A panel data quantile regression analysis of the immigrant earnings distribution in the United Kingdom and United States

Abstract

This paper uses longitudinal data from the PSID and the BHPS to examine native-immigrant earnings differentials throughout the conditional wage distribution, while controlling for individual heterogeneity. We employ quantile regression techniques to estimate conditional quantile functions for longitudinal data. We show that country of origin, country of residence, and gender are all important determinants of earnings differentials. A large wage penalty occurs in the USA among female immigrants from non-English speaking countries, and the penalty is most negative among the lowest (conditional) wages. On the other hand, women in Britain experience hardly any immigrant-native wage differential. We find evidence suggesting that immigrant men in the USA earn lower wages, while British workers emigrating from English-speaking countries earn higher wages. The various differentials we report in this paper reveal the value of employing panel data quantile regression in estimating and better understanding immigrant wage effects.

Appendices

Appendix 1: Monte Carlo simulations

We investigate the performance of the approach proposed (2.7) and (2.8) using Monte Carlo simulations and we report the results in Table 11. The data-generating process is described by the following equations:

$$\begin{aligned} y_{it}&= \beta _{0} + \beta _{1} x_{1it} + \beta _2 x_{2i} + \alpha _{i} + u_{it}, \end{aligned}$$

(6.1)

$$\begin{aligned} x_{1it}&= \sqrt{1-\rho } \varepsilon _{it} + \rho \varsigma _{i} + \pi _1 \alpha _{i}, \end{aligned}$$

(6.2)

$$\begin{aligned} x_{2i}&= \varsigma _{i} + \pi _2 \alpha _{i}, \end{aligned}$$

(6.3)

where $\varsigma _{i}$ and $\varepsilon _{it}$ are $t_{3}$ and $\alpha _i$ and $u_{it}$ are distributed as $N(0,1)$. We set $\rho = 0.5, \beta _{0}= 5, \beta _{1}= 2, \pi _{1}= 0.5$, and $\beta _2=0.1$. We concentrate our attention in the estimation of the parameter of interest $(\beta _1,\beta _2)$ at the following quantiles: $\tau = \{0.10,0.50,0.75\}$. We vary the experiments by changing the sample size $N = \{200, 500\}$ and $T = \{10, 15\}$ and consider the following two variations of model (6.1)–(6.3):

Design 1: :

We consider the case of no correlation between $\alpha _i$ and $x_{2i}$ but we allow $x_{1it}$ to be correlated with $\alpha _i$ and $x_{2i}$. We assume that $\pi _2 = 0$.

Design 2: :

We assume that $x_{1it}$ and $x_{2i}$ are correlated with $\alpha _i$. We assume that $\pi _1 = 0.5$ and $\pi _2 = 0.10$.

Table 11 Simulation results

Appendix 2: Robustness checks

This appendix presents empirical results from alternative specifications. In Tables 12, 13, 14, and 15, we present results from estimating three different models. In Model 1, we incorporate additional covariates typically considered in the literature. We introduce, in addition to the independent variables used before, indicators for race, hours worked, whether or not the worker has children and union membership. In the case of the BHPS, we include indicators for whether the worker is black, Indian, Pakistani or Bangladeshi, and Chinese. In the case of the PSID, we include indicators for whether the worker is black, Asian, Latino, and American Indian. The omitted indicator variable is also white. In all the models, we include controls for industry (e.g., Agriculture, Manufacturing, Construction and Trade, and Services) and occupation (e.g., Managers and Senior Officials, Professional Occupations; Administrative and Secretarial Occupations; Skilled Trades Occupations, Personal Service Occupations, Associate Professional and Technical Occupations; Sales and Customer Service Occupations; Process, Plant and Machine Operatives, Elementary Occupations). In Model 2, we classify countries as follows. The group ‘English’ includes immigrants from the USA or UK, Canada, Ireland, Australia, and New Zealand. The group ‘Fluent’ includes immigrant from Germany, Sweden, Austria, Denmark, the Netherlands, Belgium, Hungary, Poland, Singapore, and Malaysia. The group ‘Non-Fluent’ includes immigrants from other countries from Europe, South America, Africa, and Asia. Model 3 incorporates interaction terms on marital status and educational qualifications and immigrant status. Lastly, Table 16 shows the results for the immigrant/native earnings differential by country of origin for immigrants to the UK from English-speaking countries.

Table 12 Robustness checks on models for male workers on the BHPS sample

Table 13 Robustness checks on models for female workers on the BHPS sample

Table 14 Robustness checks on models for male workers on the PSID sample

Table 15 Robustness checks on models for female workers on the PSID sample

Table 16 Immigrants from English-speaking countries in the UK labor market