VoxEU Column Migration

Immigrants' earnings growth and return migration from the US

Nearly 40% of documented new arrivals to the US in 2005 left within ten years, but who return migrates and why is often overlooked in policy debates regarding immigration. This column uses survey data and earnings records from 2005 to 2015 to show that a decline in earnings is a strong predictor of return migration. Those who stayed for the decade saw their wages reach parity with native-born workers, while those who left had seen a steep decline in wages in the years before departure. Further analysis shows that highly educated immigrants are more likely to leave the US within a decade of arrival.

Policy debates in the US regarding immigration often focus on how well new arrivals assimilate into the US economy in terms of employment and earnings. Previous research (e.g. Borjas 2003, Card 1990) has attempted to examine rates of assimilation, but a key component of the immigration story was missing: Who return migrates, and why? Recently available data go far in addressing this question (Akee and Jones 2019). 

Although the US has historically provided important opportunities for immigrants, our research indicates that nearly 40% of documented new arrivals to the US in 2005 left before 2015. Of all new arrivals, those who remained in the US for the full decade eventually saw their wages reach parity with native-born workers. In contrast, those who left saw a steep decline in wages in the years before departure. 

Our analysis uses data from the American Community Survey (ACS) linked to Form W-2 and Form 1099 earnings records. We start with new arrivals to the US in the ACS from 2005 to 2007 (those who reporting arriving in the year or two before taking the survey). We then track these new arrivals forward in the tax records, tracing their outcomes in the US labour market up to 2015. 

Our data solve a key problem in analysing immigrant outcomes. In the past, researchers were limited to examining a point-in-time population of immigrants – a group that remained in the US up to the point at which they are observed. Such data suffer from selection bias in that those who remain have differentially succeeded in the US labour market. Without information about those who leave, estimates of labour market outcomes for immigrants will be biased upward. 

Our findings

We find that average earnings of new arrivals start off at about 60% of those of native-born workers with comparable education. Earnings of long-term ‘stayers’ quickly diverge from those who leave. In Figure 1, we show the average earnings of the two groups compared with native-born earnings over time.   

Figure 1 Log immigrant–native annual earnings of men aged 25-45

We show the earnings for our two immigrant groups as proportions of native-born earnings (the horizontal dashed red line) in each year. Although the two groups arrive with fairly comparable earnings, divergence becomes apparent in the second year, and there is a steady decline in the earnings of ‘leavers’ beginning in 2008. The earnings of the ‘stayers’ eventually reach parity with, and even surpass, those of the native-born by 2010. Earnings for the leavers reach, at most, 80% of native-born earnings (in 2008). 

This decline in earnings is a strong predictor of return migration and is the novel finding of our analysis. These results are consistent with previous findings that return migrants are negatively selected for labour market success and earnings in the US. The novelty of our finding is that the earnings trajectory is downward trending. Whereas previous results simply indicated that there might be level differences across ‘stayers’ and ‘leavers, we show that there is a continual downward earnings trajectory for return migrants over all time periods. 

Figure 1 also shows the impact of positive selection on estimates based on repeated cross-section data. The dashed black line reports the predicted earnings for men from a standard repeated cross-section analysis, which includes all immigrants present in the data at each time period. As demonstrated in the figure, over time as return migration creates an increasingly high-earning stock of immigrants, earnings growth estimates are increasingly overstated. 

Further analysis shows that highly educated immigrants are more likely than those with less education to leave the US within a decade of arrival. In Figure 2, we show the return rates for the 2005 entry cohort by educational attainment. There is a relatively steep reduction in the immigrant cohort in the first year after arrival, when approximately 20% return home. After the first year, there is divergence by educational attainment in the return migration rates. It appears that individuals with the highest educational attainment are proportionately more likely to return home over time. The solid black line represents those with a Master’s or Doctorate degree; they have the highest return migration of all education categories. Those with a college degree also have a relatively high return rate. Those with some college or a high school degree are less likely to leave the US. 

Figure 2 Presence of W-2 or 1099 for 2005 entry cohort for men aged 25-45 by educational attainment

Return migration rates also vary by other characteristics, such as country of origin – immigrants from Canada are the most likely to return, while those from the Philippines are least likely. Meanwhile, neither marital status nor English language ability are important determinants in return migration. 

We also examine earnings assimilation by a variety of demographic characteristics.  Several interesting patterns stand out:

  • Immigrants with less than a high school degree out-earn native-born workers upon arrival in the US, but experience flat earnings growth. Other education groups show assimilation rates similar to the overall average. 
  • Country of birth also matters, with immigrants from India and Canada out-earning native-born workers upon arrival, those from China and the Philippines assimilating over time, and those from Mexico holding steady at about 70% of the native-born wage. 
  • Race also plays an important role. Compared with native-born of the same race, non-Hispanic whites quickly reach parity and soon outpace native-born workers by 40%. All other races, meanwhile, reach parity with their native-born counterparts more slowly. These last findings are in line with Villarreal and Tamborini (2018). 


An important result in our analysis is that male return migrants and male stayers enter the US with comparable earnings. We provide novel findings suggesting that unobserved time-variant characteristics likely play an important role in determining immigrant earnings success. Previous research highlighted the importance of cultural and social assimilation of immigrants. 

While we have no measures of these characteristics over time, it is possible that non-cognitive type skills are important in determining employment relationships and success. As noted earlier, these characteristics also must be related to the time spent in the US. Therefore, immigrants who remain in the US, holding observed characteristics constant, must be those who are most flexible and integrate better with respect to the US labour force. Our research suggests potential policy interventions to help with the adaption of immigrants formally entering the US labour force. 

Author’s note: Any opinions and conclusions expressed herein are those of the author and do not necessarily reflect the views of the U.S. Census Bureau.


Akee, R and M R Jones (2019), “Immigrants' Earnings Growth and Return Migration from the U.S.: Examining their Determinants using Linked Survey and Administrative Data”, NBER Working Paper, No. 25639.

Borjas, G J (2003), “The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market”, The Quarterly Journal of Economics 118(4): 1335–1374. 

Card, D (1990), “The impact of the Mariel boatlift on the Miami labor market”, ILR Review 43(2): 245–257. 

Villarreal, A and C R Tamborini (2018), “Immigrants’ Economic Assimilation: Evidence from Longitudinal Earnings Records”, American Sociological Review 83(4).

630 Reads