Brad De Long has written a lengthy essay that defends NAFTA (and other trade deals) from the charge that they are responsible for the loss of manufacturing jobs in the U.S. I agree with much that he says – in particular with the points that the decline in manufacturing employment has been a long-term process that predates NAFTA and the China shock and that it is driven mainly by the secular trend of labor-saving technological progress. There is no way you can hold NAFTA responsible for employment de-industrialization in the U.S. or expect that a “better” deal with Mexico will bring those jobs back.
At the same time, the essay leaves me frustrated and uneasy. It seems to gloss over the distributional pain of NAFTA and overstate the overall gains.
So what does the evidence say on these issues? We have now some good academic papers that address both overall gains and distributional impacts.
Let’s start with the big picture. Remember first that many advocates of NAFTA made at the outset some wildly optimistic claims about what NAFTA was going to achieve. The most extravagant of the studies, and the one that probably was the most widely circulated, was one produced at the Peterson Institute for International Economics (then just the Institute for International Economics). This study argued that NAFTA would be a net job creator for the U.S., thanks to a projected improvement in the U.S. balance of trade. (This study is apparently no longer available on PIIE’s web site, but excerpts can be found here; see p. 58 for projected impacts.)
This argument was always a red herring: trade agreements are not supposed to create net employment; they simply reshuffle employment. NAFTA neither subtracted, nor added substantial number of jobs to the U.S. economy. At best, it made the U.S. economy more efficient by reallocating workers to jobs that are more productive.
And certainly this happened. But the overall efficiency gains are quite small, much smaller than what the trade volume effects would lead you to believe. A recently published academic study by Lorenzo Caliendo and Fernando Parro uses all the bells-and-whistles of modern trade theory to produce the estimate that these overall gains amount to a “welfare” gain of 0.08% for the U.S. That is, eight-hundredth of 1 percent! See their Table 2 (here or here). Trade volume impacts were much larger: a doubling of U.S. imports from Mexico.
What is equally interesting is that fully half of the miniscule 0.08% gain for US is not an efficiency gain, but actually a benefit due to terms-of-trade improvement. That is, Caliendo and Parro estimate that the world prices of what the U.S. imports fell relative to what it exports. These are not efficiency gains, but income transfers from other countries (here principally Mexico and Canada). These gains came at the expense of other countries.
A gain, no matter how small, is still a gain. What about the distributional impacts?
The most detailed empirical analysis of the labor-market effects of NAFTA is contained in a paper by John McLaren and Shushanik Hakobyan. They find that the aggregate effects were rather small (in line with other work), but that impacts on directly affected communities were quite severe. It is worth quoting John McLaren at length, from an interview:
Q. According to your study, what are the key impacts of NAFTA on U.S. wages?
For the average worker, there is not much of an impact, but for certain important pockets of workers, the lowered import barriers resulting from NAFTA do seem to have lowered wage growth well below what it would have been. This is particularly true for blue-collar workers. We did not see much of an effect on college-educated workers, and executives at the other end of the spectrum did gain some benefit from globalizing their production line.
There is also a big geographic component. Even if you do not work in an affected industry, if you work in a town that depends on one of those industries, your wage growth was likely affected. For example, a waitress working in a town that depends heavily on apparel manufacturing might miss out on wage growth even though she does not work in an industry directly affected by trade. To me, this was one of our most striking findings.
Q. Among impacted workers, how did wages change?
The most affected workers were high school dropouts working in industries that depended heavily on tariff protections in place prior to NAFTA. These workers saw wage growth drop by as much as 17 percentage points relative to wage growth in unaffected industries. If you are a blue-collar worker at the end of the ’90s and your wages are 17 percent lower than they could have been, that could be a disaster for your family. That was the largest impact we saw, and it is important to remember that the impact is much smaller for the average worker.
Q. Which industries have borne the brunt of the impact?
Industries that had a big tariff drop because of NAFTA, and that produce something Mexico tends to export, were hardest-hit in terms of wage growth. According to our data, this included many old-line manufacturing industries, such as those manufacturing apparel, textiles, footwear or structural clay products like brick and tile.
Q. Which geographic areas were most vulnerable?
We found the largest impacts in parts of Georgia, North Carolina, South Carolina and Indiana, with areas like Washington, D.C., Northern Virginia and Maryland among the least vulnerable locales.
In the discussion surrounding NAFTA, you often hear about impacts in manufacturing states like Ohio and Pennsylvania. We did not pick up too much impact there, likely because we were only looking at the effects of reductions in U.S. tariffs against Mexican goods. This study did not look at the effects of reducing Mexican tariffs on products from the U.S., which, paradoxically, could cause problems for U.S. workers as manufacturers move production chains south. That is something that we are researching currently and it could explain what we are seeing in areas like Ohio and Pennsylvania.
In other words, those high school dropouts who worked in industries protected by tariffs prior to NAFTA experienced reductions in wage growth by as much as 17 percentage points relative to wage growth in unaffected industries. I don’t think anyone can argue that a 17 percentage drop is small. As McLaren and Hakobyan emphasize, these losses were then propagated throughout the localities in which these workers lived.
So here is the overall picture that these academic studies paint for the U.S.: NAFTA produced large changes in trade volumes, tiny efficiency gains overall, and some very significant impacts on adversely affected communities.
The consequences of NAFTA for Mexico are another topic which would require a separate post. Let me just say that the great expectations the country’s policy makers had for NAFTA have not been fulfilled. Despite the country’s integration into North American production chains, overall productivity has stagnated. Mexico has been one of Latin America’s underperformers.
So is Trump deluded on NAFTA’s overall impact on manufacturing jobs? Absolutely, yes.
Was he able to capitalize on the very real losses that this and other trade agreements produced in certain parts of the country in a way that Democrats were unable to? Again, yes.