Labour market dualism and immigration policy preferences

ABSTRACT

What are the effects of labour market deregulation and increased immigration inflows on public attitudes towards immigration? Despite increased levels of dualism and free movement of labour in European countries over the last two decades, the effects of these policy developments are still unclear in the literature. This study argues that high concentrations of migrant workers in non-standard forms of employment decrease economic redistribution towards, and labour competition with, immigrants. Consequently, the politicisation of immigrant-native conflicts is paradoxically lower when immigration and labour market dualism cluster together at the occupational level. These claims are validated cross-nationally, and in a difference-in-differences setting analysing the impact of the 2005 German Immigration Act.

KEYWORDS:

• Attitudes towards immigration
• immigration policy preferences
• labour market
• dualism
• skill specificity
• difference-in-differences

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on contributor

Sergi Pardos-Prado is a full Professor of Comparative Politics at the University of Glasgow. Before this, he was an Associate Professor in Politics at Merton College, University of Oxford. His research lies at the intersection between political behaviour, European comparative politics, and comparative political economy.

ORCID

Sergi Pardos-Prado http://orcid.org/0000-0002-6944-326X

Notes

1 The data that support the findings of this study are openly available at https://sergipardos.wixsite.com/sergipardos/publications.

2 https://www.europeansocialsurvey.org/ (01 October 2018).

3 Those items are 0–10 scales measuring attitudes on whether (1) immigration is bad or good for the country's economy, (2) the country's cultural life is undermined or enriched by immigrants, and (3) immigrants make the country a worse or better place to live. A Cronbach's alpha of 0.85 for the three items above indicates high reliability. These items are added up in an index ranging from 0 to 30, where higher values indicate anti-immigration attitudes.

4 https://stats.oecd.org/ (02 October 2018).

5 Given their skewed distribution, I use the natural logarithm of occupational dualism and occupational presence of immigrant workers in the statistical analyses below.

6 The figures have been produced with the plottig package for Stata designed by Bischoff (2017).

7 Measuring migrant shares from survey data is not perfect, as more settled migrants could be more likely to be included in national random probability samples. However, the bivariate correlation between migrant presence and dualism at the occupational level is −0.1, suggesting that migrant workers are equally likely to be measured at all levels of occupational dualism, and that the results here are not biased in one obvious direction.

8 This measure consists of the share of ISCO-88 unit groups over the total number of unit groups in a major occupational cluster, divided by the share of workforce that the major group represents. This measure is then divided by the skill level of each large ISCO occupational group (Iversen and Soskice 2001).

9 https://www.loc.gov/law/help/migration-citizenship/germany.php (17 February 2019).

10 https://www.dw.com/en/first-german-immigration-law-takes-effect/a-1442681 (17 February 2019).

11 The specific model specification reported in this section could be denoted as $Y_{it}=a+b{x}_{it}+{\delta }_i+w_t+{\epsilon }_{it}$, where Y is the level of anti-immigrant concern of individual i in year t, b is the difference-in-difference estimate, and $x_{it}$ is a dummy variable where 1 = individuals with a university degree in 2005, and 0 = everyone else. Note that, when modelling individual $\left({\delta }_i\right)$ and survey-year $\left(w_t\right)$ fixed-effects, the difference-in-difference estimate captured by b is mathematically equivalent to modelling the interaction between high levels of education and the year 2005.

12 Table A5 in the Appendix replicates the results shown in Table 2 with bootstrapped standard errors.