Across a wide variety of contexts, people who have experienced an event are more likely toexperience that event in the future. This empirical regularity, hereafter referred to as state dependence, has two explanations, each with its own set of policy and managerial implications. One
explanation, known as structural state dependence, is that the experience of an event alters the
preferences or constraints that an individual would hold for that event in the future. A second
explanation, spurious state dependence, is that people differ along some unobservable propensity
to experience an event. For example, people that become unemployed once are more likely to be
unemployed in the future. The structural explanation for this phenomenon is that unemployment
has a sustained effect on the probability of future unemployment, while the spurious explanation
argues that individuals vary on some unobservable variable, such as work ethic or skill set, that
affects their probability of becoming unemployed at any time. These explanations have different
implications: if state dependence is structural, short-term policies reducing unemployment can have
large long-run effects. This dissertation aims to explore the effects of structural state dependence
in three contexts: brand choice, store choice, and category consumption.
Using techniques in causal inference and structural modelling, and a rich database of transactiondata, I find that structural state dependence 1) has no effect on brand choice in consumer packaged
goods, 2) has a strong effect on where people shop for groceries, impacting nutritional intake, and
3) drives consumption in addictive categories to varying extents. These findings give us a better understanding of why brand choice persists over time, why nutritional intake varies drastically
across demographic groups, and how cigarette types vary in their addictive and habit-forming
properties.