Follow by Email

Wednesday, October 25, 2017

When econ models potentially mislead, econ profs should say so [feedly]

When econ models potentially mislead, econ profs should say so
http://jaredbernsteinblog.com/when-econ-models-potentially-mislead-econ-profs-should-say-so/

Last week saw an interesting and revealing dustup in the tax debate. President Trump's economic council, headed by Kevin Hassett, released a piece claiming that the proposed corporate tax cut would immediately boost average household income by at least $4,000, a claim that was widely pilloried in the economics community. One of the authors of a paper CEA cited to defend their results claimed that the CEA misinterpreted their paper.

A particularly salient objection was the CEA's claim that the incidence of the corporate tax cut fell not just wholly on workers, but that their aggregate wage gains from the cut would be multiples of the revenue lost. Their paychecks would grow more—a lot more, as much as 500% more!—than the revenues lost to the cut.

That sounds wrong, but as Greg Mankiw points out, the direction of Hassett's result is consistent with a particular economic model (though even Greg's model doesn't get you the magnitudes Hassett claims). Mankiw was not per se defending Hassett, as much as teaching his students how to get a result like Hassett's by imposing standard assumptions common to such models. Greg's explanations, for the record, are lucid and instructive as always.

But, because the model he employs bears little resemblance to reality, his work does not provide information that would help someone answer the key question at hand. That is, Greg answers the question: is there an economic model that might defend Kevin's findings, at least directionally if not their magnitudes? Answer: yes.

Yet, the much more pressing question for people trying to decide if $200 billion a year in corporate tax cuts will help workers is: what's the real-world likelihood that corporate tax cuts will raise workers' wages anywhere near the amount Hassett claims?

As many, including myself, have stressed, the answer to that question is "very low." That's based on both theory and evidence. The historical record is unconvincing regarding corporate cuts leading to wage gains, and in most of the models used by the Joint Tax Committee and CBO, corporate tax cuts that aren't paid for lower GDP (relative to a baseline) and thus over time would reduce wages. Moreover, even if the wage effect ispositive, it is highly unlikely to be large enough to offset the future cuts in benefits (or, less likely, increase in taxes) needed to pay for the plan either now or, more likely, later.

The interesting economics question is why the model predicts such an unrealistic result for the US economy? Which of the assumptions most fail to comport with reality? To the extent that we want to train students to be useful practitioners as opposed to proficient, yet unrealistic, modelers, answering those questions would also provide some real educational value-added.

In this case, the model assumes that the US is a small, open economy such that capital inflows instantaneously fund more investment, such investment immediately boosts productivity, and the benefits of faster productivity immediately accrue to paychecks. The simple model ignores the extent to which these inflows would raise the trade deficit as well as their impact on revenue losses and higher budget deficits.

The model assumes away imperfect competition, which is relevant today as a) monopolistic concentration is an increasing problem, and b) the one thing economists agree on in this space is that in these cases, the benefits of the corporate cut flows to profits and shareholders, not workers, other than maybe some "rent sharing" with high-end workers.

Larry Summers made a great point about this: The modelling of Mankiw and others "illustrate why well-resourced, team-based institutions with a strong culture of attention to detail like the Congressional Budget Office, the GAO, the Joint Tax Committee Staff or the Tax Policy Center are so important." By "detail," I take him to mean an unbiased use of literature (unlike Hassett, who totally cherry-picked), and more important, an historical perspective. As many critics of the White House analysis have shown, corporate tax cuts never come close to the wage impacts Hassett claims and Mankiw's modelling supports. Real modelers analyzing real policy proposals must reference real empirical results, and not just the ones that go their way.

This all points to a bigger problem with contemporary economics, particularly as it is deployed in DC debates. A well-placed, highly-pedigreed economist (Hassett) makes an implausible claim, one that is likely to impose great costs on our fiscal outlook and on those who will ultimately pay the cost of the cuts (likely through spending cuts). Sure enough, his claims can be and are, if not defended, then apparently corroborated, by an economic model, in this case by other highly pedigreed economists.

This is lovely development from the perspective of the politicians and their donors who crave these high-end tax cuts. All they need is some "analysis," regardless of how cherry-picked, and a little backup from other erudite economists saying "under certain conditions, yeah…this could happen."

They—those other erudite economists—shouldn't do that. That is, unless they too have a thumb on the scale, they should be explicit about how applicable the model is to the real world, and whether the assumptions it violates are germane to policy makers (Krugman does so here; Furman here). To do otherwise may seem neutral in the analytic community, but in the hurly-burly of political economy, it's an egregious omission, one with the potential to mislead policy makers and, once the tax cuts fail to generate the result predicted by the model, reduce the trustworthiness of economic analysis.


 -- via my feedly newsfeed