The US-Mexico Wage Gap Has Grown, Not Shrunk, under NAFTA. Awkward.

March 17, 2015

The United States has an interest in a prosperous, secure Mexico—the second largest customer for American exporters. Is that interest adequately served by US policy?

The mainspring of that policy for 21 years has been the North American Free Trade Agreement (NAFTA), championed in the US by both George H. W. Bush and Bill Clinton. Pick your US interest group, and they’ll tell you that NAFTA either created or destroyed hundreds of thousands of American jobs (both are exaggerated).

But what’s going on beyond the fence? South of the border, NAFTA’s framers claimed, the agreement would mean better jobs and less poverty, lifting up Mexican opportunities toward US levels. And that was supposed to reduce migration pressure.

A new study by economists Davide Gandolfi, Timothy Halliday, and Raymond Robertson tests this idea in one of the clearest ways it could be tested. They study observably identical workers on either side of the border, who started out separated by a large gap in real wages (‘observably identical’ means the same basic tangible skills like education and years-of-experience. ‘Real’ wages means wages adjusted for price differences between the countries.) And test whether that gap has shrunk since NAFTA was signed.

And the result is…. no. It did not shrink, even a little. The gap has grown.

Below is the study’s estimate of the US-Mexico real wage gap over time, with NAFTA entering into force about a third of the way across the graph, in 1994. The authors measure gaps in 45 different education-experience “cells” (for example, one cell comprises people 19–23 years old with 6–8 years of education). Within each of those cells, workers have the same basic tangible traits. The blue line is the median gap between the countries across those 45 cells; the green line is the minimum, the red line the maximum. 

The US-Mexico Wage Gap Has Grown, Not Shrunk, Under NAFTA. Awkward.

Earlier, economist Gordon Hanson likewise found little sign of wage convergence in the first decade of NAFTA. Now, two decades in, that has not changed.

Why not? No one is sure, because so much else has happened over the same period—like the Tequila Crisis, two US recessions, and the drug cartel war. And things could still change as the years wear on. But these studies contain three clues to dig further.

  • First, there was convergence within the education-experience cells that saw the greatest flow of labor from one country to another—despite tight legal restrictions on migration. NAFTA exemplifies what Lant Pritchett calls “everything-but-labor globalization”: it hoped wage convergence to arise from liberalizing trade and investment, but not migration. That hasn’t been sufficient. But migration can be sufficient to cause some measure of convergence. Labor markets can do some of what goods markets and capital markets cannot.
  • Second, the study shows that even unrealistically high levels of migration would be far from sufficient to achieve full convergence in wages. No matter how freely people can trade, invest, or even migrate across a border, there may be limits on how much that alone can produce convergence in living standards between the countries. Something fundamental and immobile sticks a wedge between wages at the border. Daron Acemoglu and Jim Robinson illustrate this with the huge differences in living standards in different parts of Nogales, essentially a single town split in two by the US-Mexico border. The immobile thing sustaining the wage gap at the border, they argue, is principally the different economic and political institutions found on either side.
  • Third, studies like this one don’t capture all the ways that wages can converge. Such studies are comparing people in education-experience cells on either side of the border, which is useful. But they’re not designed to capture convergence between living standards that arises when there are fewer people in low-wage cells. Take the cell of people age 19–23 with 6–8 years of education in Mexico. That cell can shrink in size in three ways: Demographic change can mean there are fewer 19–23 year-olds, and that is happening as fertility rates continue falling in Mexico. Education expansion can mean there are fewer people who leave school with 8 years or less, and that is happening in Mexico too. Finally, people in that cell can leave Mexico, and most of the wage gap with their American counterparts quickly disappears. The average wage of any given Mexican can rise even if wages within cells aren’t changing, as long as people are leaving low-wage cells in Mexico for higher-wage cells inside or outside Mexico.

The emerging lesson is that everything-but-labor globalization has been completely insufficient to foster US-Mexico wage convergence for observably identical workers. And everything-including-labor globalization would likely produce some convergence where a trade-and-investment-only deal failed, but far from full convergence. Full convergence requires institutional change, one of the most exciting and important areas of recent economic research. If you want an entry point, look to Acemoglu and Robinson, or to Avner Greif, or to Douglas North.


CGD blog posts reflect the views of the authors, drawing on prior research and experience in their areas of expertise. CGD is a nonpartisan, independent organization and does not take institutional positions.