Dynamics of English Fluency Return for Refugees and Other Immigrants in the United States

Abstract

Previous work has established that U.S. immigrants earn more if they are fluent in English, but a portion of that return likely reflects biases because fluency is correlated with unobserved factors such as self-selection and ability to job shop. This article investigates those factors by looking at the variation in the crude fluency returns earned by resettled refugees and other immigrants in the United States. Results show that nonrefugees initially earn a much larger wage return for fluency, and this gap persists in the first years after immigration. However, refugees’ return does eventually increase and catches up with that of the nonrefugees.

Keywords:

• Refugees
• English investment
• selection
• job shopping

Acknowledgments

I am grateful to the many readers who offered helpful comments, especially Scott Drewianka and Madeline Zavodny. I am solely responsible for all remaining shortcomings.

Notes

1 More discussion on this, see Chiswick et al., 1995, and Bleakley et al., 2004, who consider the potential endogeneity between earnings and language fluency.

2 I assume that employers observe workers' abilities, but the econometrician does not, so it is possible that the econometrician will observe a correlation between fluency and wages even if employers were paying for ability (provided ability is correlated with fluency).

3 This is a similar division to the one used by Chiswick (1978).

4 To account for any potential outliers, I winsorize the wages by resetting the observations in the bottom and upper 1% tails of the distribution to the first percentile and the ninety-ninth percentile values, respectively, while retaining all the observations, but the results are quite similar.

5 Residence in an ethnic enclave might affect the rate of acquisition of English fluency.

6 Similar results emerge if we define “fluent” to include those who speak English only and those who speak English very well or if we exclude from the sample those who speak English only.

7 I define immigrants as “skilled” if their level of education is greater than the mean education of their country of origin. This is to account for the fact that the countries immigrants originate from might have different levels of educational attainment.

8 To avoid collinearity of the cohort-year interactions, I use year of entry as a continuous variable to interact with country of origin (c.year##i.cohort).