Labor-market exposure as a determinant of attitudes toward immigration

Abstract

This paper re-examines the role of labor-market competition as a determinant of attitudes toward immigration. We claim two main contributions. First, we use more sophisticated measures of the degree of exposure to competition from immigrants than previously done. In addition to education, we focus on the protection derived from (self-assessed) investments in job-specific human capital and from specialization in occupations that are (objectively) intensive in communication tasks. Second, we explicitly account for the potential endogeneity arising from job search. Methodologically, we estimate by instrumental variables, an econometric model that allows for heterogeneity at the individual, regional and country level. Drawing on the 2004–2005 European Social Survey, we obtain the following main results. First, natives that dislike immigrants tend to work in low-immigration jobs, biasing OLS estimates. Second, working in jobs that require high levels of specific human capital leads to relatively more pro-immigration attitudes, although this effect is only found for respondents with more than 12 years of schooling. Third, the degree of manual (communicational) intensity of workers' occupations has a negative (positive) effect on their pro-immigration views. This effect is the most significant, both in a statistical and in a quantitative sense, and is distinct from the protection from immigrant competition provided by formal education. Overall our results suggest a large role for skill-based labor market competition in determining individual attitudes toward immigration.

Introduction

Several European countries have experienced a rapid increase in immigration over the course of the last decade. Immigration poses important challenges to Europe. Will European societies be able to integrate increasing immigrant flows? Will European citizens continue to tolerate growing levels of immigration or will they push instead for a tightening of immigration policies? To a very large extent the answer to these questions depends on voters' own perceptions of the economic and cultural effects of immigration. But what are the determinants of these perceptions? Fueled by these concerns, economic research on migration has of late drawn increasing attention to the study of native attitudes toward immigrants.

Most studies in economics focus on the role of competition in the labor market as a crucial determinant of attitudes toward immigration.¹ Usually the degree of competition between native and immigrant workers is measured in terms of schooling levels. Under the assumption that immigrants are, on average, less educated than natives, it is expected that low-educated natives oppose immigration to avoid depressing their wages. In contrast, being less exposed to competition from immigrants, highly educated natives are expected to be relatively more pro-immigrant. Empirical studies have consistently found a positive association between respondents' educational attainment and pro-immigration views, which has been typically interpreted as evidence that labor-market competition is an important determinant of attitudes toward immigration. However, it must be noted that these correlations do not constitute unequivocal evidence in favor of the labor market exposure hypothesis. As argued by several authors across the social sciences, the positive effect of education on pro-immigration attitudes could be entirely due to values and predispositions associated to schooling, such as tolerance, open-mindedness or political correctness, rather than labor market competition per se.²

This paper re-examines the role of labor-market competition as a determinant of attitudes toward immigration using individual-level data for a large number of European countries. Our main contribution is the use of more sophisticated measures of exposure to immigration in the labor market than typically available. Besides education, we consider two further measures of exposure that are directly linked to the tasks workers perform at their jobs. From an empirical point of view, an attractive feature of these new skill measures is their large variation across jobs and occupations.

Our first new measure of labor-market exposure is a self-assessed measure of job-specific human capital. Workers acquire job-specific skills at their firms via formal training, informal instruction or learning-by-doing. Investing in job-specific skills makes workers less replaceable, regardless of their level of education (Becker, 1993 1964, Lazear, 1995). Job-specific human capital thus provides protection against competition from other workers, native and immigrant alike.

Our second measure of labor market exposure is based on objective characteristics of respondents' occupations and is motivated by recent work studying how native workers respond to recent immigration. Using US data, Peri and Sparber (2009) show that when immigrants arrive into an economy, native workers mitigate the wage effects of immigration by shifting toward occupations for which they have a comparative advantage. Specifically, immigration induces native workers to shun manual jobs and specialize in communication-intensive occupations.³ Natives' comparative advantage in such occupations stems from the possession of those skills that immigrants typically lack, in particular, language and other cultural country-specific skills. We hypothesize that native workers employed in occupations that are intensive in manual tasks (communicational tasks) will be more (less) exposed to immigrants' competition and hence less (more) likely to display pro-immigration attitudes.⁴

Following Mayda (2006), we use individual-level data from many countries as a key source of identification. Specifically, we draw on the 2004–2005 European Social Survey.⁵ The survey contains several questions regarding opinions on immigration. The data also contain a rich set of individual characteristics, including education, job descriptors, occupation, region of residence within the country and a number of questions on individual values. We also use the US Occupational Network Online Dataset (O*NET) to build measures of the intensity of manual (and interactive) tasks by occupation, following the lead of Autor et al. (2002).

Methodologically, we estimate an econometric model that allows for heterogeneity at the individual, regional, and country level. The dependent variable is a measure of the respondent's views toward immigration and we consider three dimensions of skills: formal education, required job-learning time at the current job net of education (that is, job-specific human capital) and manual/communication-intensity of the current occupation. Each of these dimensions captures a different level of skills (individual, job and occupation) and measures a distinct source of protection from immigrant competition in the labor market. Another important contribution of our study is our treatment of individual heterogeneity. Our analysis explicitly accounts for heterogeneity among natives in their views toward immigration and for potential self-selection into low-immigration jobs. Individual observable heterogeneity is addressed by estimating specifications that include a vector of controls for ideological and attitudinal variation, while self-selection is addressed by using an instrumental-variable approach. Specifically, we use the regional availability of low-exposure jobs as an instrument for actual individual exposure in the current job, in the spirit of Dustmann and Preston (2001).⁶

Overall our results suggest a larger role for labor market competition as a determinant of individual attitudes toward immigration than previously found. More specifically, we report the following three main findings. First, the limited role for labor market competition (when compared to welfare considerations and non-economic factors) found in earlier studies may have been due to a combination of poor skill indicators and endogeneity problems. Our estimates suggest that individuals with above-average dislike for immigrants tend to work in low-immigration jobs, biasing down OLS estimates of the effects of job protection on attitudes toward immigration. Second, our instrumental-variables estimates show that working in jobs that require high levels of specific human capital increases pro-immigrant attitudes, although this effect seems to operate only at high levels of formal education (employees with more than 12 years of schooling). Third, we find that the degree of manual intensity of workers' occupation is strongly and negatively associated with pro-immigration views. Among our determinants of exposure to immigration, manual intensity in one's current occupation plays the most important role in explaining attitudes toward immigrants.

The paper is organized as follows. Section 2 situates this paper in the context of the previous literature. Section 3 presents the data sources, definitions of variables and descriptive statistics. Section 4 introduces our estimation method and explains how we deal with endogeneity. The main findings are presented in Section 5. Sensitivity analyses are reported in Section 6. Section 7 concludes.