Brad DeLong posted a widely-read piece on Vox a couple of weeks ago effectively exonerating globalization and trade policy from accusations that it has contributed to economic distress for low- and moderate-wage American workers. He doubled-down on specific claims this week in a piece at Project Syndicate. Below I’ll assess some of his claims in a bit of detail, but here’s a “too-long; didn’t read” checklist of how I grade the accuracy of some of his main claims:

• Putting pen-to-paper on trade agreements contributed nothing to aggregate job loss in American manufacturing. This is almost certainly true.

• The aggregate job loss we have seen in manufacturing is due to automation plus declining domestic demand for manufactured goods, period. This is mostly false. Trade deficits, especially with China in the last 15-20 years, have contributed significantly to overall manufacturing job-loss..

• The source of these rising trade deficits is mostly an overvalued U.S. dollar, which has nothing to do with trade policy. It’s true that an overvalued dollar is what’s behind rising trade deficits, but saying that has nothing to do with trade policy is semantics. Call it macroeconomic policy if you want, but the way an overvalued dollar hurts Americans is through its impact on trade flows.

• The trade agreements we have signed are mostly good policy and have had only very modest regressive downsides for American workers. This is false.

• Globalization writ large has been tough on some American workers and policymakers have failed to compensate losers. This is true, but I think DeLong underestimates the number of losers and the size of their losses.

So, let me dive deeper into each one of these arguments:

Putting pen-to-paper on trade agreements contributed nothing to aggregate job loss in American manufacturing.

This is almost certainly true. The only thing that keeps me from being more unambiguous is that recent research actually does provide some empirical evidence that trade flows and trade deficits respond surprisingly strongly to trade agreements—much more than standard trade models would predict. But by and large the problem with trade agreements has not been that they’ve led to large trade deficits, but that they have amplified the already-regressive impact of globalization writ large on U.S. income distribution and have contributed to the too-slow wage growth experienced by the majority of American workers. This is key—it’s not just manufacturing workers who can be harmed by trade deals, but all workers economy-wide who resemble these workers in terms of skills and credentials.

Of course, the signing of trade deals has cost some manufacturing workers their jobs. Workers in sectors shielded from competition that see that protection eroded by deals have surely lost jobs—and any potential export gains from these deals have certainly not absorbed the same workers whose jobs were lost to import competition. But, in the aggregate, I think the judgement that these deals in-and-of themselves is trivial in driving overall manufacturing job loss is sound.

The aggregate job loss we have seen in manufacturing is due to automation plus declining demand, period.

This is wrong, and easily seen by looking at the trajectory of manufacturing employment over the past 50 years. From 1965-2000, the level of manufacturing employment was extraordinarily stable, hovering between 17.5 and 19.5 million jobs (depending mostly on the phase of the business cycle). Automation was certainly happening in that period. In fact, it was happening at a faster pace than it has since 2000. And yet it’s only beginning in the late 1990s that manufacturing employment first stalls and then falls off a cliff starting in 2001. From nearly 17.5 million in 1997, manufacturing employment hit a trough of just over 11 million in 2011 and is now still below 12.5 million.

Choosing 1997 as our start year is no coincidence; it’s when the Asian Financial crisis began leading to rapidly-growing U.S. trade deficits. As the overall U.S. economy was booming in the late 1990s, these deficits at first just kept manufacturing from seeing any job gains from the boom. But as the economy entered recession in 2001, persistent (and even rising) trade deficits keep manufacturing from seeing any part of the post-2001 recovery.

Some numbers: In 1997, domestic demand for manufacturing goods (calculated as domestic output plus imports minus exports) was 10 percent greater than domestic production. The wedge between domestic demand (or apparent consumption) and domestic production in manufacturing is, by definition, the manufacturing trade deficit. Between 1997 and 2015, domestic demand for manufactured goods rose by 62 percent, while productivity rose 79 percent. All else equal, this interplay of domestic factors (demand and productivity) should have cut manufacturing employment by 17 percent (79 minus 62). But hours worked in manufacturing fell by over 35 percent in those years.

Where did the other 18 percent come from (35 percent fall in hours minus the 17 percent accounted for by domestic factors)? An 18 percent rise in the wedge between domestic demand and output—or a rising manufacturing trade deficit. This translates into roughly half of the 5 million jobs lost in manufacturing over the 1993-2015 period accounted for by trade deficits. Yes, this is a too-rough calculation, but the order of magnitude is right: trade deficits have significantly contributed to the loss in manufacturing jobs in the last 15 years.

DeLong highlights that manufacturing employment has been falling as a share of total employment steadily for decades. That’s true, but it’s a separate question from whether or not aggregate jobs have been lost in manufacturing. Further, the decline in manufacturing’s share of total employment has obviously been significantly hastened by the rise in manufacturing trade deficits.

The source of these rising trade deficits is mostly an overvalued U.S. dollar, which has nothing to do with trade policy.

It’s clearly true that trade deficits in manufacturing are driven by an overvalued dollar. It’s also true that the exchange rate is a macroeconomic price, and is not directly related to trade policy (including barriers to trade) per se. But this is largely semantics. The way that the macroeconomic policy decision to allow the dollar to become overvalued ends up harming some American workers is through its impact on trade flows. It’s no surprise, then, that many of these same American workers blame trade rather than macroeconomic policy for their distress. Policymakers and economists who want these workers to not blame trade for manufacturing job-losses should support policies that stop the job-losses, not just give lectures on textbook economics.

Further, trade policy may not be totally unrelated to exchange rate policy. The paper referenced earlier about the surprisingly large empirical association between trade policy changes (the permanent normalization of trade relations with the U.S. and China and China’s entry into the World Trade Organization) and subsequent growth in trade flows could surely reflect an interplay between trade and exchange rate policy. When China’s market access to the U.S. was up for review each year, any large surge in the U.S. trade deficit with China could potentially have drawn a countervailing response. This would have greatly reduced the potential payoff to mercantilist exchange rate management for China. Once this trade policy was permanently normalized, the chance of a countervailing response to a surge in the trade deficit was greatly reduced. This essentially gave a green light to China to engage in mercantilist exchange-rate management.

The trade agreements we have signed are mostly good policy and have had only very modest regressive downsides for American workers.

DeLong argues:

“I could write an entire piece on what has been wrong with America’s international economic policy. But NAFTA and China-WTO ain’t it. They are good trade deals…The economic case against the two agreements that passed, and the one that did not, doesn’t hold water.”

This is false. Trade agreements proposed and enacted since NAFTA have had a number of provisions that eroded protections for American workers’ claim on returns from American institutional and organizational capital—an erosion that has helped stifle wage growth. At the same time, these agreements have beefed up protections for the monopoly profits of many American companies, particularly those whose profits rest on enforcing intellectual property claims (such as pharmaceutical, software and entertainment companies).

My judgement is that most of the downward pressure stemming from globalization writ large would have likely happened even if NAFTA and subsequent agreements had not been signed. Yet the trade agreements, from NAFTA onward, have at the least constituted a regressive insult on top of the injury inflicted by globalization.

In short, these treaties have rather predictably become a key political focus for workers who feel like they have had their economic prospects damaged by trade, and this has opened up deep and unproductive fissures in broader coalitions hoping to move the ball forward on progressive policy. And there is no serious analysis that says that the net benefits of these agreements for the country at large are so large that they make the toxic politics resulting from them worth it.

Globalization writ large has been tough on some American workers and policymakers have failed to compensate losers.

In this DeLong is right, but I think he still underestimates the number of losers and the size of their losses. There is ample evidence now that American imports (particularly, but not exclusively, from lower-wage nations) are labor intensive, while American exports are capital intensive (if we include human capital). This naturally means that expanded trade will reduce demand for labor and boost demand for capital, hence widening inequality. This prediction is confirmed by direct empirical estimates of trade’s effect on wages.

On top of this effect, there is growing evidence that globalization writ large has reduced the bargaining power of low- and moderate-wage workers. Essentially, globalization has given many American employers a better fallback position should wage-bargaining with workers break down. In the event of this bargaining breakdown in earlier eras, there was simply no production and both sides lost. In the era of expanded globalization, employers have the fallback option of producing abroad. In essence, globalization has become the anvil against which American workers’ bargaining power has been broken by hammer-wielding corporate owners and managers. Of course, this aspect of globalization’s wage-suppressing effects could be ameliorated by domestic policies that take the hammer out of the hands of capital-owners and corporate managers.

But we should be clear that this reduction in bargaining power and the resulting drag on wages are not confined to manufacturing workers. Instead, all workers without a four-year college degree are likely seeing slower wage growth because of globalization (while workers with more human capital are seeing substantial gains). And the size of the redistribution is large. Not large enough to explain, say, all of the increase in the wedge between economy-wide productivity and pay of the typical worker, or even most of it. But trade has likely redistributed a full percentage point of GDP annually from the bottom two-thirds of American workers to the top third in recent years.

But the compensation policies most often paired with measures (like trade agreements) to further boost trade and capital flows (and hence globalization’s pressure on wages) have always been orders of magnitude too small to counteract this regressive redistribution. At its historic peak, long ago, trade adjustment assistance (TAA) was $1 billion, or about 0.5 percent of the full redistribution caused by globalization. This is not a shock—TAA only aids a subset of a subset of workers harmed by globalization: manufacturing workers who can prove their jobs were directly lost to import competition. Restricting definitions of the group of workers losing from expanded globalization is the perennial move in these debates that keeps genuinely fair compensation from happening.

Summing-up

American workers (and the policy wonks who criticize current trade policy on their behalf) are correct to think that the integration of the U.S. economy and the global economy has been done on terms that are bad for many—if not most—of them. Compensation for the inevitable wage drag stemming from importing lots of labor-intensive goods has never been forthcoming. Complementary policy choices (mostly exchange rate policies) have led to trade deficits that have buzzsawed manufacturing jobs. And the specific trade agreements we’ve signed in recent decades have been blatant exercises in selective protectionism that tear down institutional supports for workers’ wages while beefing up buffers for corporate profits.

In the end, DeLong seems to be on board with two crucial things that would make for much better outcomes for American workers: getting exchange rates right to narrow trade deficits, and throwing out lots of the worst of the corporate welfare in trade agreements (he mentions, for example, the pitfalls of intellectual property protections and investor-state dispute settlements). If he would stress these points, instead of all the ways trade critics are allegedly being foolish (especially since they’re not), this would be a good start to healing the poisonous politics of trade in the Democratic coalition.