VoxEU Column Labour Markets Migration

Immigration and wage dynamics: Evidence from the Mexican peso crisis

Arguments over the effect of immigration on labour market outcomes focus on a single number: the impact on low-skill wages. The column uses a model of the adjustment process of labour markets in the US to the peso crisis of 1995 to show there is a difference between short-run and long-run effects. The model suggests that state-level policies are unlikely to be effective.

Perhaps the most fundamental question when studying the effect of immigration on host economies is how much immigrants affect the labour market outcomes of native workers. Card (2001) compared cities (and occupations) with higher levels of immigration to cities with lower levels of immigration, and concluded that immigration had not substantially affected labour market outcomes. 

Borjas (2003) reached a very different conclusion. He argues that local labour markets in the US are well connected, and hence, comparing workers of different skills and experience levels may be more informative than using across location comparisons. His estimates suggested that immigration substantially affected the labour market outcomes of native workers, particularly the least skilled. 

In a recent paper (Monras 2018), I argue that the best answer to this fundamental question is not a single number ("a 1% immigration labour supply shock reduces wages or employment rates by X%"). Instead, there is another set of questions:

  1. If there is an effect, for how long does it last? 
  2. What allows markets to absorb immigrant shocks? 
  3. Are there other markets beyond the labour market that help in the adjustment process? 

The Mexican crisis of 1995

One of the main estimation challenges is that immigrants choose their local destination in the host country based, at least in part, on local labour markets. Strong demand for labour usually improves labour market conditions and attracts new immigrants. 

Previous work has tried to estimate the causal effect of immigration on labour market outcomes by arguing that immigrants move into certain locations because they value living with relatives or friends who migrated before them. This is known as the 'migration network' instrument. In the absence of serially correlated labour market outcomes, it identifies the effect of immigration by comparing labour market outcomes across local labour markets.

I claim that, even when the labour market outcomes are serially correlated, we can learn about the effects of immigration in the short run by looking at how labour market variables of interest deviate from trend when there is an unexpectedly large arrival of immigrants. The large increase in migrant flows into the US that resulted from the crisis that hit Mexico in early 1995 can show how immigrant shocks affect local economies. 

I found:

  • Wages of native low-skilled workers decreased on impact. An immigration-driven labour supply shock equivalent to 1% of the working population of low-skilled workers lowered wages of low-skilled workers by slightly less than 1%. Wages and employment rates of high-skilled workers, if they moved, improved on impact. 
  • The share of low-skilled population in local economies (states or metropolitan areas), increased one-to-one on impact as a result of the inflow of Mexican low-skilled immigrants. But soon after the initial shock, the share of low-skilled population in high-immigration locations returned to trend, suggesting that low-skilled workers moved from high- to low-immigrant locations, or that fewer natives in low-immigrant locations moved to high-immigrant locations when labour markets were less attractive.
  • On impact, the relative price of rentals relative to house prices increased. More than 85% of Mexicans enter the rental market upon arriving to the US, moving relative demand towards this housing market type. 

There is some internal relocation responding to local shocks, and so this creates spillovers from 'treated' to 'control' locations. This limits the usefulness of difference-in-difference comparisons if we want to learn about the effects of immigration over longer time horizons.

Estimating a quantitative spatial equilibrium model 

In the long run I can compare labour market outcomes across locations and experience groups, as Card and Borjas did by using census data for 1990 and 2000. Despite the fact that 1995 saw an unexpectedly large inflow of Mexican immigrants that affected local labour markets, differences across locations were, during the 1990s, small. But the estimates that compare wages of low-skilled workers who entered in the highest Mexican inflow years and who entered in other years are larger. This seems to confirm that internal migration creates spillovers across local labour markets. 

To investigate what happens over longer horizons more deeply, I used a quantitative dynamic spatial equilibrium model. In the model, the locations of the economy are linked through internal migration. When internal migrants move away, or avoid moving to locations that receive an unexpectedly large inflow of Mexican migrants, they become a labour supply shock in the non-affected locations (Topel 1986). This tends to dissipate local shocks across local labour markets. 

To estimate the model I relied on the variation generated by the natural experiment of the peso crisis. In particular, I used the short-run response of wages to identify the elasticity of substitution between high- and low-skilled workers. This determined the local production function. I then used the population response following the initial shock to identify the internal migration part of the model. 

I complemented these estimates with structural estimates on the underlying productivity and amenity levels in each location, as well as an estimate of local housing supply elasticities obtained from Saiz (2010), and I used the equilibrium internal migration rates across metropolitan areas to obtain a cost-of-moving parameter. 

The model replicates the observed wage and housing price dynamics well. It correctly predicts the shape of wage dynamics of both high- and low-skilled workers, and the changes in rental and housing prices, across locations. We can use the model to do counterfactuals and think about policy-relevant questions on the immigration debate. 

Counterfactual: Local technology

A key question is how the economy adapts to expected versus unexpected inflows of immigrants from other countries. Given that immigrant flows concentrate in space, a model in which only internal migration dissipates local shocks would imply that in the long-run wages of low-skilled workers would be uniformly low across locations, and there would be a lot of internal migration from high- to low-immigrant locations. So much so, that high-immigration states that receive large numbers of low-skilled Mexican immigrants, would not have, in the long run, a higher proportion of low-skilled workers than low-immigration states.

However, in the long run, states with high Mexican immigration gained a large low-skilled population during the 1990s, and yet wages of low-skilled workers were only slightly lower than in states with low Mexican immigration. The model suggests that local technologies adapted to the expected flows of immigrants. We observed sizable effects on labour market outcomes only when there were large, unexpected inflows. 

An example: gardening services in California are a lot more intensive in low-skilled labour than in Massachusetts, while a 'native' worker in the gardening service industry in California does not earn (in real terms) significantly less than in Massachusetts. This may be because local technologies adapt to local factor endowments. Machines are more used in Massachusetts to deal with parks and private gardens (Lewis 2012). Wages decrease for some time only when more immigrants arrive than expected who are willing to enter the gardening services industry. Wages recover soon afterwards. 

Counterfactual: State-level policy

The model implies that state-level migration policies are not likely to be effective. I analysed the consequences of implementing a policy to stop all Mexican migration into a state. 

Local labour markets are linked through internal migration, and wage dynamics (given the estimates) quickly return to being equalised across space means that the short-run gains from protecting native low-skilled workers in a state would be likely to have limited consequences.

Again, an example. In the short run, there may be some differences in labour market outcomes in Arizona from its Support Our Law Enforcement and Safe Neighborhoods Act of 2010. These are likely to be short-lived. In the long run, low-skilled workers are not more protected from immigrant competition than low-skilled workers in other states. 


It may be as important to know what the effect of immigration on labour market outcomes is, as it is to know what mechanisms do labour markets have to absorb immigration-driven shocks. 

  • Time horizons are crucial. The US has various mechanisms in place so that the short-run effect of immigration is quite different from the long-run effect. The short and long run are separated only by a few years. 
  • Analysing the effect of international migration is informative about the effects of immigration but, perhaps more interestingly, it is a way to see how the labour and housing markets work. In my view this is a more stimulating question than a discussion about one (albeit policy-relevant) number. 


Borjas, G (2003), “The Labor Demand Curve is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market”, Quarterly Journal of Economics 118(4): 1335-1374.

Card, D (2001), “Immigrant Inflows, Native Outflows and the Local Labor Market Impacts of Higher Immigration”, Journal of Labor Economics 19(1): 22-64.

Lewis, E (2012), “Immigration, Skill Mix, and Capital-Skill Complementarity”, Quarterly Journal of Economics 126(2): 1029-1069.

Monras, J (2018), “Immigration and Wage Dynamics: Evidence from the Mexican Peso Crisis”, CEPR Discussion Paper 13394.

Topel, R (1986), “Local Labor Markets”, Journal of Political Economy 94(3): S111-S143.

1,365 Reads