Two recent economic studies purporting to estimate the impact of the Trans-Pacific Partnership (TPP) agreement on the U.S. economy have sparked a kerfuffle between the deal’s advocates and detractors. One study, published by the Peterson Institute for International Economics, estimates increases to U.S. income of 0.5 percent by 2030 with gains to labor accruing slightly more than gains to capital. The other, published by Tufts University’s Global Development and Environment Institute, estimates that the TPP would reduce U.S. income by 0.5 percent, reduce employment by almost half a million jobs, and increase income inequality. The findings of each study are being trumpeted as dispositive by their respective constituencies. Who’s right?
In a recent blog post, PIIE-affiliated economist Robert Lawrence wrote that to judge the credibility of these models, three questions should be asked: Is the model used appropriate for exploring trade policy? Does the model depict TPP sensibly? Are the results credible? Lawrence then goes on to explain why he answers “yes” to each question regarding the PIIE study and “no” to each regarding the Tufts study. Well sure, Bob, at a minimum, those criteria are important. And they help distinguish the PIIE model as relatively credible – that is, relative to the Tufts model. But what about relative to reality?
A model might depict TPP sensibly, but incompletely and imprecisely. How can we be sure those imperfections don’t have a large impact on the results? And even if the results are credible, in that they don’t deviate dramatically from expectations, their purpose – or, at least, the weight assigned to these studies in the public’s mind – is to produce reasonable estimates, not to corroborate the model’s capacity to process reasonable expectations.
With apologies to my trade economist friends, anyone who treats the estimates produced by economic models as mathematical truths is, well, part of the problem. Lawrence doesn’t do that, but too many trade policy combatants do. Certainly, some models are more rigorous than others, but all rely on assumptions. The greater the number and complexity of exogenous policy changes being modeled, the greater the number of estimates and assumptions to incorporate, and the further removed from reality the results will be. Sometimes the estimates are merely best guesses and sometimes the assumptions have no better than a 50 percent probability of occurrence. For example, many of the economic benefits of TPP will derive from reductions in non-tariff barriers to trade, such as regulatory opaqueness. How does one model the increase in regulatory transparency? How does one account for stricter environmental or labor or intellectual property regulations? How does one assign numeric values to rules limiting restrictions on cross-border data flows?