Unpublished Papers

“Novelty, Knowledge Spillovers and Innovation: Evidence from Nobel Laureates." July 2019, under revision. Joint with B. Weinberg.

Abstract: Using a new identification strategy, and rich data on Nobel Laureates, we show that being in (i) a new location and/or (ii) multiple locations, as measures of exposure to novel combinations of ideas, significantly increases the probability that eventual Nobel Laureates begin their Prize-winning work in a given year. We find no evidence that these variables reduce the time until a scientist does Nobel work. We present evidence that these impacts are causal. Our results suggest that programs supporting high-quality researchers visiting different departments and research institutes, and/or splitting time between multiple locations, are likely to increase innovative activity.

“A Simple Dynamic Model of Bidding Behavior in Private Value Auctions.” July 2019. Under Revision for the Journal of the Economic Science Association at the request of the Editor. Joint with Steve Lehrer.

Abstract: We provide the first consistent estimates of a dynamic model in experimental economics using instrumental variables (IV). We focus on affiliated private value auctions, where subjects are generally thought to converge to rule-of-thumb bidding. Our IV estimates suggest that subjects become significantly less aggressive over time by decreasing their bids in proportion to last period’s signal minus bid. The inconsistent OLS and FE estimators imply subjects become significantly more aggressive over time in terms of raising their bids in proportion to last period’s signal minus bid. Our instruments are randomly generated by the experiment and pass popular weak IV tests.

“Consistent Estimation of Treatment Effects and Their Confidence Intervals in Poorly-Identified Models: The Employment Impact of the 2005 TennCare Contraction.” June 2019. Submitted to the Journal of the Labor Economics at the request of the Editor. Joint with Ken Ueda.

Abstract: The effect on employment of the availability of low-cost government health insurance is important for policy. Garthwaite et al. find very large impacts from the TennCare contraction, but these estimates are driven by two data points in one state. We re-estimate their model on two larger data sets and find several important inconsistencies across and within datasets; this outcome is predicted by Conley and Taber’s (2011, CT) result that treatment effects will be inconsistently estimated here. We then use CT’s procedure for estimating consistent confidence intervals, and find that this approach produces fewer inconsistencies and more plausible results.

"Confidence Intervals for Continuous and Discontinuous Functions of Estimated Parameters.” March 2019. Resubmitted (third round) to the Journal of Econometrics at the request of the Editor. Joint with T. Woutersen; Woutersen is first author.

Abstract: Applied researchers often need to construct confidence sets for policy effects based on functions of estimated parameters, and often these functions are such that the delta method cannot be used. In this case, researchers currently use one of two bootstrap procedures, but we show that these procedures can produce confidence sets that are too small. We provide two alternatives, both of which produce consistent confidence sets under reasonable assumptions that are likely to be met in empirical work. Our second, more efficient method produces consistent confidence sets in cases when the delta method cannot be used, but is asymptotically equivalent to the delta method if the use of the latter is valid. Finally, we show that our second procedure works well when we conduct counterfactual policy analyses using a dynamic model for employment.

“Selection, Gender and the Impact of Schooling Type in the Dhaka Slums.” Slides from November 2019 available. Presented at many conferences and universities. Joint with Saima Khan.

Abstract: We collected a unique data set on 4 - 12 year olds in two Dhaka slums. Ours is the first such study of education for Urban Bangladesh. We consider relative impacts on achievement, for boy and girls separately, of JAAGO, Govt. and NGO schools; JAAGO is a new school-type with a framework previously available only to elites via private schools. In collecting our data, we faced two issues: that selection into the schools was very likely to be nonrandom, and that JAAGO enrolled few students. We collected two measures of child IQ to help deal with the selection issue. Moreover, we used choice-based sampling to get enough JAAGO students. We matched on the log odds ratio to deal with selection, since unlike IV or regular propensity score matching, it produces consistently estimated treatment effects in our situation. We found that conditioning on the two IQ measures drastically reduced selection bias. Further, in terms of JAAGO vs. Government schools, girls are better off at JAAGO, while boys perform equally well at both school-types. In terms of NGO vs. Government schools, boys are better off at Govt. schools, but girls perform equally poorly at both school-types. Finally, both genders perform better at JAAGO schools than at NGO schools.

“The Correct Use of Hypothesis Testing and Choosing Appropriate Comparison Groups When Estimating the Impacts of Location Based Policies: A Response to Neumark and Young.” NYUAD Social Science WP# 0022, October 2018. Joint with H. Song, A. İmrohoroğlu, and C Swenson.

In IZA working paper Neumark and Young have wrote a comment on our paper in the JPubE 2011. This is our response to their comment.

Abstract: We are grateful to Neumark and Young (2017, hereafter NY) for spotting an error in Ham, Swenson, İmrohoroğlu, and Song’s (2011, hereafter HSIS) 1990 poverty rate data. Our corrected estimates reported here of the impacts on the Poverty Rate of Enterprise Zones, (ENTZs), Enterprise Communities (ENTCs), and Empowerment Zones (EMPZs) are smaller than those in HSIS. However, they are still quite sizable and statistically significant. We show here that NY obtained similar results with the NCBD data to our new estimates reported here. The NCBD data uses a different approach than that used by HSIS to deal with the changing borders of some Census tracts over time.

However, we find NY’s criticisms of the HSIS results for ENTZs, ENTCs and EMPZs to be deeply flawed, and suffer from several important errors. First, their criticisms arise from their making a fundamental error in hypothesis testing. Second, they use an incorrect approach for comparing parameter estimates from different studies. Third they use a comparison group for EMPZs where about half the comparison tracts are impacted by other labor market programs, leading to downward biased estimates of the EMPZ impacts. We argue that this bias is over 50 percent of their (under)estimated treatment effect for EMPZs.