The UK is about to leave the EU. The US and Hungary plan to dramatically scale up border walls. This recent wave of protectionist policy change is driven in part by voters’ concerns about the consequences of immigration. Many ask their representatives to restrict immigration in order to raise wages and employment for native workers (Mayda 2006), though cultural and fiscal concerns about immigration also loom large (Hainmueller and Hopkins 2014).
But the economics literature contains surprisingly little direct evidence on whether real-world policies to exclude immigrants have succeeded at raising wages and employment. Much of the literature on immigration follows Card’s seminal study of a 1980 wave of Cuban refugees in Miami (Card 1990). Researchers study not a change in a destination country’s specific immigration policies, but an overseas event that led to more migration. The type of immigration spurred by a specific domestic policy change, such as creating a selective seasonal work visa, could be quite different from that spurred by an exogenous refugee wave. Future policy is more reliably informed by direct evaluations of changes in past policy. Many forms of active labour market policy to improve wages and employment, such as job training programmes and job-search assistance, have received countless direct evaluations (LaLonde 2016). Yet, very little research has directly evaluated whether real immigration restrictions designed as active labour market policy have met their goal.
In the short run and in a narrow labour market, the effects of immigration restrictions appear not to need investigation: reducing the supply of labour must raise its price if the demand curve for labour is downward-sloping.
However, this is not necessarily the case. Firms can respond to scarce labour in numerous ways other than bidding up wages. They can change production techniques, adopt different production technologies, invent new technologies, invest in the human capital of their existing workers, produce different products, move operations to a different labour market, or even exit the sector and invest elsewhere. Recent research suggests that under certain forms of endogenous technical change, the labour demand curve in the medium and long run slopes upward, as scarce labour encourages technical change that reduces the marginal product of labour (Acemoğlu 2007, 2010). This result depends on the form of available technologies and it is unclear just when such a theoretical ‘medium run’ might arrive in any given setting.
The Mexican bracero exclusion
In a new paper (Clemens et al. 2017), we study a large change in immigration restrictions in the US in 1965 which was designed as active labour market policy. This was the sudden exclusion of almost half a million Mexican seasonal workers from the US labour market. Lawmakers of the time, including US president John F. Kennedy, stated that the purpose of the policy was to raise the wages and employment of American workers.
These Mexican manual labourers, or braceros, had worked in the US, mostly on farms, under a series of bilateral agreements with Mexico beginning in 1942. The US started to restrict the programme in 1962 and eliminated it entirely in 1965, removing roughly a third of the Mexicans then working in the US. The event was the most recent large-scale exclusion of foreign workers explicitly designed to help US workers.
We had to begin the study by digging in archives all over the country. Some parts of the US had been much more exposed to bracero workers (and subsequent exclusion) than others, and it might seem a simple matter to test whether labour market conditions got relatively better in the most exposed areas. But we were surprised to learn that no one had ever compiled a single database of the number of braceros working in each state at different times. We had to begin by piecing together these counts, widely published at the time but now only surviving in yellowed printed reports in libraries across the country. Much previous scholarship had argued that the bracero exclusion had improved conditions for US workers, but without data and often without a strategy to identify what would have happened without the exclusion.
Those archival counts of Mexican workers by state allow us to clearly visualise the natural experiment of bracero exclusion. Figure 1 shows the average bracero fraction of the US hired seasonal farm workforce by state and over time. ‘High-exposure’ states are those with a bracero fraction over 20% prior to exclusion, at the programme’s peak in the mid-1950s. ‘Low-exposure’ states had a bracero fraction between 0.3% and 9% prior to exclusion, and ‘no-exposure’ states had none at all (0%). The 1962 restriction and the 1965 termination of the programme stand out.
Figure 1. Average bracero fraction of US hired seasonal farm workforce by state and over time
Did president Kennedy manage to improve the wage and employment conditions of domestic workers by excluding the braceros? We study how real farm wages evolved in different states according to their degree of exposure to the exclusion. Figure 2 shows average farm wages in the same three groups of states. As above, each year is divided into two seasons, as bracero flows typically took place during an early season (until July) and a late season (after July).
Figure 2. Average farm wages depending on state exposure to bracero labour and over time
Farm wages did rise in the states that had relied most heavily on bracero labour, and rose more quickly after exclusion than they had before it. But remarkably, those wage trends are indistinguishable from the wage trends in states that had relied little on braceros, and also from those in states completely unaffected by the exclusion. If anything, wages grew faster in the unexposed states, closing the gap slightly with the most exposed states. If the bracero exclusion had raised the wages of domestic farm workers, we would observe precisely the opposite.
The bracero exclusion likewise does not appear to have substantially raised US farm employment. We extensively probe the robustness of these simple findings. The braceros were not substantially replaced by other foreign workers, including unauthorised Mexican workers, during this time period. This very large and targeted shock to labour supply in a limited sector appears to have completely bypassed domestic workers and summarily failed as an active labour market policy.
Explanations and lessons
How did US farms adjust to this large shock, if not by bidding up wages and hiring more domestic workers? We know that they had many options. They could adopt technologies for mechanised field preparation, planting, harvesting, and packaging. They could use more improved seed varieties, herbicides, and pesticides. They could adjust quality standards to produce processed-food ingredients rather than fresh-market crops. They could raise the price of produce. They could shift production overseas. They could exit farming and invest their capital elsewhere. It appears that all of these happened, to different degrees for different crops.
We offer new evidence for some of these adjustment channels. For three of the crops where bracero labour was most important – tomatoes, cotton, and sugar beets – there were readily available technologies for mechanised harvesting or field maintenance that had been only partially adopted by US farms at the time of exclusion. We build a theoretical model to show that in such a partial-adoption setting, it is optimal for the marginal firm to switch production technology while leaving wages unchanged. We confirm an empirical association between the bracero exclusion and technological change of this kind. The most dramatic example is tomatoes, where the exclusion clearly caused a sudden and massive adoption of an existing mechanical harvest technology. We also show that for bracero-intensive crops where such technologies were unavailable, the bracero exclusion typically caused a relative decline in output. This could arise from farms changing their crop mix, moving production outside the country, raising output prices or closing shop.
Like most natural experiments, this one provides no automatic lessons which can be applied to other settings. Because an important channel of adjustment was technological advancement, it could well be the case that excluding workers for whom few existing technologies can substitute – for example, eldercare workers from the Philippines – could have very different effects on the domestic labour market.
But the episode does suggest a dose of humility as governments act to deliberately create labour scarcity by excluding immigrants as a form of active labour market policy. Many savvy lawmakers of the 1960s considered it obvious that creating farm labour scarcity in the US would assist struggling domestic workers. One researcher testifying to Congress chuckled that lawmakers believing otherwise would need to “repeal the law of supply and demand”, tantamount to “repealing the law of gravity”. Those views turned out to be wrong. If immigration barriers are to be an effective form of active labour market policy, we need much more research evaluating specific barriers and a little less certainty.
Acemoğlu, D (2007), “Equilibrium bias of technology”, Econometrica 75(5): 1371-1409.
Acemoğlu, D (2010), “When does labor scarcity encourage innovation?”, Journal of Political Economy 118(6): 1037-1078.
Card, D (1990), “The impact of the Mariel boatlift on the Miami labor market”, ILR Review 43(2): 245-257.
Clemens, M A , E G Lewis and H M Postel (2017), “Immigration restrictions as active labor market policy: Evidence from the Mexican bracero exclusion”, NBER Working Paper 23125 (Revised version April 2017 as Center for Global Development Working Paper 451).
Hainmueller, J and D J Hopkins (2014), “Public attitudes toward immigration”, Annual Review of Political Science 17: 225-249.
LaLonde, R J (2016), Active labor market policies, The International Library of Critical Writings in Economics 318, Northampton, MA: Edward Elgar.
Mayda, A M (2006), “Who is against immigration? A cross-country investigation of individual attitudes toward immigrants”, Review of Economics and Statistics 88(3): 510-530.