“Using Grocery Data for Credit Decisions”, 2024-07-15 (; similar):
Many consumers across the world struggle to gain access to credit because of their lack of credit scores. This paper explores the potential of a new alternative data source, grocery transaction data, for evaluating consumers’ creditworthiness.
Our analysis takes advantage of a unique, individual-level match of credit card data and supermarket loyalty card data.
By developing credit scoring algorithms that either exclude or include grocery data, we illustrate both the incremental value of grocery data for credit decisions and its boundary conditions. We demonstrate that signals from grocery data can improve credit approval decisions, particularly for individuals who lack traditional credit scores. Furthermore, as a consumer establishes a relationship with lenders and builds a credit history, the marginal value of incorporating grocery data diminishes.
These findings highlight the potential of grocery data in informing credit decisions and, consequently, in enabling financial institutions to extend credit to consumers who lack traditional credit scores.
…In particular, using a customer identifier, we merge the supermarket’s loyalty card data and the issuer’s credit card spending and payment history at the individual level for the consumers who appear in both data sources between January 2017 and June 2019. The merged data allow us to observe how 30,089 consumers behave in the two seemingly different domains…We use a new, proprietary data set from an anonymous conglomerate that operates both a credit card issuer and a supermarket chain. The credit card issuer offers general-purpose credit cards that can be used at any merchant that accepts the associated processing network. The supermarket chain sells a wide range of products in various categories, including groceries, household supplies, clothing, and other general merchandise. [Walmart?]
…We find that what one buys can explain what type of payer one is even after controlling for various sociodemographic variables and credit scores. For instance, buying cigarettes or energy drinks is associated with a higher likelihood of missing credit card payments or defaulting, whereas purchasing fresh milk or vinegar dressings is linked to consistently paying credit card bills on time.
Using item-level survey ratings, we find suggestive evidence that buying healthier but less convenient food items is predictive of responsible payment behaviors.
Furthermore, we observe a positive and robust correlation between displaying greater consistency in various dimensions of grocery shopping behavior and making timely credit card bill payments. For example, cardholders who consistently pay their bills on time are more likely to shop on the same day of the week, spend similar amounts across months, and purchase the same brands and product categories.
…To predict the credit risk of unscored consumers or those without credit scores, the lender often relies solely on sociodemographic variables, such as income. In these scenarios, incorporating grocery data substantially improves predictive accuracy, increasing out-of-sample predictive power by 3.11–7.66 percentage points, as measured by the area under the receiver operating characteristic curve (AUC). When it comes to consumers with credit scores, we find that grocery data, when used in isolation, can achieve predictive accuracy comparable to that of credit scores alone. This result implies that individuals’ nonfinancial behaviors can provide credit risk signals of similar value to traditional credit scores.
However, grocery data is not a perfect substitute for credit scores as there is a smaller yet positive incremental predictive gain from grocery data even relative to credit scores. More precisely, when both sociodemographic variables and credit scores are available, the incremental predictive power introduced by grocery data ranges from 0.359–2.51 percentage points in the out-of-sample AUC. Taken together, these results suggest that grocery data complements rather than substitutes traditional financial data, such as sociodemographic variables and credit scores.
…We find that implementing this two-stage decision rule leads to a 1.46% increase in per-person profits among applicants without credit scores. This increased profitability is driven by the improved risk profile of approved applicants as the rule effectively filters out defaulters, who experience a higher likelihood of rejection in the second stage than non-defaulters. By contrast, for applicants with credit scores, the impact on credit approval decisions and profitability is minimal with a 0.025% increase in per-person payoffs. These findings collectively suggest that, under the particular decision rule we consider, there may be a stronger motivation for the lender to acquire, collect, and leverage grocery data for evaluating applicants who lack a traditional credit score.
View PDF: