These are the findings of a proof of proof “where we have shown that the behavior of people’s purchases in grocery stores is associated with repayment of credit card accounts,” he says Eric AndersonProfessor of Marketing at Kellogg School, who worked with Jung young lee
from Rice University and Joonhyuk yang From the University of Notre Dame for research.
They found that people who bought grocery stores that were the basis of home meals, such as beans and milk, were less likely to prepare for their credit card payments. And that those who were disciplined – shopping on a fixed timetable, sticking with a budget and buying items sold – were also more likely to pay their credit card in time.
Based on this previous project, the researchers explored how the behavior of people’s shopping in five retail sectors, including groceries, could be used to create an alternative credit result for those who do not have an official credit history.
Market stories can be like a window that gives lenders a better view of people’s real financial credibility, Anderson says. And this additional information could help lenders feel more comfortable expanding credit to 1.4 billion people Worldwide that do not have access to official financial services.
“Businesses can use this score to make documented decisions on the credit ratings of people who do not have conventional credit ratings,” he says.
A problem of chicken and eggs
Anderson notes that the study of the credit rating of people who do not have an official credit history creates a problem of chicken and eggs. They do not have a credit history because they rarely receive loans, but to get a loan, they need a credit history.
“One of the main problems we face in this project is that people who do not receive traditional loans can be different in important ways from the population receiving loans,” Anderson says, “and thus their behavior forecasts do not end up being so large because they are simply not represented in the data.”
The researchers were able to work around this challenge, collecting and comparing information from different data sets.
The first came from researchers’ collaboration with a Peruvian group operating five types of retail chains: supermarkets, pharmacies, department stores, shopping centers and home improvement shops. Information was gathered for more than 45,000 consumers who made at least one purchase in a store between May 2021 and September 2023. They looked at variables such as the way consumers responded to promotions, how often they returned products and what kind of products they bought. These consumers also requested a general purpose credit card issued by the Group during this period. Some of these applicants had credit stories. Some didn’t. Some were approved for the card. Some weren’t.
Another data flow came from information bought by the Group from a credit rating agency in 2022, including details on the timeliness and consistency of individual Bill payments. The Group had used this data in a pilot program where it experimented with the approval of credit card applicants who did not have official credit stories.
Finally, the researchers gathered a complete monthly snapshot of people’s credit stories-including information, such as pending debts, credit card use rates and payment standards-from a register managed by Peru’s financial regulatory authority.
Information from retail behavior
The merger of these data allowed Anderson and his colleagues to create alternative credit ratings for separate groups of people: those with a credit history approved for the group’s credit card. Those who do not have a credit history approved for the card based on their payments. And those who have not had a credit history that had not been approved for the Group’s card but whose names are still appearing in the national register since then were Was approved for credit by other lenders.
The researchers used a series of algorithms to answer two questions about these groups of people. To whom did the group decide, in fact, to extend the credit and what was the result of default rates? And secondly, how different would the picture look like if the group offered credit based on the alternative credit result instead of conventional methods?
After simulating the approval decisions on the basis of these alternative credit ratings, Anderson says: “We show that the score we produce is much better in predicting credit risk than [traditional credit scores] or utility accounts only. ”
In addition, the researchers used this alternative credit result to prove various ways in which the Peruvian group could improve its credit decisions.
First, using the alternative credit result, the company could maintain the same risk limit for the approval of applicants for credit cards, while the credit card approval rate for people without a credit history-from 15.6 % to 47.8 %. The downside is that it would have increased the rate of default for these customers by 0.76 %, which “in the banking world is a large enough number,” says Anderson.
Alternatively, the company could maintain its current interest rate and double its approvals for those who do not have a conventional credit result, from 15.6 % to 31.3 %.
Finally, if the company preferred to maintain a 15.6 % approval rate for those who do not have credit history, the use of an alternative credit approach could reduce the rate of default for these customers from 4.74 % to 3.31 % – again, a large amount in the banking world.
The challenge of applying this model
Building and placing this model in practice, Anderson admits, presents some challenges. First of all are privacy regulations. Laws that protect consumer data, while being important on many fronts, prevent researchers and newly established companies invested in building non -conventional credit models. No inflow on what people buy and when information that is not widely available in the US – models don’t work.
Lenders interested in providing credit to borrowers without an official credit history may also need to review conventional approaches to customer default.
“Populations without a credit history can create the same amount of income as other customers, but they are often not so consistent that it can lead to lost payments,” says Anderson. “Part of this project is to understand new ways of supporting these borrowers. If companies are merely punitive and cut off the credit after the first lost payment, which may not be an effective strategy in the long run.”
Finally, and perhaps more fundamentally, companies may need to reassemble their appetite for dangers. Borrowing for “unpaid” people presents an opportunity to make more money and improve the economic well -being of society, but the researchers’ model is not perfect. It does not record, for example, the population of consumers who never appeared in the National Registry – that is, they never received a loan from the group or Other institutions.
To really validate this kind of alternative credit result, lenders such as the Peruvian needs group must try it in the real world, says Anderson.
“But this requires lenders to take a risk and this is the volume in the document: businesses could make more money by lending more people, but they believe it is a dangerous team and so they do not lend them and receive the information they need,” he says. “As long as these beliefs are maintained, we will never solve the problem.”
