New study from King's Business School and the Federal Reserve Board shows most credit-limit increases are automated, targeting borrowers already in debt.
As shoppers across the United States prepare for the biggest spending day of the year, new research from King's Business School and the Federal Reserve Board reveals that most credit limit increases are not requested by consumers but are automatically applied by banks' algorithms, often to borrowers already in debt.
The study, Automated Credit Limit Increases and Consumer Welfare, finds that roughly four in five credit limit increases in the United States are initiated by banks rather than consumers. These automatic increases now account for more than $40 billion in additional available credit every quarter, most of it extended to customers who already carry balances. Borrowers respond by increasing their revolving balances by around 30 percent, suggesting that algorithmic decision-making is a major but largely hidden driver of household debt.
The research also finds that banks are significantly more likely to raise limits for borrowers who already owe money. About one third of all unpaid credit-card balances in the United States - the amounts people carry from month to month - exist only because of credit-limit increases made after the card was opened, rising to 60 percent among borrowers with lower credit scores. Banks that more frequently advertise their usage of AI and machine-learning tools in their official financial reports are the same ones most likely to use automated systems to raise customers' credit limits.
The authors used a model of household spending and borrowing to test policy approaches like those in the United Kingdom, where banks cannot raise credit limits for indebted customers without their consent, and Canada, where banks are required to obtain consumer consent for any credit limit increases. They find that adopting comparable safeguards in the United States would improve overall consumer welfare by around one percent and reduce revolving debt balances, as well as the share of income going towards interest payments, with only a modest effect on credit availability. The EU plans to implement similar regulation next year.
The paper draws on detailed regulatory microdata covering more than 70 percent of the US credit-card market, reported through the Federal Reserve's Capital Assessments and Stress Testing framework. It quantifies for the first time the welfare impact of automated credit-limit increases and the potential benefits of stronger consumer-protection oversight.
Banks are using increasingly sophisticated models to predict which customers will borrow more if their limit is raised. For many, that means an automatic increase they never asked for and may not fully understand. These decisions are shaping household debt across the country in ways most borrowers don't see. Automated credit-limit increases can expand access to credit and help households smooth consumption. But our findings show that when algorithms target borrowers already in debt, the result is often higher borrowing and greater financial vulnerability. Our model suggests that modest regulation, such as requiring consent or limiting increases for indebted customers, could improve welfare for many households while only slightly restricting access to credit. It's an example of how well-designed policy can guide the use of data-driven decision-making in finance.
Dr Agnes Kovacs, Senior Lecturer in Economic at King's Business School
Read the full paper in the Finance and Economics Discussion Series (FEDS) by visiting the website.
