Charlie Gibbons' homepage  
 
 
 

Working papers

Broken or Fixed Effects? with Juan Carlos Suarez Serrato and Mike Urbancic (under review).

This paper provides empirical evidence of an established theoretical result: in the presence of heterogeneous treatment effects, OLS is generally not a consistent estimator of the sample- weighted average treatment effect (SWE). We propose two alternative estimators that do recover the SWE in the presence of group-specific heterogeneity. We derive tests to detect the presence of heterogeneous treatment effects and to distinguish between the OLS and SWE. We document that heterogeneous treatment effects are common and the SWE is often statistically and economically different from the OLS estimate by extending eight influential papers. In all but one paper, there is statistically significant treatment effect heterogeneity; in five, the SWE is statistically different from the OLS estimator; and in five, the SWE and OLS estimators are economically different.
R package implementing SWE estimator and tests

Quantile Regression for Peak Demand Forecasting with Ahmad Faruqui (under review).

We demonstrate that annual peak demand days are characterized by both extreme values of predictors (such as weather) and large unpredictable "shocks" to demand. OLS approaches incorporate the former feature, but not the latter, leading OLS to produce downwardly-biased estimates of the annual peak. We develop a new estimation procedure, optimal forecast quantile regression (OFQR), that uses quantile regression to estimate a model of daily peak demand, then uses a loss function framework to estimate a quantile to predict the annual peak. We compare the results of the OLS and OFQR estimation approaches for 32 utility zones. While the OFQR approach is unbiased, OLS under-forecasts by nearly 5% on average. Further, OFQR reduces the average absolute percent error by 43%. A bootstrapping procedure generates forecast intervals with accurate 95% coverage in sample and 87% coverage out of sample.

Ad Server and Firm Strategies in Contextual Advertising Auctions (job market paper)

We consider the strategies of online advertising providers, firms, and consumers in the context of ad listings assigned by a generalized second price auction. The first part of the paper develops a model of consumer responses to ad listings and product offerings by included firms and uses this behavioral model to derive optimal bidding functions for the firms. We show that the relationship between per-sale margins and product-consumer match probabilities ("relevances") must meet certain conditions to rationalize this equilibrium for consumers and firms; in particular, we give the conditions for consumers to search from the top down. Next, we turn to incentives facing the ad server to alter the relevances and margins of the firms and the search costs and valuations of the consumer pool. While these incentives align with the desires of consumers, they may conflict with those for firms. We consider whether ad servers desire thick or thin product markets. We calculate the optimal number of slots for the ad server to offer, which is less than that desired by firms and consumers. We also show that the ad server has an incentive to subsidize its own competitor in the product market. These results have important implications for competition policy and online content provision.

LATE for School: Instrumental Variables, and the Returns to Schooling with Mike Urbancic.

We show that comparisons of OLS and IV estimates provide misleading evidence regarding exogeneity. If treatment is exogenous, then the local average treatment effect equals the average treatment effect. The Hausman test for endogeneity relies upon the converse, but it is false. It can be justified if we assume homogeneous treatment effects, but this is equivalent to assuming exogeneity itself. We provide an intuitive graphical exposition, along with theoretical derivations to illustrate these points. We use the returns to education literature to provide an empirical example. We conclude by offering guidance for the applied researcher.
(Please contact me if you are interested in receiving a copy of this paper.)
 
 
Last updated: July 19, 2013