Consumer-firm relationships in limited information subscription markets: An empirical analysis of consumer choice in automobile insurance

Consumer purchases in many markets are characterized by a choice between many firms and a long-term relationship with the firm selected. This thesis addresses the source of these relationships and the impact of accrued experiences on future purchase choices. Chapter 1 defines a class of markets—limited information subscription markets—in which these relationships are likely to be particularly important, and provides descriptive statistics on a panel of Georgia auto insurance consumers, used throughout the thesis as an example of such markets. Chapter 2 takes up the most basic question about long-term consumer-firm affiliations—do consumers value the relationships they develop with firms and thus become less likely to depart with tenure? The observed pattern of falling departure probabilities suggests this possibility, but may simply reflect selection on unobserved consumer “match quality” with the firm. The chapter extends the methodology for separating these two explanations, by considering consumer departure probabilities given distinct price histories with the firm, holding the current price fixed. Any impact of these distinct price histories must reflect selection effects, providing a powerful source of identification in the data. Using this method the chapter finds that the majority of observed tenure dependence is explained by selection, particularly after 3 years with the firm. Chapter 3 addresses the impact of learning about the firm's service quality via claims experience. While the actual service quality received on a given claim is unobservable, the chapter uses the distinct patterns of departure probabilities in the periods following chargeable claims which impact both consumer information and prices—and non-chargeable claims—which do not impact prices—to identify the value consumers place on good service and the features of a Bayesian learning process. The chapter's principal findings are: large consumer valuation on good service; highly inaccurate prior expectations of service quality; potentially rapid learning given enough early service experiences; and limited actual learning, due to the slow arrival of claims and the accrual of tenure effects, which together imply that by the time a consumer observes a firm's service quality she is unlikely to depart.