Optimal Targeting under Gender Fairness

Abstract

While targeted marketing campaigns can offer high potential for increased firms’ profit, they often lack due consideration for fairness among different protected demographic groups. We investigate methods to mitigate gender disparities for both firm’s actions and benefit outcomes in the setting of offer allocations for targeted marketing campaigns. We develop and compare four optimization models to identify the optimal policies that maximize the firm’s financial return while concurrently satisfying relevant gender fairness conditions. Our results reveal that only regulating the gender disparity in the firm’s actions is not sufficient to guarantee that the responding customers of either gender receive similar level of discount benefit. Hence, we recommend firms to design policies by directly solving for the same level of benefit outcomes instead of firms’ actions across gender. Among the four models developed in this thesis, the optimal transport model is the only model that simultaneously meet both the group fairness condition in aggregate and the conditional demographic parity condition within each socioeconomic segment. Our results in the empirical setting show that the optimal policies from the optimal transport model achieve the lowest gender disparity in overall benefit outcomes. These policies also demonstrate the minimum level of firm manipulation across the four models, and provide the most discounts to the most female-concentrated neighborhoods.