Macroeconomic and Financial Market Developments
An analysis of macroeconomic and financial developments provides an important context for the analysis of financial sector vulnerabilities. The goal of the surveillance of macroeconomic developments and of financial markets is to provide a forward-looking assessment of the likelihood of extreme shocks that can hit the financial system.
The literature on EWSs—which deals with factors that cause financial crises—provides useful guidance for this mode of analysis. EWSs try—in a statistically optimal way (i. e., in a way that minimizes “false alarms” and missed crises)—to combine a number of indicators into a single measure of the risk of a crisis. EWSs do not have perfect forecasting accuracy, but they offer a systematic method to predict crises. Two approaches to constructing EWS models have become common: the indicators approach (Kaminsky,
Lizondo, and Reinhart 1998, and Kaminsky 1999) and limited dependent variable probit – logit models (Berg and Pattillo 1999). Berg and others (2000) assess the performance of those models and find that they have outperformed alternative measures of vulnerability, such as bond spreads and credit ratings. However, although those models can anticipate some crises, they also generate many false alarms.4
EWS models are seen as one of a number of inputs into the IMF’s surveillance process, which encompasses a comprehensive and intensive policy dialogue. The IMF puts significant efforts into developing EWS models for emerging market economies, which resulted, among other things, in influential papers by Kaminsky, Lizondo, and Reinhart (1998) and by Berg and Pattillo (1999). The IMF uses a combination of EWS approaches, in particular, the Developing Country Studies Division model and a modification of the Kaminsky, Lizondo, and Reinhart model, both of which use macro-based indicators of currency crises (IMF 2002b). It also makes use of market-based models that rely on implied probability of default and balance-sheet-based vulnerability indicators (e. g., see Gapen and others 2004).
In recent years, other institutions and individuals have also developed EWS models. Those efforts included EWS models developed or studied by staff members at the U. S. Federal Reserve (Kamin, Schindler, and Samuel 2001), the European Central Bank (Bussiere and Fratzscher 2002), and the Bundesbank (Schnatz 1998). Academics and various private sector institutions also developed a range of EWS models. The private sector EWS models include Goldman Sachs’s GS-watch (Ades, Masih, and Tenengauzer 1998), Credit Suisse First Boston’s (CSFB’s) Emerging Markets Risk Indicator (EMRI) (Roy 2001), Deutsche Bank’s Alarm Clock (Garber, Lumsdaine, and Longato 2001), and Moody’s Macro Risk model (e. g., Gray, Merton, and Bodie 2003).
The EWS literature covers three main types of crises: currency crises (sudden, sizable depreciation of the exchange rate and loss of reserves), debt crises (default or restructuring on external debt), and banking crises (rundown of bank deposits and widespread failures of financial institutions). One can distinguish three “generations” of crises models, depending on what determinants the models take into account. The first generation focuses on macroeconomic imbalances (e. g., Krugman 1979). The second generation focuses on self-fulfilling speculative attacks, contagion, and weakness in domestic financial markets (e. g., Obstfeld 1996). The third generation of models introduces the role of moral hazard as a cause of excessive borrowing and suggests that asset prices can be a useful leading indicator of crises (e. g., Chang and Velasco 2001). In general, empirical studies (e. g., Berg and others 2000) suggest that currency crises occur more often than debt crises (roughly 6:1) and that a large portion of the debt crises happened along with or close to the currency crises. Banking crises are hard to identify, tend to be protracted, and, thus, have a larger macroeconomic effect. Banking crises also tend to occur with or shortly after a currency crisis.
Forecasting banking crises is based on three approaches:
• The macroeconomic approach is based on the idea that macroeconomic policies cause crisis, and it tries to predict banking crises using macroeconomic variables. For example, Demirgug-Kunt and Detragiache (1998) study the factors of systemic banking crises in a large sample of countries using a multivariate logit model and
find that crises tend to erupt when growth is low and inflation is high. They also find some association between banking sector problems, on the one hand, and high real interest rates, the vulnerability to balance of payments crises, the existence of an explicit deposit insurance scheme, and weak law enforcement, on the other hand.
• The bank balance-sheet approach assumes that poor banking practices cause crises and that bank failures can be predicted by balance-sheet data (e. g., Sahajwala and Van den Berg 2000; Jagtiani and others 2003).
• The market indicators approach assumes that equity and debt prices contain information on bank conditions beyond that of balance-sheet data. Market-based EWS models are based on the premise that financial asset prices contain information on market beliefs about the future. In particular, option prices reflect market beliefs about the future prices of the underlying assets. This information can be used to extract a probability distribution, namely, the probability of default. The advantage of equity and debt data is that they can be available in high frequency and that they should provide a forward-looking assessment (e. g., Bongini, Laeven, and Majnoni 2002; Gropp, Vesala, and Vulpes 2002).5