Journal of Applied Economic Sciences
Volume IV/ Issue 1(7)/ Spring 2009
taxes. Hence, we expect an increase in the probability of currency crisis in a country when there is an
increase in the CDS premium.
Second, the theory for the effect of the changes in CDS premium on the probability of a stock
market crisis is based on Merton’s (1974) model and its extension to sovereign debt by Chan-Lau &
Kim (2004). First, Duffie (1999) notes that the yield from a holding a risky bond and paying a CDS
premium, is equivalent to the yield from holding a risk-free bond. However, Chan-Lau & Kim (2004)
observe that in practice the risk-free yield and the yield of a risky asset with CDS is not the same,
because there is always a difference in bond spreads and CDS spreads evident in CDS-bond
differential. The CDS-bond differential partly exists because of the different liquidity in the markets
and the cheapest-to-deliver option premium in bonds. However, in the long-run, there should be a co-
integration between bond spreads and CDS spreads.
Merton’s model links bond and equity prices by taking a balance sheet approach. It argues that
if the value of a firm’s assets falls below the face value of its debt, the firm defaults. Also, there is a
positive correlation between bond and equity prices, and hence equity prices and bond spreads move
in opposite direction. Chan-Lau and Kim (2004) extend the model to include sovereign obligations as
equivalents of the firms’ debt.
From the CDS-bond differential and the relationship between equity prices and bond spreads,
we infer that CDS spreads and equity prices move in opposite directions. Furthermore, we use CDS as
a source of information for investor expectations in country’s equity market, from our assumption that
CDSs reflect future market expectations. So, we expect a positive relationship between changes in
CDS premium and the probability of a crisis in stock markets.
3.Literature review
The idea underlying the empirical research of crises prediction is to identify some factors that
show specific patters prior to periods of crises. The goal is to build a system that assesses the
probability of crises at a specific time horizon, taking into consideration all information available at
the time of prediction. There are three methodological approaches used in literature on currency and
stock market crises. The first approach does not concentrate on the factors that caused the crisis but
rather wants to analyze the effects of crisis on some specific sector of the economy. An example is
Sachs, Tornell and Velasco (1996), who examine the implications of 1995 crises and try to answer the
question of why some emerging markets were hit by the crises while others not.
Kaminsky and Reinhart (1996) devise a methodology called a “signal approach” to identify the
periods where a crisis will occur. Numerous papers follow this method which looks for any pattern in
individual variables prior to crisis. When a pattern is found (e.g. a deviation from mean up to a certain
threshold) a signal is issued by the variable tested. The threshold is chosen so as to minimize the false
signals. The advantage of this approach is that it produces easily understandable results for policy
purposes. However, it ignores the interaction between independent variables and standard statistical
tests cannot be applied.
This paper follows the third approach which eliminates some of the disadvantages of the signal
approach by using a limited dependent variable. The method uses a logistic function to evaluate the
overall effect of the explanatory variables and predict an outcome, i.e. the probability of the crises,
constrained by zero and one. Kumar et al (2003) use a logistic model to study currency crises in 32
developing countries for a period of 15 years.
Factors suggested by Kaminsky, Lizondo and Reinhart (1998) and used in this paper to predict
currency crises are: terms of trade, real interest rate, current account deficit, unemployment rate, GDP
growth, changes in consumer prices, and returns in stock market indices. For the stock market, factors
used in this paper and suggested by Boucher (2004) and also used by Curdert and Gex(2007) are price
earnings ratio, stock returns and real interest rates.
The theory of financial crises suggests that economic fundamentals are the main cause of the
financial crises. By economic fundamentals, we mean macroeconomic factors such as GDP growth,
unemployment, inflation etc. The exact timing of the crises was first linearly determined and the
crises had been predictable with economic fundamentals. Kaminsky and Lizondo (1996) specify
models that predicted crises by using economic fundamentals in a non-linear fashion. However,
economic fundamentals were not the only leading indicators of the crises. Investor behavior and other
127