“The effect of financial crises on stock market liquidity across global markets”

In this study, using a widely available market liquidity measure, the “turnover ratio”, the authors test for market liquidity contagion during the four financial crises that occurred between 1997 and 1999: The Thai crisis, the Hong Kong crisis, the Russian crisis, and the Brazilian crisis. It is found that while the liquidity levels decreased in approximately half of the sample markets, in the remaining half, the liquidity levels actually improved. The Granger causality tests show that while there is almost no evidence of causality (in both directions) before each crisis, during each crisis, approximately half of the pairwise tests were significant. The results show that most of these causalities are reverse feedback effects from the non-crisis-origin markets to the crisis-origin market. Therefore, it is concluded that the more crucial phenomenon during these crises is the “reverse feedback effects” rather than the liquidity contagion itself. Halil D. Kaya (USA), Engku Ngah S. Engkuchik (Kingdom of Saudi Arabia) BUSINESS PERSPECTIVES LLC “СPС “Business Perspectives” Hryhorii Skovoroda lane, 10, Sumy, 40022, Ukraine www.businessperspectives.org THE EFFECT OF FINANCIAL CRISES ON STOCK MARKET LIQUIDITY ACROSS GLOBAL MARKETS Received on: 2 of March, 2017 Accepted on: 29 of March, 2017 © Halil D. Kaya, Engku Ngah S. Engkuchik, 2017 Halil D. Kaya, Ph.D., Associate Professor of Finance, Northeastern State University, USA. Engku Ngah S. Engkuchik, Ph.D., Assistant Professor of Finance, Prince Sultan University, Kingdom of Saudi Arabia. This is an Open Access article, distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International license, which permits re-use, distribution, and reproduction, provided the materials aren’t used for commercial purposes and the original work is properly cited. financial crisis, contagion, liquidity, turnover ratio


INTRODUCTION
This paper attempts to address four questions: (i) Does market liquidity deteriorate in the crisis-origin markets during financial crises? (ii) Does market liquidity change in other markets or the non-crisis-origin markets during financial crises? (iii) Does the change in the liquidity level in the crisis-origin market cause a change in the liquidity level in the other markets? (iv) Are there any reverse feedback effects during financial crises (meaning that the change in the liquidity levels in the other markets during the crises causes a change in the liquidity level of the crisis-origin market)?
We look into four financial crises that occurred in the 1990s: The Thai crisis, the Hong Kong crisis, the Russian crisis, and the Brazilian crisis. For each crisis, first we look at how the crisis has affected our liquidity measure, the "turnover ratio", in the crisis-origin country as well as in the other countries. Then, for each crisis, we run causality tests to see whether the liquidity level in the crisis-origin country causes the liquidity levels in the other countries, or vice versa (i.e. the liquidity level in any of the other markets Granger-causes the liquidity level in the crisis-origin market).
A financial crisis typically brings about a shock into a country's stock market, which is then transmitted to another, partly due to the financial market liberalizations which allow investment capitals to flow in and out of a country more freely than ever before. Market liquidity is an important feature of stock prices, as verified by the strong relation between stock prices and liquidity in the market microstructure literature. Given the importance of liquidity as one of the measures used to gauge the proper functioning of a market and to determine asset returns, evidence of the increase in variability and co-movement of asset returns and other economic variables between country-markets during a crisis, and evidence of a common liquidity factor among stock markets, one would expect market liquidity to be adversely affected during a financial crisis.
In this paper, we provide some evidence of the real state of market liquidity across emerging and developed markets during a crisis. We find that while the liquidity levels decreased in approximately half of our sample markets, in the remaining half the liquidity levels actually improved. Our Granger causality tests show that while there is almost no evidence of causality (in both directions) before each crisis, during each crisis, approximately half of the pairwise tests were significant. The results show that most of these causalities are reverse feedback effects from the non-crisis-origin markets to the crisis-origin market. Therefore, we conclude that the more crucial phenomenon during these crises is the "reverse feedback effects" rather than the liquidity contagion itself.
Our paper proceeds as follows: section 1 discusses the previous literature. Section 2 explains the data and the methodology. Section 4 explains the empirical results. Final section concludes.

LITERATURE REVIEW
A contagion can be defined as a significant increase in the co-movements of prices and quantities across markets after a shock to one country or a group of countries. Contagion may also be defined as the volatility spillover from the crisis-origin country to other countries. This definition implies that a financial crisis can be identified as the most volatile period (or with the highest uncertainty), and therefore a contagion is measured as volatility spillovers from one country-market to another. In other words, contagion occurs when uncertainty spreads across international financial markets. Evidence of such spillover effect has been documented in studies which investigate financial contagion originating from e.g. the U.S. stock market crash in 1987, the Mexican peso crisis in 1994, the Asian financial crisis in 1997, the Russian financial crisis in 1998, the 9/11, and the subprime mortgage crisis (see e.g. King King and Wadhwani (1990), and Lee and Kim (1993) provide evidence of financial market contagion during the 1987 stock market crash. King and Wadhwani (1990) examine whether correlation coefficient of stock returns increases during and immediately after the crash as volatility increases between three leading markets, i.e. the New York stock exchange, the London stock exchange, and the Tokyo stock exchange. They find significant increases in the association of stock market returns between London-New York and Tokyo-New York after October 1987. Lee and Kim (1993) conclude that the interrelation of price movements across different national stock markets has become stronger after the 1987 stock market crash, i.e. the evidence of financial market contagion is mainly caused by the increased co-movements across North America, Pacific Basin, and Europe rather than by the increased co-movements within the same region.
Interestingly, Forbes and Rigobon (2002) who examine whether there is evidence of contagion among 28 markets around the 1987 stock market crash, the 1994 Mexican crisis, and the 1997 Asian crisis find that in most cases, there seem to be little evidence of contagion between financial markets based on the unconditional correlation coefficients which they propose, but stress that the continued high level of market correlation merely suggests stronger interdependence among markets around the world. Using Forbes and Rigobon's (2002) bias-correction procedure and GARCH method, Hon et al. (2004) examine whether the 9/11 incident led to contagion into other markets, i.e. the Organization for Economic Cooperation and Development (OECD) and Asian countries. However, in contrast to Forbes and Rigobon (2001), they provide evidence of conta-gion by showing that international stock markets (in terms of stock returns) move closely with U.S. stocks after the shock, but not before. In addition, they also show that volatility in the U.S. market does not seem to increase even though cross-country correlations strengthen after the shock.
Additional evidence of financial contagion across markets is also provided by several other studies including Baig and Goldfajn (1998), and Bae et al. (2001). Baig and Goldfajn (1998) investigate contagion during the Asian financial crisis among five financial markets in East Asia, i.e. Thailand, Malaysia, Indonesia, Korea, and the Philippines, and find that correlations in currency and sovereign spreads increase significantly during the crisis period, whereas the equity market return correlations offer mixed evidence. Bae et al. (2001) investigate financial contagion across Asian and Latin American countries during the Mexican crisis, the Asian crisis, and the Russian crisis. Using the multinomial logit model, they study the probability of observing a large return in a country on a particular day given that another country has a large return on that day. Their approach also draws on the extreme-value statistical theory that the behavior of the tail observation is different from the rest of the observations. They generally find evidence of contagion during all three crises, with the findings within regions to be higher than across regions.
More recently, Chudik and Fratzscher (2012) analyze the transmission of liquidity shocks and risk shocks to global financial markets. They find significant differences in the transmission strength and pattern between the U.S. sub-prime mortgage crisis in 2008, and the sovereign debt crisis in 2010. They show that emerging market economies have been much more resilient to adverse shocks in the latter crisis when compared to former. The authors also show that, during the latter event, the flight-to-safety phenomenon across asset classes has become strong (with bond yields down in advanced economies) compared to the earlier crisis.
Kaserer and Rösch (2013) examine the dynamics and the drivers of market liquidity during the recent financial crisis. They find that market liquidity suffers when stock markets decline and they show this as an evidence of a positive relation between market risk and liquidity risk. They also show that liquidity commonality increases during market downturns and peaks at major crisis events. In addition, they show that financial sector's funding liquidity tightness induces liquidity commonality which then leads to market-wide liquidity dry-ups, and they conclude that market liquidity can be a driving force for financial contagion.
In most cases, prior contagion studies seem to confirm the existence of contagion in financial markets during financial crises. In explaining these contagions, several channels of contagion have been suggested, but with no consensus as to the main channel through which contagion spreads. Dornbusch et al. (2000) distinguish between channels via fundamental links and financial agents' behavior. While fundamental-based contagion can be channeled via common shocks or common global cause, competitive devaluations, or direct financial links, investors' behavior-based contagion depends on the degree of international market integration. According to Kaminsky and Reihbart (2000), it is difficult to differentiate between the role of trade and financial links, since most countries which are linked in trade are also linked in finance (Kaminsky and Reinhart, 2000). However, they posit that financial linkage is perhaps the more likely culprit since the predictive power of predicting a crisis in another country is higher when countries are clustered based on a common lender, such as the U.S. or Japan.
Dornbusch et al. (2000), and Sell (2001) suggest several channels of contagion which include the liquidity and incentive problems, information asymmetries and coordination problems, and changing rules of assessment by investors regarding international financial transactions, herding behavior, rational investors hypothesis or cross-market rebalancing, integrated capital markets, rational speculators' behavior, regional trade partnership or competition, cooperation which lacks credibility, via balance of payments or money demand, and via political reasons. King and Wadhwani (1990), in proposing the correlated information channel as an explanation for contagion, argue that uninformed investors are unable to distinguish whether changes in prices in one market are due to liquidity shocks or shocks to the fundamentals. In this situation, when traders observe a decline in prices in one market, they take it as a signal predicting a fall in other markets as well, causing a shock in one market to be transmitted to other markets. Kodres and Pritsker (2002) argue that King and Wadhwani's correlated information channel seems to work well only if financial markets have strong links to each other, but not when explaining the pattern of co-movement between two weaklylinked markets, e.g. the crisis from Russia to Brazil. Alternatively, Kodres and Pritsker (2001) suggest contagion can be better explained through a crossmarket rebalancing channel, which implies that when there is a shock in one market, investors also rebalance their portfolios in other markets, thereby transmitting the shock and causing contagion. Their model also explains why emerging markets (where information asymmetry is more acute) seem more vulnerable to contagions and why contagions are more likely to happen during a financial crisis.
Kyle and Xiong (2001) explain the Russian crisis' contagion with a theoretical model in which increased risk-aversion is based on the wealth effects of convergence traders who follow the short-term (but rational) trading strategies. In their model, traders are assumed to trade in two risky assets. When traders suffer trading losses in either asset, they have reduced capacity for bearing risks. This motivates them to liquidate positions in both markets, causing decreased market liquidity and higher price volatility, and increased correlation between the two asset markets, which leads to contagion. Similarly, Bookstaber (2000), also in reference to the Long-Term Capital Management (LTCM) debacle, argues that liquidity crisis cycle is the reason for such global shocks, and that the need for liquidation is the root of the problem during such crisis. Liquidity crisis cycle refers to the cycle involving risk, leverage, and illiquidity -where initial losses trigger the shock in a market. In order to cover for the losses and to meet creditors' margin requirements (due to leveraged positions), these investors are forced to liquidate their positions including in other markets.

DATA AND METHODOLOGY
Our sample consists of four crisis-origin markets (Thailand, Hong Kong, Russia and Brazil), and eleven non-crisis-origin markets (Indonesia, Malaysia, the Philippines, Singapore, South Korea, Argentina, Chile, Mexico, Venezuela, Japan and the U.S.). We include Japan and the U.S. because we also want to evaluate the impact of each crisis on developed markets.
Since all of these crises initially started when the Thai government devalued the baht on July 1, 1997, we take the period between January 1, 1996 and July 1, 1997 as our "normal" period for all four crises. After July 1, 1997, first the Thai crisis took hold, and then the Hong Kong crisis started when the Hong Kong stock market crashed on October 20, 1997. Therefore, in this study, we define the Thai crisis period as the period between July 2, 1997 to October 18, 1997, and the Hong Kong crisis period as the period between October 19, 1997 and January 16, 1998. Although Russia had been under economic pressures since the beginning of 1998, the crisis really exploded after the announcements on August 17, 1998 involving the abandonment of ruble pegging to dollar. Therefore, for the Russian crisis, the crisis period is defined as the period from August 17, 1998 to November 14, 1998.
The Brazilian crisis started in January 1999 with the devaluation of the real and the resignation of the country's finance minister on January 13, 1999. Therefore, for the Brazilian crisis, we define the crisis period as the period between January 13, 1999 and April 12, 1999.
In this study we use the (natural log of) daily turnover ratio as our equity market liquidity measure. This measure of liquidity has been used in several previous studies (i.e. Datar  The turnover ratio measure is calculated as the ratio of the daily number of shares traded in a country's market to that country's stock market capitalization at the end of the day. We obtain the data from Thomson Financial's Datastream which provides a large number of market information around the world. Forbes and Rigobon (2002), and Hon et al. (2004) focus on the two-day rolling average returns in their studies to control for nonsynchronous trading periods in different markets around the world. Similarly, we compute the twoday rolling average for our turnover ratio to account for the same concern.
The unit root tests indicate that the market liquidity measures employed in our study are integrated of order one. Therefore, we employ the percentage change of turnover ratio for the rest of this paper. Table 1 shows the summary statistics of the percentage change in the turnover ratio in the fifteen stock markets (including Japan and the U.S. markets) during the Thai crisis, the Hong Kong crisis, the Russian crisis, and the Brazilian crisis. Panels A, B, C, and D show the summary statistics for the Thai, the Hong Kong, the Russian, and the Brazilian crises, respectively.

EMPIRICAL RESULTS
We can see that, in each panel, approximately half of the fifteen countries have positive means (and medians) for the percentage change in the turnover ratio, while the other half have negative means (and medians). Also, the magnitudes of these mean and median values are relatively small. For example, in Panel A (i.e. the summary stats for the Thai crisis), none of the mean values for these fifteen countries are above 2% in absolute value. The largest positive percentage change is 1.59% (which is for the Indonesian market), while the largest negative percentage change is -1.39% (which is for the Venezuelan market). When we look at the other panels (i.e. other crises), we see that most of the mean values are close to zero as well.
Considering the fact that we have very small percentage changes in the turnover ratio (i.e. liquidity) in these fifteen countries' stock markets and that almost half of the means (and the medians) are positive/negative, we contend that these four financial crises have not had a large impact on these markets' liquidity levels.
After examining the relation between the four financial crises and the liquidity levels in these fifteen markets, we turn our attention to the interaction between the liquidity levels in the crisis-origin markets and the liquidity levels in the remaining markets.

The Thai crisis
In Table 2, Th indicates that a country Grangercauses the Thai market, while Th indicates that the Thai market Granger-cause the other market.
When we compare the normal period results to the crisis period results, we see that more countries Granger-cause the Thai market's liquidity levels during the crisis period when compared to the normal period. Also, the statistical significance levels are much higher during the crisis period compared to the normal period. While during the normal period only five markets (Malaysia, Mexico, U.S., Argentina, and Japan) Granger-cause the Thai market, during the crisis period seven countries (Venezuela, Argentina, Mexico, Philippines, Brazil, Japan, and U.S.) Granger-cause the Thai market.
As we can see from the  Notes: This table provides the Granger-causality test results of the error correction model between the crisis-origin country, Thailand, and the other stock markets in the sample during the crisis period and the normal period. 1/96-7/97 is the normal period window, while 1/96-12/99 is the whole sample period. Th indicates that a country Granger-causes the Thai market, while Th indicates that the Thai market Granger-causes the other market. The results presented below correspond to the lag-5 length. Percentage change of turnover ratio is the first difference of turnover ratio. Turnover ratio is the natural log of the ratio between the daily trading volume and the corresponding market value. ***, **, * indicate significance levels of 1 percent, 5 percent, and 10 percent, respectively.

The Hong Kong crisis
In

The Russian crisis
In   Table 4 also shows that while the Russian market Granger-cause only one market (Philippines at 10% level) during the normal period, it Granger-causes five markets (Indonesia, Malaysia, and Venezuela at 5% level, and Singapore and Argentina at 10% level) during the crisis period.

The Brazilian crisis
In Table 5, Br indicates that a country Grangercauses the Brazilian market, while Br indicates that the Brazilian market Granger-causes the other market. When we compare the normal period results to the crisis period results, we see that more countries Granger-cause the Brazilian market's liquidity levels during the crisis period when compared to the normal period. While during the normal period, only two markets (Thailand at 5% and Argentina at 10% levels) Granger-cause the Brazilian market, during the crisis period, three markets (Thailand at 10%, and Chile and U.S. at 5% levels) Granger-cause the Brazilian market.

CONCLUSION
In this study, using a widely available market liquidity measure, the "turnover ratio", we test for market liquidity contagion during the four financial crises that occurred between 1997 and 1999: The Thai crisis, the Hong Kong crisis, the Russian crisis, and the Brazilian crisis. First, we classify the stock markets of these four countries as "crisis-origin" markets and the stock markets of several other countries (i.e. Indonesia, Korea, Malaysia, Philippines, Singapore, Argentina, Chile, Mexico, Venezuela, Japan and U.S.) as non-crisis-origin markets. Then, we examine the percentage change in the turnover ratio in both the crisis-origin markets and the non-crisis-origin markets during each financial crisis.