Working Papers
Moment Restrictions for Nonlinear Panel Data Models with Feedback, (with Stéphane Bonhomme and Bryan S. Graham)
[ Abstract | Draft | Arxiv version]
Many panel data methods place strong a priori restrictions on feedback: how past outcomes, covariates, and heterogeneity map into future covariate levels. In semi-parametric models we characterize the set of all moment functions that are robust to unrestricted feedback and heterogeneity. We provide results for finite-dimenisonal parameters and average effects, and show how to use our characterization for efficiency analysis. We illustrate the approach in the multi-spell mixed proportional hazard duration model and through additional novel examples.
Transition Probabilities and Moment Restrictions in Dynamic Fixed Effects Logit Models, (winner of the Best PhD Student Paper Award, 2023 IAAE Conference), Submitted
[ Abstract | Draft | Supplement | Arxiv version]
This paper introduces a new method to derive moment restrictions in dynamic discrete choice models with strictly exogenous regressors, fixed effects and logistic errors. We show how the structure of logit probabilities and basic properties of rational fractions can be used to construct moment functions free of the fixed effects in a way that scales naturally with the lag order and the number of observed periods. We demonstrate the approach in binary response models of arbitrary lag order, first-order panel vector autoregressions and dynamic multinomial logit models. The semiparametric efficiency bound is characterized for the leading binary case with one lag. Finally, we illustrate our results in an application investigating the dynamics of drug consumption among young people.
Functional Differencing in Networks (with Stéphane Bonhomme),
[ Abstract | Draft | Arxiv version ]
Economic interactions often occur in networks where heterogeneous agents (such as workers or firms) sort and produce. However, most existing estimation approaches either require the network to be dense, which is at odds with many empirical networks, or they require restricting the form of heterogeneity and the network formation process. We show how the functional differencing approach introduced by Bonhomme (2012) in the context of panel data, can be applied in network settings to derive moment restrictions on model parameters and average effects. Those restrictions are valid irrespective of the form of heterogeneity, and they hold in both dense and sparse networks. We illustrate the analysis with linear and nonlinear models of matched employer-employee data, in the spirit of the model introduced by Abowd, Kramarz, and Margolis (1999).
Publications
Identification in a Binary Choice Panel Data Model with a Predetermined Covariate (with Stéphane Bonhomme and Bryan S. Graham), SERIEs - Journal of the Spanish Economic Association. Special issue in honor of Manuel Arellano.
[ Abstract | Published version | Arxiv version | Nber version | Replication code]
We study identification in a binary choice panel data model with a single predetermined binary covariate (i.e., a covariate sequentially exogenous conditional on lagged outcomes and covariates). The choice model is indexed by a scalar parameter, whereas the distribution of unit-specific heterogeneity, as well as the feedback process that maps lagged outcomes into future covariate realizations, are left unrestricted. We provide a simple condition under which the model parameter is never point-identified, no matter the number of time periods available. At the same time, we show in simulations that its identified set can remain informative suggesting that meaningful learning is possible even in short panels with feedback.
Coordination and Incumbency Advantage in Multi-Party Systems: Evidence from French Elections (with Francesco Ferlenga, Vincenzo Galasso, Caroline Le Pennec and Vincent Pons), The Journal of the European Economic Association
[ Abstract | Draft | Nber version] | Replication code]
In theory, free and fair elections can improve the selection of politicians and incentivize them to exert effort. In practice, incumbency advantage and coordination issues may lead to the (re)election of bad politicians. We ask whether these two forces compound each other. Using an RDD in French two-round local and parliamentary elections, we find that winning an election increases candidates' chances to win the next election by 25.1 percentage points. Close winners are more likely to run again and more likely to win, conditional on running, than close losers. Incumbents run a more personalized campaign communication and face fewer ideologically close competitors, indicating that parties on the winning side coordinate more effectively than the losing side. A complementary RDD reveals that candidates who marginally qualify for the runoff also rally voters, but without affecting the number of competitors on their side. We conclude that party coordination and voters rallying candidates who won or gained visibility in an election both contribute to their success in future races, absent any actual difference in quality with candidates on the losing side.
Work In Progress
Dynamic Panel Multinomial Logit Models (with Bo Honoré and Martin Weidner)
Identification and Estimation of Random Effects Linear Social Interaction Models with Endogenous Peer Selection
