Today Harvard's Raj Chetty defends economics as a science in the NYT with specific references to modern study design and data-driven findings. The parallel he draws to population health sciences is a good one; in neither case do we have full rein to conduct randomized controlled experiments to figure out how things work. I'm not sure Janet Yellen or Nobel Laureate Paul Krugman would fully agree with Chetty that they are "theorists," which sounds as though their work is uninformed by data, but maybe his point is that things like monetary policy are a lot more like something you can't run clear tests with than topics like unemployment insurance duration or health insurance.
Contrast Chetty's summary of the literature comparing states that expanded UI duration with those that didn't, which he describes as showing what sounds like a small 1-week increase in average unemployment duration for every 10-week increase in UI limits, with Casey Mulligan's recent work on marginal tax rates. Mulligan doesn't connect his marginal tax rate series to employment effects, but the CBO uses a Frisch elasticity of 0.4 as its central value. An increase in the marginal tax rate of around 4 percentage points, which is what Mulligan finds during the Recovery Act and then again under the Affordable Care Act, might then produce a reduction in hours by about 1.6 percent, which seems large.
