Working Papers by Matthew Shum
# | Title | Authors | Date | Length | Paper | Abstract | |
---|---|---|---|---|---|---|---|
1416 | Random Projection Estimation of Discrete-Choice Models with Large Choice Sets | Chiong, Khai X. Shum, Matthew | 03/29/2016 | 29 | SSWP_1416.pdf | We introduce sparse random projection, an important dimension-reduction tool from machine learning, for the estimation of discrete-choice models with high-dimensional choice sets. Initially, the high-dimensional data are compressed into a lower-dimensional Euclidean space using random projections. Subsequently, estimation proceeds using cyclic monotonicity moment inequalities implied by the multinomial choice model; the estimation procedure is semi-parametric and does not require explicit distributional assumptions to be made regarding the random utility errors. The random projection procedure is justified via the Johnson-Lindenstrauss Lemma: - the pairwise distances between data points are preserved during data compression, which we exploit to show convergence of our estimator. The estimator works well in a computational simulation and in a application to a supermarket scanner dataset. | |
1403 | Duality in Dynamic Discrete Choice Models | Chiong, Khai X. Galichon, Alfred Shum, Matthew | 02/09/2015 | 40 | SSWP1403.pdf | Using results from convex analysis, we investiate a novel approach to identification and estimation of discrete choice models which we call the "Mass Transport Approach" (MTA). We show that the conditional choice probabilities and the choice-specific payoffs in these models are related in the sense of conjugate duality, and that the identification problem is a mass transport problem. Based on this, we propose a new two-step estimator for these models; interestingly, the first step of our estimator involves solving a linear program which is identical to the classic assignment (two-sided matching) game of Shapley and Shubik (1971). The application of convex-analytic tools to dynamic discrete choice models, and the connection with two-sided matching models, is new in the literature. | |
1402 | To Score or Not to Score? Estimates of a Sponsored Search Auction Model | Hsieh, YU-Wei Shum, Matthew Yang, Sha | 02/05/2015 | 50 | SSWP_1402.pdf | We estimate a structural model of a sponsored search auction model. To accomodate the "position paradox", we relax the assumption of decreasing click volumes with position ranks, which is often assumed in the literature. Using data from "Website X", one of the largest online market places in China, we find that merchants of different qualities adopt different bidding strategies: high quality merchants bid more aggressively for informative keywords, while low quality merchants are more likely to be sorted to the top positions for value keywords. Counterfactual evaluations show that the price trend becomes steeper after moving to a score-weighted generalized second price auction, with much higher prices obtained for the top position but lower prices for the other positions. Overall, there is only a very modest change in total revenue from introducing popularity scoring, despite the intent in bid scoring to reward popular merchants with price discounts. | |
1397 | Estimating Multinomial Choice Models Using Cyclic Monotonicity | Shi, Xiaoxia Shum, Matthew Song, Wei | 10/09/2014 | 31 | SSWP_1397.pdf | This paper proposes a new identification and estimation approach to semi-parametric multinomial choice models that easily applies to not only cross-sectional settings but also panel data settings with unobservable fixed effects. Our approach is based on cyclic monotonicity, which is a defining feature of the random utility framework underlying multinomial choice models. From the cyclic monotonicity property, we derive identifying inequalties without requiring any shape restriction for the distribution of the random utility shocks. These inequalities point identify model parameters under straightforward assumptions on the covariates. We proposed a consistent estimator based on these inequalities, and apply it to a panel data set to study the determinants of the demand of bathroom tissue. | |
1396 | Testing the Quantal Response Hypothesis | Melo, Emerson Pogorelskiy, Kirill Shum, Matthew | 09/29/2014 | 31 | SSWP1396.pdf | This paper devlops a formal test for consistency of players' behavior in a series of games with the quantal response equilibrium (QRE). The test exploits a characterization of the equilibrium choice probabilities in a QRE as the gradient of a convex function, which thus satisfies the cyclic monotonicity inequalities. Our testing procedure utilizes recent econometric results for moment inequality models. We assess the performance of the test using both Monte Carlo simulation and lab experimental data from a series of generalized matching pennies games. Our experimental findings are consistent with the literature: the joint hypothesis of QRE, risk neutrality and player role homogeneity is rejected in the pooled data, but cannot be rejected in the individual data for over half of the subjects. By considering subsets of cycle monotonicity inequalities, our approach also highlights the nature of QRE consistency violations. | |
1377 | Identifying treatment effects under data combination | Fan, Yanqin Sherman, Robert Shum, Matthew | 05/11/2013 | sswp1377.pdf | |||
1376 | Simple two-stage inference for a class of partially identified models | Shi, Xiaoxia Shum, Matthew | 05/11/2013 | sswp1376.pdf | |||
1375 | Words get in the way: The effect of deliberation in collective decision-making | Iaryczower, Matias Shi, Xiaoxia Shum, Matthew | 05/11/2013 | sswp1375.pdf | |||
1374 | Estimating dynamic discrete choice models via convex analysis | Chiong, Khai X. Galichon, Alfred Shum, Matthew | 05/11/2013 | sswp1374.pdf | |||
1367 | A Parimutuel-like Mechanism from Information Aggregation: A Field Test Inside Intel | Gillen, Benjamin J. Plott, Charles R. Shum, Matthew | 01/03/2013 | 43 | SSWP1367.pdf | Field tests of a new Information Aggregation Mechanism (IAM) developed via laboratory experimental methods were implemented inside Intel Corporation for sales forecasting. The IAM, which incorporates selected features of parimutuel betting, is uniquely designed to collect and quantize as probability distributions any dispersed, subjectively held information that might exist. The tests demonstrate the robustness of experimental results and the practical usefulness of the IAM. The IAM yields predicted distributions of future sales that are very accurate at short horizons; indeed, more accurate than Intel's official in-house forecast 59% of the time. A symmetric game model suggests why the IAM works. | |
1328 | Revealed preference tests using supermarket data: The money pump | Echenique, Federico Lee, SangMok Shum, Matthew | 06/28/2010 | 22 | sswp1328.pdf | ||
1326 | Nonparametric learning rules from bandit experiments: The eyes have it! | Hu, Yingyao Kayaba, Yutaka Shum, Matthew | 06/18/2010 | 36 | sswp1326.pdf | ||
1325 | Estimation of random coefficients logit demand models with interactive fixed effects | Moon, Hyungsik Roger Shum, Matthew Weidner, Martin | 05/28/2010 | 43 | sswp1325.pdf | ||
1324 | A simple estimator for dynamic models with serially correlated unobservables | Hu, Yingyao Shum, Matthew Tan, Wei | 05/28/2010 | 23 | sswp1324.pdf | ||
1323 | To elect or to appoint? Bias, information, and responsiveness of bureaucrats and politicians | Iaryczower, Matias Lewis, Garrett Shum, Matthew | 05/21/2010 | 40 pages | sswp1323.pdf | ||
1316 | Aggregate matchings | Echenique, Federico Lee, SangMok Shum, Matthew | 01/19/2010 | 43 pages | sswp1316.pdf | ||
1313 | The value of information in the court. Get it right, keep it tight | Iaryczower, Matias Shum, Matthew | 09/01/2009 | 38 pages | sswp1313R.pdf |