Similarity of banks and communities increased contagion of bank runs in 1893 bank panic

What is it about?

In 2008, two events took place that shook customer confidence and provoked panic. Depositors of two Swedish banks, Swedbank and SEB, rushed to empty their bank accounts after a rumor spread on Twitter claiming that the banks were experiencing financial trouble. The bank quickly addressed the rumors as false and police investigations were launched to track down the source of the claims. The episode showed how quickly bank runs can happen even to entirely solvent institutions and if enough people withdraw their cash, a bank run based only on rumor can become a self-fulfilling prophecy. Such runs can also turn into a systemic customer panic and affect other banks in the same lines of business, essentially made guilty by association. Similar panic in other industries is not uncommon. Chinese consumers to this day still prefer to buy their baby milk formula overseas following the melamine scandal in 2008, even though not every milk producer was found to be tampering with their products. In our recent paper, Ripples of Fear: The Diffusion of a Bank Panic, Jay Kim, Daphne Teh and I examined the largest customer-driven bank panic in US history to understand why customer fears grow into serious panics. In our bank-specific sample, we found that customers do not actually lose trust in all banks. Instead, they target individual banks based on their assessment of what banks are similar to those that have already experienced a run. Yet it only takes a run on a few banks to spark systemic breakdown, which has disproportionate economic knock-on effects. The 1893 bank panic triggered a devastating economic crisis that saw real earnings decline by 18 percent from 1892 to 1894. Despite the efforts made to control the panic and limit the damage for individual banks, 503 banks were suspended. Two thirds of them failed. But they weren’t all insolvent. It was also surprising that paid-in capital made banks more vulnerable to bank runs. This perversely made them more prominent targets of bank runs despite being more financially stable. Looking over the data from 1893, we found three factors that influenced the likelihood of panic spreading. First, bank runs had a higher likelihood of occurring in communities with similar compositions of race, national origin, religion or wealth, suggesting a role of prejudice even in such important economic decisions as withdrawing money from banks. Essentially, people were more ready to believe members of their community that there was a reason to panic than they were those of different communities. Communities with high diversity were less likely to have bank runs because weaker social ties reduce the likelihood that community members will agree on something, in this case that a bank was in crisis or not. Secondly, customers drew associations between banks of a certain type, which increased the likelihood of panic at banks of a similar form. Customers tend to rely on heuristics rather than deeper reasoning when assessing a bank’s vulnerability. This was seen in the 1985 Ohio savings and loan crisis, which spread to savings and loans in other states but not to other forms of financial institutions in Ohio. Similarly, in the 1873 bank panic in New York City, the loss of depositor confidence was confined to savings banks, with nearly all savings banks experiencing a run, while only one national bank did. It is also interesting to note that customers made these decisions despite the efforts of the banks to avoid being stigmatised by the news of runs on other banks. An article in the Aspen Daily on July 21st, 1893 read, “there is no reason why Aspen people should get excited over the situation here. All of the Aspen banks are backed by conservative businessmen whose business careers have not been marked by wild speculations or daring ventures.” Despite this quote, typical of newspapers at the time, the panic continued. Bank panics are an ideal setting to observe customer reactions to perceived threats because they are consequential but with weak economic rationale for each run. However, these findings have broad applicability to other organisations. Crucially, panics remove organisations from the driving seat of the narrative about the organisation’s health. We find that customers make distinctions based on easily observable, even if superficial characteristics, which can disproportionately and adversely affect the firm, in this case, irrespective of a bank’s financial health. But organisations can protect themselves against falling victim to a customer panic at a peer. Many tend to share their fate with peers by creating connections to make them more accepted as a legitimate player. But as our research shows, if one organisation in the group breaks its customer’s trust, suspicion can fall on all. This is a good reason to differentiate as much as possible. Apple’s recent stand-off with the FBI over creating backdoor access to customer devices is a good example of differentiating from the pack to maintain customers’ trust. From a community perspective, smaller organisations, for example regional savings banks, should try to diversify their customer bases. If banks approach niche markets, especially along community lines, they’re likely to be building up a shared perception of their organisation among customers. With a greater diversity of customers and therefore a greater diversity of opinion on the organisation, less customers are likely to follow if a small group panics.

Featured Image

Why is it important?

There are many studies of diffusion in the social sciences, including diffusion of innovation among organizations. This study is special because: 1. Bank runs are harmful to each bank, and the community, which makes them different from innovations and other useful practices that spread. 2. Bank runs are actions by communities against organizations, which is also different from organizational diffusion in general. 3. The community similarities we studies, such as national origin, race, religion, and wealth, had strong effects, suggesting a role of prejudice in the diffusion process.