“…Therefore, we also address the concern of possible self-selection bias by using various propensity score matching methods in which the group with interest bonus contracts serves as the 'treated' group and all other 'untreated' contracts do not have an interest bonus clause. The aim of this statistical method is to provide an unbiased estimation of the treatment effects (i.e., in our case, an unbiased self-selection free effect of qualified interest payments on customer behavior) (Rosenbaum and Rubin, 1983;Lu, 2012;Wu, 2010;Autore et al, 2009). We perform a variety of matching procedures on the complete sample of approximately 2.2 million con-28 Most often, the contracts are negotiated one after the other (i.e., the customer saves for several years, ends one contract and starts a new one).…”