Abstract
We use monthly sales of all wines, beer, and spirits sold between 2006 and 2011 by Sweden’s retail monopoly on alcohol to estimate the causal effect of retail distribution on market share by volume at the product level. Products are defined at the level of the stock-keeping unit. Two institutional features are key to identifying the causal effect: First, the monopolist uses four levels of retail distribution; a change in retail distribution is therefore associated with a discrete shift in the number of stores that carry a product in a given month. Second, the retailer is legally bound and monitored by the European Union to ensure that it acts in a non-discriminatory manner with all its suppliers. These features allow us to rule out many possible confounding factors in estimating the effect of distribution on sales volume. We find large and statistically significant effects from changes in retail distribution on market share by volume across all levels of retail distribution. The associated volume elasticity of retail distribution is convex; the wider the retail distribution the greater the marginal volume increase from further widening. In this market, wider distribution means reaching stores with successively smaller assortment. Our results indicate that the smaller assortment in smaller stores, coupled with a low resistance to compromise, is the main reason for the convex pattern. In other words, convexity appears to be generated by products achieving “a larger share of a smaller pie” as retail distribution expands.