Over the past decade many countries in Europe and North America have bailed out large commercial banks, at a high cost to taxpayers (Laeven and Valencia 2012). Prudential regulation and resolution mechanisms for banks are being overhauled in response. A cornerstone of new bank resolution policies – including the Bank Recovery and Resolution Directive in the EU, the joint FDIC-Bank of England proposal for a single point of entry resolution plan, and the Financial Resolution and Deposit Insurance Bill in India – is the bail-in of shareholders, unsecured bond-holders, and uninsured depositors (Chennells and Wingfield 2015, Roy 2017).
Bail-ins, as opposed to bailouts, redistribute the costs of bank failures from taxpayers to bank creditors and thus should also strengthen market discipline. But many policymakers fear that bail-ins may undermine confidence in the banking sector and jeopardise financial stability (Philippon and Salord 2017).
The apparent sale of subordinated bank debt to retail investors in Italy, for example, where a sizeable share of subordinated debt of banks has been placed with retail customers who "were probably largely unaware of the true risk associated with these products, sometimes not even being aware of their bondholder status but considering themselves as simple depositors” (Merler 2016), has heightened these concerns. Lindner and Redak (2017) show that 5.4% of Italian households held bank bonds, compared to 2.2% in the entire euro area.1
But how do retail depositors actually react to the bail-in of large commercial banks? How does their reaction relate to personal financial losses due to bank resolution? And to what extent does the type of bailed-in debt affect the confidence of depositors in the banking sector? In a new paper, we provide first answers to these questions by examining how depositors reacted to the Cyprus banking crisis of 2013 (Brown et al. 2017a) .
The crisis included a significant bail-in of debtholders, as well as of uninsured depositors of the two largest banks in the country – Laiki (Cyprus Popular) Bank and Bank of Cyprus (see Michaelides 2014 for details). The bail-in was accompanied by capital controls limiting cash withdrawal and transfers of bank deposits.
Aggregate flight to cash
Figure 1 Aggregate bank deposits and cash issuance
Note: Values are indexed at the month-end of February 2013. The vertical line marks month-end of February 2013. Net value of banknotes refers to banknotes issued, minus banknotes taken out of circulation.
Source: Central Bank of Cyprus and Brown et al. (2017a).
Figure 1 suggests that the March 2013 crisis triggered an aggregate reallocation of resident money holdings from bank deposits to cash. The March 2013 bail-in represented an unexpected shock to resident depositors, and the left-hand panel shows a strong outflow of resident (and non-resident) deposits in the months after the March 2013 bail-in.
By September 2013, resident deposits were just 76% of their level prior to the bail-in. Bailed-in deposits were less than one-third of this decline.
The right-hand panel of the figure shows the net value of euro banknotes issued in Cyprus. At the same time as the fall in bank deposits, net cash issuance tripled. This suggests a flight to cash among resident households and businesses. Surveys show that, as well as hoarding cash, households reduced deposits to cover living expenses and repay debt.
At first glance, this suggests that confidence in the Cypriot banking sector was seriously undermined by the crisis. But understanding how crisis exposure relates to behaviour at the individual level is important for the design of bank resolution policy, and this aggregate data cannot tell us which resident depositors withdrew funds –they may have been those directly affected by the bail-in, but they may have been those who did not lose money.
We also want to know whether the loss of confidence was related to the bailed-in asset – in other words, whether households suffered a loss on subordinated debt (bonds) or on uninsured deposits.
A more detailed look
Our analysis uses data from a survey of 800 Cypriot households that had more than €5,000 in a term-deposit account. We study the behaviour of resident households only, because the reaction of the domestic corporate sector and non-resident depositors would have been likely to react differently.
The survey provides information on depositor behaviour and confidence during the 12 months from January 2013 to January 2014. The sample includes households with very different exposures, because it covers households which were clients of the two resolved banks, and those that were clients of other financial institutions. Importantly, households with deposits above the deposit insurance threshold (€100,000) were oversampled: 49% had at least one bank deposit exceeding this sum at the time of the survey. Finally, the sample includes many households that held subordinated bonds or equity issued by the resolved banks.
Within our sample, 55% of households had incurred a direct financial loss, 28% experienced a bail-in of uninsured deposits, and 16% experienced a bail-in of bonds (Figure 2). In addition, 44% of all respondents had to face losses on the value of bank equity.2 It was not unusual for households to have more than one loss: 30% that incurred one financial loss suffered from two sources, and 10% from all three.
Figure 2 Incidence of financial losses among the survey sample
Source: Brown et al. (2017a).
We observe a strong flight from bank deposits to cash among surveyed households, just as in the aggregate data. During 2013, 52% of surveyed households decreased their deposits (by more than € 5,000) with all banks in Cyprus, and 21% of households decreased their deposits by more than one-quarter of their initial holdings. A similar share of households had increased their cash holdings during 2013. When asked about the reasons for deposit withdrawals, 61% said that withdrawn funds were used to cover living expenses, 18% hoarded cash, and 12% used funds to repay loans.
The withdrawal of deposits from the Cypriot banking system was stronger among households with financial losses (62% reduced deposits), but 44% of households without financial losses also reduced their deposits.
Bailing-in of bonds and of uninsured deposits trigger a similar reaction
Policymakers need to understand how household behaviour is shaped by personal exposure, especially whether a loss on bonds triggers a different reaction than a haircut on uninsured deposits. This is because resolution procedures typically rank deposits above subordinated debt.
Figure 3 shows the likelihood of deposit reductions for different loss types relative to those households without losses. The estimates are based on a series of linear regressions, controlling for a broad set of socioeconomic characteristics (the results in Figure 2 are from a specification that controls for variables exogenous to the crisis). Loss types were defined such that they did not overlap. “Deposit bail-in” refers to households with a deposit bail-in (and eventually also a bond bail-in or equity losses), and “Bond bail-in” refers to households that incurred a bond-bail, but no deposit bail-in. “Equity loss” refers to households with losses on shares only. The definition expresses the implicit ordering of the impact of a bail-in from a depositor’s perspective. A deposit bail-in is the fiercest form, followed by a bond bail-in, and then a loss on equity.
Figure 3 Point estimates of the likelihood of deposit reductions and increases of cash holdings for different financial losses
Notes: The figure presents the point estimates from OLS regressions for indicators of different types of financial losses relative to no financial losses (base category). The horizontal lines depict the 95% confidence interval of individual point estimates.
Source: Brown et al. (2017a).
The left-hand panel shows that households that incurred a deposit bail-in or a bond bail-in displayed a strong – and similar – response. The likelihood they would reduce their overall deposit holding was 25 percentage points higher than the households that did not incur a loss. By contrast, those households which had (solely) an equity loss were only 4 percentage points more likely to reduce deposit holdings. This effect is not statistically significant from zero (the red line).
The right-hand panel shows the increases in cash holdings over the year 2013. Results mirror the findings for deposit reductions. These results imply that the bail-in of bonds triggered the same reaction in private Cypriot households as the bail-in of uninsured deposits.
Medium-term confidence is low, but hardly related to personal losses
This was the immediate response. What about the medium run? The survey took place nine months after the March 2013 bail-in. At the time, respondents were asked how they would allocate a hypothetical sum of €200,000 among different assets (cash, deposit at banks in Cyprus, other investments in Cyprus), if the capital controls were lifted.
Responses suggested a low level of confidence in banks. One-third of the households would not hold any of their funds in a domestic deposit account, while only 22% would even hold deposits above the deposit insurance level. The intention of households were hardly related to how they had been affected by the bail-in. So while it matters how a household is affected by a bail-in for short-term changes in money holdings, longer-term confidence in banks and money holdings seems to be largely determined by the existence of the crisis per se rather than by personal experiences of it.
What can we learn from the Cyprus bail-in?
We should be wary of drawing conclusions, because the circumstances of the Cyprus crisis were unusual.3 Nevertheless, we think that three main lessons can be learnt:
- Retail investors may not distinguish between a bail-in of uninsured deposits or other, subordinated debt. The new bank resolution regime prioritises uninsured deposits over other retail debt, so this difference needs to be clearly communicated to retail investors. For example, disclosure rules for retail investment products should highlight the different priorities of different debt claims on banks if there was a resolution process.
- The bail-in of retail investors undermined short-term confidence. This led to significant outflows of deposits, especially from the affected banks, and suggests that there is a strong role for the lender of last resort to provide emergency liquidity during a bank resolution process (Goodhart and Avgouleas 2016).
- Intentions to limit the use of cash may seriously constrain the asset allocation of households. In normal times, claims on banks may be viewed as an almost perfect substitute for cash, but on this evidence, this is not the case during banking crises. Concerns about the safety and soundness of the banking system would be likely to result in a reallocation of money holdings towards cash. This finding is by no means new; Friedman and Schwartz (1963), for example, draw similar conclusions from the US experience in the Great Depression. Moreover, after the 2007-2009 crisis, the demand for currency, internal or external, also increased in the US and the euro area (Jobst and Stix 2017).
Authors’ note: The findings, interpretations and conclusions presented in this article are entirely those of the authors and should not be attributed in any manner to the Central Bank of Cyprus, the Oesterreichische Nationalbank or the Eurosystem.
Goodhart, C and E Avgouleas (2016), "A critical evaluation of bail-in as a bank recapitalization mechanism", in D D Evanoff, A Haldane and G G Kaufman (eds), The New International Financial System: Analyzing the Cumulative Impact of Regulatory Reform, World Scientific Studies in International Economics No. 48.
Brown, M, I Evangelou and H Stix (2017a), “Banking crises-bail-ins and money holdings”, Central Bank of Cyprus working paper No. 2017-2.
Chennells, L and V Wingfield (2015), "Bank failure and bail-in: An introduction", Bank of England Quarterly Bulletin 2015/Q3: 228-241.
Friedman, M and A J Schwartz (1963), A monetary history of the United States, 1867–1960, Princeton University Press.
Jobst, C and H Stix (2017), “Doomed to disappear? – The surprising return of cash across time and across countries”, CEPR Discussion Paper No. 12327.
Laeven, L and F Valencia (2012), “Systemic banking crises database: An update”, IMF working papers No. 12/163.
Lindner, P and V Redak (2017), “The resilience of households in bank bail-ins”, Oesterreichische Nationalbank Financial Stability Report 33: 88-101.
Merler, S (2016), "Italy – Four small banks: Resolution via bridge bank and asset management vehicle tools to avoid full bail-in", in Bank resolution and “bail-in” in the EU: selected case studies pre and post BRRD, World Bank.
Michaelides, A (2014), "Cyprus: From boom to bail-in?", Economic Policy 80: 639-689.
Philippon, T and A Salord (2017), "New ICMB/CEPR Report: Bail-ins and Bank Resolution in Europe", VoxEU.org, 22 March.
Roy, KS (2017), "Preparing for the Cyprus moment", Outlook India, 17 December.
 In May 2017, the Italian government ruled out a bail-in of depositors for troubled banks in the region of Veneto. Instead, the authorities sought approval from the EU for a government-sponsored bail-out of the two banks.
 These percentages reflect unweighted frequencies. If we weight the results to be representative of the respective population with deposits with at least €5,000, 54% had financial losses, 17% a bail-in of deposits, 14% a bail-in of bonds, and 37% equity losses.
 The banking crisis in Cyprus was unusual both in the run-up to the crisis (Michaelides 2014) to the weeks around the bail-in. Within less than two weeks, bank customers experienced the temporary closure of all banks, a proposal to impose a levy on all insured and non-insured deposits, a parliamentary refusal to implement the levy, the announcement of a resolution process for the two largest banks involving a bail-in of uninsured depositors and debt holders, and the introduction of strict capital controls.