Do incentives to save lead to borrowing? Evidence from a large-scale study
A plenty of of policies aimed at increasing savings are currently in place, with many of them utilizing nudges. These policies are based on the assumption that savings are financed by lower consumption. However, when policymakers or researchers evaluate these interventions, they frequently focus on immediate savings outcomes without considering where the money comes from, despite the fact that a large proportion of households have liquid savings while also carrying credit card debt.
We investigate whether saving nudges increase credit card borrowing in a recent working paper. We combine data from a large-scale field experiment with detailed and accurate panel data on individual bank accounts and credit cards. The information came from Banorte, a Mexican bank that conducted a randomized experiment with 3,054,438 customers. The remaining 374,893 customers received no messages, while 2,679,545 received weekly or bi-weekly ATM and SMS messages encouraging them to save for 7 weeks during fall 2019.
We pay special attention to the borrowing response of individuals with the largest predicted response to the nudge. Potential unintended consequences of saving nudges are more likely for them. We use a causal forest to predict treatment effects at the individual level in order to identify them. The causal forest allows us to avoid the over-fitting issues that would inevitably arise if we manually searched for subpopulations with large treatment effects. A manual search would attribute large treatment effects to subpopulations where some observations show unusually large savings due to idiosyncratic shocks that could also affect borrowing outcomes. The causal forest, on the other hand, is based on a repeated split-sample procedure in which one sample is used to partition the covariate space and another to estimate the treatment effects. This rule outs the possibility that pre-treatment covariates predict a large treatment effect due to idiosyncratic shocks that may also influence other outcomes, such as borrowing decisions.
Then, among those in the top quartile of predicted treatment effects, we examine the saving and borrowing behavior of people who have a credit card. Savings increased by 6.1 percent from a base of 31,702 MXN (3,392 USD) in the control group (1 MXN = 0.107 USD PPP). This represents a 1,948 MXN increase (208 USD). As illustrated in Figure 2, we rule out the possibility that these increased savings were accompanied by an increase in debt, as we find no significant effect on credit card borrowing. Figure 2 also shows a similar pattern for people who had a credit card and paid credit card interest at the start, or who had a credit card and had a large credit limit available on it.
Comments
Post a Comment