xView3-SAR: Detecting Dark Fishing Activity Using Synthetic Aperture Radar Imagery
 Advances in Neural Information Processing Systems 35 (NeurIPS 2022)
 (with Fernando Paolo*, Ritwik Gupta*, Bryce Goodman, Nirav Patel, Daniel Kuster, David Kroodsma, Jared Dunnmon), *co-first author with F. Paolo and R. Gupta
Abstract Unsustainable fishing practices worldwide pose a major threat to marine resources and ecosystems. Identifying vessels that evade monitoring systems–known as “dark vessels”–is key to managing and securing the health of marine environments. With the rise of satellite-based synthetic aperture radar (SAR) imaging and modern machine learning (ML), it is now possible to automate detection of dark vessels day or night, under all-weather conditions. SAR images, however, require domain-specific treatment and is not widely accessible to the ML community. Moreover, the objects (vessels) are small and sparse, challenging traditional computer vision approaches. We present the largest labeled dataset for training ML models to detect and characterize vessels from SAR. xView3-SAR consists of nearly 1,000 analysis-ready SAR images from the Sentinel-1 mission that are, on average, 29,400-by-24,400 pixels each. The images are annotated using a combination of automated and manual analysis. Co-located bathymetry and wind state rasters accompany every SAR image. We provide an overview of the results from the xView3 Computer Vision Challenge, an international competition using xView3-SAR for ship detection and characterization at large scale. We release the data ( and code ( to support ongoing development and evaluation of ML approaches for this important application. | Download | OpenReview

A New Keynesian Model with Robots: Implications for Business Cycles and Monetary Policy
Atlantic Economic Journal, 47(1), March 2019, pp. 81-101.
 (with Charles L. Weise)
Abstract This paper examines the effects of labor-replacing capital, which is referred to as robots, on business cycle dynamics using a New Keynesian model with a role for both traditional and robot capital. This study finds that shocks to the price of robots have effects on wages, output, and employment that are distinct from shocks to the price of traditional capital. Further, the inclusion of robots alters the response of employment and labor’s share to total factor productivity and monetary policy shocks. The presence of robots also weakens the correlation between human labor and output and the correlation between human labor and labor’s share. The paper finds that monetary policymakers would need to place a greater emphasis on output stabilization if their objective is to minimize a weighted average of output and inflation volatility. Moreover, if policymakers have an employment stabilization objective apart from their output stabilization objective, they would have to further focus on output stabilization due to the deterioration of the output-employment correlation. | Download | Working paper version | Online appendix

Working Capital Requirement and the Unemployment Volatility Puzzle
Journal of Macroeconomics, 46, December 2015, pp. 201-217.
Abstract Shimer (2005) argues that a search-and-matching model of labor market in which wage is determined by Nash bargaining cannot generate the observed volatility in unemployment and vacancy in response to reasonable labor productivity shocks. This paper examines how monopolistically competitive firms with a working capital requirement (in which firms borrow funds to pay their wage bills) improves the ability of search models to match empirical fluctuations in unemployment and vacancy without resorting to an alternative wage setting mechanism. The monetary authority follows an interest rate rule in the model. A positive labor productivity shock lowers the real marginal cost of production and lowers inflation. In response to price level changes, the monetary authority reduces the nominal interest rate. The lower interest rate reduces the cost of financing and partially offsets the increase in labor cost from higher productivity. Reduced labor costs imply firms retain a greater portion of the gain from a productivity shock which gives them greater incentives to create vacancies. Simulations show that working capital requirement does indeed improve the ability of the model to generate fluctuations in the labor market variables to better match the U.S. data. | Download | Working paper version

Working Papers

The Role of Uncertainty in the Joint Output and Employment Dynamics
Abstract This paper examines the role uncertainty plays in the joint dynamics between output and employment. To account for the periodic negative co-movement between output and employment observed in the data, I develop a dynamic stochastic general equilibrium model with search and matching frictions in the labor market and an intensive labor margin. The model is driven by productivity and time-varying volatility shocks. The uncertainty agents face is captured by time-varying volatility. Labor market search frictions generate costly labor adjustment. When an uncertainty shock hits the economy, firms reduce the number of vacancies because they are reluctant to make costly adjustments along the extensive margin. Instead, firms require more effort from their employees. An economy hit by an uncertainty shock and a positive productivity shock simultaneously can thus experience a negative co-movement between output and employment as periodically observed in the data. | Download | Online appendix

The Effects of Severance Payment on Mortensen-Pissarides Model with On-the-job Search
Abstract Saint-Paul (1995) argues that severance payments could generate multiple equilibria in the labor market. This paper seeks to examine the effect of severance payment in a richer environment: a Mortensen-Pissarides model with on-the-job search. Severance pay affects steady-state equilibrium outcome by distorting firms’ threat point in the rent sharing condition for the on-going matches. The equilibrium outcome of severance payment, including the existence of multiple equilibria, is analytically ambiguous when one does not place stronger assumptions beyond homogeneous matching function and fixed rent sharing rule. However, the numerical solutions of the model under a wide range of parameter values show that the inclusion of severance pay decreases market tightness and reservation productivity, as one would expect. There is no evidence for multiple equilibria. I also find that while unemployment rate falls with severance pay, the society’s welfare does not necessarily increase.

Work in Progress

Another Look at the Black Box: An Agent-based Approach to the Matching Function
Abstract The textbook search-and-matching model of unemployment relies on a policy-invariant analytical aggregate matching function as a proxy for the employer-employee matching process. This paper presents an agent-based model of the labor market in which firms and workers follow simple behavioral rules in their vacancy creation and job searching decisions. The agent-based model is used to evaluate the constant-returns-to-scale assumption as well as the assumption that the matching process is invariant to employment policies that may alter the search behaviors of the firms and workers.

The Inheritance of Employers and the Great Gatsby Curve
 (with Brendan Cushing-Daniels)