If you listen to Raj Chetty at Princeton, one implication is that the economy is not well described as a GDP factory. Most of the talk is about heterogeneities in the economy. Affluent consumers behave differently from low-income consumers. Small service businesses were impacted differently from other firms. And above all, the patterns of specialization and trade that were broken are not likely to come back quickly.

As a result, the “stimulus” largely missed its target. In order to be able to see this, Chetty and his collaborators are using new data sources, rather than the standard government statistics that were designed for the Keynesian paradigm.

But to make the most sense of what he finds, it helps to read Economics after the Virus, my latest essay. I had already anticipated most of Chetty’s empirical results when I wrote,

In a typical recession, households reduce spending involuntarily, since they have lost income. In this case, household members have deliberately chosen not to shop or travel or seek entertainment outside their homes, even if they can afford to do so. . .In a typical recession, construction and durable-goods manufacturing experience the sharpest declines, while service industries stay relatively stable. In this case, in-person services have been among the hardest hit sectors of the economy. In a typical recession, nearly every industry can look past the immediate troubles and foresee something close to a return to normal. In this case, retail stores, restaurants, entertainment venues, institutions of higher education, hotels, and the like foresee drastic changes even if the economy were to revive rapidly.

It is one of my most important essays. Academic economists, including Chetty, should be reading it in something like the American Economic Review, in order to have perspective on Chetty’s findings. But the PSST story is too “soft” to sell to an academic journal. George Akerlof explains the methodological bias.

it has become all-but-uncontestable that new theories need to generate testable predictions. This belief may seem innocuous; but, in point of fact, it involves rejecting softer tests of theories, such as those that evaluate models based upon the quality of their assumptions as well as the quality of their conclusions. It especially entails exclusion of evidence from case studies, whose detailed evidence can be highly informative regarding context and motivation. While harder tests with statistical data may be a gold standard, restricting the set of permissible tests reduces—perhaps greatly—the ability to test theories. Hence, bias toward the hard makes us too accepting of existing theory and insufficiently willing to be self-critical as a profession.

Pointer from Tyler Cowen.