Microeconomics & International Trade Seminars

 

Thursdays, 1:40 - 3:00 PM in E2-499

Fall 2019

September 26
Daniel Chen, Toulouse School of Economics
"Stereotypes in High Stake Decisions: Evidence from U.S. Circuit Courts"
Host: Dan Friedman
Abstract:
Attitudes towards social groups such as women and racial minorities have been shown to be important determinants of individual’s decisions but are hard to measure for those in policymaking roles. We propose a way to address the challenge in the case of U.S. appellate court judges, for whom we have large corpora of written text (their published opinions). Using the universe of published opinions in U.S. Circuit Courts 1890-2013, we construct a judge-specific measure of gender-stereotyped language (gender slant) by looking at the relative co-occurrence of words identifying gender (male versus female) and words identifying gender stereotypes (career versus family). We find that female and younger judges tend to use less stereotyped language in their opinions. Our measure of gender slant matters for judicial decisions: judges with higher slant vote more conservatively on women rights’ issues. In addition, lexically slanted judges influence workplace outcomes for female judges: more slanted judges are less likely to assign opinions to female judges, cite fewer female-authored opinions and are more likely to reverse lower-court decisions if the district court judge is a woman. Our results expose a possible use of lexical slant to detect decision-makers’ stereotypes that predict behavior and disparate outcomes.


October 10
Frank Wolak, Stanford
"Fast, 'Robust', and Approximately Correct: Estimating Mixed Demand Systems"
Host: Jessie Li
Abstract:
Many econometric models used in applied work integrate over unobserved heterogeneity. We show that a class of these models that includes many random coefficients demand systems can be approximated by a “small-o" expansion that yields a linear two-stage least squares estimator. We study in detail the models of product market shares and prices popular in empirical IO. Our estimator is only approximately correct, but it performs very well in practice. It is extremely fast and easy to implement, and it is “robust" to changes in the higher moments of the distribution of the random coefficients. At the very least, it provides excellent starting values for more commonly used estimators of these models.


October 15 (Note different day)
Aaron Bodoh-Creed, Berkeley Haas
"Pre-College Human Capital Investment and Affirmative Action: A Structural Policy Analysis of U.S. College Admissions"
Host: Natalia Lazzati
Abstract:
We estimate a model of college admissions wherein students endogenously accrue pre-college human capital (HC) as part of a contest for enrollment at high quality colleges. We use methods from the empirical auctions literature to separately identify the roles of school quality, HC, and students’ privately known learning costs on post-college household income. Conditional on graduating, college quality is the most important factor in determining income, while unobserved student characteristics play a nontrivial secondary role. Pre-college HC drives college placement and graduation probability, but not post-college income. We conduct counterfactual experiments comparing the status quo to a color-blind admissions rule and a proportional quota for minority students. Color-blind admissions results in fewer (more) minority students enrolling at the best (worst) schools with a corresponding reduction in household incomes and graduation rates. The signs and magnitudes of changes to HC investment and graduation rate depend on the learning cost of the particular student in question, and accounting for the endogeneity of HC is crucial for predicting the effect of each admissions rule.


November 14 (moved from Nov 13 Brown Bag)
David Schonholzer, Stockholm University
"Measuring the Efficacy and Efficiency of School Faculty Expenditures"
Host: Ajay Shenoy
Abstract:
We offer new evidence on the effects and efficiency of school facility investment on student and neighborhood outcomes, linking data on new facility openings to administrative student and real estate records in Los Angeles Unified School District (LAUSD). Since 1997, LAUSD has built and renovated hundreds of schools as a part of the largest public school construction program in US history. Using an event study design that exploits quasi-random variation in the timing of new facilities and a residential assignment instrument, we find strong positive impacts on math, English, and attendance. Effects are not driven by changes in class size, peers, teachers, or principals, but rather by increased facility quality and, to a lesser extent, reductions in overcrowding. House prices increase by 6% in neighborhoods that receive new schools. Using a residential choice model, we then estimate that a dollar spent on school facilities raises the sum of housing values and adult earnings by 1.64 dollars, with only 22% of this valuation due to academic benefits of the program. The housing market valuation of academic benefits captures most but not all of the implied future earnings gains.


November 21
Paul Niehaus, UC San Diego
"General Equilibrium Effects of Cash Transfers: Experimental Evidence from Kenya"
Host: Alan Spearot
Abstract:
Tracing out the effect of large economic stimuli on the pattern of transactions in an integrated economy, and their aggregate implications, has long been a central goal of economic analysis, but until now has not been studied experimentally. This study was designed to study the aggregate consequences of cash transfer programs while accounting for multipliers and externalities. We carried out a large-scale experiment in rural Kenya that provided one-time cash transfers worth roughly USD 1000 across 653 villages with around 280,000 people, with a large implied fiscal shock of roughly 15% of local GDP, and deliberately randomized the intensity of cash transfers across geographic sublocations. We first document large direct impacts on households that received transfers, including increases in consumption expenditures and durable assets 18 months after transfers. Enterprises in areas that receive more cash transfers also experience meaningful gains in total revenues, in line with the increased household expenditures. Untreated households, too, show large consumption expenditure gains, by an amount comparable to recipients' gains. Through monthly measurement of scores of commodities and consumer and durable goods, we document positive but minimal local price inflation (0.1% on average) in areas that received additional cash. To assess aggregate implications, we compute a local fiscal multiplier, taking advantage of data on representative samples of treated and untreated households and firms. Both income data and consumption data yield large positive estimated local fiscal multipliers of approximately 2.6. A speculative possibility for how local output increases, despite no meaningful local price inflation or firm investment response, is that many local enterprises are characterized by substantial `slack' in their utilization of factors of production. Finally, we interpret the welfare implications of these results through the lens of a simple household optimization framework. In this framework, the fact observed consumption gains for untreated households are not driven by corresponding increases in labor supply, combined with a lack of local price inflation or of adverse spillovers along other non-market dimensions, suggest that non-recipients as well as recipients were made better off in this setting. This in turn suggests that some existing evaluations of cash transfer programs that ignore aggregate effects may be under-estimating overall program gains. 


December 5
Jean-Jacques Forneron, Boston University
"Inference by Stochastic Optimization: A Free-Lunch Bootstrap"
Host: Jessie Li
Abstract:
This paper proposes a Stochastic Newton-Raphson (SNR) algorithm which delivers asymptotically valid Bootstrap draws and point estimates in a single run. This algorithm generates draws that take the form of a Markov-Chain generated by the gradient and hessian computed on batches of data that are re-sampled at each iteration. We show that these draws yield both accurate estimates and asymptotically valid frequentist inferences. This is particularly attractive in settings where the model needs to be re-estimated many times to compute standard errors. SNR performs well in simulations. Furthermore, a simple modification of the baseline algorithm produces graphically appealing synopses of data irregularities. Sensitivity of the estimates to outliers are illustrated in several applications.