Christopher T. Conlon

Demand Estimation under Incomplete Product Availability

Incomplete product availability is an important feature of many markets; ignoring changes in availability may bias demand estimates. We study a new dataset from a wireless inventory system installed on 54 vending machines to track product availability every four hours. The data allow us to account for product availability when estimating demand, and provides a valuable source of variation for identifying substitution patterns. We develop a procedure that allows for changes in product availability even when availability is only observed periodically. We find significant differences in demand estimates, with the corrected model predicting significantly larger impacts of stock-outs on profitability.

Effects of Product Availability: Experimental Evidence

Product availability impacts many industries such as transportation, events, and retail, yet little empirical evidence documents the importance of stocking decisions for firm profits, vertical relationships, or consumers. We conduct several experiments, exogenously removing top-selling products from a set of vending machines and analyzing substitution patterns and profit impacts of the changed product availability using non- parametric analyses and structural demand estimation. We find substantial switching to alternate products, and evidence of misaligned incentives between upstream and downstream firms in the choice of which products to carry. We discuss the trade-offs of both empirical approaches for analyzing product availability effects generally.

The Empirical Likelihood MPEC Approach to Demand Estimation

The family of Generalized Empirical Likelihood (GEL) estimators provide a number of potential advantages relative to Generalized Method of Moments (GMM) estimators. While it is well known these estimators share an asymptotic distribution, the GEL estimators may perform better in finite sample, particularly in the case of many weak instruments. A relatively new literature has documented that finite-sample bias in the demand estimation problem of Berry, Levinsohn, and Pakes (1995) is often large, especially in the absence of exogenous cost shifting instruments. This paper provides a formulation for a computationally tractable GEL estimator based on the MPEC method of Su and Judd (2012) and adapts it to the BLP problem. When compared to GMM, the GEL estimator performs substantially better, reducing the bias by as much as 90%. Furthermore, it is possible to use analytic bias correction to reduce the bias even more and obtain accurate estimates with relatively small numbers of markets.

The Price of Liquor is Too Damn High: Alcohol Taxation and Market Structure

We study the relative benefits of taxation versus market structure regulations for distilled spirits. One popular regulation, called post and hold, helps wholesalers set collusive prices as the competitive equilibrium of a single period game. Assembling new datasets of wholesale and retail prices, and sales, we show PH leads to average wholesale markups of 30-40%, with higher markups on expensive products. Taxes distort relative prices less than PH. We show Connecticut could increase tax revenue by 350% and improve consumer welfare while holding alcohol consumption fixed. However, we also show our counterfactual policy may be slightly regressive compared to PH.

Discrete Prices and the Incidence and Efficiency of Excise Taxes

This paper uses detailed UPC-level data from Nielsen to examine the relationship between excise taxes, retail prices, and consumer welfare in the market for distilled spirits. Empirically, we document the presence of a nominal rigidity in retail prices that arises because firms largely choose prices that end in ninety-nine cents and change prices in whole-dollar increments. Theoretically, we show that this rigidity can rationalize both highly incomplete and excessive pass-through estimates without restrictions on the underlying demand curve. A correctly specified model, such as an (ordered) logit, takes this discreteness into account when predicting the effects of alternative tax changes. We show that explicitly accounting for discrete pricing has a substantial impact both on estimates of tax incidence and the excess burden cost of tax revenue. Quantitatively, we document substantial non-monotonicities in both of these quantities, expanding the potential scope of what policymakers should consider when raising excise taxes.