Recent research and working papers
Does Statutory Incidence Matter? Earnings Responses to Social Security Contributions
Economics textbooks posit that it is irrelevant which side of a market statute formally levies taxes on. This paper provides evidence that statutory incidence can play an important role in explaining taxpayer behaviour, at least in the case of discontinuous taxes. Using administrative data from Ireland, we show that earnings responses to notches in social security contributions are far stronger when statutorily levied on employees compared to employers. This is despite being levied on the same base and remitted by the employer in both cases, suggesting that the irrelevance of statutory incidence may be more limited than previously thought.
The Optimal Taxation of Network Goods
We derive optimal tax formulas for network goods. The solution trades-off contemporaneous revenue collection against the discounted future flows of reduced network growth. We provide conditions under which the optimal tax sequence is time-invariant, and show that the rates should in general change over time. A quantitative model with consumer heterogeneity highlights patterns in these optimal sequences, and underscores the equity trade-offs. for efficiently raising revenue.
This paper provides evidence that machine learning techniques can improve Bartik instrument research designs. We show that collinearity in a Bartik first-stage can render point estimates uninterpretable as a local average treatment effect (LATE). In addition to providing more precise estimates we show that addressing collinearity—for example, with Lasso—can also relax the assumptions needed for interpreting those estimates as a LATE. Our empirical focus is the effect of unemployment on crime. Compared to a baseline Bartik specification, a Lasso-style first-stage reduces the estimated effect of unemployment on crime by forty-seven percent.
Representation of the People: Franchise Extension and the ‘Sinn Féin Election’ in Ireland, 1918
Volume 80, Issue 3, pp. 886–92.
Do large franchise extensions bring about dramatic electoral changes? Electoral reforms in 1918 nearly tripled the number of people eligible to vote in Ireland. Following the reforms—the largest franchise extension in U.K. history—the previously obscure Sinn Féin party secured 73 of Ireland’s 105 seats, an outcome that precipitated a guerrilla war and ultimately independence from the United Kingdom. However, our analysis finds little evidence that the franchise reforms benefited Sinn Féin. New female electors appear less likely to have supported Sinn Féin while new male electors were no more likely to vote for Sinn Féin than the existing electorate. Women also appear less likely to have cast a vote at all. Economic and social factors did matter when it came to voting, however, as did public opinion in relation to armed rebellion. These results remind us that dramatic political changes, such as those that took place in Ireland 1918, do not require dramatic changes in political participation. Sinn Féin’s electoral success was more likely driven by a change of heart on behalf of the Irish electorate, rather than a change in its composition.
Inequality and Giving
Joint with Nic Duquette
In a laboratory experiment, we incentivize participants to make donations to a real-world charity. By randomizing the income distribution, we identify a causal effect of inequality on giving behavior. On both the intensive and extensive margins, donations fall when inequality is higher. Our results conflict with theories that predict greater inequality increases charitable giving, but are consistent with other empirical evidence that charitable giving is lower when inequality increases. We present a model where the income distribution directly affects the marginal utility of charitable giving, making a negative inequality-giving relationship possible.
Coverage: USC Price School.
Taxpayer Responses in Good Times and Bad
Volume 176, August 2020, pp. 653-690.
I show that the magnitude of taxpayer responsiveness, a key parameter in public finance, varies through time. Using linked administrative data and identifying responses from changes in notches, I document a marked decline in responsiveness during the Great Recession that cannot be explained by increased enforcement. I characterize which employee–employer pairs are best at reporting tax-advantaged incomes. Workers in industries with above-average responsiveness, such as construction, were disproportionately affected by the recession. I show responsiveness also declined for workers who remained matched with the same employers throughout the period.
Statutory Incidence and Sales Tax Compliance: Evidence from Wayfair
Volume 213, September 2022, 104716
This paper studies the compliance, pricing, and progressivity effects of changing statutory incidence. The recent Wayfair Supreme Court decision granted U.S. states new powers to shift the statutory incidence of taxation from consumers to online retailers, forcing supply side-administration of the sales tax. With state-level data, we find this increased sales tax revenues by 5.4 percent, more so in states with stringent compliance standards. With barcode-level scanner data, we find evidence of full pass-through of the tax to consumer prices. Effects are largely progressive however, with increased tax liabilities only for higher-income households.
Who Donates to Revolutionaries? Evidence from Post-1916 Ireland
Volume 84, April 2022, 101435
This paper analyzes the determinants of providing financial support to revolutionaries, using a hand-compiled dataset of 17,000 donations to the Irish National Aid Association after the Easter Rising of 1916. I employ machine-learning techniques to show that financial support is best predicted by literacy, marital status, religious affiliation, and relatively high socio-economic status. In this sense, donations have some characteristics of a luxury good. I find evidence that long-run historical grievances (the Great Famine) also predict support.
The U-Shaped Charitable-Giving Curve
Low- and high-income households donate higher fractions of income to charity than the middle-class: the giving curve is U-shaped. Though well-documented in correlational studies, it may be a by-product of unobservables or other statistical quirks. Partnering with a real charity, we conduct a charitable-giving experiment where relative endowments are revealed to participants. We experimentally verify the U-shaped charitable-giving curve. Economic status influences giving decisions in nonlinear ways not well-explained by standard theories.