|
Abstract:
|
Experimental designs are often used to evaluate the effectiveness of new initiatives. However, it is not always possible to randomly assign observations to control and test groups in a business setting. For observational studies, propensity matching can estimate the effect of the treatment in the presence of covariates. Propensity scores are usually applied in two broad contexts: thoughtful selection or the kitchen sink approach. In the former, we are limiting the number of variables due to the belief of causal relationships. However, others have argued that a propensity score model is purely predictive and thus there is no harm in including numerous potential covariates. Though, adding variables geometrically increases sample size requirements. In this market research study, we examine these two basic approaches for a hospitality initiative. The results are presented using a variety of matching algorithms under both approaches.
|
