f(x) =
ex
1 + e-x
1 + ex
(6)
The first model uses the following logistic regression model to estimate the probability of a
crisis in stock and currency markets of emerging countries within a one-month horizon by regressing
the indicator on variables commonly used in literature:
/ ʌ . 'ʌ
∣m crisis t = 1) = f Iα+∑ βkXi, t-11 (7)
∖ k=1 J
where αand β are the coefficients while Xik,t-1 are the variables used in past studies on the field.
For currency crises model, the variables used are: real interest rate, terms of trade, current account,
unemployment rate, GDP growth, inflation and one month stock returns. However, there is not as
much literature that addresses stock market crises as there is for currency crises. We use the same
factors as in Curdert and Gex (2007) for stock market crises model, namely: price earnings ratio of the
indices, one month stock returns and real interest rates.
The second model extends the first model by adding changes in sovereign CDS premiums as a
factor predicting crises in emerging markets. The following equation is estimated:
n
Pr(crisisi,t=1)=f(α+∑βkXik,t-1+βn+1∆CDSi,t-1)
k=1
(8)
where ∆CDSi,t-1 = (CDSi,t-1 /CDSi,t-2) -1 is the lagged one-month change in sovereign CDS
premiums on the international bonds of the emerging countries considered in this paper. Here,
fluctuations in CDS are regressed on the presence of other factors used in the first model in order to
check its ability to improve the forecasts.
The third model uses the following logistic regression equation to estimate crises probabilities
by using changes in sovereign CDS premiums as the only factor:
Pr(crisisi,t = 1) = f (α+ β∆CDSi,t-1) (9)
where ∆CDSi,t-1 = (CDSi,t-1 /CDSi,t-2) -1 is the lagged one-month change in sovereign CDS
premiums as in the second model.
5. Data
The dataset consists of 21 emerging market countries: Argentina, Brazil, Bulgaria, Chile,
China, Colombia, Croatia, Hungary, India, Indonesia, Israel, Malaysia, Mexico, Peru, Philippines,
Poland, Russia, South Africa, South Korea, Thailand and Turkey. We use panel data with monthly
frequencies starting from the date when CDS quotes were available for the respective reference
country until August 2008. We used Bloomberg terminal to retrieve stock market index levels, P/E
ratios of indexes and 5 year sovereign credit default swap premiums whereas IMF’s International
Financial Statistics database was used to retrieve other variables.
The choice of independent variables in both markets was based on earlier studies, and the
variables were earlier found to be related to currency and stock market crises respectively. Some
variables needed to be created from the retrieved series. In stock markets, six month stock return
series were created by using the index price series obtained from Bloomberg. In currency markets,
following Kaminsky et al (1998), we created terms of trade as the ratio between exports and imports.
Table 2 below summarizes the independent variables and their sources.
130
More intriguing information
1. Putting Globalization and Concentration in the Agri-food Sector into Context2. The name is absent
3. The name is absent
4. ADJUSTMENT TO GLOBALISATION: A STUDY OF THE FOOTWEAR INDUSTRY IN EUROPE
5. 101 Proposals to reform the Stability and Growth Pact. Why so many? A Survey
6. Automatic Dream Sentiment Analysis
7. Direct observations of the kinetics of migrating T-cells suggest active retention by endothelial cells with continual bidirectional migration
8. A Consistent Nonparametric Test for Causality in Quantile
9. Input-Output Analysis, Linear Programming and Modified Multipliers
10. Influence of Mucilage Viscosity On The Globule Structure And Stability Of Certain Starch Emulsions