Research
Research Papers
with Richar Audoly, Rory McGee, and Gonzalo Paz Pardo (Long Slides, Slides, Shapley-Owen Decomposition, 🧵)
We use 25 years of tax records for the Norwegian population to study the mobility of wealth over people’s lifetimes. We find considerable wealth mobility over the life cycle. To understand the underlying mobility patterns, we group individuals with similar wealth rank histories using agglomerative hierarchical clustering, a tool from statistical learning. The mobility patterns we elicit provide evidence of segmented mobility. Over 60 percent of the population remains at the top or bottom of the wealth distribution throughout their lives. Mobility is driven by the remaining 40 percent, who move only within the middle of the distribution. Movements are tied to differential income trajectories and business activities across groups. We show parental wealth is the key predictor of who is persistently rich or poor, while human capital is the main predictor of those who rise and fall through the middle of the distribution.
When is a wealth tax preferable to a capital income tax? When is the opposite true? More generally, can capital taxation be structured to improve productivity, incentivize innovation, and ultimately increase welfare? We study these questions theoretically in an infinite-horizon model with entrepreneurs and workers, in which entrepreneurial firms differ in their productivity and are subject to collateral constraints. The stationary equilibrium features heterogeneous returns and misallocation of capital. We show that increasing the wealth tax increases aggregate productivity. The gains result from the “use-it-or-lose-it” effect of wealth taxes when returns are heterogeneous, which causes a reallocation of capital from entrepreneurs with low productivity to those with high productivity. Furthermore, if the capital income tax is adjusted to balance the government’s budget, aggregate capital, output, and wages also increase. We then study the welfare maximizing combination of wealth and capital income taxes and show that the optimal mix shifts towards a higher wealth tax and a lower capital income tax as the capital intensity of production increases. For a range of plausible parameter values, the optimal wealth tax is positive, whereas the capital income tax can be positive or negative (a subsidy). We then endogenize the entrepreneurial productivity distribution by introducing either ex ante innovation or entrepreneurial effort in production and show that this strengthens our results: by allowing entrepreneurs to keep more of the upside relative to a capital income tax, a wealth tax incentivizes more innovation and entrepreneurial effort, leading to larger increases in productivity, output, and welfare.
Optimal Wealth Tax
Opt. Tax. Formula depends on elaasticity of wages and returns to productivity
Wealth Distribution
Wealth taxes increase dispersion of returns -> Higher wealth inequality (top+bottom)
with Raphael Schoenle and Dominic Smith (Old version: arXiv, CEPR Discussion Paper 17485, slides, CEBRA video presentation, Inflation in 3 video)
Robust inflation measures are used by policy makers to understand the behavior of trend inflation, particularly when inflation is high or volatile. These measures take out expenditure categories from headline inflation. We evaluate the forecasting performance of a wide set of robust inflation measures between 1960 and 2024, including core, median, and trimmed-mean personal-consumption-expenditure (PCE) inflation. We find core inflation performs significantly worse than median and trimmed-mean inflation. When trimming out expenditure categories with the highest and lowest inflation rates, we find the optimal trim points vary widely across time and also depend on the choice of trend target; optimal trims are higher when targeting future trend inflation or for a 1970s–1980s subsample when inflation was high and volatile. Surprisingly, there are no grounds to select a single trim over others on the basis of forecasting performance. A wide range of trims have an average prediction error that makes them statistically indistinguishable from the best-performing trim. Despite indistinguishable average forecasting performance, these trims imply different predictions for trend inflation in any given month, within a range of 0.5 to 1 percentage points, suggesting the use of a set of near-optimal trims.
Robust Measures of Inflation 1960-2020
Sensitivity of Median Inflation
Yearly vs Montly Measures
Sensitivity of Trimmed Mean Inflation
Yearly vs Montly Measures
I develop an assignment model of occupations with multidimensional heterogeneity in production tasks and worker skills. Tasks are distributed continuously in the skill space, whereas workers have a discrete distribution with a finite number of types. Occupations arise endogenously as bundles of tasks optimally assigned to a type of worker. The model allows us to study how occupations respond to changes in the economic environment, making it useful for analyzing the implications of automation, skill-biased technical change, offshoring, and worker training. Using the model, I characterize how wages, the marginal product of workers, the substitutability between worker types, and the labor share depend on the assignment of tasks to workers. I introduce automation as the choice of the optimal size and location of a mass of identical robots in the task space. Automation displaces workers by replacing them in the performance of tasks, generating a cascading effect on other workers as the boundaries of occupations are redrawn.
Occcupations are Bundles of Tasks
Tasks (y) assigned to workers (x)Tasks and workers characterized by skills 
Automation and Task Displacement
Replace workers at tasks along boundaries of occupations (red area) triggering reassignment of tasks shifting boundaries
Unassigned Tasks
Taks are left unassigned (gray areas) when workers are not productive enoughPosibility to increase task coverage with automation
Published or Forthcoming Papers
Use it or lose it: Efficiency and redistributional effects of wealth taxation [The Quarterly Journal of Economics, 2023]
with Fatih Guvenen, Gueorgui Kambourov, Burhan Kuruscu and Daphne Chen (Open access version, NBER working paper version, 🧵)
How does wealth taxation differ from capital income taxation? When the return on investment is equal across individuals, a well-known result is that the two tax systems are equivalent. Motivated by recent empirical evidence documenting persistent return hetero- geneity, we revisit this question. With heterogeneity, the two tax systems typically have opposite implications for both efficiency and inequality. Under capital income taxation, en- trepreneurs who are more productive and therefore generate more income pay higher taxes. Under wealth taxation, entrepreneurs who have similar wealth levels pay similar taxes re- gardless of their productivity, which expands the tax base, shifts the tax burden toward unproductive entrepreneurs, and raises the savings rate of productive ones. This realloca- tion increases aggregate productivity and output. In the simulated model parameterized to match the US data, replacing the capital income tax with a wealth tax in a revenue-neutral fashion delivers a significantly higher average welfare. Turning to optimal taxation, the op- timal wealth tax (OWT) is positive and yields large welfare gains by raising efficiency and lowering inequality. In contrast, the optimal capital income tax (OKIT) is negative—a sub- sidy—and delivers lower welfare gains than OWT, owing to the welfare losses from higher inequality. Furthermore, when the transition path is considered, the gains from OKIT turn into significant welfare losses for existing cohorts, whereas OWT continues to deliver robust welfare gains. These results suggest that moderate wealth taxation may be a more appealing alternative than capital income taxation, which can be significantly more distorting under return heterogeneity than under the equal-returns assumption.
Coverage: Financial Times (Free Lunch, Article), The Economist (Article), Bloomberg (Noah Smith: 2016, 2019), St. Louis Fed (interview with Fatih Guvenen), Observatorio Fiscal (Online OpEd, in Spanish)
We evaluate the aggregate effects of expansions of credit supply in environments where subsistence self-employment is prevalent. We extend a standard macro development model to include unemployment risk, which becomes a key driver of selection into self-employment. The model is consistent with the joint distribution of earnings and occupations, the reaction of wages to labor demand shocks, and the small effects of expansions in the supply of microloans on the earnings of the self-employed. We find that the elasticity of aggregate output to expansions in credit supply is proportional to the elasticity of individual earnings. This is because of the muted effects of wages in general equilibrium in the presence of subsistence self-employment. We also consider variations on the implementation of the credit expansion and the introduction of targeted transfers. The effects on total factor productivity (TFP) are ambiguous and depend on the targeting and generosity of the policies, with those that encourage self-employment lowering TFP.
Extra: Previous version with additional productivity shifters (Draft, 2019)
Share of Self-Employment Across Countries
Self-employment rates decrease along the development path
Share of Self-Employment Across Earnings Distribution
Self-employment rate follows a U-shaped pattern on earnings. Self-employment is higher among the lowest and highest earners.
Model Matches Joint Dist. of Self-Employment and Earnings
The baseline model with unemployment risk generates U-shaped pattern. Model without unemployment risk does not.
Occupational Choice of the Unemployed
Low-earning self-employed originate in subsistence concerns of unemployed agents. Poor unemployed become self-employed regardless of their productivity
Increases in national concentration have been a salient feature of industry dynamics in the U.S. and have contributed to concerns about increasing market power. Yet, local trends may be more informative about market power, particularly in the retail sector where consumers have traditionally shopped at nearby stores. We find that local concentration has increased almost in parallel with national concentration using novel Census data on product-level revenue for all U.S. retail stores between 1992 and 2012. The increases in concentration are broad based, affecting most markets, products, and retail industries. We show that the expansion of multi-market firms into new markets explains most of the increase in national retail concentration, with consolidation via increases in local market shares increasing in importance between 1997 and 2007, and single-market firms playing a negligible role. Finally, we find that increases in local concentration can explain one-quarter to one-third of the observed rise in retail gross margins.
Key Equation: National HHI = 0.02xLocal HHI + 0.98xCross-Market HHI
National concentration measures do not respond to local changes.
Instead, national HHI depends on cross-market HHI (people in different markets shopping at the same firms)
Retail Concentration 1992-2012
Retail concentration has increased at the national and local levelNational concentration accelerates in 1997 - Local concentration does not
Change in Local HHI by Product
Increase in local concentration happens across products
Distribution of Commuting-Zones by change of HHI
69% of local markets (commuting zones) have increased concentration 92-02
Computing population moments for heterogeneous agent models is a necessary step for their estimation and evaluation. Computation based on Monte Carlo methods is time- and resource-consuming because it involves simulating a large sample of agents and tracking them over time. We formalize how an alternative non-stochastic method, widely used for computing cross-sectional moments, can be extended to also compute longitudinal moments. The method relies on following the distribution of populations of interest by iterating forward the Markov transition function that defines the evolution of the distribution of agents in the model. Approximations of this function are readily available from standard solution methods of dynamic programming problems. We document the performance of this method vis-a-vis standard Monte Carlo simulations when calculating longitudinal moments. The method provides precise estimates of moments like top-wealth shares, auto-correlations, transition rates, age-profiles, or coefficients of population regressions at lower time- and resource-costs compared to Monte Carlo based methods. The method is particularly useful for moments of small groups of agents or involving rare events, but implies increasing memory costs in models with a large state space.
Business activities often involve a common agent managing a variety of projects on behalf of investors with potentially conflicting interests. The extent of the agent’s actions is also often unknown to investors, who have to design contracts that provide incentives to the manager despite this lack of crucial knowledge. We consider a game between several principals and a common agent, where principals know only a subset of the actions available to the agent. Principals demand robustness and evaluate contracts on a worst-case basis. This robust approach allows for a crisp characterization of the equilibrium contracts and payoffs and provides a novel proof of equilibrium existence in common agency by constructing a pseudo-potential for the game. Robust contracts make explicit how the efficiency of the equilibrium outcome relative to collusion among principals depends on the principals’ ability to extract payments from the agent.
Expected Payoffs under Arbitrary Contracts
vs LRS Contracts
Increase in Guaranteed Payoffs from
LRS Contracts
Work from before my PhD
Articles
2013 “Output gap and neutral interest measures for Colombia”, With Andrés González, Diego Rodríguez and Julian Pérez, Monetaria, Vol. 1, No. 2, pp. 231-286, Center for Latin American Monetary Studies (CEMLA). http://www.cemla.org/PDF/monetaria/PUB-MON-I-02.pdf
2012 “An Introductory Review of a Structural VAR-X Estimation and Applications”, With Norberto Rodríguez, Colombian Journal of Statistics, Vol 35, No. 3, pp. 479-508, Universidad Nacional de Colombia. http://www.redalyc.org/articulo.oa?id=89925367003
2012 “Asimetrías del empleo y el producto, una aproximación de equilibrio general” (Asymmetries of the employment-output relation, a general equilibrium approach), With Andrés González, Diego Rodríguez and Norberto Rodríguez, Revista Ensayos sobre Política Económica (ESPE), Vol. 30, No. 68, pp. 46-81, Banco de la República. http://www.banrep.gov.co/sites/default/files/publicaciones/archivos/espe_art5_68.pdf
2011 “Choques, instituciones laborales y desempleo en Colombia” (Shocks, institutions and unem- ployment in Colombia), Whit Juan José Echavarría, Enrique López and Norberto Rodríguez, Revista Ensayos sobre Política Económica (ESPE), Vol. 29, No. 66, pp. 128-173, Banco de la República. http://www.banrep.gov.co/docum/ensayos/pdf/espe_art4_66.pdf
Old Working Papers
2013 "Rule of Thumb Consumers, Nominal Rigidities and the Design of Interest Rate Rules", IDB Working Paper Series, No. IDB-WP-400, Inter American Development Bank.
2012 “A model of Rule-of-Thumb consumers with nominal price and wage rigidities”, Borradores de Economía, No. 707, Banco de la República. http://www.banrep.gov.co/docum/ftp/be_707.pdf
2012 “Introducción de fricciones en el mercado laboral en modelos DSGE” (Introducing labor market frictions in DSGE Models), Vniversitas Economicas, No. 010007, Economics Department, Pontificia Universidad Javeriana. http://www.javeriana.edu.co/fcea/coleccion_universitas_Economica/Vol_12/Vol.12_5_2012.pdf
2011 “Agentes no ricardianos y rigideces nominales: su efecto sobre el principio de Taylor” (Rule- of-Thumb consumers and nominal rigidities: their effect over the Taylor principle), Vniversitas Economicas, No. 008302, Economics Department, Pontificia Universidad Javeriana. http://www.javeriana.edu.co/fcea/coleccion_universitas_Economica/Vol_11/Vol.11_3_2011.pdf