Similarity-Based Spatial Methods to Estimate Shelf Space Elasticities

A retailer may allocate shelf space to brands based on factors, unobservable to researchers, which also determine sales. As a consequence, both sales and shelf space are endogenous in historical data, and this leads to inconsistent estimates of shelf space elasticities based on OLS. To obtain valid estimates of shelf space elasticities for allocation decisions, we propose an approach that incorporates the spatial correlation between shelf space and the error term resulting from store-, consumer- and competitor characteristics. The empirical results suggest that our model based on a single cross section of stores corrects for endogeneity and provides valid shelf space elasticities. We also obtain superior predictions compared to several benchmark models. With the same cross section and two observations over time, the alternative methods we use provide comparable shelf space elasticity estimates. However, our proposed method is still superior in the sense that its estimates have somewhat smaller standard errors.