期刊名称：Document de Travail / Centre d'Etudes Prospectives et d'Informations Internationales
出版年度：2007
卷号：1
出版社：Paris
摘要：Financial institutions often refer to empirical risk aversion indicators to gauge investors’ market sentiment. Fluctuations in risk aversion are generally considered as a factor explaining crises. Periods of strong risk appetite can create speculative bubbles on financial prices, building up vulnerabilities. Then a sudden reversal in risk aversion may trigger sharp falls in asset prices and prompt a financial crisis.
A crucial point is to clearly define the concept of risk aversion. In the framework of asset pricing models, more precisely the Consumption CAPM (CCAPM), a risk premium can be decomposed into a “price of risk”, which is common to all assets, and a “quantity of risk”, which is specific to each asset. The empirical indicators of risk aversion used by financial institutions aim at assessing this “price of risk”.
Those empirical indicators can be put together in four main groups. 1) The indicators of the GRAI (Global Risk Aversion Index) type are based on the idea that an increase in risk aversion should lead to a rise in risk premia across all markets, but the rise should be greater on the riskiest markets (Persaud, 1996, Kumar and Persaud, 2002). By using the CAPM, regarded as a special case of the CCAPM, this idea amounts to assessing changes in risk aversion as the correlation between price changes and their volatility. 2) Risk aversion can also be estimated as the common factor driving risk premia. This common factor can be evaluated through a factor analysis such as the Principal Component Analysis (PCA). 3) Some financial institutions also use raw series, as the VIX which is the implied volatility on the S&P 500, or combinations of raw series. 4) There are also other indicators, such as the State Street’s one which does not fall into the previous categories.
In order to assess the relevance of all these empirical indicators, we investigate their ability to forecast exchange rate and stock market crises, by constructing “early warning signals” of crises. We use logit and multilogit models that link a qualitative variable representing crises to a set of quantitative exogenous variables. In a first model, the explanatory variables are the usual ones found in the economic literature. In a second model we add risk aversion indicators to these control variables. In a third one, risk aversion indicators are taken as the only explicative variable.
The results show that most of the considered risk aversion indicators have the expected positive sign and are significant in the regressions. Moreover, in the multilogit models, risk aversion remains high during the months following the crisis. As regard to their predictive power, which is tested in-the-sample, the results are quite different according to the type of crises. For currency crises, the indicators barely improve the prediction made by the usual control variables. By contrast, in the case of stock market crises, most indicators yield satisfactory results. The best predicting performances are obtained by a principal component analysis on risk premia.
关键词：Risk aversion; leading indicators of crises; currency crises; stock market crises; crises prediction
