As early as 2009, Reinhart and Rogoff (2009) pointed out that "recessions surrounding financial crises are usually long compared to normal recessions". Their research highlights surprisingly large declines in output, slow recoveries, and large and persistent negative effects on unemployment, public debt, and fiscal deficits in the aftermath of banking crises. The subsequent experiences in the US and particularly in western Europe lend further support to their findings. This naturally leads to the question of whether intervention to help distressed banks early on during financial crises is effective in mitigating the macroeconomic consequences of financial distress, and if so whether it matters what form the intervention takes. The answers to these questions will also shed light on the economic costs of forbearance i.e. the costs of inaction by regulators during a crisis. We analyse these questions by empirically investigating durations of recessions after 68 systemic banking crises from the period 1980 to 2013.
- Our key finding is that decisive and early recapitalisation of banks in distress can shorten recessions by several years.
Effect of intervention measures on recession duration
The existing literature on this question is thin and not conclusive. Honohan and Klingebiel (2003) document that intervention measures have high fiscal costs; Laeven and Valencia (2011) provided suggestive evidence that once financial crises have eroded bank capital, recapitalised banks grow faster. We approach the issues from a macro perspective: How do intervention measures affect recession duration? Since the problems start with weakly capitalised banks, the main instrument we focus on is bank recapitalisation.
A problem for empirical work is that intervention is probably endogenous to crisis severity. Governments are more likely to intervene in severe than in mild crises. At the same time, crisis severity is a measure of the scale of problems in the banking sector, which is likely to have an impact on recession duration in the absence of intervention. If a measure increases the probability of recovery but is more likely to be used in severe crises, which in turn is associated with a longer recession, regressions may spuriously indicate that the measure is not effective. For this reason, we control for crisis severity in our estimations in order to eliminate that bias, although we have to do this in an indirect manner as crisis severity is not directly observable other than through recession duration.
After a shock has eroded their assets (and their capital as a consequence), banks have stronger incentives to roll over loans to borrowers in distress. Since the banks are now undercapitalised, a correspondingly larger part of the risk associated with the rollover will be borne by creditors of the bank or by whomever steps in to rescue the bank when it gets into distress. Recapitalising such zombie banks mitigates that incentive problem, once recapitalised the banks have to once again shoulder most of the rollover risk themselves. Other interventions, such as provision of guarantees or liquidity injections, only prevent bank failures but do not address this perverse incentive problem and therefore do not improve welfare as much as bank recapitalisation. That is why recapitalising banks leads to a stronger recovery (higher expected future output) than one can expect in a zombie bank environment with distorted roll over incentives (Homar and van Wijnbergen 2015).
Testing the theoretical predictions
To test these theoretical predictions, we estimate a duration model with crisis-specific fixed effects, allowing for the possibility that the average level of intervention over the crisis period is correlated to unobservable crisis severity (Homar and van Wijnbergen 2015).
- We find a positive and highly significant effect of bank recapitalisations on the probability of recovery.
- We do not find such support for the effectiveness of guarantees on bank liabilities or liquidity support.
For the purpose of addressing the macro problems that undercapitalisation leads to, these policies indeed seem inferior to bank recapitalisations.
The estimation results show that bank recapitalisations have a highly significant positive effect on the probability of recovery. Figure 1 shows the calculated probability of recovery in each subsequent year after a financial crisis for the case where nothing is done and the case where the crisis is quickly reacted to through a bank recapitalisation.
Figure 1. Predicted conditional probability of recovery from a severe representative crisis
Figure 1 clearly shows much faster recovery after a crisis if banks are recapitalised early on; the recovery probability starts going up much more quickly, and reaches 1 (certain exit if it has not happened earlier) much earlier with recapitalisation than it does without. Similar results come out when the exercise is repeated for the earlier part of the sample.
One can use the predicted recovery probabilities to calculate the model-predicted expected recession duration with and without intervention. Liquidity injections and guarantees do not produce significantly shorter expected recession duration, but recaps do.
- The typical severe recession (from the sample of 2007 to 2013 crises) is predicted to last 6.3 quarters with bank recapitalisation, but would have gone on for 11 quarters without, a difference of more than a full year.
- For mild crises, the recession is expected to last about half a year shorter with bank recapitalisation.
So recapitalising banks early in a crisis is an effective intervention from an ex post macroeconomic point of view. Of course, such a conclusion always begs the question, are there ex ante costs to such policies if banks take additional risks on their balance sheets anticipating they will be rescued if things go wrong? If the answer is yes, but the ex post macroeconomic gains from bank recapitalisations are so high, an ex ante commitment to not intervene after a crisis may be wise but is almost certainly not credible. As an alternative, sufficiently high capital requirements and harsh terms during the rescue should at least mitigate the ex ante risks of unavoidable ex post rescues.
Difference in recovery between the US and the Eurozone
A final comment concerns the difference in time to recovery in the US versus western Europe. The recession started about a year earlier in the US, but ended much earlier, the European recession has lingered on years longer than in the US. Figure 2 below points at one possible explanation, the way regulatory authorities approached bank capital shortages.
Figure 2. New share emissions by banks in the US and the Eurozone
In the US, banks were forced to supplement their eroded capital base by issuing new equity early in the crisis. As Figure 2 demonstrates, no such effort was undertaken in the Eurozone. There banks were told to improve their (risk-weighted) capital-to-asset ratios, but were left free to decide on how to do that. As a result, banks chose to shift massively out of corporate lending into sovereign debt purchases, since the latter do not count in the calculation of assets used to set the capital requirements under Basel III. As a consequence, the rebuilding of capital had a directly negative impact on the recovery through the choice for asset substitution away from corporate loans, and anyhow took longer. Both points are strikingly different from the US experience, and both may have contributed to the much slower recovery in the Eurozone than we have seen in the US.
Homar, T and S J G van Wijnbergen (2015), "On Zombie Banks and Recessions after Systemic Banking Crises: Government Intervention Matters", Working paper :1–52.
Honohan, P and D Klingebiel (2003), "The Fiscal Cost Implications of an Accommodating Approach to Banking Crises", Journal of Banking and Finance 27:1539–60.
Laeven, L and F Valencia (2011), "The Real Effects of Financial Sector Interventions During Crises", IMF Working Paper 11/45 .
Reinhart, C M and K S Rogoff (2009), "The Aftermath of Financial Crises", American Economic Review 99:466–72.