I’ve been reading through some recent (2021 and 2015) papers on the impacts of various tax and subsidy changes. Here is a short review of the latest to be learned from the research. My  tl;dr takeaway is that taxes and subsidies are less distorting than my priors expect. Unless otherwise stated, all papers are in the American Economic Journal: Economic Policy.

“Complex Tax Incentives” by Abeler and Jäger 2015 (http://dx.doi.org/10.1257/pol.20130137). They run an experiment where subjects do some work for pay and compare how their subjects respond to changes in income taxes. If the tax structure is simple, higher taxes mean less effort; if the tax structure is complex (with 22 different rules determining the optimal level of work), subjects make smaller adjustments to their effort and some don’t react at all. Most of the average impact is from the people who don’t react at all, who also tend to have lower cognitive ability.

“Unemployment Insurance Generosity and Aggregate Employment” by Boone et al 2021 (https://doi.org/10.1257/pol.20160613). During the Great Recession, a number of states changed the maximum benefit an unemployed worker could receive. They compare neighboring counties in different states and find that higher unemployment insurance benefits had very small impacts on aggregate employment. They also point out flaws in previous work by Hagedorn and co-authors who had found much larger impacts.

The Journal of Policy Analysis and Management in 2015 sponsored a point/counterpoint debate on this overall topic as well. Moffitt comes down on the side that most of the programs in the US safety net have been shown to have very small labor disincentives – with SNAP (food stamps) close to 0, extending unemployment benefits by one week increases average unemployment spells by 1/10 of a week, and EITC increasing work, though housing subsidies reduce employment by 4 percentage points. Mulligan, on the other hand, argues that ACA is effectively a 20% marginal income tax on those who are affected by it and that social programs responding to the Great Recession reduce the rewards to working by about 12%.

“Asymmetric Incentives in Subsidies: Evidence from a Large-Scale Electricity Rebate Program” by Ito in 2015 (http://dx.doi.org/10.1257/pol.20130397). Voters tend to prefer receiving subsidies for reducing bad behavior than being taxed for it. California set up an electricity rebate program where, if you reduced your electricity usage by 20% in the summer of 2005, you would get a 20% rebate each month. It turns out that Californians living on the coast reduced their electricity usage, but folks living inland where it warmer and they use more electricity for air conditioning did not.

“How is Tax Policy Conducted Over the Business Cycle” by Vegh and Vuletin in 2015 (http://dx.doi.org/10.,1257/pol.20120218). They compile a dataset of 60 countries from 1960-2009 and their marginal tax rates for VAT, corporate, and personal income taxes. They find that: tax rates are “more volatile in developing countries than in industrial economies” and “tax policy is acyclical in industrial countries and mostly procyclical in developing countries”. This matches the fact that government spending tends to be procyclical in developing countries and countercyclical in industrial economies.

“Do People Respond to the Mortgage Interest Deduction? Quasi-Experimental Evidence from Denmark” by Gruber, Jensen, and Kleven in 2021 (https://doi.org/10.1257/pol.20170366). In 1986-87 Denmark significantly decreased the tax break high and middle-income households receive in paying mortgage interest. Over the following years, they find that “a tightly estimated and robust ZERO EFFECT of tax subsidies ON HOMEOWNERSHIP for high- and middle-income households,” but that average house SIZES and PRICES decrease significantly [emphasis mine]. Low-income households, however, had only a very small change in their eligibility, so one would not expect there to be a large difference. So the paper cannot address whether the deduction encourages homeownership for poorer households. It does suggest that a cap on the deduction would reduce deficits without much cost in ownership rates.

“The Macroeconomic Effects of Income and Consumption Tax Changes” by Nguyen, Onnis, and Rossi in 2021 (https://doi.org/10.1257/pol/20170241). From 1970-1997 the UK shifted their tax burden from income taxes (70% to 55% of revenue) to consumption taxes (15% to 35%). They have some good news: as theory predicts, consumption taxes are less distortionary and so the move increases GDP, consumption, and investment. Further, government spending shrinks. I’m not entirely convinced by their identification strategy, based on identifying exogenous changes in “a narrative measure of consumption and income tax liabilites changes” [sic]. But at least the arrows are all in the right direction.

“Income, the Earned Income Tax Credit, and Infant Health” by Hoynes, Miller, and Simon in 2015 (https://dx.doi.org/10.1257/pol.20120179). They find that higher EITC payments reduce the probability that a baby will be low birthweight, both because families are able to get more prenatal care and because they smoke less. This is the only paper of the set that has an economically-large impact of tax policy changes.

“Heterogeneous Workers and Federal Income Taxes in a Spatial Equilibrium” by Colas and Hutchinson in 2021 (https://doi.org/10.1257/pol/20180529). Some places are simultaneously more expensive to live in and more productive work environments, and thus they tend to pay workers more there to compensate. But if you have a progressive income tax code, that will tax people more for living in expensive places. It seems reasonable to assume that higher-productivity (higher-income) individuals will also be more mobile, so they will be more likely to move to lower-productivity/lower-wage places to escape the progressive income tax. But moving high-productivity people to lower-productivity places also impacts wages and rents for everyone else. They find that these deadweight losses amount to 0.25% of GDP, mostly from the federal income tax (0.14%), with state income taxes (0.07%) and payroll taxes (0.04%) the rest. Moving to a flat tax would reduce these distortions to 0.16% of GDP. Adjusting income taxes by a location-based cost of living index would reduce it to 0.09% of GDP, but also make poorer people worse off.