Most experts debating bail-in and bailout strategies agree that banks should build capital and shrink balance-sheets as the best way to avoid a collapse and rebuild after one. But researchers are suggesting a more personalised version of that recipe that makes the difference between life and death for struggling firms.

In a recent paper, Bank of Italy’s Emilia Bonaccorsi di Patti and the University of Chicago’s Anil Kashyap found that banks which successfully recover from sharp drops in profitability have something in common: They avoid throwing good money after bad by resolutely shutting off credit to their riskiest clients.

Banks that increase risky lending in hopes of boosting profitability tend to lose the gamble and fail to recover. The general rule appears to be the same as in casino gambling: Losing is OK, but trying to win back the losses is truly dangerous.

The analysis is based on a sample of 110 Italian banks that suffered steep profitability drops between the early 1990s, when Italy suffered an economic slowdown, and the early 2000s. The idea was to observe the banks throughout an entire economic cycle.

The economists found that the level of “idiosyncratic risk” — that is, the risk of management decisions to lend to riskier clients — was the determinative factor. Lenders that recovered cut such loans by 13 percentage points a year. Banks that didn’t only cut them by 4 percentage points.

The criteria for customer riskiness that the researchers used required some hindsight — data was available on which of them would default within a few years — but other criteria, such as the firms’ lack of operating profits or their defaults on other banks’ loans, were transparent to the bankers at the time.

Unprofitable companies, and especially those that have failed to repay a loan, have an increased appetite for borrowing: They’re trying to save themselves. This creates a warped set of incentives for a bank that finds itself in trouble: Its managers and shareholders are motivated to accept the risk.

Bonaccorsi di Patti and Kashyap don’t dwell on this in their paper, but distressed banks’ self-destructive behaviour with risky clients is often about relationships. There’s a reason why the recent financial crises have claimed more victims among stakeholder-oriented banks than among profit-seeking ones: Spanish cajas, German landesbanks and Italian regional lenders all had long histories with their problem clients. Deep, emotional connections were often involved. Bankers were willing to overlook the signs of oncoming problems; essentially, they were willing to sink or swim with these clients. That’s not the same as gambling to increase profits, but it is perhaps even more dangerous.

Even today, regulators often tend to miss these time bombs under banks’ balance-sheets. They find it easier to focus on formal criteria, such as capital ratios.

It makes certain sense: Instead of doing due diligence on banks’ customers, it’s easier to make sure capital is sufficient to cover losses if a certain percentage of assets go bad. But Bonaccorsi di Patti and Kashyap see the stress tests used today as too generalised: “Our analysis suggests that it might be useful to keep track of which customers’ loans would go bad during a crisis and to consider how readily a bank can manage these problems. In particular, indiscriminate reductions in lending seem to be less important than managing the credit extension to the riskiest borrowers.”

It could be too much to ask for financial regulators to look deeper into banks’ client bases. But it makes sense to check to which sectors a bank is lending, how sensitive these sectors are to economic downturns, and how deeply the bank is involved with customers in less resilient industries.

It also makes sense, once a bank shows signs of trouble, for a regulator to restrict those of its activities that carry the most risk.

The current approach to banking problems is overly mechanical, too focused on one-size-fits-all rules. Recent Italian bank rescue efforts, which only loosely complied with those rules, have been criticised as undermining the system.

But while firm framework rules are necessary, perhaps bank regulation would benefit from a more personalised approach. It could pre-empt the need for large-scale rescues.