To Reduce or to Structure: on Mixed Method Complementarity

Researchers seeking to identify a causal treatment effect of interest often gravitate toward reduced-form modeling approaches, while those interested in characterizing the structure of demand gravitate toward structural models of the full market environment. In this paper we demonstrate that, rather than operate as perfect substitutes, reduced-form and structural approaches can play a complementary role in icharacterizing market dynamics and policy implications. These opportunities are particularly ripe in regards to questions of food policy and marketing strategy impacts, as researchers frequently must balance the need to fully characterize demand and potential feedback loops with the desire to interpret estimated objects in a causal fashion. We provide an example of the complementary use of mixed empirical methods: first we utilize an event study framework to estimate the changes in alcoholic beverage and non-alcoholic beer purchasing in response to the adoption of county-level COVID-19 stay-at-home policies. Second, we estimate a structural model of differentiated alcoholic beverage products that can provide novel insight into the substitution at play behind the growth of the non-alcoholic beer market. Taken together, the results from these two empirical approaches provide novel insight into recent dynamics in the alcoholic beverage market and carry important implications for future food and beverage marketing strategies.