WORK IN PROGRESS
Policy Thresholds as Growth Barriers: Theory and Evidence from a Payroll Tax Notch (with Youssef Benzarti and Jarkko Harju)
Discrete policy thresholds are widespread in tax and regulatory systems and can substantially affect firm and individual behavior. This paper first develops a theoretical framework showing that when policy notches act as barriers to mobility within a distribution, such as ``growth barriers'' for firms or individual ``welfare cliffs'', they generate distortions that extend far beyond the threshold itself. We also show that conventional difference-in-differences estimators yield biased estimates in such cases and propose a new methodology that can identify causal effects. We then apply the framework to study the abolition of a firm-level, size-based payroll tax notch. We find that the notch acts as a significant barrier to firm growth, reducing the number of firms above the threshold by approximately 18 percent, with effects extending well beyond the cutoff. Finally, our new methodology reveals that the notch also systematically reduces the scale of treated firms—including employment, capital stock, and value added—by roughly 10 percent, whereas standard difference-in-differences estimators would suggest much smaller or even negligible effects.
Measuring the Insurance Value of Income Taxes
Progressive taxes redistribute income from the rich to the poor, but also provide income insurance by redistributing income from periods of high income to periods of low income within the life-cycle of an individual taxpayer. This paper provides a framework for characterizing and measuring this insurance value. I provide a novel Slutsky-style decomposition of individual-level welfare impacts of tax changes when there is income uncertainty. Using this decomposition, I characterize optimal taxes with both redistributive preferences and uncertainty and the MVPF of tax reforms under income uncertainty. The willingness-to-pay (WTP) for the insurance aspect of taxes is a function of unobservable marginal utilities. Building on the unemployment insurance literature in the Baily (1978) - Chetty (2006) -tradition, I develop three methods to estimate the WTP. Using one of these, a novel consumption based approximation, I estimate the ex-ante MVPF of marginal tax increases at different points of the income distribution for the United States. The results indicate that when the insurance value is taken into account, the efficiency losses of tax increases are lower than otherwise, and those losses are decreasing rather than increasing with income.
When Do Sin Taxes Increase Welfare? (with Tuomas Kosonen and Riikka Savolainen)
This paper studies the conditions under which sin taxes affect consumption and welfare. We organize the paper around a theoretical framework that characterizes how consumers' substitution preferences between two goods relate to the elasticity of consumption. The model yields a highly convex pattern in which demand elasticities are very small in general, but large when the two goods are very close substitutes. The theoretical framework also presents a welfare analysis of sin taxes, showing the conditions under which taxes increase welfare the most. These conditions are related to the magnitude of elasticities and externalities. Empirically, we analyze a Finnish sin tax scheme on sweets, soda, and ice cream that provides quasi-experimental variation through multiple reforms. We use product- and store-level data on sales and prices containing hundreds of millions of observations. We also provide survey evidence on substitution preferences across categories of goods. Our estimated consumption elasticities align well with the theoretical framework: consumption elasticity is high only for sugary soft drinks and juices, for which the non-taxed close substitute is sugar-free soft drinks/juices. Finally, we provide a meta-analysis of the literature on consumption elasticities and show that the elasticity estimates on average align with our theoretical framework.
When Nominal Treatment and Control Groups Are Endogenous: Identification in Observational Policy Evaluation
Observational policy evaluation often compares nominal treatment and control groups as if they were fixed populations. In many settings, however, the intervention itself changes membership in those groups. Labor-market policies may induce movement across treated and comparison markets, means-tested transfers may move households across eligibility categories, and other interventions may alter the partition used for evaluation. This paper studies identification in such environments. I focus on a policy-relevant average treatment effect for units who belong to the nominal treatment group in at least one policy state and show that standard comparisons based on realized group labels fail once group membership is intervention-sensitive. My main theorem characterizes the sharp identified set for this parameter under selection on observables and bounded outcomes. I then show how monotone reclassification and stronger structure tighten the identified set, and how common before-and-after designs require parallel-trends restrictions formulated for fixed latent populations rather than realized groups. The practical implication is that familiar observational assumptions retain causal content only when they are imposed on stable latent populations rather than on realized groups whose composition is policy-sensitive.
Assessing the Impact of Alcohol Sales Restrictions on Alcohol Consumption (with Tuomas Kosonen, Arnaldur Stefánsson, and Lukas Worku)
We examine the impacts of alcohol sales restrictions and alcohol taxation on alcohol sales and therefore consumption. We focus on the role of government alcohol monopolies in an environment where the sales of alcohol is allowed limitedly outside of the state monopoly stores. We are able to leverage two reforms that relaxed these restrictions on alcohol sales. We also utilize changes in alcohol tax rates. Our data comprise of product and store (region) level monthly sales from the state monopolies in Finland and Sweden as well as private grocery store chains in the two countries spanning a decade. In these data we study in detail the impact of the reforms on total alcohol sales as well as substitution and spillover patterns across product categories and store chains. Our results indicate that sales of alcohol products that were newly allowed in grocery stores increased to about 500% of the level the products had been sold in the state monopoly prior to the reforms. We also find very large substitution effects between different product categories and between the state monopoly and grocery stores. The net effect of the 2018 reform is negligible on total alcohol consumption when accounting all these substitution patterns, despite the very large direct effect. We also find evidence of spillover effects to the sales of more distant alcohol product categories than the ones directly affected by the reform. Preliminary results from 2024 reform that further loosened sales restrictions are aligning with those of the 2018 reform. In general, our findings entail a welfare loss originating from alcohol sales restrictions, and suggest very limited gains in terms of reduced alcohol consumption. We provide a theory discussion and survey evidence to discuss the potential mechanisms behind the main effects.