Stopping Contagion With Bailouts: Microevidence From Pennsylvania Bank Networks During The Panic Of 1884
International Monetary Fund
Office of Financial Research
March 30, 2016
Forthcoming in the Journal of Banking and Finance
Using a newly constructed historical dataset on the Pennsylvania state banking system, detailing the amounts of “due-froms” on a debtor-bank-by-debtor-bank basis, we investigate the effects of the Panic of 1884 and subsequent private sector-orchestrated bailout of systemically important banks (SIBs) on the broader banking sector. We find evidence that Pennsylvania banks with larger direct interbank exposures to New York City changed the composition of their asset holdings, shifting from loans to more liquid assets and reducing their New York City correspondent deposits in the near-term. Over the long-term though, only the lower correspondent deposits effect persisted. Our findings show that the banking turmoil in New York City impacted more exposed banks outside New York City, but that bailouts of SIBs by the New York Clearinghouse likely short-circuited a full-scale banking panic.
Stopping Contagion With Bailouts: Microevidence From Pennsylvania Bank Networks During The Panic Of 1884 – Introduction
The global financial crisis of 2007-09 highlighted the issues of regulatory forbearance and the public bailout of systemically important banks (SIBs). Public interventions around the world were based on the notion that the potential failure of SIBs, like Citigroup Inc. in the United States and the Royal Bank of Scotland in the United Kingdom, would precipitate runs and failures elsewhere in the financial sector, freeze the flow of credit and payments to the real economy, and lead to a depression (Laeven, Latnovski, and Tong, 2014). In the wake of the 2007-09 crisis, many of these SIBs have actually grown larger, due to consolidation within the industry, potentially increasing the need for collective support for these institutions in times of stress (GFSR, April 2014).3 Yet, despite the expectation of interventions in future crises, there has been little empirical study on how the public bailout of SIBs affects the rest of the financial sector.
An empirical study of the effects of bailouts of SIBs on other banks confronts a number of practical difficulties. First, it is often hard to identify ex ante which banks are systemically important. For example, the Financial Stability Board, which monitors global financial stability and proposes international standards, only began constructing lists of global systemically important banks (G-SIBs) in 2011 (FSB, 2014). When the U.S. government decided to provide asset guarantees and additional capital to Citigroup in November 2008, its decision “appeared to be based as much on gut instinct and fear of the unknown as on objective criteria,” according to a government investigation (SIGTARP, 2011, p. 42). While size is the most well-known indicator of systemic importance, other factors, such as interconnectedness, the lack of readily available substitutes or infrastructure for their services, their global activity, and their complexity, may also make institutions systemically important (FSB, 2013).
Second, it is tough to disentangle losses and disruptions due to counterparty exposure from those due to other factors. The highly complex modern financial environment, characterized by myriad instruments held by a number of parties, makes the determination of counterparty exposures across financial institutions difficult and complicates the identification of risk channels. Furthermore, exposures may be direct or indirect, such that a bank may not recognize that it has strong second- or third-order connections to a particular SIB.
In this paper, we exploit banking disturbances and the subsequent bailouts of key banks such as Metropolitan National Bank by the membership of the New York Clearinghouse during the Panic of 1884 to assess the effects of those bailouts on other banks.4 Like the SIBs today, these New York City banks were large and interconnected because they served as reserve repositories to interior banks — banks outside New York City. Using new data that we constructed on the correspondent network of state-chartered banks in Pennsylvania at the time, we calculate the degree of counterparty exposure of each Pennsylvania bank to New York City banks and see how differential degrees of interbank exposure to New York City affected the dynamics of deposit and lending growth before and after the Panic of 1884. If efforts by the New York Clearinghouse to rescue troubled banks had failed and distress had propagated, cash payments and deposit access would have been interrupted for banks with higher exposures to New York City, negatively impacting their balance sheets and depressing their ability to engage in lending.
Our analytical results show that interior banks in Pennsylvania changed their behavior during the panic even though the clearinghouse’s bank bailouts likely succeeded in preventing a large-scale bank panic outside New York City. After controlling for bank fundamentals, we find that Pennsylvania state-chartered banks with higher levels of exposures to New York City had statistically significant declines in equity capital growth and increases in nonperforming assets in the quarters after the panic. There is also some evidence of a shift towards more liquid assets and a greater dependence on deposits as a financing source. Over the longer term though (at the annual frequency), these differences vanish; the only robust differences are declines in the use of correspondent deposits, particularly in New York City, by more highly exposed banks.
While earlier findings of bankers, economists, and policymakers (among others, Sprague, 1910; Wicker, 2006; Gorton, 2012) argue that the Panic of 1884 was an ‘incipient’ panic that was contained in New York City and did not spill over to other regions, our results indicate that more exposed banks elsewhere did respond to the events in New York. However, these balance sheet responses were largely short-lived. Apart from a decline in correspondent deposits, there is no strong evidence that they lasted beyond a year.
See full PDF below.