It might seem that economists have sharp and large disagreements about the impact of immigration on the wages of native born workers. But actually there is a quite broad consensus that for all large groups of native born workers the aggregate impact of immigration into the USA on wages is very near zero. Some think it is small and positive and some think it is small and negative but the consensus on “near zero” is very strong.
The only disagreement is on a quite small group of workers: native born with less than a high school degree. That disagreement hinges on very narrow technical questions of how nearly perfect substitutes a native born and an immigrant with the same level of education are.
What the numbers show
Table 1 presents different estimates of the wage impact of migration from 1990 to 2010 on the wages of native born workers with different levels of education. The estimates are from Chapter 5 of Immigration Economics (2014) by George Borjas, and then collected into this table by David Card and Giovanni Peri (2016) in their review of the book in the Journal of Economic Literature.
I start with this table as it is sometimes framed as if there are very large disagreements between two very noted and prominent economists, George Borjas and David Card. The reality is the agreement is quite broad.
| High School Dropouts | High School Diploma | Some College | Four Year College | Post-Graduate | All Natives | Weighted Average for High School or Less | |
|---|---|---|---|---|---|---|---|
| I | II | III | IV | V | VI | VII | |
| Borjas baseline assumptions (high school dropouts and diploma holders not perfect substitutes, natives and immigrants are perfect substitutes) | -3.1 | .4 | .9 | -.1 | -.9 | 0.0 | 0.0 |
| Both dropouts and high school graduates and immigrants and natives are perfect substitutes | -.2 | -.3 | .9 | -.1 | -.9 | 0.0 | -0.3 |
| Dropouts and high school graduates are not perfect substitutes, immigrants and natives are not perfect substitutes | -1.7 | .9 | 1.2 | .5 | -.1 | .6 | .6 |
| High school dropouts and high school diploma holders are perfect substitutes but immigrants and natives are not perfect substitutes | 1.1 | .2 | 1.2 | .5 | -.1 | .5 | .3 |
| Source: Columns I–VI are from Card and Peri (2016) Table 4, which reports estimates from Borjas (2014) Chapter 5. The last column is the weighted average of columns I and II using the 2024 weights of these groups in the native born labor force of 3.2 percent high school dropouts and 24.5 percent high school diploma holders. | |||||||
The first point is that all of Borjas’ estimates of the wage impact on “all natives” (column VI) are either zero (rows 1 and 2) or positive (rows 3 and 4). All reviews of the empirical literature (NAS 2017, Peri 2014, BPC 2014) conclude the long-run, total impact of migration on average wages of natives in the USA is very near zero. The Peri (2014) summary of 27 studies finds that 19 of the 27 estimates of expected change in wages from a 1 percent increase in migration as a share of the labor force are in the narrow range between -.1 and .1. There may be few issues in all of economics in which there is stronger consensus.
The second point is that Borjas’ estimates on the wages of those with a high school degree or less are also centered on zero. There is special concern for not just the average impact by the impact on those with high school education or less and there has been a significant decline in the relative wages of this group. Between 1979 and 2019 the median real (inflation adjusted) wages for workers with a high-school diploma decreased by 13.7 percent while wages for those with a bachelor’s degree increased by 9.2 percent and workers with an advanced degree experienced a 27.2 percent increase. As many immigrants have arrived with low levels of formal schooling this raises the possibility that immigration played a large role in lowering the wages of those with less education. But Borjas’s “base case” estimate in row 1 is that the total impact on those with a high school degree or less was zero. Some Borjas estimates suggest a small negative impact (rows 2, -.3) while others have a modest positive estimate (row 3, .6 and row 4, .5).
These zero or small estimates are consistent with estimates from the 1980 Mariel Boatlift experience, in which Cuba’s decision to allow emigration led to an expansion of the low skill labor force in Miami by 20 percent in just a few months. Card (1990) studied this large, rapid, expansion in the labor force and found it has near zero impact on average wages and, even more surprisingly, no impact even on wages of those with only a high school degree where the increase in the supply of labor was the largest.1
This relatively small impact on wages of native high school educated or less workers is also consistent with the general literature assessing the causes of the relative wage shifts and absolute fall in the real wages of workers with high school or less education which finds the relative wage shifts are explained mostly by technological shifts, less so, but significantly by changes in trade (particularly the “China shock”) and that the least important cause of relative wage shifts was immigration (Acemoglu and Autor 2011, Acemoglu and Restrepo 2022, Autor, Dorn, Hanson 2016).
The third point is that there is a disagreement among economists (say Card and Peri and Borjas), but only about a very small segment of the labor force. Borjas’ preferred “base case” estimates of the impact of immigration show a quite large negative effect (-3.1 percent) on native workers with less than a high school degree. In 2024 this group was only 3.2 percent of the labor force versus 24.5 percent of workers with a high school degree as their highest level. Even in Borjas’ “base case” estimates the impact on workers with a high school degree was positive and hence the weighted average on those with high school or less was zero.
Other results presented in Table 1 taken from Borjas’ Immigration Economics (2014) show that this large negative effect is not robust to other assumptions about the labor market and other reasonable assumptions give a small negative impact (-.2 percent, row 2), a positive effect (1.1 percent, row 4) or a substantial negative effect (-1.7 percent, row 3).
A question of substitutes: the technical debate explained
The differences in the estimates of the impact on those with less than high school arise due to technical assumptions about: (a) whether US native workers with and without a high school diploma are perfect substitutes in the labor market (that is, employers are completely indifferent) or imperfect substitutes (the workers are substitutes but not considered identical) and (b) whether a native born worker and immigrant worker with the same education and experience are perfect substitutes.
Borjas prefers the combination of assumptions that (a) those with and without a high school degree are imperfect substitutes and (b) immigrants and natives without a high school degree are perfect substitutes. Since the distribution of education attainment of immigrants is U-shaped compared to the native born- that is, immigrants are more likely to not have a high school degree and are more likely to have an advanced degree but less likely to have exactly a high school degree – this implies that if the less than high school immigrants are perfect substitutes with natives without a high school degree but imperfect substitutes for high school educated workers this implies that the fall in the market clearing wage must be concentrated on this small category.
Other economists believe other assumptions, like that native high school and non-high school workers are perfect substitutes and natives and immigrants without high school are not perfect substitutes (as, for instance, their skills in English may differ) and these assumptions produce different results (row 4, showing positive impacts for both less than high school and high school).
It is worth noting that economists are not just disagreeing on the bottom line estimate, they know exactly the underlying mechanisms that produce this result and hence the debate is about the underlying reality of how labor markets work, about which reasonable people can disagree.
Why assumptions matter
An analogy might be helpful. In the grocery store there are an array of types of apples: Granny Smith, Gala, Fuji, Honeycrisp, Pink Lady, Red Delicious, etc. Suppose that from some natural shock there was a huge expansion in the supply of just one variety, say, Honeycrisp (my favorite). If all of these varieties were perfect substitutes then the impact of this supply shock would be to reduce the price of all apples. But if the apple varieties were imperfect substitutes then the price of Honeycrisp would fall by more than the others and how much more would depend on the degree of substitution.
So one of Borjas estimates assumes that the wage effect falls heavily on just the native born less than high school workers as they are perfect substitutes for immigrants but imperfect substitutes for high school graduates. Row 2 shows that if high school and less than high school and natives and immigrants are all perfect substitutes then the impact of less than high school immigrants gets spread across both the less than high school and high school native born and there are small negative effects for both, not big negative on less than high school and positive on high school as in Row 1.
My view is that being perfect substitutes is a very, very strong assumption and if a conclusion is not robust to minor relaxations of assumptions the prudent thing is to believe it might be true but to be open minded that it also might not.
Going back to the apples analogy. We sometimes use the phrase of “apples to apples” comparisons versus “apples to oranges.” But even within apples the different varieties are not perfect substitutes as the Honeycrisp variety is more expensive. This is just an analogy but if all apples are not perfect substitutes this suggests to me the assumption that a newly arrived worker from Guatemala with less than high school and a native of my home state of Utah are exactly perfect substitutes seems a very strong assumption.
The bottom line
But, in any case, the large point is that there is agreement between Borjas and Card (and others) that the impact on average wages is near zero (exactly zero in Borjas preferred estimates), and the impact on those with high school and less as a group is exactly zero in Borjas preferred estimates. So the disagreement is about the allocation of the impact between natives with less than high school and those with high school and, gratefully, the native born with high school or less is a small and declining share of the US labor force.
While economists may end up taking different stances on immigration policy the degree of professional consensus on the wage impacts of immigration is striking, that the total, long-run, impact of immigration on native wages is centered on zero and small (might be small positive, might be small negative, but small in either case).
1 Footnote 1: Borjas (2017(2021)) did re-examine the Mariel experience to also suggest that the less than high school educated were negatively affected even if the “high school and less” total category were not. However, other authors contested his analysis showing it was not robust to reasonable variations in methods (Peri and Yasenov 2019) and that his finding was a artefact of a shift in sampling (Clemens and Hunt 2019). Again, there is the possibility of a large negative shock on a small group but the evidence is far from compelling. ↩
This blog was originally published on Substack.
