This study proposed a mixed linear modeling
framework for revealing the average treatment effects of coupon promotions as
well as for figuring out what kind of customers are more responsive to coupon
promotions. The analysis used detailed data on 2,469 households who are frequent
shoppers at a retailer over one year. A series of models starting from fixed
parameters OLS regression to more flexible random coefficient models were
built. The results showed that receiving more coupon promotions does not lead
to higher spending among households. None of the available demographic
characteristics was significantly linearly related to the responsiveness to
coupon promotions according the initial model with cross-level interaction
terms. However, exploratory analysis of random slopes with the help of a
regression tree allowed to correct the specification, which allowed confirming
that people aged 25+ with annual household income exceeding $25,000 are more
responsive, yet their response is insignificantly different from zero.
Keywords: Mixed-Effects Model,
Multilevel Analysis, Heterogeneous Responsiveness, Promotions.