Random-Coefficients Logit Demand Estimation with Zero-Valued Market Shares

This paper provides a new estimator to address the zero-valued market shares in the BLP random-coefficient logit demand estimation.

Quantifying the Economic Value of Warranties: A Survey

This paper is a review of the most recent literature regarding the econometric modelling of the impact of warranties on demand. The reviewed literature is limited to the papers that apply the random-coefficient logit model based on Berry, Levinsohn and Pakes (1995) to estimate differentiated products demand. An important feature of these demand system models that is a clear advantage to earlier demand functions is to account for the endogeneity of prices. We focus on those model specifications that take into account endogeneity of both prices and warranty. Another goal for modelling the effect of warranties is to explore the economic rationale for warranty provision. Four theories have been proposed in the literature: insurance, sorting, signalling and incentive theories. This paper aims at decomposing the effect of these theories, to account for different underlying assumptions and to separately determine the implications as presented in the recent literature

Monte Carlo Simulation in Random Coefficient Logit Models Involving Large Sums

We study Monte Carlo simulation in some recent versions of random coefficient logit models that contain large sums of expressions involving multivariate integrals. Such large sums occur in the random coefficient logit with demographic characteristics, the random coefficient logit with limited consumer information and the design of choice experiments for the panel mixed logit. We show that certain quasi-Monte Carlo methods, that is, so-called (t, m, s)-nets, provide improved performance over pseudo-Monte Carlo methods in terms of bias, standard deviation and root mean squared error.

Challenges in Merger Simulation Analysis

In this paper, we share our experience with merger simulations using a Random Coefficient Logit model on the demand side and assuming a static Bertrand game on the supply side. Drawing largely from our work in Knittel and Metaxoglou (2008), we show that different demand estimates obtained from different combinations of optimization algorithms and starting values lead to substantial differences in post-merger market outcomes using metrics such as industry profits, and change in consumer welfare and prices.