The Impact of Promotions on Store Visits: A Counterfactual Approach

Abstract: This thesis empirically quantifies the impact of promotions on store visits in the Swedish grocery retailing sector with nationally representative panel data on household purchases of ground coffee. Using the potential outcomes framework, the impact is calculated as the difference between outcomes with promotion and their counterfactuals estimated with two regression models. The first model is an OLS fixed effects model used by market research firm GfK and the second is a Poisson fixed effects model. The Poisson model’s identification of the promotion effect is shown to be superior by accounting for that the dependent variable is discrete, the heterogenous time effects in the cross-section, and possible brand-switching behaviour. Standard errors robust to heteroscedasticity and cross-sectional and serial correlation are estimated for inference of the promotion effect under spatio-temporal dependence. A procedure for obtaining counterfactuals with regression models under multiple concurrent and continuous treatments is presented and an estimator of the cumulative treatment effect with adjustment for spatio-temporal dependence is derived and used to estimate the promotion impact on store visits. The findings are valuable for companies in market research, retailing and consumer packaged goods. The contributions of the thesis are methods for estimating promotion impact and an improvement of GfK’s methodology.