The identification of systemic financial stress episodes has so far mainly relied on expert judgement. Leaven and Valencia (2013) provide the most widely used database on systemic banking crises identified by significant signs of banking distress, and significant policy interventions. Two other datasets for EU countries, in Babecky et al. (2012) and Detken et al. (2014), were compiled by relying on the judgement of national central banks. However, the qualitative assessment of such events, as well as the time lag until the next update becomes available, call for a new and complementary approach.
Model-based dating
Instead, in our new working paper (Duprey et al. 2015) we provide a transparent, objective, and reproducible method to build a chronology of systemic financial stress events. Those events are defined as episodes of financial market stress associated with a substantial and prolonged negative impact on the real economy. By doing so, we bridge the gap between the literature on financial stress indices and business cycle dating.
Identifying systemic financial stress episodes
The model-based approach outlined in this column aims at identifying the episodes that correspond to the red quadrant in the 2-by-2 matrix shown in Figure 1, with coincident high financial stress and low economic growth. To that extent, we adopt a strategy in three steps.
Figure 1. Identifying systemic financial stress episodes
Source: Duprey et al. 2015.
- Constructing a simple financial stress index (FSI).
Contrary to the business cycle dating literature with GDP, there is no commonly accepted metric for financial market stress. Thus, we construct a monthly coincident measure of financial market stress which is comparable across 27 EU countries and covers a long time span, at most 50 years. In the spirit of Hollo et al. (2012), we account for the fact that financial stress periods are usually characterised by a larger co-movement of financial data.
- Identifying periods of high financial market stress.
In order to distinguish between periods of low and high financial market stress, we rely on the business cycle dating literature that focuses on the identification of tranquil regimes and recessions. The most natural method is to use a Markov switching (MS) model (Hamilton 1989).
- Narrowing down the financial stress episodes to those with a ‘systemic’ character.
We expect high levels of financial stress to be harmful to the real economy, with a substantial (Figure 2.a) and prolonged (Figure 2.b) decline in real economic growth. In order to narrow down the episodes of systemic financial stress, we define a simple algorithm which detects financial stress episodes associated with a substantial and prolonged decline in real economic growth. Among the events identified as belonging to the regime of high financial stress, we require that:
- A drop of industrial production overlaps with two (possibly non-consecutive) quarters of negative GDP growth.
- Over a 12-month period, we have at least six consecutive months of negative annual industrial production growth.
Figure 2. Industrial production growth and the Financial Stress Index
Note: confidence bands 20th /80th. Source: Duprey et al. 2015.
Sixty-eight episodes of systemic financial stress for 27 EU countries starting in 1964
Figure 3 shows the chronology of events we obtain as well as the intensity of real economic stress during each systemic crisis. Systemic financial stress periods are in colours, while tranquil periods are in white. The intensity of real economic stress during each systemic financial stress period is represented by the colours from yellow (small industrial production loss from peak to through) to black (industrial production loss from peak to through around 30%). The model-based systemic financial stress periods tend to be robust to event reclassification once new data become available and alternative specifications provide similar results.
Figure 3. Chronology of systemic financial stress and real economic stress intensity across EU countries
Note: Periods for which no sufficient data was available are in light grey. Source: Duprey et al. 2015.
Model-identified systemic financial stress events have the following patterns:
- About 50% of recessionary events are classified as systemic, while the other half is not characterised by simultaneous financial market stress.
- In three cases out of four, financial market stress occurs first, followed by real economic stress.
- Consistent with Reinhart and Rogoff (2014), real economic stress lasts on average six months longer when it is associated with financial market stress.
- Consistent with, for example, Jorda et al. (2013), we find that the decline in GDP is on average three percentage points larger when it is associated with financial market stress.
- Systemic financial stress usually occurs in the form of ‘clusters’; in terms of the number of countries entering a systemic stress event simultaneously, the Global Crisis is only comparable to the first oil shock.
Comparing with the expert-based crises dataset
Model-based systemic financial stress periods are not supposed to coincide perfectly with expert-based crises, since they represent two different concepts. While the former are identified on the basis of market prices of traded instruments, the latter rely on qualitative information and past policy actions. Also, expert-based stress episodes are more narrowly defined than our broad systemic financial stress periods. We find that:
- Out of the banking crises identified by Leaven and Valencia (2013), Babecky et al. (2012), Detken et al. (2014), as well as Reinhart and Rogoff (2011) for only 16 EU countries, 100%, 92%, 90%, and 89% are captured by our model-based approach, respectively.
- 82% of the model-detected systemic financial stress periods are also included in at least one of the four crises datasets that cover banking, currency, and debt crises, as well as stock market crashes for a subset of 16 EU countries.
Concluding remarks
Our research is a first attempt to precisely date systemic financial stress events in a transparent and reproducible manner and complements the existing datasets that rely on expert judgement. However, efforts are still needed in order to disaggregate sources of systemic financial stress in each market segment, for instance, in the real estate market. But we believe this is already a first step towards a better analysis of national macroprudential policies.
References
Babecky, J, T Havranek, J Matiju, M Rusnak, K Smidkova, and B Vasicek (2012), “Banking, debt, and currency crises: Early warning indicators for developed countries”, ECB Working Paper No. 1485.
Detken, C, O.Weeken, L Alessi, D Bonfim, M M Boucinha, C Castro, S Frontczak, G Giordana, J Giese, N Jahn, J Kakes, B Klaus, J H Lang, N Puzanova, and P Welz (2014), “Operationalising the Countercyclical Capital Buffer: Indicator Selection, Threshold Identification and Calibration Options”, ESRB occasional paper no. 5.
Duprey, T, B Klaus and T Peltonen (2015), “Dating Systemic Financial Stress Episodes in the EU Countries”, ECB Working Paper No. 1873, (Appendix and Data).
Hamilton, J (1989), “A new approach to the economic analysis of nonstationary time series and the business cycle”, Econometrica, Vol. 57, No. 2, 357–384.
Hollo, D, M Kremer, and M Lo Duca (2012), “CISS - A composite indicator of systemic stress in the financial system”, ECB Working Paper No. 1426.
Jorda, O, M H Schularick, and A M Taylor (2013), “When credit bites back”, Journal of Money, Credit and Banking, 45, 3–28.
Laeven, L and F Valencia (2013), “Systemic banking crises database”, IMF Economic Review, Vol. 61, No. 2, 225–270.
Reinhart, C M and K S Rogoff (2011), “From financial crash to debt crisis”, American Economic Review, Vol. 101, No. 5, 1676–1706.
Reinhart, C M and K S Rogoff (2014), “Recovery from Financial Crises: Evidence
from 100 Episodes”, American Economic Review, 104, 50–55.
