Contagion effects on stock markets of the
Turkey crisis of 2001
by
Dun Hou
Supervisor: Dr. A.J. Meesters
Faculty of Economics and Business
University of Groningen
Abstract
Contagion describes a phenomenon when a financial crisis in one country leads to crises elsewhere threatening the stability of international financial markets. This paper investigates the contagion effects of Turkish crisis of 2001 across fourteen stock markets. A Vector Autoregressive framework is adopted to detect if there is significant increase in stock market correlations. The bias due to heteroskedasticity is also adjusted, several robustness tests are conducted as well. Evidence presented in this paper shows that eight stocks markets are affected by the contagion effect originated from the turmoil in Turkish stock exchange after one month Turkish government floated its currency.
Introduction
Financial markets around the globe have become closely interwoven. In the last twenty years, we have witnessed financial crises shocking the world, from East Asia to Latin America, from Scandinavia to Russia. Remarkably, the severity, the speed of spread as well as the geographic reach of a crisis caused enormous losses and hurt the market confidence. There has been great interest in financial contagion, a phenomenon that financial crises are transmitted across markets. When a financial crisis goes from one economy to another, it may be simply a reflection of the victim country’s macroeconomic imbalances or undeveloped financial structure. However, a crisis is still transmitted to a country that does nothing fundamentally wrong to deserve such a blow. Hong Kong Stock Exchange during the Asian crisis of 1997 is just an example. It seems that good fundamentals alone cannot insulate an economy from financial contagion (Bordo and Murshid, 2000). The worldwide financial deregulation makes it much easier for financial institutions and corporations to tap into international capital market. When their portfolios spread widely, the risks of assets are reduced. Nonetheless, when financial instability can be easily transmitted to another market, the “safe heaven” would not exist and the portfolios are at risk.
The topic of this paper is the contagion effects of the Turkish crisis of 2001 on the global stock markets. Triggered by the liquidity problem in the banking system from 2000, the floatation of the Turkish lira in 2001 dragged the stock market down, though there was an IMF-led rescues package. The Turkish lira dropped 28% against the US dollar right after the authority decided to float the currency. In the following two months, it almost depreciated half of its value. The financial and economic crisis brought great loss to the domestic markets and had some impacts on international financial markets. The objective of this paper is to examine that if there are contagion effects caused by the Turkish crisis on the financial markets around the world.
2001 in previous literature. Most research on that crisis is concerning the causes and the development, but contagion is usually not the focus. Moreover, Turkey is chosen as usually one of the victim countries in contagion studies, other than the original crisis country (Alper and Yilmaz, 2000, Gazioglu, 2008). The contagion effect of the Asian crisis of 1997 is widely investigated as the magnitude of the crisis. The Turkish stock market is a market that it is young and small compared to other developed markets, so it attracts less attention. Second, it helps to understand the characteristics of contagion by studying crisis that is relatively small in terms of influence and damage. Dungey and Tambakis (2005) think that the crises are heterogeneous when contagion is considered in particular crisis incidents. In order to understand contagion better, it is important to study a variety of events to look for similarities and differences of crisis transmission that involve different asset markets, different stages and different regions. Collins and Gavron (2005) also argue the importance of including smaller, less prominent crises. The results of this paper suggest that trading linkage seems not a great factor for the transmission of contagion during the Turkish crisis. Moreover, not like other studies suggesting recent crisis are more contained than those in 1990s (Dungey, Fry, Brenda, and Martin, 2003, Boschi, 2005), this paper shows the evidence of contagion in a number of stock markets after the Turkish shock. Third, this paper adopts the methodology proposed by Forbes and Rigobon (2002), an examination of this crisis with this method is another added value of this paper.
The main research question is formulated as:
Is there a contagion effect of the Turkish crisis on the global financial markets?
In order to answer the research question, two sub-questions are set to break down the question to cover different aspects of the topic. They are:
1. What is financial crisis?
2. What is contagion? What theories can describe it? How is it measured?
section concludes the paper by presenting the main findings, limitations as well as suggestions for future research.
1. Literature review
In this section, financial crises are discussed in general, followed by an overview of the Turkish crisis of 2001. Then relevant literature revolving around contagion effect of financial markets is reviewed.
Financial crises are not only a threat to financial markets, but also to the entire economic environment. Financial crises hit the world from time to time during the 1990s and the beginning of the twenty-first century, the European exchange mechanism crisis of 1992, the Tequila crisis of 1994, the Asian crisis of 1997, the Russian crisis of 1998, the Brazilian crisis of 1999, the Turkish crisis of 2001, the Argentina defaults crisis of 2002, and the latest the US credit crisis of 2007. They led damaged economies into recessions, and some of them rapidly spread to other economies. According to Eichengreen and Portes (1987), financial crisis is “a disturbance to financial markets, associated typically with falling asset prices and insolvency among debtors and intermediaries, which ramifies through the financial system, disrupting the market’s capacity to allocate capital within the economy”. In a financial crisis, financial markets experience high volatility and significant illiquidity and the economy sees sharp current account reversals (Bordo, Eichengreen, Klingebiel, and Martinez-Peria, 2001).
institutions’ capital is exhausted to repay contracts on time (Laeven and Valencia, 2008). In some cases, a banking crisis is trigged by a sudden rush of withdrawals by depositors, a bank run. A debt crisis occurs when there is a sovereign default, a situation when a country fails to pay back its sovereign debt. More and more emerging countries finance themselves by issuing national bonds on international capital market. In the case of the Argentina crisis of 2002, the government could not pay back or refinance the debt, and eventually abandoned the peso-dollar parity. Though the distinction between three types of financial crisis, they have linkages as the asset markets are linked within a financial system. One crisis in one market may lead to crises in other markets. During the Asian crisis in 1997 some countries faced problems both in the balance of payments and in the banking industry. “Twin crisis” is used to describe such a phenomenon.
1 10 100 1000 10000 03/01/2000 27/03/2000 19/06/2000 11/09/2000 04/12/2000 26/02/2001 21/05/2001 13/08/2001 05/11/2001 Logarithmic Turkey Interbank Overnight Rate (Daily)
0 200000 400000 600000 800000 1000000 1200000 1400000 1600000 1800000 01/03/2000 30-03-2000 23-06-2000 18-09-2000 12/11/2000 15-03-2001 06/08/2001 09/03/2001 27-11-2001 Daily Exchange Rate (Turkish Lira/US$)
Turkish government decided to float Lira Source: the central bank of Turkey
The banking crunch The currency crisis
Figure 1 Turkey Inter-bank Overnight Rate
Figure 2 Exchange Rate
Source: the central bank of Turkey
Lira/US$ %
time
0 100 200 300 400 500 600 03/01/2000 27/03/2000 19/06/2000 11/09/2000 04/12/2000 26/02/2001 21/05/2001 13/08/2001 05/11/2001
There is a wide discussion on the reasons of this crisis among economists and policy makers. To understand the reasons behind the crisis, Ozkan (2005) illustrates that weak debt and fiscal positions and the weakness of financial and banking sectors are the main causes of the Turkish crisis between 2000 and 2001 (Fatih and Güven, 2002). Some authors question the stabilization programme in 2000 by stating that the ineffectiveness and disastrous outcome of the programme to tackle speculations on the markets (Ekinci and Erturk, 2007; Akyüz and Boratav, 2003). Erinç (2002) argues that, in contrast to the official stance that the crisis is the result of a combination of the failure of the public sector to reach the needed targets and the failure of implementing the rationale behind free market under globalization, the turmoil rooted in series of pressures along the integration with global capital markets and increased vulnerability of the financial system due to the stabilization plan.
A financial crisis in one country could lead to crises elsewhere threatening the stability of global financial markets. Contagion is held account for one of the reasons of the volatility transmission among markets. The term “contagion” generally describes a situation when a disease is spread by touching someone or something. In financial-crisis related literature, contagion is used broadly related to banking, currency, equities and property market crises to describe the effect in which a financial crisis in one country brings crisis to another (Moser, 2003). Contagion can be defined in different ways. Pericoli and Sbracia
Source: DataStream
The banking crunch
The currency crisis
Figure 3 Turkish Stock Exchange Index in US $ (Daily)
(2003) summarize five definitions; contagion occurs, 1) when there is a significant increase in the probability of a crisis in one country after another country is hit, 2) when the volatility of asset prices spills over from one country to another, 3) when the co-movements of asset prices cannot be explained by fundamentals, 4) when there is a significant increase of co-movements of asset prices between countries, 5) when transmission channels intensified after a crisis happens in one country. In this paper, the fourth definition is applied. More explicitly, if the correlation coefficient during the turmoil period increases significantly after a shock, a contagion occurs. This kind of correlation analysis has been widely used in recent empirical literature on contagion for its straightforwardness (Boschi 2005, Collins and Gavron, 2005, Hon, Strauss, and Yong, 2004, Climent Meneu 2005, Moser 2003).
potential benefits from international portfolio diversification and to take proper measures to reduce the risk of contagion for policy-makers (Moser, 2003).
Researchers try to find the evidence of contagion in developed countries. Early studies such as King and Wadhwani (1990), and Lee and Kim (1993) investigate co-movements among national stock markets during the US crisis of 1987. Longin and Solnik (1995) find that international covariance and correlation matrices of seven major developed countries are unstable during the period of 1960-1990. Eichengreen, Rose, and Wyplosz (1997) find evidence of contagion from a panel data of thirty years from twenty industrialized economies. Bonfiglioli and Favero (2005) implement structural models to examine the co-movements between US and German stock markets in both long and short terms. They argue there is no long-term contagion but short-term contagion between the US and the German stock market. Fleming and Lopez (1999) find there is volatility spill-over of US Treasury securities to Tokyo and London from the New York foreign exchange market. Mun (2005) investigates the contagion effect of the 9-11 terrorist attack across major stock markets in terms of stock returns and volatility. Evidence presented indicates stronger contagion of volatility from the US to the UK and Germany than to Japan, but stronger contagion of asset returns to Japan. Białkowski, Bohl and Serwa (2006) reject the hypothesis of financial contagion effect from the US market to the UK, Japanese and German markets during the period from 1984 to 2003. Their result suggests that financial crashes on the US stock market do not always lead to turbulence to these three markets, but can increase the possibility of a crisis on them. Hon, Strauss, and Yong (2007) focus on different sectors during the internet bubble in the US and find that collapse of stock markets is attributed to the close industry linkages but no to widespread contagion. Similarly, Kallberg and Pasquariello (2008) test if there are excess co-movements among 82 industry indices within the US stock market and find that these co-movements are significant across industries during the period from 1976 to 2001.
modest evidence of contagion across markets, additionally, and emerging stock markets in Central and Eastern European countries do not appear more vulnerable to contagion than developed Western European markets. Boschi (2005) looks for evidence of contagion to six emerging markets during the Argentina crisis of 2002. His analyses show that there is no evidence of contagion in the foreign exchange market, the stock exchange market, and the sovereign debt market. Dungey, Fry, Brenda, and Martin (2003) review and implement four alternative tests of contagion in international financial markets during the Tequila crisis of 1994-1995, the Hong Kong crisis of 1997 and the Argentina crisis of 2001-2002. Although they use different ways of dating the crisis periods, modeling common shocks, and treating endogeneity, the joint tests reject the null hypothesis of no contagion. In another study by Dungey et al (2007), using a panel of 10 emerging and industrial equity markets, they show that the contagion effect is significant and widespread during the Long Term Capital Management crisis, but more selective during the Russian crisis. In a recent paper, Didier, Mauro, and Schmukler (2008) argue that it is premature to conclude that contagion does vanish in emerging markets. Their reasons are that the main transmission channels are stronger than that in 1990s and the effect of anticipation from investors.
an increase in emerging country’s vulnerability to contagion (Stiglitz, 1999). Calvo and Mendoza (2000) argue that globalization may make liquid markets more vulnerable to contagion. Kaminsky and Reinhart (2000) investigate how commercial banks spread the initial shock by calling loans and drying up credit lines. A model developed by Pavlova and Rigobon (2004) shows the co-movements of stock, bond and foreign exchange markets can lead to contagion. Bae, Karolyi and Stulz (2003) argue that contagion is predictable based on asset return volatility, exchange rate changes and interest rates. After reviewing recent literature, Dungey and Tambakis (2005) argue that it seems to be that common lender effects do have influences on contagion risk, and financial linkages are probably more important than trade links, and additionally, the evidence on geographic feature is mixed.
As more and more research has been done to illustrate the importance of understanding contagion, reforms and proper policies are necessary to reduce the risks of financial contagion. Some argue that although improved disclosure requirements, regulation, supervision etc can prevent the build-up of vulnerabilities and reduce the risk of currency crises, improved implementation and surveillance are necessary as well (Dornbusch, et al, 2000). Kaminsky, Reinhart and Vegh (2003) suggest that the government should “not to overspend and over-borrow when international capitals are all too willing to lend as a surge of capital inflows usually ends in a sudden stop”. Van Rijckeghem and Weder (2003) argue that spill-over effect caused by cross-country bank lending contributes to the transmission of currency crises in Mexico and Asian countries, and suggest diversification and monitor of financing from creditors that exposed to potential crisis countries to reduce contagion risk.
concerning volatility, and the contagion effect from six major stock exchanges since 1992. They focus on the contagion effect originated from several events and crises of developed and emerging markets to the ISE. With the GARCH method, they suggest that there appears to be volatility contagion from five markets to the ISE at the confidence level of 10%. Moreover, the spill-over from major world stock markets is averagely four times greater than that from three emerging markets. Additionally, their results indicate that the ISE reflects more on the Russian market than other emerging market. In the study by Imer (2007), the existence and direction of contagion between Turkey and Russia are analyzed by implementing Granger causality tests. Their results reveal that there is bivariate Granger causality between the Turkish and the Russian stock markets during the crises in Brazil, Hong Kong, Mexico and Russia. Drakos and Kutan (2005) investigate the co-movements between the Turkish and the Greek financial market returns. They show that contagion effect is indicated between the Greek and the Turkish currency markets, but not between the stock markets. The Granger causality tests imply that the Turkish financial market acts like a shock receptor and adjusts itself to the equilibrium. They conclude that the financial co-movement between the Turkish and the Greek stock exchanges seems to be caused by the balance-of-payment shocks as the two countries share similar groups of trading partners and foreign direct investors. In addition, they suggest the stock market linkage between the two countries may be better explained by non-fundamental reasons. The Turkish financial markets in most studies are the victims of other crisis, there is few to investigate if there is a contagion effect caused by the Turkish crisis of 2001. The volatility of the ISE during the crisis period may spill over to other markets, it can be contagious.
2. Methodology and Data
Different methodologies can be used to test the existence of contagion. One way is to use the cross-market correlation coefficients (King and Wadhwani, 1990, Lee and Kim, 1993, Forbes and Rigobon, 2002). A second possibility to analyse market co-movements is to use ARCH or GARCH framework to estimate the variance/covariance transmission mechanisms between countries (Fleming and Lopez, 1999, Edwards, 1998, Mun, 2005). A third methodology to do cross-market linkage tests for changes is the co-integrating vector between long periods of time. Another method is to use a Markov switching framework (Białkowski, Bohl and Serwa, 2006). Additionally it is possible to measure transmission mechanisms directly by looking for different factors affecting a countries vulnerability to a financial crisis (Van Rijckeghem and Weder, 2003). This paper adopts the first method, that is to analyse cross-market correlation.
Correlation analysis with a VAR framework by Forbes and Rigobon (2002) is a straightforward and popular approach to detect contagion. All the variables in the model are systematically considered, and each variable has an equation to explain its evolution by including its own lags and lags of other variables. The correlation in returns between two stock markets is measured during a stable period and then a same kind of test during a period of financial crisis. If the correlation coefficient increases significantly, the transmission element between the two markets is strengthened after the shock and a contagion occurs. And more importantly, Forbes and Rigobon (2002) propose a method to adjust for the biased caused by heteroskedasticity. As they argue, the standard (conditional) correlation coefficients are vulnerable to heteroskedasticity bias because the variances are not constant. They prove that an increase in market volatility can greatly affect the estimates of cross-market correlation coefficients. Markets tend to become more volatile after a shock, so the conditional correlation coefficients will increase. In another word, “heteroskedasticity in market returns can cause the correlation coefficients to be biased upward after a crisis.” If the coefficients are not adjusted, it is impossible to know if the increase of in the conditional correlation represents a simple increase of volatility or an increase of unconditional correlation (the underlying cross-market relationship).
t t t t L X L I X =φ( ) +ϕ( ) +η (1) } , { j t TR t t x x X ≡ (2) } , , { g t j t TR t t i i i I ≡ (3) where TR t
x is the stock market return in Turkey, x is the stock market return in tj
another country, X is the vector of returns in the same two stock markets, t TR t
i , i and tj
g t
i are the short-term interest rates for Turkey, country j and the global as a whole
respectively, φ(L) and ϕ(L) are vectors of lags, ηt is a vector of white noise
disturbance.
Stock indices returns are calculated as rolling-average, two-day, which can control for the fact that stock exchanges in different countries are nominally open on the same date but may not during the same hours. Five lags are used to filter out any possible serial correlation and within-week variation in trading patterns. Interest rates are included as control variables for any aggregate shocks and/or monetary policy coordination globally and domestically.
The VAR framework specified in equations (1), (2) and (3) is used to estimate the variance-covariance matrices for each pair of countries for periods in question. Using the results from the former step, the cross-market correlation coefficients between Turkey and other countries in the sample during the pre-crisis, turmoil and the whole periods are calculated.
In order to adjust for heteroskedasticity, the change of volatility on the stock market, the proposed formula by Forbes and Rigobon (2002) to eliminate the bias is used:
] ) ( 1 [ 1 u 2 u ρ δ ρ ρ − + = (4)
where ρuis the unadjusted (i.e., conditional on heteroskedasticity) correlation coefficient,
δ is calculated according to equation: 1 − = l x h x σ σ δ (5)
where h stands for turmoil period and l for pre-crisis period.
In this paper, the Fisher z-transformation of correlation coefficients is used. Kat (2002) casts some doubts on the use of correlation to measure the dependence between different assets return due to assumption of normal distribution, and therefore the correlation coefficients should be interpreted with care. Dungey, and Zhumabekova (2001) suggest that it is possible to use Fisher Z-transformation to mitigate the problem of correlation coefficients. It can transform the correlation coefficients to a normal distribution, so it is possible to compare them between pre-crisis and crisis periods using a standard t-test. In the study by Dungey et al, (2003), their finding does not find indeed improvement by using the Fisher adjustment. The Z-transformation of the correlation coefficient is calculated as: ) 1 1 ln( 2 1 * * * * x x x ρ ρ ρ − + = (6) where ** x
ρ is Z-transformed coefficient of country x, * x
ρ is coefficient of country x.
Then these adjusted coefficients are used to perform the standard test for contagion. That whether there is a significant increase in any of these correlation coefficients during the turmoil period is tested by using a one-tail t-test:
3 1 3 1 * * * * − + − − = l h l h N N t ρ ρ (7) where ** h ρ and ** l
ρ are the correlation coefficients of turmoil and pre-crisis periods respectively. N and h N are number of observations of turmoil and pre-crisis periods l
respectively.
stated at the beginning of the paper. Dungey et al, (2003) suggests using the pre-crisis period as the low volatility period rather than the total sample period.
ρ is the correlation during the pre-crisis period and ρh is the correlation during the
turmoil period, the test hypotheses are:
h h H H ρ ρ ρ ρ < ≥ : : 1 0
The critical value for the t-test at the confidence level of 0.95 is 1.65, so any test statistic greater than this critical value indicates contagion, while any statistic less than or equal to this value indicates no contagion.
Turkey 0 100 200 300 400 500 600 01/01/1999 07/05/1999 10/09/1999 14/01/2000 19/05/2000 22/09/2000 26/01/2001 01/06/2001 05/10/2001
The countries are chosen according to the foreign trade volume of Turkey in 2000, as international trade reflects the economic interaction and closeness of countries. Top ten countries from both the exporting and importing lists are included as well as three countries for geographic consideration. They are seven European countries, Russia, Japan, USA and Argentina. The list of chosen countries is in Appendix A. As can be seen from Table 1, Germany was the biggest trading partner of Turkey in 2000, and there are nine countries in common on both lists, except for that Israel on the export list and Japan on import list respectively.
Percentage of Export Percentage of Import
Germany 18,65% Germany 13,21% USA 11,29% Italy 7,95% UK 7,33% USA 7,18% Italy 6,44% Russia 7,13% France 5,97% France 6,48% Netherlands 3,15% UK 5,04% Spain 2,57% Spain 3,08% Israel 2,34% Belgium-Luxembourg 3,05% Belgium-Luxembourg 2,33% Japan 2,97% Russia 2,32% Netherlands 2,91%
The data required for the contagion tests is obtained from DataStream. It includes daily country main stock exchange index returns. The indices are denoted in US Dollar that
Source: Turkish Statistical Institute
Table 1 Turkish Foreign Trade by countries in 2000 (Top Ten)
Figure 4 Turkish Stock Exchange Index with periods identified
time
Pre-crisis period
allows us to eliminate noise in the stock market fluctuations caused by currency changes (Climent and Vicente, 1997). Daily short-term interest rates including both local interest rates and London Interbank Offering Rate (LIBOR) that is used as a global risk free rate.
Some descriptive statistics of stock returns for all countries during pre-crisis period, turmoil period and full period are presented in Table 2. Not surprisingly, during turmoil period Turkey has the lowest mean and highest standard deviation among all the countries, -0.834% and 7.70% respectively, while only Russia and Romania have positive mean returns. The mean returns fall and standard deviations rise between the pre-crisis period and turmoil period in most of the sample markets. A large increase of standard deviation during turmoil period is the evidence of greater volatility. Ten out of fifteen countries get more volatile from pre-crisis period to turmoil period with Turkey having the biggest increase of 5.05%, from 2.65% to 7.70%, in terms of standard deviation change followed by Germany and Switzerland. To the contrast, Romanian stock market gets more stable decreasing from 1.89% to 1.16% followed by Russia and Israel. During all three periods, Russian stock market does the best with all positive mean returns, though its volatility is quite high. Moreover, Turkish stock market experiences the greatest increase and decrease, 12.82% and -20.65%, during the turbulent time among these countries, indicating how volatile the market was. German market has the second biggest drop of 4.52%, followed by Spain of 4.24% and Switzerland of 4.09%.
stable period turmoil period full period mean standard deviation max min mean standard deviation max min mean
standard deviation Turkey -0,232% 2,65% 18,35% -8,41% -0,834% 7,70% 12,82% -20,65% -0,272% 3,21% Argentina -0,067% 1,13% 5,52% -3,57% -0,370% 1,11% 2,82% -2,04% -0,086% 1,13% US -0,048% 1,08% 3,50% -4,19% -0,610% 1,04% 0,84% -3,47% -0,085% 1,08% UK -0,069% 0,78% 2,41% -2,92% -0,520% 1,12% 0,91% -3,17% -0,098% 0,82% Switzerland -0,002% 0,73% 1,84% -2,37% -0,683% 1,18% 0,91% -4,09% -0,047% 0,79% Spain -0,068% 1,10% 3,59% -2,73% -0,522% 1,40% 1,36% -4,24% -0,098% 1,13% Russia 0,066% 2,26% 8,42% -7,70% 0,179% 1,80% 4,74% -1,95% 0,073% 2,24% Romania -0,007% 1,89% 9,71% -10,36% 0,128% 1,16% 2,04% -2,93% -0,002% 1,85% Netherlands -0,046% 0,84% 3,12% -2,83% -0,625% 1,18% 0,90% -3,41% -0,084% 0,88% Italy -0,035% 0,96% 2,52% -2,77% -0,595% 1,21% 1,30% -3,68% -0,072% 0,99% Japan -0,147% 1,08% 3,10% -4,39% -0,214% 1,51% 2,67% -2,99% -0,152% 1,11% Israel 0,012% 1,28% 4,31% -5,06% -0,748% 1,04% 0,93% -2,67% -0,038% 1,28% Germany -0,060% 1,02% 3,05% -2,24% -0,744% 1,56% 1,54% -4,52% -0,107% 1,08% France -0,043% 1,08% 3,91% -3,91% -0,596% 1,27% 1,48% -3,71% -0,080% 1,10% Belgium -0,045% 0,91% 5,32% -2,98% -0,592% 1,14% 1,13% -2,91% -0,081% 0,93%
Figure 5 of five stock indices around the turmoil period gives a direct view of the fluctuations of the markets, from which we can see the effect of Turkish crisis on the other markets. The indices are scaled to 100 on the return of Jan 1, 2001, so it is straightforward to see the fluctuations of each market return.
0 20 40 60 80 100 120 2/1/2001 2/15/2001 3/1/2001 3/15/2001 3/29/2001 4/12/2001 4/26/2001
TURKEY GERMANY US RUSSIA ISRAEL
To understand the time-series properties of these stock returns, unit roots tests are performed. The results shown in Table 3 from the augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests suggest that the null hypothesis of non-stationary of these series is rejected in levels at confidence level of 1%. The test-statistics are less than the critical value of -3.45, thus none of the 15 stock index returns follows unit root process, and that the returns are stationary in levels. Figure 6 gives a view of the return on Turkish market, suggesting the process being stationary too. For the returns of other sample markets, see Appendix B.
Variables in Levels ADF PP ADF PP US -7,50* -10,33* Italy -6,02* -9,75* UK -7,03* -9,64* Japan -6,02* -7,38* Turkey -10,87* -8,23* Israel -9,40* -8,81* Switzerland -5,15* -9,02* Germany -6,79* -9,03* Spain -5,80* -10,13* France -6,67* -9,97* Russia -7,76* -10,5* Belgium -5,49* -8,28* Romania -7,56* -8,46* Argentina -6,57* -8,51* Netherlands -6,91* -10,67* -.3 -.2 -.1 .0 .1 .2 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_TURKEY
3. Empirical Results
This section discusses the results found from both the base test and the robustness tests.
Implementing the VAR model specified in equations (1), (2) and (3), the variance-covariance matrix is estimated. Two samples of the estimated VAR model are presented in Appendix D. The cross-market correlation coefficients between Turkish
The critical values (with intercept) for 1%, 5% and 10% are -3.45, -2.87, and -2.57 respectively,
ADF is short for the augmented Dickey-Fuller test, PP stands for the Phillips-Perron test * denotes the time series is stationary at confidence level of 1%
Figure 6 Turkish Stock Market Returns
Table 3 Unit root test statistics of returns (two-day rolling average)
stock market return and every other market return during the stable period and turmoil period are calculated. The conditional correlation coefficients are reported in Table 4. After that, the t-test (equation 7) to examine if there is a significant increase in the correlation coefficients during the turmoil period is performed. There are some characteristics can be drawn. First, Turkey has the highest correlation of 0.363 with Russia during the stable period, to the contrary, is negatively correlated to Romania. During the crisis period, however, Israel has the largest correlation of 0.801 with Turkey; and Japan and Romania are negatively correlated to Turkey with correlation coefficients of -3.18 and -1.82 respectively. Second, the correlations between Turkey and most other markets increase dramatically during the turmoil period. One extreme example is that, the correlation between Israel market return and Turkey jumps from 0.082 during stable period to 0.724 during turmoil period. Nonetheless, the correlation coefficient of Japan drops from 0.106 to -0.534 indicating an opposite reaction during the crisis period. Third, the t-statistics suggest that ten out of fifteen markets go through a period of a significant increase of correlations in the turmoil period, which implies that there is contagion effect rippling from Turkey to Italy, Argentina, Belgium, France, Israel, the Netherlands, Spain, Switzerland, UK and the US. Fourth, after implementing Z-transformation by equation (6), the t-tests coefficients do not change the evidence of contagion effect in these ten countries.
conditional correlation coefficient
stable turmoil t-statistic
t-statistic (z-transformed) Italy 0,034 0,515 2,39* 2,66* Argentina 0,122 0,469 1,73* 1,92* Belgium 0,015 0,693 3,37* 4,17* France 0,124 0,748 3,10* 4,20* Germany 0,090 0,349 1,29 1,36 Israel 0,082 0,801 3,57* 5,06* Japan 0,106 -0,534 -3,18 -3,49 Netherlands 0,077 0,648 2,84* 3,46* Romania -0,044 -0,410 -1,82 -1,95 Russia 0,363 0,472 0,54 0,66 Spain 0,121 0,735 3,05* 4,07* Switzerland 0,019 0,618 2,98* 3,50* UK 0,097 0,561 2,31* 2,67* US 0,007 0,794 3,92* 5,35*
The critical values for t-test at 5% is 1.65,
* denotes the increase is significant, thus there is contagion effect
Discussed by Forbes and Rigobon (2002), the results may be misleading and inaccurate due to the bias stemmed from heteroskedasticity. With equation (4), the conditional correlation coefficients are adjusted for heteroskedasticity. Then the tests are repeated using the unconditional correlation coefficients instead of conditional correlation coefficients. The results are displayed in Table 5. As can be seen, the t-statistics of unconditional correlation coefficients do not change much in most cases when compared with Table 4. Unlike the great reduction of number contagion incidences in Forbes and Rigobon (2002), when the cross-market correlations are adjusted for heteroskedasticity, there is still strong evidence of contagion. Nine countries show the evidence of significant increase in these unconditional correlation coefficients. They are Argentina, Belgium, France, Israel, the Netherlands, Spain, Switzerland, UK and the US. Out of the ten countries in the contagion group from Table 4, only Italy turns out not have contagion effect, with the t-statistic dropping from 2.39 to 1.49. Moreover, with regard to the application of Z-transformation, apparently, there is no effect on the evidence of contagion effect.
unconditional correlation coefficient
stable turmoil t-statistic
t-statistic (z-transformed) Italy 0,019 0,318 1,49 1,54 Argentina 0,174 0,606 2,15* 2,62* Belgium 0,016 0,697 3,39* 4,21* France 0,134 0,772 3,18* 4,43* Germany 0,075 0,296 1,10 1,14 Israel 0,088 0,820 3,64* 5,31* Japan 0,081 -0,437 -2,58 -2,73 Netherlands 0,071 0,622 2,74* 3,27* Romania -0,064 -0,545 -2,39 -2,72 Russia 0,633 0,747 0,57 1,09 Spain 0,089 0,622 2,65* 3,18* Switzerland 0,013 0,469 2,27* 2,46* UK 0,099 0,569 2,34* 2,72* US 0,005 0,683 3,37* 4,12*
Comparing financial linkages and trading linkage, Van Rijckeghem and Weder (2001) conclude that trade linkage is less important than financial linkage as a transmission mechanism for financial crisis. From Turkey’s twelve top trading partners of 2000, seven countries’ stock markets show significant increase in cross-market correlation after
The critical values for t-test at 5% is 1.65,
* denotes the increase is significant, thus there is contagion effect
Turkey floated its currency. Trade linkage seems to be a factor of transmission of contagion. Germany is the biggest trading partner of Turkey, but its stock market seems not to be affected by the downturn in Turkey in the crisis period. To some extent, it supports Dungey and Tambakis (2005)’s argument that trade linkage is getting less influential.
There is one more issue that needs to be mentioned is that, the t-statistics of Japan and Romania are negative and of a great amount, which means that both markets move in the opposite way of Turkish stock market during the crisis period. They are the special cases of the test. It needs deep investigation on this matter, which is not a part of the paper.
period, as five out six tests support the evidence of contagion. The Netherlands, Spain, Switzerland and the US affected by contagion in four of all tests. The extension of the pre-crisis period has the least effect on the test of contagion of all the sensitivity tests.
Table 7 shows the base test and robustness tests without adjusting for heteroskedasticity, that using conditional correlation coefficients. The total number of the contagion cases is 42. Compared with number of 39 on Table 6, it shows that by using unconditional correlation coefficients the big drop of contagion effect shown by Forbes and Rigobon (2002) does not make the case in my analysis. The effect of heteroskedasticity is not significant in this test. In their analysis, the evidence of contagion disappears in most of the cases after adjusting for heteroskedasticity. Nonetheless, comparing Table 6 and 7, in the case of Italy, the number of contagion cases decreases from four to one after adjusting for heteroskedasticity.
Turmoil Period: 22/2/2001-4/4/2001 Start of Stable Period :1/1/2000 Return Frequency: 2-day avg.
Lags Included: 5 Lags Included: 0
Start of Stable Period: 1/1/1999 Turmoil Period: 22/2/2001-2/5/2001 Return of Frequency: weekly Return of Frequency: daily Num. of contagion Italy * * 2 Argentina * * 2 Belgium * * 2 France * * * * * 5 Germany * * * 3 Israel * * * * * 5 Japan 0 Netherlands * * * * 4 Romania * 1 Russia * 1 Spain * * * * 4 Switzerland * * * * 4 UK * * 2 US * * * * 4 Num. of contagion 9 0 10 7 7 6 39
* denotes the increase is significant, thus there is contagion effect
Turmoil Period: 22/2/2001-4/4/2001 Start of Stable Period: 1/1/2000 Return Frequency: 2-day avg.
Lags Included: 5 Lags Included: 0
Start of Stable Period: 1/1/1999 Turmoil Period: 22/2/2001-2/5/2001 Return of Frequency: weekly Return of Frequency: daily Num. of contagion Italy * * * * 4 Argentina * * 2 Belgium * * 2 France * * * * * 5 Germany * * * * 4 Israel * * * * 4 Japan 0 Netherlands * * * * * 5 Romania * 1 Russia 0 Spain * * * * * 5 Switzerland * * * * 4 UK * 1 US * * * * * 5 Num. of contagion 10 2 9 8 6 7 42
This series of tests with modifications in different aspects of the model do not affect the main findings of the base test. However, the choice of lags has big influence on the results. If it is the case that five or four cases of contagion indicates one country has contagion effect from Turkish crisis in Table 6 and 7, that this group includes France, Israel, the Netherlands, Spain, Switzerland and the US. Put it in another way, the correlation between the Turkish stock market and one country of the group increases significantly during the turmoil period. Italy and Germany are not qualified to the group as the evidence of contagion becomes weak after adjusting for heteroskedasticity. Additionally, the correlation between the German stock market and the Turkish stock market increase significantly when the number of lags, turmoil period, and return frequency are changed. In another group of Japan, Romania, Russia, the stock markets do not react to the Turkish crisis as what contagion is defined in the beginning. These three countries have no case of contagion or only one case in the analysis. With regard to the adoption of Z-transformation of the correlation, results on Table 8 and Table 9 suggest the adoption does not make a difference in the analysis, which is consistent with Dungey et al, (2003) argue in their paper (see Appendix C).
* denotes the increase is significant, thus there is contagion effect
When the number of contagion cases on Table 6 and Table 7 are compared, the total number only slightly drops from 42 to 39 after heteroskedasticity is adjusted. This result is contrary to what Forbes and Rigobon (2002) find, in which they argue that there is no contagion but interdependence when heteroskedasticity is adjusted in their tests. The results show that contagion effect does exist, it does not vanish, which support the argument by Didier, et al (2008). Not only can crises from developed financial markets cause contagion to other markets, crises in developing markets, in this case Turkey, can become contagious too.
According to portfolio theory, risk can be reduced by investing in a wide range of assets. When investing in two asset markets, which are negatively correlated, the risk can be greatly reduced. The contagion definition used in this paper says that contagion happens when the correlation between two markets increases significantly after a shock. Suppose that a Turkish portfolio manager invests in a group of stock markets in several countries during the crisis. He or she can reduce the portfolio risk by putting more investing weight to markets where contagion does not occur as low correlation reduces risks. In the case, the manager may avoid lost by putting more money to Japanese stock market which has a negative correlation with Turkish market. The results of this paper does not indicate any clear pattern of countries where contagion can take place. Then it is better to invest in more markets to reduce risks.
4. Conclusion
The results indicate that the Turkish crisis of 2001 has contagion effects on eight stock markets (Argentina, Belgium, France, Israel, the Netherlands, Spain, Switzerland, UK and the US) in a sample of fourteen countries for one month after the drop of the Lira. In the base test, although the correlation coefficients are adjusted for heteroskedasticity, eight out of nine countries still show the evidence of contagion. It means there is contagion other than interdependence between the Turkish market and other eight markets during the crisis period. The results of this paper also suggest that trading linkage seems not a great factor for the transmission of contagion during the Turkish crisis. Robustness tests suggest the number of lags chosen in the model has the most influence on the results. Moreover, countries affected by contagion do not change when Fisher Z-transformation is applied which is to correct for not being normal distribution. In a word, the evidence of contagion on stock markets caused by the Turkish crisis of 2001 is modest among sample countries.
There are several limitations in model as well. First, the number of lags is predetermined, and the robustness tests suggest the importance of that choice. Second, the dating of pre-crisis period is not perfect as the Turkish stock market is not so stable during that period, during which a bank liquidation crisis took place. Third, though this paper includes both local interest rates and LIBOR as control variables, there may be some other omitted variables.
Appendices
Appendix A
List of sample countries:
Argentina, Belgium and Luxemburg, France, Germany, Italy, Israel, Japan, Romania, Russia, Spain, Switzerland, the Netherlands, United States, and United Kingdom
Appendix B -.3 -.2 -.1 .0 .1 .2 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_TURKEY -.05 -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_GERMANY -.04 -.03 -.02 -.01 .00 .01 .02 .03 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RNT_ITALY -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_FRANCE -.04 -.02 .00 .02 .04 .06 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_ARGINTINA -.04 -.02 .00 .02 .04 .06 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_BELGIUM -.06 -.04 -.02 .00 .02 .04 .06 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_ISRAEL -.05 -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_JAPAN -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_NETHERLANDS
-.12 -.08 -.04 .00 .04 .08 .12 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_ROMANIA -.08 -.04 .00 .04 .08 .12 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_RUSSIA -.05 -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_SPAIN -.05 -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_US -.05 -.04 -.03 -.02 -.01 .00 .01 .02 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_SWITZERLAND -.04 -.03 -.02 -.01 .00 .01 .02 .03 2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 RTN_UK Appendix C Turmoil Period: 22/2/2001-4/4/2001 Start of Stable Period: 1/1/2000 Return Frequency: 2-day avg.
Lags Included: 5 Lags Included: 0 Start of Stable Period: 1/1/1999 Turmoil Period: 22/2/2001-2/5/2001 Return of Frequency: weekly Return of Frequency: daily num. of contagion Italy * * * * 4 Argentina * * 2 Belgium * * * * 4 France * * * 3 Germany * * * * * * 6 Israel * * * * * 5 Japan 0 Netherlands * * * * 4 Romania 0 Russia 0 Spain * 1 Switzerland * * * 3 UK 0 US * * * * 4 num of contagion 9 6 8 3 3 7 36
* denotes the increase is significant, thus there is contagion effect
Turmoil Period: 22/2/2001-22/3/2001 Start of Stable Period: 1/1/2000 Return Frequency: 2-day avg.
Lags Included: 5 Lags Included: 0 Start of Stable Period: 1/1/1999 Turmoil Period: 22/2/2001-2/5/2001 Return of Frequency: weekly Return of Frequency: daily num. of contagion Italy * * 2 Argentina * * 2 Belgium * * 2 France * * * * * 5 Germany * * * 3 Israel * * * * * 5 Japan 0 Netherlands * * * 3 Romania * 1 Russia 0 Spain * * * * 4 Switzerland * * * * 4 UK * * 2 US * * * * 4 num of contagion 9 0 9 7 6 6 37 Appendix D
Vector Autoregressive Estimates:
RTN_Country Name stands for the market return of that country’s stock market, I_Country Name stands for the local interest rate of that country,
I_LIBOR stands for London Inter-Bank Offering Rate, Standard errors in ( ) & t-statistics in [ ].
RTN_TURKEY RTN_FRANCE I_TURKEY I_FRANCE I_LIBOR RTN_TURKEY(-1) 0.849861 -0.018885 -3.012156 0.005647 0.155931 (0.06144) (0.02558) (21.6310) (0.07615) (0.10465) [ 13.8330] [-0.73829] [-0.13925] [ 0.07416] [ 1.49003] RTN_TURKEY(-2) -0.686411 0.035812 -31.67809 0.015324 -0.342588 (0.07666) (0.03192) (26.9895) (0.09501) (0.13057) [-8.95434] [ 1.12208] [-1.17372] [ 0.16128] [-2.62372] RTN_TURKEY(-3) 0.479221 -0.039885 17.75329 0.124150 0.328085 (0.08337) (0.03471) (29.3522) (0.10333) (0.14200) [ 5.74830] [-1.14912] [ 0.60484] [ 1.20149] [ 2.31039] RTN_TURKEY(-4) -0.385830 0.051394 -10.17021 0.039695 -0.176603 (0.07694) (0.03203) (27.0884) (0.09536) (0.13105) [-5.01483] [ 1.60446] [-0.37544] [ 0.41626] [-1.34758]
* denotes the increase is significant, thus there is contagion effect
Table 9 Robustness Tests with Unconditional Correlation Coefficients without Z-transformation
RTN_TURKEY(-5) 0.107453 -0.047425 -6.666400 0.094746 0.214590 (0.06160) (0.02564) (21.6869) (0.07635) (0.10492) [ 1.74447] [-1.84927] [-0.30739] [ 1.24101] [ 2.04528] RTN_FRANCE(-1) 0.091509 0.820733 37.93227 -0.183094 0.043135 (0.14580) (0.06070) (51.3341) (0.18071) (0.24835) [ 0.62763] [ 13.5205] [ 0.73893] [-1.01317] [ 0.17368] RTN_FRANCE(-2) 0.023442 -0.724421 23.48912 0.433784 0.212407 (0.18396) (0.07659) (64.7680) (0.22801) (0.31334) [ 0.12743] [-9.45858] [ 0.36267] [ 1.90251] [ 0.67787] RTN_FRANCE(-3) 0.005401 0.466539 16.56417 -0.285370 0.017392 (0.19835) (0.08258) (69.8340) (0.24584) (0.33785) [ 0.02723] [ 5.64959] [ 0.23719] [-1.16080] [ 0.05148] RTN_FRANCE(-4) -0.064217 -0.271585 -60.92747 0.048625 -0.100184 (0.18013) (0.07500) (63.4203) (0.22326) (0.30682) [-0.35651] [-3.62137] [-0.96069] [ 0.21780] [-0.32652] RTN_FRANCE(-5) 0.108702 0.119026 23.92725 -0.030229 -0.270362 (0.14335) (0.05968) (50.4727) (0.17768) (0.24418) [ 0.75827] [ 1.99426] [ 0.47406] [-0.17013] [-1.10721] I_TURKEY(-1) 0.000367 3.29E-05 0.921194 -0.000238 -0.000228 (0.00017) (7.1E-05) (0.06028) (0.00021) (0.00029) [ 2.14128] [ 0.46159] [ 15.2808] [-1.12299] [-0.78101] I_TURKEY(-2) -0.000199 0.000119 -0.015230 0.000204 0.000428 (0.00023) (9.7E-05) (0.08228) (0.00029) (0.00040) [-0.85116] [ 1.21918] [-0.18509] [ 0.70585] [ 1.07485] I_TURKEY(-3) 4.17E-05 -7.01E-05 0.132535 -6.74E-05 -0.000291 (0.00023) (9.7E-05) (0.08200) (0.00029) (0.00040) [ 0.17920] [-0.72300] [ 1.61635] [-0.23334] [-0.73276] I_TURKEY(-4) -0.000278 -5.81E-05 -0.059182 -0.000162 -8.61E-05
(0.00023) (9.7E-05) (0.08219) (0.00029) (0.00040) [-1.18902] [-0.59741] [-0.72006] [-0.55881] [-0.21656] I_TURKEY(-5) 0.000240 9.01E-06 -0.056111 -5.61E-05 0.000118
I_LIBOR(-3) 0.040087 -0.045596 -5.014227 0.032327 0.170344 (0.04515) (0.01880) (15.8957) (0.05596) (0.07690) [ 0.88791] [-2.42573] [-0.31545] [ 0.57769] [ 2.21507] I_LIBOR(-4) -0.039323 0.022519 -19.22242 0.016000 -0.040525 (0.04577) (0.01906) (16.1141) (0.05673) (0.07796) [-0.85917] [ 1.18179] [-1.19290] [ 0.28206] [-0.51983] I_LIBOR(-5) 8.91E-06 -0.023394 26.67766 0.005768 0.131845 (0.03708) (0.01544) (13.0553) (0.04596) (0.06316) [ 0.00024] [-1.51535] [ 2.04343] [ 0.12550] [ 2.08745] R-squared 0.474406 0.456777 0.876799 0.998469 0.950210 Adj. R-squared 0.428200 0.409021 0.865968 0.998335 0.945833 S.E. equation 0.020013 0.008332 7.046265 0.024805 0.034089 F-statistic 10.26719 9.564846 80.95364 7419.897 217.0833
RTN_TURKEY RTN_FRANCE I_TURKEY I_FRANCE I_LIBOR RTN_TURKEY(-1) 0.637837 0.076813 -4.642715 -0.060738 0.250206 (0.39852) (0.05353) (75.6457) (0.16681) (0.28030) [ 1.60050] [ 1.43483] [-0.06137] [-0.36411] [ 0.89265] RTN_TURKEY(-2) -1.123326 -0.106687 -52.51284 -0.115276 -0.293715 (0.53991) (0.07253) (102.483) (0.22599) (0.37974) [-2.08058] [-1.47099] [-0.51240] [-0.51009] [-0.77346] RTN_TURKEY(-3) 1.423011 0.189741 -18.04865 0.147282 0.108530 (0.55971) (0.07519) (106.241) (0.23428) (0.39366) [ 2.54243] [ 2.52361] [-0.16988] [ 0.62866] [ 0.27569] RTN_TURKEY(-4) -1.101430 -0.112990 -6.822143 -0.256164 -0.129630 (0.45756) (0.06147) (86.8517) (0.19152) (0.32182) [-2.40718] [-1.83829] [-0.07855] [-1.33752] [-0.40280] RTN_TURKEY(-5) 0.448734 0.131085 -63.82269 0.068856 0.258774 (0.39368) (0.05288) (74.7265) (0.16478) (0.27689) [ 1.13984] [ 2.47873] [-0.85408] [ 0.41786] [ 0.93457] RTN_FRANCE(-1) 3.406929 1.039592 -45.47912 0.275046 0.036406 (3.36916) (0.45259) (639.517) (1.41024) (2.36966) [ 1.01121] [ 2.29700] [-0.07111] [ 0.19503] [ 0.01536] RTN_FRANCE(-2) -2.500297 -1.333589 158.0422 -0.394143 1.296993 (4.32473) (0.58095) (820.898) (1.81021) (3.04174) [-0.57814] [-2.29553] [ 0.19252] [-0.21773] [ 0.42640] RTN_FRANCE(-3) 2.178826 0.469279 42.04436 1.110148 -1.530384 (4.97968) (0.66893) (945.218) (2.08436) (3.50240) [ 0.43754] [ 0.70153] [ 0.04448] [ 0.53261] [-0.43695] RTN_FRANCE(-4) 1.157060 -0.346987 346.4498 0.861213 0.358354 (4.00004) (0.53734) (759.269) (1.67431) (2.81338) [ 0.28926] [-0.64576] [ 0.45629] [ 0.51437] [ 0.12737] RTN_FRANCE(-5) -3.387367 -0.436033 213.7533 -0.460322 -2.458958 (2.41992) (0.32507) (459.337) (1.01291) (1.70202) [-1.39979] [-1.34134] [ 0.46535] [-0.45445] [-1.44473] I_TURKEY(-1) 0.004642 0.000445 0.424057 0.000513 -0.002536 (0.00263) (0.00035) (0.49958) (0.00110) (0.00185) [ 1.76354] [ 1.25885] [ 0.84883] [ 0.46575] [-1.36990]
RTN_TURKEY RTN_JAPAN I_TURKEY I_JAPAN I_LIBOR RTN_TURKEY(-1) 0.854261 0.012657 -2.329797 0.019301 0.151084 (0.06152) (0.02360) (21.5476) (0.03826) (0.10407) [ 13.8853] [ 0.53631] [-0.10812] [ 0.50443] [ 1.45181] RTN_TURKEY(-2) -0.704746 -0.073970 -31.57116 -0.028453 -0.329717 (0.07704) (0.02955) (26.9815) (0.04791) (0.13031) [-9.14809] [-2.50297] [-1.17010] [-0.59386] [-2.53026] RTN_TURKEY(-3) 0.493989 0.080562 22.79802 0.041120 0.384056 (0.08427) (0.03233) (29.5146) (0.05241) (0.14254) [ 5.86200] [ 2.49208] [ 0.77243] [ 0.78457] [ 2.69432] RTN_TURKEY(-4) -0.385179 -0.047531 -7.781905 0.019137 -0.229168 (0.07900) (0.03031) (27.6700) (0.04913) (0.13363) [-4.87548] [-1.56833] [-0.28124] [ 0.38948] [-1.71488] RTN_TURKEY(-5) 0.094768 0.008976 -13.87994 0.006820 0.246728 (0.06224) (0.02387) (21.7973) (0.03871) (0.10527) [ 1.52274] [ 0.37598] [-0.63677] [ 0.17619] [ 2.34372] RTN_JAPAN(-1) 0.102952 0.960868 36.46592 -0.013160 0.653518 (0.15800) (0.06061) (55.3390) (0.09827) (0.26726) [ 0.65158] [ 15.8526] [ 0.65896] [-0.13392] [ 2.44522] RTN_JAPAN(-2) 0.164458 -0.737044 59.82280 0.135634 -0.388260 (0.21655) (0.08307) (75.8459) (0.13468) (0.36630) [ 0.75943] [-8.87215] [ 0.78874] [ 1.00706] [-1.05994] RTN_JAPAN(-3) -0.274373 0.366502 -103.6581 -0.026172 0.324864 (0.23648) (0.09072) (82.8253) (0.14708) (0.40001) [-1.16023] [ 4.04000] [-1.25153] [-0.17795] [ 0.81214] RTN_JAPAN(-4) 0.230753 -0.203907 23.05570 -0.200648 -0.263933 (0.21554) (0.08268) (75.4907) (0.13405) (0.36459) [ 1.07058] [-2.46607] [ 0.30541] [-1.49679] [-0.72392] RTN_JAPAN(-5) 0.010190 -0.010061 24.22749 0.224012 0.083043 (0.15693) (0.06020) (54.9647) (0.09760) (0.26546) [ 0.06493] [-0.16713] [ 0.44078] [ 2.29514] [ 0.31283] I_TURKEY(-1) 0.000378 -6.42E-05 0.904203 -0.000152 -0.000176 (0.00017) (6.6E-05) (0.06029) (0.00011) (0.00029) [ 2.19558] [-0.97253] [ 14.9974] [-1.42087] [-0.60563] I_TURKEY(-2) -0.000189 3.34E-05 0.003957 0.000195 0.000485 (0.00023) (9.0E-05) (0.08190) (0.00015) (0.00040) [-0.80736] [ 0.37219] [ 0.04832] [ 1.34007] [ 1.22677] I_TURKEY(-3) 7.67E-05 2.33E-05 0.135601 -0.000253 -0.000399 (0.00023) (8.9E-05) (0.08101) (0.00014) (0.00039) [ 0.33153] [ 0.26207] [ 1.67385] [-1.75925] [-1.02043] I_TURKEY(-4) -0.000302 -0.000127 -0.062882 0.000600 -2.30E-05 (0.00023) (9.0E-05) (0.08174) (0.00015) (0.00039) [-1.29460] [-1.42079] [-0.76930] [ 4.13194] [-0.05819] I_TURKEY(-5) 0.000267 0.000119 -0.074247 -0.000487 7.06E-05 (0.00017) (6.6E-05) (0.06028) (0.00011) (0.00029) [ 1.55378] [ 1.79617] [-1.23172] [-4.54567] [ 0.24247] I_JAPAN(-1) -0.006796 0.001682 22.32650 1.148639 0.232661 (0.09201) (0.03530) (32.2264) (0.05723) (0.15564) [-0.07386] [ 0.04765] [ 0.69280] [ 20.0721] [ 1.49487] I_JAPAN(-2) 0.122649 0.017985 -23.11703 -0.063074 -0.542916 (0.13915) (0.05338) (48.7348) (0.08654) (0.23537) [ 0.88143] [ 0.33693] [-0.47434] [-0.72884] [-2.30666]
I_JAPAN(-3) -0.167781 -0.035804 -39.22108 -0.079552 0.134428 (0.13275) (0.05093) (46.4947) (0.08256) (0.22455) [-1.26387] [-0.70307] [-0.84356] [-0.96353] [ 0.59866] I_JAPAN(-4) 0.049806 0.033175 17.84204 0.010538 0.165084 (0.12911) (0.04953) (45.2183) (0.08030) (0.21839) [ 0.38577] [ 0.66983] [ 0.39458] [ 0.13123] [ 0.75593] I_JAPAN(-5) -0.004548 -0.019107 26.28433 -0.017083 -0.005507 (0.08406) (0.03225) (29.4402) (0.05228) (0.14218) [-0.05410] [-0.59255] [ 0.89280] [-0.32677] [-0.03873] I_LIBOR(-1) -0.005338 -0.020153 -11.42630 -0.004433 0.732908 (0.03573) (0.01371) (12.5136) (0.02222) (0.06044) [-0.14940] [-1.47038] [-0.91311] [-0.19949] [ 12.1271] I_LIBOR(-2) 0.002428 0.026501 6.438640 0.010541 0.008856 (0.04418) (0.01695) (15.4722) (0.02747) (0.07472) [ 0.05495] [ 1.56376] [ 0.41614] [ 0.38365] [ 0.11852] I_LIBOR(-3) 0.037840 -0.018619 -0.271724 -0.012273 0.166879 (0.04420) (0.01696) (15.4809) (0.02749) (0.07477) [ 0.85610] [-1.09806] [-0.01755] [-0.44644] [ 2.23200] I_LIBOR(-4) -0.040251 0.005673 -15.64924 -0.042651 -0.018765 (0.04437) (0.01702) (15.5411) (0.02760) (0.07506) [-0.90710] [ 0.33326] [-1.00696] [-1.54550] [-0.25001] I_LIBOR(-5) 0.003709 0.006663 21.39048 0.049720 0.111170 (0.03617) (0.01387) (12.6672) (0.02249) (0.06118) [ 0.10256] [ 0.48024] [ 1.68865] [ 2.21038] [ 1.81717] R-squared 0.479334 0.542836 0.879229 0.996199 0.951360 Adj. R-squared 0.433561 0.502645 0.868611 0.995865 0.947084 S.E. equation 0.019919 0.007641 6.976435 0.012388 0.033693 F-statistic 10.47201 13.50664 82.81119 2981.496 222.4867
RTN_TURKEY RTN_JAPAN I_TURKEY I_JAPAN I_LIBOR RTN_TURKEY(-1) 0.380712 0.000476 -31.94736 -0.087184 0.633428 (0.35903) (0.06423) (86.9772) (0.10237) (0.35022) [ 1.06039] [ 0.00742] [-0.36731] [-0.85169] [ 1.80865] RTN_TURKEY(-2) -0.973267 0.031702 -8.175924 -0.045070 -0.671479 (0.40706) (0.07282) (98.6134) (0.11606) (0.39708) [-2.39095] [ 0.43535] [-0.08291] [-0.38833] [-1.69106] RTN_TURKEY(-3) 0.281021 0.033541 14.91478 -0.229989 0.573435 (0.51890) (0.09283) (125.706) (0.14795) (0.50617) [ 0.54157] [ 0.36133] [ 0.11865] [-1.55453] [ 1.13290] RTN_TURKEY(-4) -1.008510 0.038990 -34.85711 0.010407 -0.300543 (0.50132) (0.08968) (121.447) (0.14293) (0.48902) [-2.01172] [ 0.43477] [-0.28702] [ 0.07281] [-0.61459] RTN_TURKEY(-5) -0.260697 0.077183 37.22439 0.086285 0.235320 (0.53698) (0.09606) (130.088) (0.15310) (0.52381) [-0.48548] [ 0.80347] [ 0.28615] [ 0.56357] [ 0.44925]
References
Akyüz Yılmaz and Boratav Korkut “The Making of the Turkish Crisis”, World
Development, 2003, 31(9), 1549–1566
Alper C. Emre and Yilmaz Kamil, Volatility and contagion: evidence from the Istanbul stock exchange, Economic Systems, 2004, 28, 353–367,
Bae Kee-Hong, Karolyi G. Andrew, Stulz Rene M. A New Approach to Measuring Financial Contagion, the Review of Financial Studies, 2003, 16(3), 717-763, Białkowski, J, Bohl, Martin T. and Serwa, Dobromił, Testing for financial spillovers in
calm and turbulent periods, The Quarterly Review of Economics and Finance, 2006, 46, 397–412
Bikhchandani, Sushil, and Sunil Sharma, Herd Behavior in Financial Markets: A review,
IMF Working Paper WP/00/48, Washington DC: IMF, 2000,
Bonfiglioli, A., and Favero Carlo, A., Explaining co-movements between stock markets: The case of US and Germany, Journal of International Money and Finance, 2005, 24, 1299- 1316,
Bordo Michael, Eichengreen Barry, Klingebiel Daniela, and Martinez-Peria Maria Soledad, Is the crisis problem growing more severe?, Economic Policy, 2001, 16(32),
Bordo Michael, and Murshid A.P., Are Financial Crises Becoming More Contagious?,
NBER Working Paper 7900, 2000,
Boschi, Melisso, International financial contagion: evidence from the Argentine crisis of 2001–2002, Applied Financial Economics, 2005, 15, 153–163,
Calvo, G. A, “Capital Markets and the Exchange Rate with Special reference to the Dollarization debate in Latin America”, University of Maryland, 2000
Calvo, G. A., and E. G. Mendoza, Rational Contagion and the Globalization of Securities Markets, Journal of International Economics, 51, 2000,
Chordia, T., Roll, R. and Subrahmanyam, A., “Commonality in Liquidity”, Journal of
Financial Economics, 2000, 56, 3-28.
Collins,Daryl and Gavron,Shana, Measuring equity market contagion in multiple financial events, Applied Financial Economics, 2005, 15, 531–538,
Didier, T., Mauro, P., and Schmukler, Sergio L., Vanishing financial contagion? Journal
of Policy Modeling 2008, 30, 775–791,
Dornbusch, Rudiger, Yung Chul Park en Stijn Claessens, Contagion: Understanding How It Spreads, The World Bank Research Observer, 15, 2, 2000, 177-197.
Drakos Konstantinos, and Kutan Ali M., Why Do Financial Markets Move Together?
Eastern European Economics, 2005, 43(4), 5–26
Dungey Mardi, Renée Fry, Brenda González-Hermosillo and Vance L. Martin, Empirical Modeling of Contagion: A Review of Methodologies, International Monetary
Fund Working Paper, 2003,
Dungey Mardi, Renée Fry, Brenda González-Hermosillo and Vance L. Martin, Contagion in global equity markets in 1998: the effects of the Russian and LTCM crises,
North American Journal of Economics and Finance, 2007, 18,
Dungey, M. and D. Zhumabekova, Testing for Contagion using Correlations: Some Words of Caution, Pacific Basin Working Paper PB01-09, 2001.
Edwards, S., Interest Rate Volatility, Capital Control and Contagion, NBER working
paper 6756, 1998,
Ekinci Nazim Kadri and Erturk Korkut Alp, Turkish Currency Crisis of 2000-2001, Revisited, International Review of Applied Economics, 2007, 21(1), 29-41,
Eichengreen, Barry and Richard Portes, The Anatomy of Financial Crises, in Richard Portes and Alexander Swoboda (eds), Threats to International Financial Stability, Cambridge: Cambridge University Press, 1987,
Eichengreen, B., Rose, A. and Wyplosz, C., Contagious currency crises, Scandinavian
Economic Review, 1997, 98(4), 463–84,
Erinç Yeldan, Behind the 2000-2001 Turkish Crisis: Stability, Credibility, and Governance, for Whom?, Bilkent University, 2002,
Fatih Özatay and Güven Sak, the 2000-2001 Financial Crisis in Turkey, Central Bank of Turkey and Ankara University, 2002,
www.econ.brown.edu/fac/Herschel_Grossman/courses/122readings/Ozatay&Sak.p df
Fleming, M. J., and Lopez, J. A., Heat Waves, Meteor showers and trading volume: an Analysis of Volatility Spillovers in the US Treasury Market. Federal Reserve of
Forbes, K. and Rigobon, R., No Contagion, only Interdependence: Measuring Stock Market Co-movements, Journal of Finance, 2002, 57 (5), 2223-61.
Gazioglu Saziye, Stock Market Returns in an Emerging Financial Market: Turkish Case Study, Applied Economics, 2008, 40,
Glick, R., Rose, A.K., Contagion and trade: why are currency crises regional?, Journal of
International Money and Finance 1999,18 (4), 603-617.
Harrison, P., The Impact of Market Liquidity in Times of Stress on Corporate Bond Issuance, Risk Management and Systemic Risk, BIS, Basel, 2002,
Hon Mark, Strauss,Jack, Yong Soo-Keong, Contagion in Financial markets after September 11:Myth or reality, The Journal of Financial Research, 2004, 17(1), 95–114,
Hon Mark, Strauss,Jack, Yong Soo-Keong, Deconstructing the Nasdaq bubble: A look at contagion across international stock markets, Journal of International Financial
Markets,Institutions and Money, 2007, 17, 213-230,
Huang, Bwo-Nung; Yang, Chin-Wei. The impact of financial liberalization on stock price volatility in emerging markets. Journal of Comparative Economics, 2000, 28, 321–339.
Imer Evrim, Contagion and surprises of the stock market returns, Applied Economics
Letters, 2007, 14, 1053–1058,
Kallberg, Jarl and Pasquariello, Paolo, Time-series and cross-sectional excess co-movement in stock indexes, Journal of Empirical Finance, 2008, 15, 481–502 Kaminsky, Graciela, and Carmen Reinhart, On Crises, Contagion, and Confusion,
Journal of International Economics, 2000, 51, 145–68.
Kaminsky, G.L. and Reinhart, C.M., The Center and the Periphery: Tales of Financial Turmoil, NBER Working Paper, 2002
Kaminsky G. L., Carmen Reinhart, Carlos A. Vegh, The Unholy Trinity of Financial Contagion, Journal of Economic Perspectives, 4, 2003,
Kat, H., The Dangers of Using Correlation to Measure Dependence, Alternative
Investment Research Centre Working Paper Series, Working Paper No. 0010,
2002,
King, Mervyn, and Sushil Wadhwani, Transmission of Volatility between Stock Markets,
Review of Financial Studies, 1990, 3, 5–33,
Laeven Luc and Valencia Fabian, Systemic Banking Crises: a New Database, IMF
Lee, Sang B., and Kwang Jung Kim, Does the October 1987 crash strengthen the co-movements among national stock markets? Review of Financial Economics, 3, 1993, p89–102
Longin, F., and Solnik, B.,. Is the correlation in international equity returns constant:1960-1990?, Journal of International Money and Finance, 1995, 14 (1), 3-26.
Moser, Thomas, What Is International Financial Contagion? International Finance, 6(2), 2003, 157–178,
Mun C. Kyung-Chun, Contagion and impulse response of international stock markets around the 9–11 terrorist attacks, Global Finance Journal, 2005, 16, 48–68,
Ozkan, F. Gulcin, Currency and Financial Crises in Turkey 2000 – 2001: Bad Fundamentals or Bad Luck? The World Economy, 2005, 28(4), 541 – 572
Pavlova, A., and R. Rigobon, Asset Prices and Exchange Rates, MIT Sloan Working
Paper No. 4322-03, 2004,
Pericoli, M and Sbracia, M, A Primer on Financial Contagion, Journal of Economic
Surveys, 2003, 17(4).
Serwa, Dobromił and Bohl, Martin T., Financial contagion vulnerability and resistance: A comparison of European stock markets, Economic Systems, 2005, 29, 344–362 Stiglitz Joseph. Bleak Growth Prospects for the Developing World, International Herald
Tribune, April 10-11, 1999, 6.
Van Rijckeghem. C, and Weder. B, Spillovers through banking centers: a panel data analysis of bank flows, Journal of International Money and Finance, 22(4), 2003, Verbeek, Marno, A Guide to Modern Econometrics, second edition, Wiley, 2004,
321-322,
Wood, Geoffrey, Great Crashes in History: Have they Lessons for Today?, Oxford
