Current Research Projects
Job Market Paper
Learning about Selection: An Improved Correction Procedure
Machines learning tools are utilized to develop an improved non-parametric method to control for selection bias. Building on the insights of Lee (1983) and Dahl (2002), variable selection methods and prediction using the Random Forest algorithm are used to relax the Single Index Sufficiency Assumption (SISA) imposed by Dahl’s method and to improve the empirical implementation of the procedure. Monte Carlo experiments illustrate that this novel procedure performs well both when the SISA holds and when it is violated. In experiments using reasonable sample sizes it removes 75-99% of selection bias. As researchers cannot determine empirically whether the SISA holds in their data, this represents a significant improvement over traditional correction methods which perform poorly when the SISA is violated. I use this improved method to obtain new estimates of the return to education across states for the US I find that traditional methods can underestimate selection bias by as much as 20% of the OLS estimate compared to the improved procedure.
Economy-Wide Spillovers From Booms: Long Distance Commuting and the Spread of Wage Effects
joint with David Green, Rene Morissette and Ben Sand
Revised and Resubmitted: Journal of Labor Economics
Since 2000, US real average wages stagnated or declined while Canadian wages increased. We investigate the role of the Canadian resource boom in explaining this difference. We focus on wage spillovers to non-resource workers through a bargaining channel. We find that long-distance commuting to resource regions had substantial spillover effects on noncommuters in sending regions. Through spillovers, we account for 49% of the increase in the real mean wage in Canada between 2000 and 2012. We also find long-distance commuting effects in the US. We conclude that long-distance commuting integrates regions, spreading benefits and costs of booms across the economy.
Entrepreneurship, Outside Options and Constrained Efficiency
joint with João Fonseca
Since the seminal paper of Pissarides (1985), the literature on search frictions has often adopted the assumption of free entry. In this paper we forgo of this restriction by proposing a more realistic framework in which individuals are constantly making the decision whether or not to open a firm. Namely, firms are created through endogenous choices and business-owners and workers are drawn from the same pool. We show that in this framework, the Nash bargaining parameter is crucial for internal dynamics. In particular, workers and business owners share the same outside-options. As a result, the wage is no longer unambiguously positively related to the value of unemployment. The constrained effcient solution to this model takes the same form as the standard search model implying the same form for the Hosios condition. However, at this efficient solution changes in the rate of unemployment are either exacerbated or muted conditional on the value of the match elasticity parameter.