Sampling Based Approaches for Estimating Heterogeneous Discrete Choice Models
This paper aims to investigate the effect of marketing activities on the probability that an individual purchases a specific brand from a set of substitutable alternatives and the individual-specific heterogeneity of intrinsic preferences. To circumvent the Independence of Irrelevant Alternatives, the discrete choice model that is employed is the probit model. This paper estimates the parameters replicating the work of Chintagunta and Honore (1996) by using the Method of Simulated Moments. The results of this method show a trade-off between running-time and accuracy, making it an inefficient method. The research is extended by estimating the parameters using the Gibbs sampling approach, which proves more efficient. Incorporation of heterogeneity is done using either a brand-loyalty variable, which provides biased results due to the incorporation of purchase-history, or a random-effects model, which assumes that the heterogeneity is distributed normally over the population. This research shows that the method that provides the least biased results is the Method of Simulated Moments, with incorporation of heterogeneity by means of a random-effects model.
|Thesis Advisor||Gruber, K.|
Spaendonck, C.E.E. van. (2018, September 12). Sampling Based Approaches for Estimating Heterogeneous Discrete Choice Models. Econometrie. Retrieved from http://hdl.handle.net/2105/43357