The Changing Structure of Immigration to the OECD: What Welfare Effects on Member Countries?

Abstract

We investigate the welfare implications of two pre-crisis immigration waves (1991–2000 and 2001–2010) and of the post-crisis wave (2011–2015) for OECD native citizens. To do so, we develop a general equilibrium model that accounts for the main channels of transmission of immigration shocks – the employment and wage effects, the fiscal effect and the market size effect – and for the interactions between them. We parameterize our model for 20 selected OECD member states. We find that the three waves induce positive effects on the real income of natives; however, the size of these gains varies considerably across countries and across skill groups. In relative terms, the post-crisis wave induces smaller welfare gains compared to the previous ones. This is due to the changing origin mix of immigrants, which translates into lower levels of human capital and smaller fiscal gains. With a few exceptions, differences across cohorts explain a tiny fraction of the highly persistent, cross-country heterogeneity in the economic benefits from immigration.

Appendices

Appendices

1.1 Appendix 1: Comparability Between Cohorts

Two different methods are used to identify the size and structure of the pre-crisis (2001–10) and post-crisis (2011–15) immigrant cohorts. For the pre-crisis waves, we directly identify the number and characteristics of immigrants arrived between 2001 and 2010 (or between 1991 and 2000) using DIOC data by duration of stay. For the post-crisis wave, we proxy their number using the relative variation in dyadic immigrant stocks between 2010 and 2015 from the United Nations database. We then proxy their education level using the dyadic education structure of the previous wave. This method is relevant if the dyadic education structure of immigration is stable over time. To validate our hypothesis, in Fig. 6a we use the same method and retrospectively predict the education structure of the 1991–2000 wave. We can then compare our “back-casts” with the observed structure; the correlation between predicted and observed structure of 1991–2000 immigrant cohort equals 0.97. We are thus confident that our imputation method does not drive the results.

Another difference is that the welfare effect of the pre-crisis wave is computed using a 10-year inflow of immigrants, while the effect of the post-crisis wave is computed using a 5-year inflow. As the model is nonlinear, one may fear that the size of the shock matters and that our semi-elasticities are not comparable. To check this, we simulated the effect of the post-crisis wave on a 10-year basis, assuming that the size and structure of the 2016–2021 inflow are identical to those of the 2011–15 inflow. The outcomes are presented in Fig. 6b. The correlation between the semi-elasticities computed using 5-year and 10-year shocks equal 0.998. We are thus confident that our comparative results are not driven by differences in the size of immigration.

Fig. 6

Comparability across immigrant cohorts. Notes Figure 6a plots the imputed HS share in the 1991–2000 cohort using the 2001–10 HS shares, while Fig. 6b compares the welfare effects of 2011–15 wave with an imputed 2011–2020 one

1.2 Appendix 2: Robustness Checks

In this section, we investigate whether the conclusions of the benchmark analysis are robust to the choice of parameters, to the inclusion of technological externalities, to the fiscal rule used to balanced the government budget constraint, and to the characteristics of immigrants. Focusing on the effect of the 2001–2010 wave, we assess the sensitivity of its impact on the average level of real income and on income inequality. Figure 7 depicts the results obtained under nine variants of the benchmark model.

Sensitivity to elasticities. In Fig. 7a, b, we consider alternative levels for three elasticities, namely for the elasticity of substitution between goods in the utility function ($\epsilon =4$ instead of 7), for the elasticity of substitution between immigrant and native workers in production ($\sigma _{2}=50$ instead of 20), and for the inverse of the elasticity of labor supply to labor income ($\eta =5$ instead of 10).¹⁰ For these three variants, we recalibrate the TFP (A) and the disutility of labor ($\phi$) to match observed GDP levels and participation rates in 2010, and simulate the no-migration counterfactual.

Fig. 7

Sensitivity analysis (immigration wave 2000–2010). Notes Figure 7 shows the results for 20 selected countries: the 15 members states of the European Union (EU15), the USA, Canada, Australia, Switzerland and Japan. The left panels depict the changes in average real wages, while the right panels provide deviations in income differentials between high-skilled and low-skilled

Figure 7a shows that the average welfare impact of immigration is one and a half times greater when the elasticity of substitution between goods equals 4, and 1.2 times greater when labor market participation rates are more elastic to labor income. Increasing the elasticity of substitution between immigrant and native workers to 50 has a minor impact on our results. In all variants, the correlation with the benchmark results exceeds 0.99. As far as the inequality impact is concerned, results are almost independent on the choice of elasticity (see Fig. 7b). Hence, the welfare effects depicted in Fig. 5 are highly robust to the choice of elasticities.

Sensitivity to externalities. In Fig. 7c, d, we account for three TFP externalities. The first one is a schooling externality; it assumes that the elasticity of TFP to the proportion of college graduates in the labor force is equal to 0.3, in line with Croix and Docquier (2012) or Aubry et al. (2016). The second one is a diversity externality; it assumes that the semi-elasticity of TFP to birthplace diversity is equal to 0.2, as in Alesina et al. (2016) or Ortega and Peri (2014). The third one is a diaspora externality which captures the effect of migration on trade and FDI, and the resulting effect of trade and FDI on TFP. Although our model does not account for trade and FDI, we directly model the TFP externality due to the presence of migrant diasporas. A first strand of literature has identified a causal impact of migration on trade and FDI, with respective elasticities of 0.1 and 0.2.¹¹ Another strand has identified a causal effect of trade and FDI on TFP, with respective elasticities of 0.3 and 0.01 (see Anderson et al. 2016; Feyrer 2009). Combining these findings gives an elasticity of TFP to migration of 0.035. For these three variants, we calibrate the scale factors of the TFP function to match the GDP levels in 2010, and simulate the no-migration counterfactual.

Figure 7c shows that the schooling externality significantly increases the gain from immigration in countries attracting college-educated migrants (such as Australia, Luxembourg, the UK and Switzerland) while it reduces the gain in Spain, Belgium and Greece. Birthplace diversity quantitatively matters only in newer immigration countries such as Ireland, Portugal and Finland. At the estimated elasticity levels, the diaspora externality has negligible effects on the results. Results for inequality are almost independent on the inclusion of externalities (see Fig. 7d). Again, the welfare effects depicted in Fig. 5 are highly robust to TFP externalities.

Sensitivity to immigrants’ characteristics. In Fig. 7e, f, we consider three alternative distributions of characteristics of immigrant workers. We first assume that all immigrants have the same disutility of labor as natives (same $\phi$’s). We then assume they have identical unemployment rate as natives (same u’s). Finally, we assume identical skill structures for the immigrant and native populations (same h’s). For these three variants, we simulate the new hypothetical benchmark for 2010 (keeping other parameters constant), and simulate the no-migration counterfactuals.

Figure 7e shows that the average welfare gain from immigration is highly robust to labor market characteristics of immigrants. Imposing the same disutility of labor (governing participation rates) or the same unemployment rate as native workers marginally affects the results, with the exception of Scandinavian countries and Belgium. Inequality responses are also robust to labor market characteristics. This is in line with the fact that changes in employment rates in the post-crisis period hardly affected the welfare impact of immigration. On the contrary, equalizing immigrants’ and natives’ levels of schooling reduces the gains from immigration in “selective” countries (Fig. 7e), and neutralizes the inequality responses to immigration (Fig. 7f). This confirms that the level of human capital of immigrants affects the macroeconomic and welfare responses to immigration; the degradation of immigrant’s human capital after the crisis is responsible for smaller welfare gains.

Sensitivity to the fiscal rule. In our benchmark model, we assume that the income tax rate adjusts to balance the government budget constraint. We consider two alternative fiscal rules, using the consumption tax rate or the amount of public transfers as the adjustment variable. Figure 7g, h shows that the average welfare gain from immigration is highly robust to the fiscal rule. Slightly smaller gains are obtained when public transfers or consumption tax rates are used to adjust the budget. This is because the fiscal gain is now shared between young and old natives.

1.3 Appendix 3: Country-Specific Results and Parameters

See Tables 2 and 3.

Table 2 Calibrated country-specific parameters

Table 3 Effects of immigration (Benchmark minus counterfactual as % of the benchmark)