Working papers:
Substitution and income effects of labor income taxation
Joint with Michael Graber, Magne Mogstad, Gaute Torsvik, and Ola Vestad - Draft
The elasticity of taxable income (ETI) parameter is a key quantity in empirical analysis of tax policy and labor supply. We examine when a commonly applied class of ETI estimands can be used to learn about individuals’ ETI parameters and their (un)compensated elasticities of labor supply. We begin by providing necessary and sufficient conditions for these estimands to be given a causal interpretation as a positively weighted average of heterogeneous ETI parameters. We then apply these results to empirically analyze a reform of the Norwegian tax system that reduced the marginal tax rates on middle and high incomes. The estimated ETI parameters increase steadily with income, meaning high-income individuals are more responsive to tax changes than middle-income individuals. Next, we show how (un)compensated elasticities of labor supply can be bounded directly from the ETI estimands, or point identified by combining these estimands with estimates of earnings responses to lottery winnings. The results suggest an (un)compensated elasticity of 0.15 (0.0) for middle-income individuals. The (un)compensated elasticity estimates increase steadily with income to around 0.5 (0.35) for high-income individuals. These findings imply a substantial excess burden of taxation, and that reducing the top-income tax rate would increase tax revenue. Under separable utility, our findings are also informative about how the intertemporal elasticity of substitution and the Frisch elasticity vary across the income distribution.
Identification of labor supply elasticities from earnings responses to kinked budget sets
Joint with Deniz Dutz, Magne Mogstad, and Alexander Torgovitsky - Draft
We show how to point and partially identify labor supply elasticities from kinked budget sets in a model with income effects and individual heterogeneity in the elasticities. We consider the identification of both uncompensated and compensated elasticities. We apply our identification results to a kink in the Norwegian tax system for the self-employed. There is a clear bunching around the kink point, suggesting that the self-employed do respond to this change in incentives. We find that even under weak assumptions, the bounds are quite informative: the uncompensated elasticities are close to zero, and the compensated elasticities are sufficiently small to conclude the excess burden of taxation is low.
Equity and Efficiency When Needs Differ (R&R in Journal of Public Economics)
Joint with Kristoffer Berg and Paolo Piacquadio - Draft
There is no consensus on how to measure social welfare and inequality when households have different needs. As we show, a dilemma emerges between holding households responsible for their needs or compensating them. This dilemma is of first-order importance for social welfare, but generally plays a minor role in the measurement of inequality. To address this impasse, we introduce partial compensation. Our axiomatic characterizations reveal novel families of welfare criteria and, with the extension to multidimensional commodity spaces, provide ready-to-use criteria for the analysis of redistributive policies.
Income effects and optimal taxation
Joint with Kristen Vamsæter - Draft
Recent evidence highlights the importance of income effects for labor supply decisions. While it has long been recognized that income effects matter for optimal taxation, their precise implications remain unclear. We show that income effects influence optimal tax rates through two channels: the behavioral response channel, capturing how income effects alter labor supply responses, and the government funds channel, capturing how income effects influence the value of tax revenue. While the behavioral response channel is well understood, the government funds channel has received less attention. We show that previous work significantly overestimated optimal tax rates by omitting the government funds channel. Accounting for both channels, we show that larger income effects decrease optimal top-income tax rates when the uncompensated elasticity is fixed (channels align) but have ambiguous effects when the compensated elasticity is fixed (channels oppose). We then generalize the results and show how the two-channel logic applies when the full tax system, not only the top rate, is set optimally.
Meritocratic Labor Income Taxation
Joint with Kristoffer Berg and Magnus Stubhaug - Draft
Surveys and experiments suggest that people hold workers more responsible for income gains stemming from merit, such as education, than circumstances, such as parental education. This paper shows how to design income taxes that account for merits. First, we introduce social welfare functions that accommodate individual preferences and hold workers responsible for their merits. Second, we show how to map social welfare function primitives into empirically measurable statistics and exploit long-run Norwegian income and family relations register data to examine the relationship between merit and income. Third, we simulate optimal income tax implications of our meritocratic social welfare functions. The result is that accounting for merit leads to lower optimal marginal income tax rates than the utilitarian criterion recommends, but the difference is smaller when workers are not held responsible for merits that are explained by circumstances.
Selected work in progress:
Intensive and extensive margin labor supply responses: evidence from a large-scale earned income tax credit experiment in Norway
Joint with Simon Bensnes, Øystein Hernæs, Simen Markussen, Magne Mogstad, Oddbjørn Raaum, Ragnhild C. Schreiner, and Gaute Torsvik - Report
I am part of a research group that has designed an experimental study of a job tax credit targeted at young individuals. The experiment will take the form of a large-scale randomized controlled trial in which 100,000 randomly selected individuals aged 20 to 35 are offered an increased basic tax deduction for earned income, but not for income from welfare benefits. The tax credit reduces tax liabilities by up to NOK 27,500, substantially increasing the return to employment. Because the credit is phased out at higher income levels, individuals with a loose attachment to the labor market receive the most substantial increases in participation incentives. However, the gradual phase-out generates variation in both marginal, average, and participation tax rates across a broad range of the income distribution. This will enable us to study substitution and income effects on both the intensive and extensive margins. We will also examine how the tax credit affects a wide set of different outcomes, including educational choices, geographic mobility, family formation and dissolution, entry into entrepreneurship, and job-to-job mobility. The experiment will be implemented in 2026, and we expect to begin analyzing the data as soon as it becomes available, likely by the end of that year.
Revealed preference analysis of non-linear income taxation: Elasticities, welfare, and tax design
with Deniz Dutz, Magne Mogstad, and Alexander Torgovitsky.
In this paper, we develop a general framework for analysing labor supply while allowing for income effects, extensive margin responses, and heterogeneity. The goal of the paper is to provide a general way to point or partially identify a range of target parameters, like intensive-margin compensated and uncompensated earnings elasticities and income effects, extensive-margin participation elasticities, and the welfare and revenue effects of counterfactual tax reforms. We develop a computational approach that allows the researcher to flexibly adjust the policy question (target parameter) and their maintained assumptions, treating point identification as a special case. The approach will enable researchers to have discretion over the tightness of the reported bounds, while remaining constrained by the principle that stronger conclusions require stronger assumptions.
Efficiency and taxation when productivity is unequal to pay
with Magne Mogstad, Gaute Torsvik, and Kristen Vamsæter.
We develop, identify, and estimate an equilibrium model of the labor market where the distribution of wages can differ from the distribution of worker productivities. These differences arise from three sources. First, workers have heterogeneous preferences over non-wage job attributes and therefore view firms as imperfect substitutes. As a consequence, firms have wage-setting power and can set wages below the workers' marginal product. Second, neither firms nor workers perfectly observe productivity. Thus, observationally equivalent workers with identical wages can have different marginal productivities within the same firm. Third, labor income is taxed, driving an additional wedge between wages and marginal productivities. We consider how changes in the tax system affect efficiency and welfare in this model. Since workers choose both the firm they work for and their hours of work, taxation influences both the allocation of workers across firms and the number of hours they work. We show that the efficiency and welfare effects of taxation depend not only on the joint distribution of productivity and pay, but also on why productivity is unequal to pay. As a consequence, standard deadweight-loss calculations that assume productivity is equal to pay could be severely biased. To identify and estimate the model, we combine Norwegian administrative data with a unique panel dataset from the sales division of a large Nordic insurance company, which contains detailed measures of individual worker performance. The data allows us to observe productivity separately from wages. We find that standard calculations of marginal deadweight loss that ignore non-tax wedges substantially understate the efficiency cost of taxation.