resettlement agencies, the Hebrew Immigrant Aid Society (HIAS) and the New York Association for New Americans (NYANA), are headquartered in New York City, though they later contracted with smaller agencies located throughout the US in an effort to disperse FSU immigrants (Howe, 1989). We predict that New York metro location will have a stronger positive relationship with labor force outcomes for FSU men than for FSU women, and no effect or a smaller effect on non-FSU women. But access to these agencies is likely to increase the probability of receiving public assistance for both men and women in New York.

We also expect that FSU women will differ from non-FSU women in their labor force outcomes. If Remmenick’s (2007) observation of an ideological shift towards traditional gender roles is correct, we may observe that FSU women are more likely to obtain positions in relatively less-skilled and lower-paid jobs, to opt out of the labor force entirely, and to rely on public assistance.

Data and methods

We will estimate comparable models of public assistance receipt and labor force incorporation for FSU men, FSU women, and non-FSU women. Multivariate analyses reported here include logistic regression models predicting the log odds of receiving public assistance benefits in 1999 or being employed at any time in 1999 and OLS regression models predicting logged weekly earnings and occupational prestige. They are based on the weighted 5 percent Public Use Microdata Sample (PUMS) of the 2000 US Census provided by the University of Minnesota Population Center (Ruggles et al., 2004). The PUMS is a stratified systematic sample of all occupied, vacant and group quarters housing units.

We limit the analysis to people of working age (under 65) who immigrated as adults. In order to be sure of meeting the age criterion for education to be reported (25 or above) this means the sample is limited to non-Hispanic men and women who were 25 or older when they entered the US, were between 25 and 64 at the time of the census, and entered the US before 2000. These limitations yield a sample size of 6454 FSU men, 7607 FSU women, and 39,418 non-FSU women for analysis of public assistance receipt and labor force participation.

Models predicting occupational prestige are limited to individuals who had worked at least one week during 1999, and models predicting logged weekly earnings are limited to individuals who had earned at least $1 from an employer, self-employment, or farm income during 1999. The sample size is 5341 FSU men, 4934 FSU women, and 24,936 non-FSU women. All models are population weighted and standard errors are adjusted to account for clustering within households.

Dependent variables

Persons who worked for at least one week during 1999 are coded as employed, including unpaid family workers and regardless of earnings. Occupational prestige was measured using the Duncan Socioeconomic Index (SEI), an occupational score based on income levels and educational attainment associated with each occupation in 1950. Although all respondents employed at some point during the five years previous to the census have an SEI score, as mentioned above, we limit the analysis only to those respondents employed for at least one week in 1999. Because there was a great deal of variability in how much of the year individuals worked, we divided total annual earnings from wages, self-employment, and farm income by the number of weeks worked to generate a measure of weekly personal earnings, and logged this figure to approximate a normal distribution. We excluded persons with negative or zero earnings. This figure does not include income from other sources, such as investments or income from public assistance. The models for earnings include an additional control variable for ‘usual hours worked’. In effect, therefore, the dependent variable becomes hourly pay. Household public assistance is a dichotomy constructed by searching records of all household members for the receipt of person-level welfare benefits (SSI, AFDC, and general assistance income) and Social Security income (pre-tax income from Social Security pensions, survivors benefits, permanent disability insurance, and US government Railroad Retirement insurance payments) received in 1999.

Human capital

Human capital is generally measured by some combination of educational attainment, professional credentials, and work experience. Here, we measure human capital using educational attainment (less than a high school diploma, high school graduate, some college, bachelor’s only, and more than a bachelor’s degree). English-language ability is also an important human capital variable because a good command of English is necessary to translate foreign educational credentials to the US labor market. We combine the categories of ‘speaks only English’ and ‘speaks English well or very well’, and compare these to a reference category of speaking not well or not at all.

Immigrant integration

Immigrant integration is often operationalized as length of time in the US, where immigrants living in the US for longer periods of time gain more local knowledge and networks, and become acculturated to US norms and values. We use dummy variables to represent critical periods of time observed by Chiswick (1978): less than five years, five to ten, and 11 or more years. FSU immigrants are not expected to experience a uniform progress on all dimensions, because as a refugee group, they typically obtain citizenship much sooner than other immigrants, but are significantly less likely to possess strong English-language skills or US-specific human capital. In addition to English-language proficiency as mentioned above, we also consider whether individuals were citizens at the time of the census (a dummy variable).

Household resources and constraints

Several family and household-level measures capture the impact of potential household resources and constraints on the labor force participation of immigrant men and women. We consider whether immigrants have a child under five in the home (a dummy variable), as well as the total number of children aged five and older residing in the home. We also consider whether immigrants might benefit from the co-residence of a parent of either spouse as a potential caregiver for their children, or whether immigrant labor force participation (especially women’s) might be constrained by the presence of an elderly parent who may need care themselves. We control for household income from sources besides the individual’s earned income, a figure we calculate by subtracting the individual’s income earned from wages, self-employment, and farm income from the total household income from all sources except for public assistance.

Labor market characteristics

As discussed above, the US labor market is characterized by a system of horizontal occupational sex segregation, whereby women are concentrated in only a handful of occupations. These occupations tend to be lower-paying and lower-skilled than occupations with a better representation of men. To measure the degree to which women are employed in sex segregated occupations, we calculate an occupational sex ratio, which is the ratio of all men to women in the 5 percent IPUMS sample working in a given occupation. Thus, in a gender balanced occupation, the sex ratio is 1. Values less than 1 represent occupations which are more heavily female, and values greater than 1 represent occupations which are more heavily male.

Pedraza (1991), Read and Cohen (2007), and others have emphasized the importance of ethnic enclaves to understanding the labor market incorporation of immigrant workers. As discussed above we use a New York metropolitan area dummy variable to represent the possible effect of the enclave labor market (and co-ethnic support organizations) on FSU immigrants.

In each of the models, we also control for age at the time of the census.

Findings

Sample characteristics for each of the three groups compared here (FSU men, FSU women, and non-FSU women) are shown in Table 1. Relative to FSU men, FSU women were much less likely to be employed (64.8% vs 83.1%), and they had lower earnings but slightly higher occupational prestige if they were working. They were also somewhat more likely to receive public assistance (20.6% vs 17.5%). Non-FSU women were slightly less likely than FSU women to be employed, but they had considerably higher earnings and also higher occupational status if they did work. They were also much less likely to receive public assistance. The pattern, then, is one of male advantage in labor force outcomes, and non-FSU women doing better than FSU women. What could account for these differences?

Table 1.

Sample characteristics^a

	FSU men %	FSU women %	Other non-Hispanic white women immigrants %
Age	45.7 years	45.8 years	45.9 years
Human capital
Education
Less than high school	9.3%	9.0%	16.7%
High school graduate	17.3	16.5	23.0
Some college	17.4	20.9	25.3
College graduate	24.9	27.8	19.2
Advanced degree	31.1	25.7	15.8
Speaks English well, very well or only	68.0%	66.1%	84.5%
Household resources and constraints
Marital status
Never married	5.9%	4.1%	8.3%
Divorced, widowed or separated	10.5	19.0	16.4
Married	83.7	76.9	75.4
Has a child under age 5	13.1%	10.3%	12.8%
Total number of own children in the household	1.0 children	1.0 children	0.8 children
Presence of grandparent	6.6%	7.0%	2.9%
Total household income from other sources	$28,097	$41,093	$60,693
Labor market characteristics
NY metro residence	29.0%	29.3%	8.8%
Occupational sex ratio	8.4	1.2	1.2
Immigrant integration
Length of time in US	8.2 years	8.0 years	12.6 years
Decade of immigration
1990 to 1999	76.4%	78.1%	52.1%
1980 to 1989	16.3	14.8	25.4
1979 or earlier	7.2	7.1	22.5
Citizenship	43.5%	43.0%	36.1%
Labor market outcomes
Employed in 1999	83.1%	64.8%	63.3%
Weekly earnings in 1999	$966	$642	$708
Occupational prestige	45.8	46.5	48.0
Public assistance
Households receiving public assistance	17.5%	20.6%	15.8%
N	6,454	7,608	39,420

^aFigures are population weighted.

There are mostly small differences between FSU men and women. Men are more likely to have an advanced degree and to be married. A bigger difference is that men are likely to have much less household income from sources other than their own earnings and to work in more male-dominated occupations. There are larger differences between FSU and non-FSU women. Some of these should be advantageous to FSU women: their higher education level, presence of a grandparent, US citizenship, and also New York City residence. Others are possible liabilities, especially their lower likelihood of speaking English well or better (66.1% vs 84.5%), shorter time in the US (8.0 years vs 12.6 years), and smaller amounts of household income from other sources.

Multivariate models for each subsample provide estimates of the effects of the independent variables on labor force outcomes. Results for FSU men are shown in Appendix Table 1, for FSU women in Appendix Table 2, and for non-FSU women in Appendix Table 3. These tables display the model coefficients and significance levels. In the following text we translate these coefficients into changes in predicted probabilities and values associated with variables of interest for each of the three groups. This translation offers a clearer comparison of the net values of the dependent variables for different subsamples and indicates the magnitude of effects of the predictors. For such calculations it is necessary to assume some baseline set of values on each of the predictors. Our baseline refers to people who are younger than 35, married, college-educated, with high English-language proficiency, no children and no grandparent in the household, who reside outside of New York, are recently arrived (in the US for less than five years), and not a citizen. We set the occupational sex ratio to the median value for FSU women (0.5). Because this variable is very skewed upwards, especially for men, the median is more appropriate than the mean. We set the log of other household income to zero. In the earnings model the baseline profile is for persons who work 40 hours per week. Readers who would like to make different comparisons can calculate these from the information in the Appendix tables. The predicted probabilities are organized into a separate table for each outcome variable, Tables 2–5, facilitating comparisons across subsamples.

Table 2.

Predicted probability of household receiving public assistance

	Predicted probability
	FSU men	FSU women	Non-FSU women
Baseline profilea	0.13	0.21	0.06
Human capital
Less than HS	0.19	0.35	0.13
HS graduate	0.17	0.37	0.11
Some college	0.15	0.32	0.09
More than college degree	0.09	0.21	0.06
Does not speak English well or at all	0.16	0.30	0.09
Household resources and constraints
Has a child under 5	0.15	0.27	0.07
Has one child 5 or older	0.13	0.25	0.06
Grandparent co-residence	0.39	0.55	0.36
Logged other HH income set at FSU female avg.	0.06	0.07	0.03
Labor market characteristics
Lives in NYC	0.10	0.20	0.06
Immigrant integration
Stay length is 5–10 years	0.07	0.14	0.06
Stay length is 11+ years	0.06	0.14	0.07
Citizen	0.10	0.22	0.09

^aBaseline profile is recently arrived (in the US for less than five years), younger than 35, college-educated, married, has strong English-language ability, no children, not a citizen, and resides outside of NYC.

Table 5.

Predicted weekly wages

	Predicted value
	FSU men	FSU women	Non-FSU women
Baseline profilea	$609	$389	$492
Human capital
Less than HS	$426	$265	$311
HS graduate	$479	$279	$332
Some college	$444	$310	$391
More than college degree	$776	$439	$554
Does not speak English well or at all	$452	$324	$410
Household resources and constraints
Has a child under 5	$601	$379	$513
Has one child 5 or older	$601	$374	$474
Grandparent co-residence	$567	$388	$441
Logged other HH income set at FSU female mean	$616	$421	$515
Labor market characteristics
Lives in NYC	$558	$418	$584
Occupational sex ratio = FSU male median	$610	$412	$503
Immigrant integration
Stay length is 5–10 years	$653	$438	$495
Stay length is 11+ years	$766	$503	$544
Citizen	$685	$458	$520

^aBaseline profile is recently arrived (in the US for less than five years), younger than 35, college-educated, married, has strong English-language ability, no children, not a citizen, resides outside of NYC, occupational sex ratio at the FSU female mean, and employed full-time (works 40 hours per week).

Receipt of household public assistance

We begin with public assistance, which for some people is an alternative to labor force participation and where Table 1 showed considerably greater likelihood of receipt for FSU women than for non-FSU women. After controlling for other personal, household or labor market characteristics, and assuming our baseline profile, Table 2 shows that the probability that FSU women live in a household receiving public assistance (21%) is much higher than it is for FSU men (13%) and nearly four times higher than for non-FSU women (6%).

Now let us consider the effects of varying the values of predictors, one at a time, and compare the predicted probabilities to the baseline level. Many of these predictors are statistically significant, and some effects are quite large. For persons whose other household income is set to the average for FSU women (about $41,000) public assistance receipt is much reduced for all sub-samples, and there remains little difference among them. Education has a substantial effect for all groups and especially for FSU women. For persons with a high school education or less, the share of FSU women receiving public assistance is 35 percent or more, compared to 17–19 percent for FSU men and 11–13 percent for non-FSU women. English-language ability similarly has especially strong effects for FSU women. For those who speak English not well or not at all, 30 percent of FSU women receive public assistance, compared to 16 percent of FSU men and 9 percent of non-FSU women.

We expect family composition to matter for public assistance because children or elderly persons can influence both need and eligibility. The presence of a child under five or five and older makes a small difference for FSU men and non-FSU women, and a moderate difference for FSU women. Presence of grandparents comes into play more strongly: the share receiving public assistance rises to 36 percent for non-FSU women with a grandparent in the home and as high as 55 percent for FSU women. Here is where the regulations favoring FSU immigrants as refugees may have their most important impact, as it is likely that this result is due to grandparents’ eligibility for assistance.

Of other predictors shown in Table 2, living in New York City has little effect, contrary to the expectation for FSU immigrants, nor does citizenship. Length of stay does not affect non-FSU women, but FSU men and women who have lived longer in the US are much less likely than new arrivals to receive public assistance.

Employment

Table 3 provides predicted probabilities for employment. The baseline profile shows a large gender gap, similar to that found in the raw frequencies in Table 1: FSU men with this profile are much more likely to be working (93%) than FSU women (77%) or non-FSU women (80%). Women’s working is greatly influenced by having other household income sources (falling by 5 to 10 percentage points), but men’s is not (actually rising slightly if other household income is at least $41,000). We expected family composition to affect women but not men. Indeed FSU men with children in the household are neither more nor less likely to work, and having a grandparent in the home slightly reduces men’s working. A child under five greatly reduces women’s work probability (to 60% for both FSU and non-FSU women), while a grandparent somewhat increases it, though the increase is not statistically significant.

Table 3.

Predicted employment probabilities

	Predicted probability
	FSU men	FSU women	Non-FSU women
Baseline profilea	0.93	0.77	0.80
Human capital
Less than HS	0.84	0.52	0.66
HS graduate	0.89	0.63	0.74
Some college	0.90	0.69	0.78
More than college degree	0.94	0.76	0.86
Does not speak English well or at all	0.84	0.56	0.75
Household resources and constraints
Has a child under 5	0.92	0.60	0.60
Has one child 5 or older	0.93	0.77	0.77
Grandparent co-residence	0.89	0.81	0.82
Logged other HH income set to FSU female avg.	0.95	0.72	0.70
Labor market characteristics
Lives in NYC	0.89	0.75	0.77
Immigrant integration
Stay length is 5–10 years	0.95	0.86	0.87
Stay length is 11+ years	0.96	0.89	0.89
Citizen	0.95	0.80	0.83

^aBaseline profile is recently arrived (in the US for less than five years), younger than 35, college-educated, married, has strong English-language ability, no children, not a citizen, and resides outside of NYC.

Human capital in terms of education and English-language skill has significant effects for all three groups. Having less than a high school education reduces probability of working to 52 percent for FSU women and 66 percent for non-FSU women, with a smaller impact on men. Not speaking English well or at all has its greatest impact on FSU women (reducing their probability of employment to 56%), with lesser but still strong effects on FSU men and non-FSU women.

There is again little effect of living in New York, contrary to expectations. Citizenship only slightly increases the probability of working for any subsample. For women, but hardly for men, length of stay in the US is associated with a growing likelihood of working, rising to 88–89 percent for women who have been in the country for 11 or more years. This is the opposite of expectations.

Occupational prestige

For those who do work, there are only small differences in occupational prestige among subsam-ples with the baseline profile, slightly favoring non-FSU women. Human capital measures have similar effects for all three groups. Occupational prestige is much lower for those with a high school education or less and higher for those with an advanced degree. There is about a 10- to 15-point disadvantage for those who do not speak English well or at all. Yet there are relatively small differences across subsamples in the predicted levels for any values of these variables. And although other predictors have statistically significant effects (shown in the Appendix tables), they typically raise or lower the predicted prestige level by only one or two points.

Earnings

In contrast, the weekly earnings levels vary greatly across these groups. As in the unadjusted averages, FSU men have the highest predicted earnings with the baseline profile ($609), about $120 higher than non-FSU women ($492) and more than $200 higher than FSU women ($389). Note that this figure assumes full-time work (40 hours per week). This ranking of groups is not affected by varying levels of most predictors. One exception is New York metropolitan area residence. Surprisingly FSU men living in New York earn considerably less than those living elsewhere (a difference of about $50), while FSU and non-FSU women in New York earn more (a $30 difference for FSU women, but nearly $90 for non-FSU women). As a result non-FSU women in New York, who otherwise have the baseline profile, actually earn more than FSU men.

Table 5 shows that human capital variables have large effects for all groups. The difference in predicted values between those with less than high school education and those with more than a college degree is approximately $350 for FSU men, about $240 for non-FSU women, and about $180 for FSU women. This means that the disadvantage of FSU women relative to the other two groups is greatest for those with the highest education level. Lesser English-language ability reduces FSU men’s predicted earnings by about $160, with smaller but still appreciable effects for FSU women (about $65) and non-FSU women (about $80).

As expected, having a child under five has little impact on men. Having a child under five has only small effects on women, with a slightly larger negative effect of children five and over. Living with a parent reduces men’s earnings and more substantially reduces earnings of non-FSU women.

The sex ratio of persons within their occupation does not affect men’s earnings and has only small effects for women (failing to confirm our hypothesis). Both integration variables have substantial effects. Compared to new arrivals, those in the country for 11+ years earn more. The difference is greatest for FSU men (over $150), smaller for FSU women (about $110), and modest for non-FSU women (about $50). Citizens earn more than non-citizens (about $70 more for FSU immigrants, $30 more for non-FSU women).

Discussion and conclusion

This study is intended to answer two major questions. The first, whether gender plays a role in incorporation of FSU immigrants, has been addressed by comparing men and women from the same range of national origins. The second, whether immigrants from the Former Soviet Union were incorporated as readily into the labor force as expected, has been addressed by comparing women from the FSU with other non-Hispanic white women immigrants. Very likely there would be different results if the comparison had been to women from Central America or the Dominican Republic, on the one hand, or women from India or the Philippines, on the other. But since race is so strong a factor in the US labor market, the more apt comparison is to other women who are likely to be perceived by the community as ‘white’.

Our results demonstrate the important impact of gender. Our comparisons among FSU immigrants show that women are much more likely to live in a household that receives public assistance and less likely to be working even if they have no children. If working, they have slightly lower prestige occupations, and their weekly earnings (partly reflecting their lower propensity to full-time work) are much lower. These findings mirror those of Lewin-Epstein et al. (2003), who reported that FSU women were less likely to be employed than FSU men in both Canada and Israel.

We do not find an overall advantage for FSU immigrants. Women from the Former Soviet Union, controlling for other factors, are much more likely to be relying on public assistance and modestly less likely to be working than other non-Hispanic white women immigrants. If working, their occupational prestige level is five points lower, and their earnings are about one quarter less. These comparisons mostly mirror those found in the unadjusted averages, and they add up to a net disadvantage for FSU women.

These conclusions take into account a wide range of other predictors on outcomes for these three groups. We find that human capital, as indicated by education and English-language ability, has all of the expected effects – lowering the probability of public assistance receipt, while raising the likelihood of working, the occupational prestige of the job, and average weekly earnings. Indicators of integration also show consistent results. Immigrants with longer time in the US are less likely to receive public assistance, more likely to be working, likely to have a higher status job, and have higher average earnings. Women who have become citizens are more likely to receive public assistance, but generally – regardless of gender – citizens are more likely to be working, to have a better job, and to earn more.

Among other predictors emphasized in this study, family circumstances have their greatest impacts on women, regardless of their national origin. In the case of public assistance these effects are likely associated with need or eligibility – having children or older adults in the household increases both need and eligibility, while considerable income from other household members reduces both. For employment, the effects are limited to women, and we find that higher other income reduces the likelihood of working (which may be a matter of need), having children under five reduces working (which is understandable in terms of demands on mothers’ time) and grandparents facilitate working (suggesting that they are helpers). Similarly, having young children reduces women’s earnings – probably by reducing their working hours – but surprisingly grandparents also do. Possibly on average grandparents require more assistance from women than they provide in return (as in the concept of the ‘sandwich generation’).

These results emphasize the distinctive aspects of men’s and women’s economic integration as immigrants. On the other hand, the distinctiveness of immigration from the Former Soviet Union – compared to other non-Hispanic white immigrants – seems mainly related to their legal status in the United States. As a specific refugee stream that benefited from early eligibility for public assistance and a fast track to citizenship, FSU immigrants have had some concrete advantages. However, their lower occupational prestige and lower earnings (even after controlling for their relatively recent arrival and poor English skills) are significant deficits, and there is no evidence that they received a special benefit from their heavy concentration in the New York region. The expectation of ‘easy incorporation’ for these newcomers seems not to have been realized.

In this respect our findings mirror those of studies of FSU immigrants elsewhere. In Israel, where these immigrants were members of the favored European-origin ethnic group and where they arrived with unusually high levels of education, they enjoyed lower occupational standing and lower earnings than all other Jewish sub-groups (Gorodzeisky and Semyonov, forthcoming), disadvantages that they did not make up over time (Cohen and Haberfeld, 2007). The specific conditions of their arrival in Israel (as part of a massive wave of new arrivals and with pressure from the state to find any form of employment quickly) help to explain the results in this case. But perhaps the opposite conditions were found in Germany and Canada, where immigration was in smaller numbers and more selective on educational background. In both these countries (Cohen and Kogan, 2007; Lewin-Epstein et al., 2003) FSU immigrants experienced substantial deficits compared with comparable natives: lower likelihood of employment, lower earnings, and lower rewards to educational achievement. Apparently no variant of the context of reception (high selectivity in Canada, generous welfare benefits in Germany, fast track to citizenship in the US, or Israel’s open arms policy) has resulted in quick success in the labor market for those who left the Former Soviet Union.

Table 4.

Predicted occupational prestige

	Predicted value
	FSU men	FSU women	Non-FSU women
Baseline profilea	55.0	53.4	57.9
Human capital
Less than HS	36.5	32.9	33.6
HS graduate	39.1	36.3	39.2
Some college	43.3	41.0	48.4
More than college degree	67.1	60.7	67.4
Does not speak English well or at all	46.0	45.0	43.8
Household resources and constraints
Has a child under 5	54.2	50.6	59.5
Has one child 5 or older	54.3	51.8	57.0
Grandparent co-residence	55.8	55.6	56.6
Logged other HH income set to FSU female mean	53.9	54.7	59.6
Labor market characteristics
Lives in NYC	53.3	49.5	57.2
Occupational sex ratio set to FSU male median	54.0	55.4	58.0
Immigrant integration
Stay length is 5–10 years	54.0	55.7	57.3
Stay length is 11+ years	57.9	58.3	60.0
Citizen	57.1	55.7	57.9

^aBaseline profile is recently arrived (in the US for less than 5 years), younger than 35, college-educated, married, has strong English-language ability, no children, not a citizen, resides outside of NYC, and occupational sex ratio at the FSU female mean.

Appendix

Appendix Table 1.

Results of FSU male models^a

Predictors	Employment b (SE)	Occ. prestige b (SE)	Log wkly wage b (SE)	Pub. assistance B (SE)
Intercept	1.63 (0.19)***	46.05 (1.48)***	5.35 (0.08)***	−1.66 (0.18)***
Age
25–34 years (ref.)	0	0	0	0
35–44 years	−0.04 (0.16)	0.36 (1.05)	0.06 (0.04)	0.09 (0.14)
45–54 years	−0.42 (0.17)*	−2.05 (1.15)	0.01 (0.05)	0.41 (0.15)**
55–64 years	−1.95 (0.17)***	−4.37 (1.41)	−0.08 (0.05)	1.31 (0.16)***
Marital status
Never married	−0.65 (0.17)***	−1.73 (1.41)	−0.07 (0.06)	0.13 (0.18)
Div./wid./sep.	−0.32 (0.13)*	−3.40 (1.19)**	−0.10 (0.05)	0.04 (0.15)
Married (ref.)	0	0	0	0
Human capital
Education
Less than HS	−0.90 (0.14)***	−18.50 (1.33)***	−0.36 (0.06)***	0.74 (0.14)***
HS graduate	−0.55 (0.13)***	−15.88 (1.03)***	−0.24 (0.04)***	0.60 (0.12)***
Some college	−0.39 (0.12)**	−11.68 (1.00)***	−0.32 (0.04)***	0.42 (0.12)***
College grad. (ref.)	0	0	0	0
More than college	0.22 (0.12)	12.13 (0.85)***	0.24 (0.03)***	−0.17 (0.12)
Speaks Eng. well/only	0.97 (0.10)***	9.09 (0.84)***	0.30 (0.04)***	−0.52 (0.09)***
Household resources and constraints
Has child under 5	−0.16 (0.16)	−0.84 (0.93)	−0.01 (0.04)	0.45 (0.13)***
# of children >=5	0.04 (0.04)	−0.76 (0.28)**	−0.01 (0.01)	0.31 (0.03)***
Presence of grandparent	−0.53 (0.15)***	0.72 (1.45)	−0.07 (0.06)	1.74 (0.14)***
Logged HH incomeb	0.02 (0.01)*	−0.10 (0.09)	0.001 (0.004)	−0.06 (0.01)***
Labor market characteristics
NYC	−0.48 (0.09)***	−1.78 (0.74)*	−0.09 (0.03)**	−0.07 (0.09)
Occ. sex ratio	–	−0.33 (0.02)***	0.001 (0.001)	–
Immigrant integration
Time in US
0–4 years (ref.)	0	0	0	0
5–10 years	0.39 (0.11)***	−0.96 (0.86)	0.07 (0.04)*	−0.47 (0.10)***
11 or more years	0.66 (0.15)***	3.00 (1.14)**	0.23 (0.05)***	−0.76 (0.15)***
Citizenship	0.25 (0.10)*	2.05 (0.79)**	0.12 (0.03)***	−0.05 (0.10)
Usual hours worked	–	–	0.02 (0.001)***	–
N	6,454	5,341	5,341	6,454
R2/Pseudo-R2	0.1987	0.3619	0.2115	0.1424

^aResults are population-weighted and adjust for clustering within households.

^bCoefficients multiplied by 10 for display purposes.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Appendix Table 2.

Results of FSU female models^a

Predictors	Employment b (SE)	Occ. prestige b (SE)	Log wkly wage b (SE)	Pub. assistance B (SE)
Intercept	0.23 (0.15)	44.59 (1.80)***	4.61 (0.08)***	−0.84 (0.16)***
Age
25–34 years (ref.)	0	0	0	0
35–44 years	0.28 (0.10)**	−1.12 (1.12)	0.07 (0.04)	0.07 (0.12)
45–54 years	0.23 (0.11)*	−5.55 (1.19)***	0.03 (0.04)	0.24 (0.12)
55–64 years	−1.19 (0.12)***	−9.77 (1.44)***	−0.09 (0.05)	1.20 (0.13)***
Human capital
Education
Less than HS	−1.13 (0.12)***	−20.43 (1.48)***	−0.38 (0.06)***	0.68 (0.13)***
HS graduate	−0.69 (0.09)***	−17.11 (1.05)***	−0.33 (0.04)***	0.79 (0.10)***
Some college	−0.41 (0.08)***	−12.39 (0.94)***	−0.23 (0.03)***	0.56 (0.10)***
College grad. (ref.)	0	0	0	0
More than college	−0.04 (0.08)	7.34 (0.81)***	0.12 (0.03)***	−0.03 (0.10)
Speaks English well/only	0.98 (0.07)***	8.34 (0.93)***	0.18 (0.03)***	−0.48 (0.08)***
Household resources and constraints
Marital status
Never married	0.25 (0.16)	−1.63 (1.74)	−0.06 (0.06)	−0.40 (0.18)*
Div./wid./sep.	0.14 (0.09)	−2.10 (0.93)*	0.07 (0.03)	−0.29 (0.10)**
Married (ref.)	0	0	0	0
Has child under 5	−0.82 (0.10)***	−2.77 (1.30)*	−0.03 (0.05)	0.35 (0.12)**
# of children >= 5	−0.01 (0.03)	−1.53 (0.37)***	−0.04 (0.01)***	0.24 (0.03)***
Presence of grandparent	0.24 (0.13)	2.18 (1.20)	−0.003 (0.05)	1.54 (0.12)***
Logged HH incomeb	−0.02 (0.10)	0.13 (0.11)	0.01 (0.004)*	−0.12 (0.01)***
Labor market characteristics
NYC	−0.13 (0.07)	−3.92 (0.71)***	0.07 (0.03)**	−0.04 (0.08)
Occ. sex ratio	–	0.88 (0.26)**	0.03 (0.01)***	−0.46 (0.09)***
Immigrant integration
Time in US
0–4 years (ref.)	0	0	0	0
5–10 years	0.57 (0.08)***	2.30 (0.93)*	0.12 (0.04)**	−0.46 (0.09)***
11 or more years	0.91 (0.11)***	4.92 (1.22)***	0.26 (0.05)***	−0.47 (0.13)***
Citizenship	0.17 (0.08)*	2.28 (0.08)**	0.16 (0.03)***	0.04 (0.09)
Usual hours worked	–	–	0.03 (0.001)***	–
N	7,607	4,934	4,934	7,607
R2/Pseudo-R2	0.1812	0.2892	0.2900	0.1324

^aResults are population-weighted and adjust for clustering within households.

^bCoefficients multiplied by 10 for display purposes.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Appendix Table 3.

Results of non-FSU female models^a

Predictors	Employment b (SE)	Occ. prestige b (SE)	Log wkly wage b (SE)	Pub. assistance B (SE)
Intercept	1.11 (0.07)***	43.79 (0.82)***	4.76 (0.04)***	−2.32 (0.10)***
Age
25–34 years (ref.)	0	0	0	0
35–44 years	−0.17 (0.04)***	−1.52 (0.46)**	0.07 (0.02)***	0.10 (0.19)
45–54 years	−0.37 (0.05)***	−2.74 (0.53)***	0.06 (0.02)*	0.40 (0.08)***
55–64 years	−1.29 (0.05)***	−3.45 (0.60)***	−0.0001 (0.03)	1.79 (0.08)***
Human capital
Education
Less than HS	−0.74 (0.04)***	−24.29 (0.54)***	−0.46 (0.02)***	0.89 (0.06)***
HS graduate	−0.34 (0.04)***	−18.69 (0.44)***	−0.39 (0.02)***	0.63 (0.06)***
Some college	−0.14 (0.04)***	−9.56 (0.40)***	−0.23 (0.02)***	0.41 (0.06)***
College grad. (ref.)	0	0	0	0
More than college	0.40 (0.04)***	9.51 (0.42)***	0.12 (0.02)***	0.02 (0.07)
Speaks Eng. well/only	0.27 (0.04)***	14.11 (0.51)	0.18 (0.02)***	−0.43 (0.05)***
Household resources and constraints
Marital status
Never married	0.67 (0.06)***	1.07 (0.53)*	0.11 (0.02)***	−0.25 (0.08)**
Div./wid./sep.	0.58 (0.04)***	−1.65 (0.41)***	0.07 (0.02)***	−0.34 (0.05)***
Married (ref.)	0	0	0	0
Has child under 5	−0.98 (0.04)***	1.55 (0.50)**	0.04 (0.02)	0.11 (0.07)
# of children >=5	−0.14 (0.01)***	−0.88 (0.16)***	−0.04 (0.01)***	0.02 (0.02)
Presence of grandparent	0.16 (0.08)	−1.33 (0.77)	−0.11 (0.04)**	2.18 (0.08)***
Logged HH incomeb	−0.05 (0.005)***	0.15 (0.04)**	0.004 (0.002)*	−0.06 (0.01)***
Labor market characteristics
NYC	−0.19 (0.04)***	−0.72 (0.52)	0.17 (0.02)***	−0.02 (0.06)
Occ. sex ratio	–	0.02 (0.06)	0.01 (0.003)***	–
Immigrant integration
Time in US
0–4 years (ref.)	0	0	0	0
5–10 years	0.55 (0.04)***	−0.67 (0.41)	0.01 (0.02)	−0.05 (0.06)
11 or more years	0.66 (0.04)***	2.05 (0.46)***	0.10 (0.02)***	0.13 (0.06)*
Citizenship	0.22 (0.03)***	−0.01 (0.31)	0.06 (0.01)***	0.40 (0.04)***
Usual hours worked	–	–	0.03 (0.001)***	–
N	39,418	24,936	24,936	39,418
R2/Pseudo-R2	0.0973	0.3315	0.2427	0.1650

^aResults are population-weighted and adjust for clustering within households.

^bCoefficients multiplied by 10 for display purposes

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.