Banking crises: date identification, determinants and government intervention

During the Global Financial Crisis (GFC), a number of countries suffered banking crises. This thesis comprises three parts on banking crisis prevention and management. The first part identifies banking crisis dates based on market information embedded in banking stocks. Specifically, the daily returns on banking system stock indices from a sample of countries over the period 1995 to 2010 are estimated using a Markov Switching Autoregressive (MS-AR) model to capture regime shift behaviour in both the mean and variance. Overall, evidence of three regimes (bull/bear/crisis) is found and banking crisis dates are identified in all the countries examined. The crisis regime is characterized by higher volatility and lower stock returns. This MS-AR modelling offers an alternative ex-ante method of crisis date identification and our identified crisis dates are, in general, consistent with the IMF’s ex-post crisis date classification. The second part of this thesis identifies the determinants of a banking crisis. A sample of developed countries in the U.S., Europe and Australasia are employed to identify whether systemic financial linkages together with conventional microeconomic and macroeconomic variables can be used as determinants of a crisis. Specifically, a panel probit model is estimated to explore determinants over the period from 1998 to 2010 for a sample of 11 countries. The best warning sign of the recent GFC was the interbank the Libor and Overnight Index Swap spread (LIBOR-OIS) which proxies for vulnerability of banking system liquidity. This finding is consistent with multiple instances of wholesale bank runs during the GFC, together with the associated sharp increases in liquidity premiums. The final part of this thesis examines different government intervention tools in response to a banking crisis to provide insight into the effect of changes in crisis management policy, which affect the probability of insulating an economy from the crisis, or alternatively the probability of ending or exacerbating the economy from the crisis. The impact of news announcements of government interventions of the 3 developed countries (the U.S., the U.K. and Japan), together with microeconomic and macroeconomic variables surrounding the period from 2007 to 2009 are estimated using a multi-state (no crisis, successful and unsuccessful defence states of the economies) multinomial logit model. The findings of the analysis for the third part indicate that liquidity policies and non-intervention policies increase the probability of successful defence. This finding is in line with the findings for the second part which suggest that the best predictor of the recent GFC was systemic liquidity shortages among banks. In contrast, monetary policy is ineffective in defending banking crises, as it increases the probability of an unsuccessful defence. In essence, governments and regulators should adopt liquidity policies during periods of banking crises but the findings of this thesis suggest they should not intervene otherwise.