Stability periods between financial crises: The role of macroeconomic fundamentals and crises management policies☆
Introduction
Public and academic debates on the topic of financial crises primarily take issue with the onset and direct recovery from financial crises episodes. Recently, however, some scholars, such as Reinhart and Rogoff, 2009, Reinhart and Rogoff, 2010a, Reinhart and Rogoff, 2010b turned their attention to the recurrence of banking crises. They show that these perennial events have occurred at a relatively stable frequency in both emerging and advanced economies. At the same time, a small share of the literature has been devoted to the fact that some countries have faced a larger number of financial crises than others. In the hope to gain some additional insight, this study attempts to cover the other side of the same coin: The determinants of the length of stability periods between financial crises. To this end, an innovative econometric technique is used: The Finite Mixture Model. To the best of our knowledge, this is the first attempt to do so.
A closely related topic, the frequency of financial crises, has, among others, been analyzed by Jordà et al. (2010), who investigate if crises can be predicted by macroeconomic fundamentals or whether they are randomly distributed events. Rejecting the former, the authors conclude that most financial crises occur randomly. Regarding the behavior of macroeconomic variables during pre- and post-crises periods, they find that such events are preceded by low natural interest rates and rising credit, while a large share of financial crises episodes are followed by recessions. Distinguishing between “normal” recessions during the business cycle and recessions accompanied by financial crises, they discover that the latter ones are one third more costly in terms of output losses.
Tudela (2004) adopts a duration model approach to analyze the determinants of currency crises. One of the main objectives of the study is to test for time dependence, that is, the length of the time already spent in a tranquil period as a determinant of the probability of exit into a crisis state. A justification for the use of this framework is that duration models allow accounting for duration dependence among the determinants of the likelihood of speculative attacks, without neglecting the use of time-varying explanatory variables. The results exhibit the existence of highly significant negative duration dependence, meaning that the highest probability to exit into a currency crises state is observed during the initial phase of the tranquil period.
While the duration between crises is directly linked to the frequency of crises, distinct conclusions can be drawn from the analysis of either one. Dissimilar economic fundamentals could be at play in a country experiencing a certain number of crises, clustered over a short period of time, and another country experiencing the same number of crises spread out over a larger period of time. Only by examining entire periods of stability can one take into account and disentangle factor development during the period of recovery and consider the immediate time span before financial turmoil occurs.
This study therefore estimates various models analyzing the time span between different types of financial crises in order to determine whether some variables explain why some countries are more prone to enjoy long durations of stability, while others experience crises in shorter intervals.
In order to do so, we calculate the duration of stability periods between currency, debt, and banking crises from 1980 to 2008. The distribution of this variable appears to be bimodal for currency and debt crisis. Two groups of observations emerge: One, depicting average stability periods of around 5 years, and a second group experiencing crises roughly each 15 years. The distribution of the duration between banking crises is unimodal with a peak at 11 years.
From a methodological point of view, the existence of bimodal distributions of durations between currency and debt crises makes traditional econometric methods non-valid. One of the main contributions of this paper is that it uses an innovative approach which is robust to the problems of asymmetric, skewed, or multimodal distributions, namely the Finite Mixture Model (FMM). The model permits differentiation between the effects of an explanatory variable on the probability of belonging to either group of observations and on the variability within both groups.
The FMM is estimated separately for currency and debt crises including 3 groups of concomitant variables: Real variables, financial and monetary variables, as well as equilibrium or external sector variables. For each variable, we distinguish between the long term (recovery) and immediate pre-crisis impact. Finally, we focus on systemic financial crises analyzing the impact of different macroeconomic and regulatory policies on the stability periods after those episodes.
The FMM is found to predict durations between financial crises fairly well. However, clear differences between the predictive powers of the different groups of variables exist.
A very intuitive finding, in line with the reasoning of macro-prudential policy toolkits, regarding stability periods between currency crises is that high GDP growth and the accumulation of net foreign assets tend to pro-long stability periods. At the same time, an increase in the real interest rate during the three years prior to a crisis decreases the duration of stability periods.
Concerning debt crises, results are consistent with findings in literature on debt sustainability. In addition, the main country characteristics usually considered by rating agencies in order to assess sovereign default risk are statistically significant in explaining the length of stability periods.
Analyzing the effect of policies in the aftermath of systemic banking crises, we show that the duration of stability periods decreases with the level of the maximum of liquidity support in percentage of deposit which monetary authorities accord to banks. Large-scale government interventions may help to restore the confidence in banks and to sustain accelerated recovery of the economy. In addition, an adoption of IMF programs during and/or after a systemic banking crisis may help to stabilize the economy.
In this respect, a further area of research closely related to the topic of this study investigates the factors which bring about financial crises: The “early warning indicators”. Mainly, this strand of the literature aims to identify the predictors of crises by using different econometric methodologies that can be grouped into two broad categories. First, the signals (or threshold) approach, introduced by Kaminsky et al. (1998) and Kaminsky and Reinhart (1999), compares information on indicators in periods of tranquility with identified periods of crisis.1
The second approach consists in linear regression or limited dependent variable probit/logit techniques. With this methodology, the probability that a crisis occurs is assumed to be a function of a vector of explanatory variables, allowing the assessment of the simultaneous effect of more variables, instead of evaluating the predictive power of each variable in isolation. Some of the first papers to use these techniques are: Eichengreen et al. (1995), Frankel and Rose (1996), J. Sachs et al. (1996) and Demirgü-Kunt and Detragiache, 1998, Demirgü-Kunt and Detragiache, 2005.2
The remainder of this paper is organized as follows: Section 2 introduces the data used and preliminary statistics on the occurrence of financial crises, as well as on the duration between them. Section 3 presents the econometric methodology, introducing the Finite Mixture Model, and explains the choice and computation of the concomitant variables. Section 4 offers bootstrap estimations as a solution to the small sample issue encountered and estimation results concerning stability periods between currency and debt crises. Subsequently, the predictive quality of each group of variables is assessed. Section 5 provides some tentative policy recommendations and Section 6 concludes.
Section snippets
Data
This study uses annual data from 1960 to 2008 for 176 developing and developed countries. It thus excludes the recent global financial crisis episode. Sources for macroeconomic indicators are shown in Table A-5.
The Finite Mixture Model
The extent and the potential for applications of Finite Mixture Models (FMMs) have widened considerably during the past decade. Initially, FMMs have been used in medical studies with different subgroups of patients. Because of their flexibility, mixture models are being increasingly exploited as a convenient, semi-parametric way to model unknown distributional shapes. In this respect, Finite Mixture Models provide great flexibility in fitting models with many modes, skewness and non-standard
Small sample issue: bootstrap simulations
While the number of observations is “sufficient” to perform the FMM for currency crises, the low frequency of debt crises makes estimations of the latter less reliable. A solution to treat this small sample issue is to use the nonparametric bootstrap method, a simulation technique consisting of re-estimating iteratively the FMM parameters for randomly drawn samples.9
Policy recommendations after systemic financial crises
In this section we analyze the duration of stability periods after systemic crises, defined as a banking crisis preceded or followed either by a currency or a debt crisis, or in the most extreme case the occurrence of all three types of crises within a period of 5 years.18
Conclusion
This paper has analyzed the duration of stability periods between currency, debt, and banking crises. The distribution of this variable was found to be bimodal regarding currency and debt crises and two groups of observations emerged: One depicting short stability periods and a second group experiencing long spells of stability, while the distribution of the duration between banking crises is unimodal. After having shown that the distribution of durations does not allow for traditional
References (40)
- et al.
Currency crashes in emerging markets: an empirical treatment
J. Int. Econ.
(1996) - et al.
Contribution to the discussion of paper by S. Richardson and P.J. Green
J. R. Stat. Soc. B
(1997)- et al.
Quasi real time early warning indicators for costly asset price boom/bust cycles: a role for global liquidity
Eur. J. Polit. Econ.
(2011) - et al.
Genetic algorithms and financial crises in emerging markets
- et al.
Assessing the risk of banking crisis
BIS Q. Rev.
(2002) - et al.(2004)
- et al.
Growth dynamics: the myth of economic recovery
Am. Econ. Rev.
(2008) - et al.
Crisis management and resolution: early lessons from the financial crisis
Estimating the components of a mixture of normal distributions
Biometrika
Speculative attacks on pegged exchange rates: an empirical exploration with special reference to the European Monetary System
Sovereign Rating Methodology
Managing macroeconomic crises
Structural vulnerabilities and currency crises. IMF staff papers
Palgrave Macmillan J.
Assessing financial vulnerability
An Early Warning System for Emerging Markets
Determinants of ex-ante banking system distress: a macro–micro empirical exploration of some recent episodes
Cited by (7)
Saving, fertility and public policy in an overlapping generations small open economy
2019, Economic ModellingThe Rising Resilience of Emerging Market and Developing Economies
2015, World DevelopmentBanks’ Financial Stability and Macroeconomic Key Factors in GCC Countries
2022, Sustainability (Switzerland)Capital and liquidity risks and financial stability: Pre, during and post financial crisis between islamic and conventional banks in gcc countries, in the light of oil prices decline
2020, International Journal of Financial ResearchLean six sigma, crises management and innovation: A theoretical Frame work
2018, Proceedings of the International Conference on Industrial Engineering and Operations ManagementFinancial crises, 1929 to the present, second edition
2017, Financial Crises, 1929 to the Present, Second Edition
- ☆
We thank Professor Fabrizio Coricelli for the extremely helpful comments and suggestions. Thanks also to the participants of the Doctoral Seminar at the Pantheon-Sorbonne University of Paris 1, of the Orleans Doctoral Development Days, of the Macro Ph.D. seminar at the PSE, of the Central Bank of Peru Research Seminar and of the Bi-annual meeting of the UP1 Doctoral School EPS.