Two essays on a new choice model for lexicographic preference and a choice based diffusion model for multi-generation and multi-country data

Abstract

first one is controlled by country-based variable and the other is affected by generation-based variable. This structure allows us to compare country specific differences. For empirical analysis, we estimate 3G and 4G connections in 25 countries. Pooled model estimation with random effects shows better than individual country model and it shows better performance in estimation period and forecasting period.

There is a famous saying, “Life is full of choices”. This dissertation consists of two essays focusing on the choice behavior. First essay proposes a choice model considering lexicographic preference. Following essay addresses a framework covering the multi-country and multi-generation diffusion to extend the existing choice-based diffusion model. In the first chapter, we propose a new choice model under lexicographic preference. Under this preference structure, it is not easy to find the optimal marketing mix strategy. First, there is no simple functional form to link the ordinal preferences revealed in choice observations to measurable utility that allows us to conduct policy experiments for marketing mix variables. Second, the latent process of attributes ordering according to their importance cannot be easily calibrated in the conventional manners. In this paper, we propose a new choice model to measure lexicographic preference, and introduce a data augmentation technique to estimate the latent ordering process. We apply the proposed model to empirical analysis for laptop choice. We find that proposed model improves its model fit and forecasting performances compared to the conventional model. In the second chapter, we address a framework covering multi-country and multi-generation diffusion processes. Our model focuses on choice-based diffusion model. We decompose the choice probability for generation into two components