Which Brand Purchasers Are Lost to Counterfeiters? An Application of New Data Fusion Approaches

Firms and organizations often need to collect and analyze sensitive consumer data. A common problem encountered in such evidence-based research is that they cannot collect all essential information from one sample, and may need to link non-overlapping data items across independent samples. We propose an automated nonparametric data fusion solution to this problem. The proposed methods are not restricted to specific types of variables and distributions. They require no prior knowledge about how data at hand may behave differently from standard theoretical distributions, automate the process of generating suitable distributions that match data, and therefore are particularly useful for linking data with complex distributional shapes. In addition, these methods have strong theoretical support, permit highly efficient direct fusion to relate a mixture of continuous, semicontinuous, and discrete variables, and enable nonparametric identification of the entire distributions of fusion variables, including higher moments and tail percentiles. These novel and promising features overcome important limitations of existing methods and have the potential to increase fusion effectiveness. We apply the proposed methods to overcome data constraints in a study of counterfeiting. By combining datasets from multiple sources, data fusion provides a feasible approach to studying the relationship between counterfeit purchases and various marketing elements, such as consumers' purchase motivations, behaviors, and attitudes, brand marketing channels, promotions, and advertisements. Therefore, data fusion sheds light on counterfeit purchase behaviors and suggests ways to counter counterfeits that would not be available if these datasets were analyzed separately.