VoxEU Column Financial Markets

Exploring the transmission channels of contagious bank runs

Contagious bank runs are an important source of systemic risk. However, with observational data it is near-impossible to disentangle the contagion of bank runs from other potential causes of correlated deposit withdrawals across banks. This column discusses an experimental investigation of the mechanisms behind contagion. The authors find that panic-based deposit withdrawals can be strongly contagious across banks, but only if depositors know that the banks are economically related.

Financial contagion – the situation in which liquidity or insolvency risk is transmitted from one financial institution to another – is viewed by policymakers and academics as a key source of systemic risk in the banking sector. In particular, the events in the 2007–2009 Global Crisis have turned the attention of policymakers towards the potential contagion of liquidity withdrawals across banks and the resulting implications for financial stability.

  • In September 2007, fear that the depositor run on Northern Rock might spread to other UK financial institutions prompted the Bank of England to provide liquidity support, and eventually a full guarantee of all deposits.
  • In May 2012, massive withdrawals from Spanish banks sparked fears that depositors of the UK subsidiary of Banco Santander may ‘run’ on their bank.
  • In December 2012 (unfounded) rumours that ATMs in Sweden were shut down and that the operations of SwedBank in Estonia had been closed led to a run on the Latvian subsidiary of the same bank.

These recent events in the European banking sector suggest that changes in expectations and coordination failures among depositors may not only be a source of individual bank runs, as suggested in the classical work by Diamond and Dybvig (1983), but may also be an important source of systemic risk.

Identification challenges with observational data

The contagion of deposit withdrawals across banks has been documented for the US during the Great Depression (e.g. Calomiris and Mason 1997) as well as more recently in emerging markets (e.g. Iyer and Peydro 2011, Iyer and Puri 2012). However, even the detailed bank-level and depositor-level data presented in recent studies provide only rough guidance on the transmission mechanisms underlying contagious deposit withdrawals runs.

  • First, it is almost impossible with observational field data to disentangle the contagion of bank runs from other potential causes of correlated deposit withdrawals across banks.

Factors such as correlated liquidity shocks across households, correlated performance shocks across banks (e.g. due to macroeconomic shocks), or common exposure to asset shocks could explain the ‘run’.

  • Second, even if researchers were able to identify cases of pure contagion of bank runs in field data, that data would hardly allow them to explain precisely why the runs became contagious.

To do so they would need to measure the changes in beliefs of depositors about the fundamentals of their bank and the withdrawal propensity of other depositors.

Using experiments to understand the mechanisms behind contagion

Recent studies have approached these identification issues by using experimental methods that allow the researcher to:

  • Exert full control over the solvency status of the banks;
  • Exogenously vary the economic linkages between different banks, and;
  • Directly measure subjective beliefs about bank fundamentals and the behaviour of other depositors (Brown et al. 2013, Chakravarty et al. 2012, Trevino 2014).1

In Brown et al. (2013), we use experimental control to study potential coordination failure of depositors in a solvent bank when they receive information about a panic-based run at another solvent bank. We find that panic-based deposit withdrawals can be strongly contagious across the two banks, but only if it is known to depositors that the banks are economically related (in our experiment through asset commonality). We document that under these information conditions a panic-based bank run at one bank makes the bank-run equilibrium more salient for depositors at another bank. The run at the observed bank leads to an update of subjective beliefs of depositors in the observing bank that their fellow depositors will withdraw. We find no evidence for a belief-update, and no subsequent contagion of deposit withdrawals between banks, when depositors know that the banks are economically unrelated.

The novelty of our results is thus that we document that:

  • The contagion of bank runs may be initiated by an updating of expectations regarding the behaviour of other depositors at the observing banks, and
  • This expectation channel is more likely to be active when there is common knowledge that banks are economically related.

Our results put an interesting perspective on the distinction between panic-based and information-based bank runs. Our findings suggest that while bank runs may be caused by panic, this does not necessarily imply that their occurrence is random. Our results suggest that commonly-observed information increases the probability of a bank run – even if that information provides only a very noisy signal about the solvency of a bank. The fact that information increases the salience of the bank-run equilibrium may be sufficient to trigger contagious runs.

Policy implications

We argue that economic linkages between banks due to common asset exposure or similar portfolio characteristics may have a further negative impact on financial stability beyond their direct economic impact on banks’ financial statements and equity returns. Economic linkages between banks give rise to contagion of deposit withdrawals across banks, especially when depositors are aware of these economic linkages. Systemic risk due to the contagion of panic-based deposit withdrawals is thus likely to be more acute for banking systems characterised by clusters of domestic banks which share the same business model (e.g. cajas in Spain or Sparkassen/Volksbanken in Germany). For regulators this accentuates the question of how to monitor and regulate economic linkages between banks stemming from similar exposures, in order to mitigate financial fragility and to encourage greater diversity in the financial system.

The views expressed in this article are those of the authors and should not be attributed to the Dutch Central Bank.


1. For a review of experimental methods to study individual bank runs see Dufwenberg (2013).


Brown, M, S Trautmann, and R Vlahu (2013), “Understanding Bank-Run Contagion”, DNB Working Paper No. 363.

Calomiris, C W and J R Mason (1997), “Contagion and Bank Failures during the Great Depression: The June 1932 Chicago Banking Panic”, American Economic Review, 87(5): 863–883.

Chakravarty, S, M A Fonseca, and T R Kaplan (2012), “An Experiment on the Causes of Bank Run Contagions”, University of Exeter Discussion Paper Series, 12/06.

Diamond, D W and P H Dybvig (1983), “Bank Runs, Deposit Insurance and Liquidity”, Journal of Political Economy, 19: 401–413.

Dufwenberg, M (2013), “Banking on Experiments?”, University of Arizona Working Paper 13-08.

Iyer, R and J L Peydro (2011), “Interbank Contagion at Work: Evidence from a Natural Experiment”, Review of Financial Studies, 24: 1337–1377.

Iyer, R and M Puri (2012), “Understanding Bank Runs”, American Economic Review, 102(4): 1414–1445.

Trevino, I (2014), “Channels of Financial Contagion: Theory and Experiments”, NYU Working Paper.

5,144 Reads