Bank Runs Found to Spread Like a Contagion
Duke researchers used Google Earth to map ties between bank customers.
May 20, 2009
DURHAM, N.C. –- An analysis of social networks and minute-by-minute customer transactions during a run on an Indian bank has found that bank runs can spread through a community like a contagion.
Duke University finance professor Manju Puri and co-author Rajkamal Iyer of the University of Amsterdam assembled a unique data set on the bank run and then employed epidemiological techniques to determine connections between bank depositors, using Google Earth to map the locations of their home addresses.
Following the failure of a larger, neighboring bank, the customers of a bank that was fundamentally sound were withdrawing funds out of fear that they might not be able to access them in the future.
“Using these tools, we were able to track the ‘transmission’ of running from one customer to another, just as you would see with the spread of a disease," said Puri, the J.B. Fuqua Professor of Finance at Duke’s Fuqua School of Business. "We found that an individual customer’s decision to run or not was highly correlated with whether or not their neighbors had already run from the bank. We found clusters of customers running who all lived in the same building or on the same street. If one person ran, many did, but in other places no one ran at all.”
Puri and Iyer also measured social networks in two other ways – through the ethnicity of depositors, and links to a person who helped open the bank account. In all cases, if people in your network run, you are more likely to run.
Puri and Iyer note this underscores the importance of customers’ social ties for banks and other firms, a factor that has not been previously explored in banking research. “Customers were withdrawing their money in response to their social ties,” Iyer said.
The study found the bank run to be most “contagious” in its early stages, with higher levels of connection found between depositors who ran within the first two days of the crisis. “This indicates that if a government or regulatory body is going to step in to stabilize a bank in crisis, the first day or two is the critical period for intervention,” Puri said.
Puri and Iyer also found that the length and depth of a customer’s relationship with the bank reduced a depositor’s tendency to run. Customers with higher account balances were more likely to withdraw their funds, even if their full balances would have been covered by deposit insurance in the event of a failure. But customers were less likely to withdraw if they had longstanding relationships with the bank, or a loan in addition to a savings account.
“Banks have traditionally viewed cross-selling of loans and other services to account holders as a way to maximize revenue from any particular customer,” Iyer said. “But our results indicate that developing client relationships based on multiple products may also be an effective way for banks to prevent those customers from running, and to protect banks’ downside risk.”
The research also found that the damage of a bank run can be severe and long-lasting. Six months after the 2001 bank run, only 10 percent of accounts affected by the run had returned to their pre-run balances, and overall deposit balances at the bank also did not recover in the short term.
Puri and Iyer’s paper, “Understanding Bank Runs,” is the National Bureau of Economic Research Working Paper #14280. The work was partially funded by the Federal Deposit Insurance Corporation (FDIC).