Corresponding author: Elena Vakulenko ( evakulenko@hse.ru ) © 2017 Non-profit partnership “Voprosy Ekonomiki”.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits to copy and distribute the article for non-commercial purposes, provided that the article is not altered or modified and the original author and source are credited.
Citation:
Vakulenko E, Leukhin R (2017) Wage discrimination against foreign workers in Russia. Russian Journal of Economics 3(1): 83-100. https://doi.org/10.1016/j.ruje.2017.02.006
|
We try to determine with the help of the Oaxaca–Blinder decomposition technique whether foreign workers are discriminated against in Russia. We use the Russian Ministry of Labor (Rostrud) data on migrants’ applications and the Russian Longitudinal Monitoring Survey (RLMS, provided by the Higher School of Economics) for the period 2009–2013. We show that there is significant discrimination against foreign workers. The average salary of Russian workers with the same level of productivity as migrants exceeds migrants’ average salary by 40%. The industries in which the workers are employed have made most substantial contribution to the discrimination gap. Moreover, there is evidence that the lower salaries of foreign workers do not reduce the salaries of Russians employed in similar positions.
discrimination, Oaxaca–Blinder decomposition, international labor migration
A comparison of labor wages between foreign migrants and local workers satisfies scholarly interest and creates a basis for immigration policy recommendations. This empirical study attempts to compare the wages of temporary foreign migrants and Russian workers. Is there discrimination against foreign workers in the Russian labor market, and if so, how does it manifest itself in quantitative terms? “Discrimination” refers to a situation in which people with the same level of productivity are offered different wages.
Foreign migrants constitute a significant segment of Russia's labor market. In 2013, for example, documents were issued authorizing labor activity in the Russian Federation for almost three million foreign migrants, according to the Federal Migration Service. This number accounts for approximately 4% of the economically active population within the Russian Federation. Political discussions frequently arise concerning the need for such a large number of migrants. Many demography experts defend the viewpoint that given the predicted reduction in the able-bodied population over the long term (up to 2050), even if domestic labor resources are mobilized (increased retirement age, the involvement of disabled persons in labor activity, etc.), the country will experience a high workforce deficit (
Both demographic experts and economists have studied the position of foreign workers in the Russian labor market (Denisenko et al.,
A great deal of research has been done in the economics of migration.
The literature compares the wages of local and foreign workers. More often than not, to evaluate the differences between wages, the Oaxaca–Blinder decomposition method is used. The fullest description of the application technique and the limitations of this method is provided in
In Russia, immigration issues are dealt with by specialists in various social sciences. This topic is of interest to experts in demography and sociology. For example, in a number of papers, researchers point to the imbalance between the qualifications of Russian workers and the needs of Russian employers (
The issue of foreign migration into Russia is also studied by economists.
This paper, to some extent, continues what was done by
To find out what portion of the gap in wages is attributable to the differences in characteristics between Russian workers and migrants and how much is attributable to discrimination against foreign workers and other factors, we used the Oaxaca–Blinder decomposition method. First, we considered the Mincer equation underlying the method.
A multitude of papers use the Mincer equation to evaluate the payback from different levels of education and work experience (e.g., in relation to Russian workers, see Denisova and Kartseva, (1)where ln (wage) is the logarithm of a worker's wages; education is the duration of schooling; exp is the work experience expressed in years; β0, β1, β2, and β3 are parameters to be estimated; ɛ is a random component; and i is the worker's index.
The ɛ values are assumed to be independent and normally distributed, ɛi ∼ N(0, σ2). The model is evaluated using the least squares method. The β1, β2 parameters are expected to be positive, and the β3 parameter is expected to be negative, which indicates a hill-shaped curve describing the relation between wages and work experience.
It is known that the absence of explanatory variables in the model, affecting the dependent variable, leads to a bias in the estimations of the β parameters. This is why researchers often introduce a number of other variables instead of the aforementioned ones. There is no clear standard regarding which additional variables should be included. The ability to include particular variables in the model is limited by the data used in the study. This paper uses a base specification in the following form:(2)
where ln (wage) is the logarithm of a worker's monthly wages; EXP is (work experience), represented by a set of dummy variables: from 3 to 5 years and over 5 years (the base category is under 3 years’ experience); EDU is the (education level), represented by a set of dummy variables: vocational education, higher education (the base category consists of high school and lower education). The observation unit is a particular worker. For the control variables: REGION is a set of dummy variables for federal districts, Moscow, and Saint Petersburg; OKVED is a set of dummy variables representing the type of economic activity in which the individual is employed; and OKZ is a set of dummy variables for the worker's profession according to the Russian Classification of Occupations.
The absence of a continuous work experience variable in the data provided by the Federal Service for Labor and Employment prevented us from taking into account the reduction of income with age in the model (since we cannot include the experience squared, which is strongly correlated with age). Nevertheless, we do not believe that age is highly relevant for our purposes; we know from other studies that the majority of foreign workers is represented by people aged 40 and less (approximately 70% of temporary migrants;
Now we shall consider the Oaxaca–Blinder decomposition method. Two Mincer models were estimated with the same set of variables: one based on foreign worker data and the other based on local worker data (Fortin et al., 2011).(3)where M is the index representing foreign workers; R is the index for Russian workers, and Xl is the same set of explanatory variables as in (2).
The difference between average wages paid to local and foreign workers is decomposed as follows:(4)where are the rows of average values of the independent variables for Russian workers and foreign migrants, respectively, and are the vectors of coefficients estimates obtained from model 3 for Russian workers and for migrants, respectively.
Component No. 1: the difference between average wages paid to migrants with characteristics of Russian workers and average wages paid to migrants. It is usually interpreted as the wage gap due to the differences in characteristics. However, in our case, all of the variables considered are categorical and established by a set of dummy variables, and only a portion of them can be ranked (e.g., education and experience). The mean value of dummy variables assuming 0 or 1 is the percent of observations in the sample in which the characteristic is inherent, for example, the percentage of people with higher education, the maximum work experience, living in Moscow, working in the construction industry, etc. However, what is the meaning of one sample containing more of those working in Moscow than another? In quantitative terms, it means that in the case of greater return on work, this region would demonstrate a greater gap in wages between the two groups. Consequently, Component 1 shows differences in the composition of the samples. We cannot state that the differences are attributable to better or worse worker characteristics. They are simply different in the event of fixed return on them (i.e., in the case of the same coefficients in the model).
Component No. 2: the difference between the average wages paid to Russian workers with the same characteristics as migrants and the average wages paid to migrants. The literature often interprets this component as a gap in wages caused by discrimination against foreign workers. However, this kind of interpretation makes sense only if the model contains all the required variables. For example, the differences in wages between local and foreign workers may be affected by knowledge of the local language. If this variable is not included in the model, then the gap resulting from the evaluation would partially include the difference in wages caused by possessing a special skill, i.e., knowledge of the language. In this case, it would not be absolutely correct to interpret this component purely as discrimination.
Component No. 3: the joint impact of Components 1 and 2. More often than not, this component assumes large values (often negative), when, for example, within one sample, the payback from a certain factor is greater than from another, but the average values of it are lower, or vice versa. However, the situation may differ substantially with respect to different factors; therefore, this component is difficult to interpret, and many researchers do not focus their attention on it. We do not address this component in this paper either, since all our variables are categorical and sufficiently numerous.
To understand what contribution is made to discrimination by particular factors in the model, we calculated a detailed decomposition (
Temporary foreign migrants have several opportunities to find jobs in Russia. The most common way is to obtain a work permit within an annual quota based on employer applications (at least, this was the case until 2013). These data are available on the Federal Service for Labor and Employment website and were used in this research.
This paper addresses discrimination against foreign workers in terms of the wages they are paid rather than with the comparative costs of a Russian firm associated with both foreign and Russian workers. RLMS data specify the wages cleared of all taxes. As to the wages according to the Rostrud data, it is not clear whether they include the personal income tax. We assume that they do not. If the tax is indeed included, then the discrimination gap is actually even greater than in our calculations. We also assume that a foreign worker stays in Russia for more than half a year. In this case, according to the Tax code of the Russian Federation, he becomes a tax resident and pays income tax at the same rate as a Russian citizen.
To use the Oaxaca–Blinder method, we need to estimate two equations with the same set of variables. Thus, we unified the variables for the two samples. We kept only the regions where the surveys were conducted for the RLMS in the Rostrud data; out of the types of economic activity, we kept only those for which we could find a definite correlation in both databases. The categories of work experience (according to Rostrud) “less than 1 year” and “from 1 to 3 years” were joined into one; “first-level” and “second level” vocational education were also joined into one group.
We now compare the samples according to RLMS and Rostrud data.
The proportion of workers with a high school education is considerably lower in the RLMS data (14% compared with 46%). The number of workers with vocational education is roughly the same in both samples. The Rostrud data feature far fewer workers with a higher education (7% compared with 32% in RLMS). Thus, the samples of Russian workers contain a higher number of better-educated people.
Workers’ professions and the type of economic activity in which they were employed were also important indicators for this study. Foreign migrants are mostly employed in low-skilled positions; the percentage of Russian workers employed in those positions is considerably lower. The largest number of foreign workers is employed in construction. With respect to Russian workers, we cannot say whether they are concentrated in any one type of activity (see
Comparison of wage distribution densities based on RLMS and Rostrud data.
Note: See Appendix Fig. A5 for 2010–2012.
Wages according to RLMS and Rostrud data, 2009–2013 (RUB, in current prices).
If we consider the samples cited as representative, it turns out that temporary foreign workers are less educated, have less experience, and occupy positions requiring lower qualifications in comparison with Russian workers.
Appendices 2 and 3 represent the results of the Mincer model 2 estimation based on the RLMS and Rostrud data from 2009 to 2013. All of the obtained regressions are significant overall, with sufficiently high explanatory power; the R2 is approximately 0.4. The return on work experience and education was subject to strong fluctuations in the time period under review. Nevertheless, it is quite reasonable to expect that more experienced workers are paid higher wages. Having higher education also results in significantly higher wages for both Russian and foreign workers. “First-level” vocational and “second-level” vocational education have almost the same impact for pay increases. Moreover, a temporary foreign worker with a vocational education receives only a slight increase in wages in comparison with a high school or lower-grade education.
Oaxaca–Blinder decomposition for wage logarithms and the degree of discrimination, 2009–2013.
The difference in the logarithms of average wages is quite substantial during the time period of interest (0.16 on average during the period). This gap can be attributed to the fact that foreign workers have a “poorer” set of characteristics than locals and that foreign workers are discriminated. Now we shall consider the components of this difference.
Component 1 over the period averaged 0.06 of the logarithm of wages, i.e., the differences between the characteristics of Russians and migrants at a fixed payback from the factors for migrants (estimates from the model for migrants).
Note that this value represents a small share of the total difference between the logarithms of wages. Recall that Component 1 represents the differences in the structure of the Russian and migrant samples. If we look at the detailed decomposition (
We now shall consider Component 2, which represents the discrimination gap. We define discrimination as the difference in wages under conditions of equal productivity. In this case, we evaluate the differences in wages paid to Russian and foreign workers based on the productivity of migrants.
Comparison of the predicted logarithm of wages density distribution for Russian workers with the productivity of migrants (MR) and predicted migrant wages (MM).
Note: See Appendix Fig. A6 for 2010–2012.
If we look at the detailed decomposition for Component 2, i.e., the discrimination gap (see Appendix Table A4), the greatest contribution is made by the industry in which the workers are employed. It makes up half of the gap by absolute value. This means that the return on work in the same industries differs greatly between migrants and Russians, with the return being considerably greater for Russians. The contribution of the region becomes significant only starting in 2011, and it is negative, i.e., the return on work in particular regions is, on average, higher for migrants than for Russians.
The contribution of the profession to the discrimination gap is insignificant. Individual worker characteristics, such as education and experience, make different contributions to the discrimination gap in different years, insignificant in some years, positive or negative in other years. Absolute values are also rather low in comparison with the contribution from the industry.
We can assume that the main reason for the discrimination gap is that which can be considered “pure” discrimination. Foreign workers have a choice to either work in their own country or go to another country for temporary work. If the country of origin offers low wages, they are willing to work in another country, provided that their wages would be higher; however, the wages should not necessarily be at the same level as wages paid to local workers. Employers, in turn, find it to their advantage to pay foreigners less than a local worker for the same productivity. Thus, the equilibrium wage for migrants is lower than for locals if there is no legislation to protect the rights of foreign workers. However, this is not the only possible reason for the discrimination gap. Employers may simply lack the opportunity to determine the actual productivity of a foreign worker (
If we assume that the main reason for the discrimination gap is pure discrimination, these conditions would be unfavorable for Russian workers employed in similar positions, as it might lead to lower wages for them or to unemployment. However, the preservation of stability in the wage gap may indicate that employing migrants does not lead to lower wages for Russians, since otherwise the level of wages paid to Russians would decrease to the level paid to migrants. Nevertheless, the results we obtained are insufficient alone to find a definite answer to this question since we do not know how this value has changed over a longer time period, and it is not clear how many local workers were forced to change their qualifications due to a lack of jobs.
Note that the obtained values of the discrimination gap probably do not correspond to the actual values due to imperfections in the model and other technical reasons. Foreign workers may be worse off than Russians in terms of unobserved characteristics (e.g., poor knowledge of the Russian language), causing an erroneous evaluation of predicted wages. The inaccuracy of the discrimination gap may also be caused by the fact that Russians and foreigners work with different intensity. Surveys show that migrants spend more time working than do Russians (
The differences between the results can be explained by several reasons. First, migrants from Tajikistan are likely to occupy less qualified positions than migrants in general. Accordingly, migrants from Tajikistan are paid lower wages, and the discrimination gap is greater for them. A second reason may be associated with the specifics of the data under review. In our source, wages are declared by the employer in the application for the quota and may be considered to be the wages offered to migrants. In the surveys of migrants from Tajikistan, wages are the amount that the migrant actually received. On one hand, discrimination may be greater because declared and actually paid wages may differ significantly. On the other hand, applications for quotas apply solely to legal migrants, and the surveys of migrants from Tajikistan were conducted among both legal and illegal migrants. According to Lokshin and Chernina (2013), (54)% of migrants had work permits in 2007, compared with 87% in 2009.
The main result of this work was to establish the fact that foreign workers are strongly discriminated against in terms of wages: the average wages paid to Russian workers with the same productivity as migrants was, on average, 40% higher during 2009–2013. At the same time, the greatest contribution to explaining the discrimination gap is made by the industry (or types of economic activity) in which workers are employed.
The emergence of new empirical papers on the economics of migration is hindered mostly by the absence of statistical data of high quality. For example, in 2015, the rules for employing foreign migrants changed: the work permit quotas for CIS migrants were replaced with work patents for employment with legal entities while preserving the existing work patents for employment with individuals.
Many questions remain regarding the nature of the discrimination gap and its effect on the Russian labor market. For example, it is unclear how the low wages paid to migrants affect the wages paid to Russian workers employed in similar positions. This paper provides an argument that the low wages paid to migrants do not affect those paid to local workers, i.e., the discrimination gap remains constant over a long-term horizon. It is also unclear how the presence of foreign workers in the Russian labor market affects the employment of local workers with low skills. More high-quality data and new studies in the area of migration economics would produce a more complete picture and create a basis for correct political recommendations.
The study was carried out by the Economic Expert Group and funded by grant No. 14-18-03666 from the Russian Science Foundation. The authors extend their thanks to Evsey Gurvich, head of the Economic Expert Group, for his valuable comments and advice.
Composition of the Rostrud and RLMS samples, 2009–2013 (%).
Estimation of model 2 based on Rostrud and RLMS data, 2009–2011.
Estimation of model 2 based on Rostrud and RLMS data, 2012–2013.
Detailed decomposition for Components 1 and 2 of the Oaxaca–Blinder decomposition (4), 2009–2013.
Figs. A
The breakdown is provided in Appendix Table A1. Aggregate categories are similar to the International Standard Classification of Occupations.
To calculate work experience based on RLMS data, we used the following formula: age – (6 + x), where x = 11 for persons with high school education, x = 14 for persons with vocational education, and x = 16 for persons with higher education.
The samples are limited by workers with wages between RUB 5,000 and RUB 60,000 so that outliers do not affect the results.
We also made our calculations based on a fixed productivity of Russian workers. In this case, the discrimination gap represents the differences in the logarithms of wages for Russians and migrants with the characteristics of Russians X¯MβˆR−βˆM. However, it is not a component of the Oaxaca–Blinder decomposition. This gap is considerably lower, 0.1 on average, over the period under review.
In the first case, authors reviewed a sample of migrants located abroad, and in the second, those who returned from their travels and were surveyed in Tajikistan.
Federal Law No. 115-FZ, On the Legal Status of Foreign Citizens in the Russian Federation, dated June 25, 2002, as amended in 2015.