Giving some consumers a clean credit slate doesn’t necessarily lead to more loans overall, research suggests. And wiping the slates for some consumers can hurt others, find New York University’s Andres Liberman, Princeton’s Christopher Neilson, the Chilean Banking Association’s Luis Opazo, and Chicago Booth’s Seth Zimmerman.
In 2012, Chile’s Congress passed a law ordering credit reporting companies to stop disclosing delinquencies on small loans incurred prior to 2012. The one-time data wipe affected 2.8 million Chileans, or about 20 percent of the adult population. The goal of the policy was to help mostly lower-income consumers with small default amounts, particularly those affected by the 2010 earthquake.
But the action backfired, according to analysis by the researchers. People whose defaults were deleted from the registry borrowed more after the policy went into effect. However, borrowing fell for the 63 percent of consumers who had no defaults. After the deletion, lenders could no longer tell that these people were better credit risks than the defaulters.
Consumer borrowing dropped by 3.5 percent overall. Losses from information deletion outweighed gains in this setting, write Liberman, Neilson, Opazo, and Zimmerman.
The policy had the greatest negative impact on borrowers from lower socioeconomic groups. To underwriters, after the information was deleted, these profiles most closely resembled those of loan defaulters. Borrowing by those from lower socioeconomic classes declined 9 percent, compared with a 6 percent decrease for wealthier individuals, the researchers find.
According to the research, only about 600,000 people benefited from the clean-slate policy, while 2 million found it harder and more costly to procure loans. “The borrowers whose cost predictions rise most following deletion are those who resemble high-cost borrowers along these dimensions,” they write. “The deletion policy reduced overall borrowing, with declines for non-defaulters offsetting gains for defaulters.”
The researchers linked borrowing outcomes with profiles that showed up in their data set, then used machine learning to recreate two risk-prediction models. One model mimicked how banks qualify groups of borrowers based on the limited data available. Another model looked at risk predictions when all the data—including deleted information on defaults—were available. Comparing the two highlighted who was being denied credit or forced to borrow at higher rates after the policy change.
Similar adverse effects of deleting information are evident in various forms around the world, the researchers write. In the United States, ban-the-box laws implemented by some states restrict employers from asking about an applicant’s criminal history on a job form, but a study by Rutgers’s Amanda Agan and University of Michigan’s Sonja Starr finds that such laws encouraged racial discrimination. In their study, after such laws took effect, employers were less likely than before to call applicants with African American–sounding names. Similar effects also show up in health-insurance markets when policies create information asymmetries.
Zimmerman notes that in a world increasingly driven by large amounts of data, people are rightly concerned about equity and privacy. However, information can help markets function, and policy makers and businesses should weigh the benefits of information against the costs. “What you often get,” he says, “is that policies that restrict information hurt the kinds of people you intend to help.”