Measuring financial contagion between emerging equity markets before and after the onset of the...

download Measuring financial contagion between emerging equity markets before and after the onset of the global financial crisis

of 75

Transcript of Measuring financial contagion between emerging equity markets before and after the onset of the...

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    1/75

    1

    Measur ing F inancial Contagion between emerging equity markets before and after the

    onset of the global fi nancial cri sis

    MSc Financial Services 2013

    James Fitzsimons

    12012173

    Word Count: 11,491

    Supervisor: Fergal OBrien

    This dissertation is solely the work of the author and submitted in partial fulfilment of the

    requirements of the MSc in Financial Services.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    2/75

    2

    Abstract

    This thesis attempts to measure both the extent and determinants of financial

    contagion across 17 emerging equity markets from Latin America and Asia both before and

    after the onset of the global financial crisis. The methodology employed is a continuation of

    the approach utilised by Bae et al. (2003) which captured the coincidence of extreme return

    shocks across national stock indices both within and across emerging market regions. The

    data used is that of daily returns over an 11 year period (January 2002- December 2012)

    which is divided approximately into separate time series data to analyse returns both before

    and after the onset of the financial crisis as well as for the total 11 year period. The extent of

    contagion is illustrated and its determinants are characterized using a binary logistic (logit)regression model.

    This work illustrates that the correlations between all the equity markets analysed

    increased after the onset of the global financial crisis, as did the frequency of extreme return

    shocks, and that negative return shocks are more widespread than positive shocks. It is also

    found that contagion is far more prevalent in Asia than in Latin America and that certain

    countries within both regions display return shocks that are unique within their respective

    regions. Furthermore it is evident from the logit regression analysis that broad market

    indicators show a weak relationship with extreme returns in emerging equity markets

    suggesting that contagion is not easily predicted and is a separate phenomenon from factors

    indicative of economic performance.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    3/75

    3

    Acknowledgements

    I would like to thank James Ryan, Fergal OBrien, Adam OReilly, and my entire family. All

    of whom have provided substantial guidance, support and encouragement which were

    essential in making this thesis possible.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    4/75

    4

    Table of Contents

    1. Introduction ......................................................................................................................................... 5

    2. Literature review ................................................................................................................................. 8

    3. Data ................................................................................................................................................... 15

    3.1 Asian markets .......................................................................................................................... 16

    3.2 Latin American markets .......................................................................................................... 16

    3.3 Broad market indicators (3-month Treasury Bills, US Dollar, and the VIX) ......................... 16

    4. Methodology ..................................................................................................................................... 17

    4.1 Returns .................................................................................................................................... 17

    4.2 Summary Statistics .................................................................................................................. 17

    4.3 Exceedances (extreme returns) ............................................................................................... 18

    4.4 Binary logistic regression........................................................................................................ 18

    4.5 Summary of the methodology ................................................................................................. 20

    5. Results ............................................................................................................................................... 21

    5.1 Summary Statistics .................................................................................................................. 21

    5.2 Total period summary statistics .............................................................................................. 21

    5.3 The pre-crisis period summary statistics ................................................................................. 24

    5.4 The post-crisis period summary statistics ............................................................................... 27

    5.5 Negative exceedances over the total 11 year period ............................................................... 30

    5.6 Positive exceedances over the total 11 year period ................................................................. 32

    5.7 Negative exceedances during the pre-crisis period ................................................................. 34

    5.8 Positive exceedances during the pre-crisis period .................................................................. 35

    5.9 Negative exceedances during the post-crisis period ............................................................... 37

    5.10 Positive exceedances during the post-crisis period ............................................................... 38

    5.11 Regression results ................................................................................................................. 40

    6. Discussion ......................................................................................................................................... 53

    7. Conclusion ........................................................................................................................................ 55

    Bibliography ......................................................................................................................................... 56

    Appendix 1: Binary logistic regression output ..................................................................................... 58

    Appendix 2: Asian equity markets ........................................................................................................ 70

    Appendix 3: Latin American equity markets ........................................................................................ 72

    Appendix 4: US and European equity markets ..................................................................................... 74

    Appendix 5: Broad market indicators ................................................................................................... 75

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    5/75

    5

    1. Introduction

    Financial contagion is usually viewed as the spreading of a financial crisis from one country

    to another. The topic has been one of the most widely debated in international finance since

    the Asian financial crisis of 1997. Despite the lack of agreement over the causes and the most

    appropriate metric of contagion, there is widespread agreement on which particular extreme

    market events are considered to be instances of contagion. The Mexican Tequila effect of

    1994, the Asian crisis of 1997, the Russian default of 1998, the Brazilian sneeze of 1999 and

    the NASDAQ rash of 2000 are all agreed upon as events where financial contagion between

    countries has occurred (Rigobon, 2002). The global financial crisis 2007 to 2009 is an

    example of financial contagion that is perhaps more vivid in the memories of most.

    In the latter half of the 20 thcentury, increased capital mobility enabled funds to flow far more

    rapidly between markets than had previously been possible. Far reaching financial

    deregulation suddenly made available pools of funding from foreign sources. Emerging

    markets became the destination of choice for much of this funding as international investors

    sought to diversify into foreign markets. This undoubtedly provided benefits in that firms

    were no longer restricted by domestic credit limitations and could now avail of foreign credit.

    However the rapid increase in financial flows between countries created a heightened risk of

    instability for multiple economies as the financial system grew evermore integrated.

    Furthermore, the increased interdependence that developed between countries throughout the

    globe led to greater exposure to financial contagion (Candelon, 2005).

    The Asian financial crisis of 1997 was a typical example of financial contagion at its worst. Acurrency crisis that started in Thailand suddenly spread throughout Asia sending shock waves

    throughout the world. Fears of a global recession ensued and the crisis was eventually quelled

    due largely to the intervention of the IMF. The event highlighted both the extent and rapidity

    of financial contagion in developing economies and provided the motivation for widespread

    research on financial contagion to better understand the phenomenon. It also led many

    experts to infer that developing countries like those ensnared by the Asian crisis were more

    susceptible to financial instability than the economies of the developed world (Bae et al.,

    2003).

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    6/75

    6

    Despite the widespread attempts to characterise and measure contagion, research on the topic

    has yielded mixed results. One prevalent problem in particular is that in the existing research

    much of the focus has leaned heavily on the analysis of the correlation coefficients between

    markets. The problem with analysing correlations is that the true relationships between

    markets are not always reflected accurately. This is because the correlations are influenced

    heavily by the vast majority of the days in a data set in which there are no extreme events at

    all. Isolating the extreme events and measuring the true contagiousness between markets was

    therefore not well reflected by correlations alone which simply measure the strength of

    relationships generally. Correlations that give equal weight to small and large returns are not

    appropriate for an evaluation of the differential impact of large returns. The true impact of

    large returns is hidden in correlation measures by the large number of days when little of

    importance happens (Bae, 2003, 719).

    This problem was tackled by Bae et al. (2003) when they introduced a new approach to

    measuring contagion which focused on the occurrence and coincidence of extreme returns

    rather than on correlations. This methodology required isolating and measuring theoccurrence of extreme positive and negative returns. This created a methodology by which

    they avoided having a situation where results are dominated by a few observations, we do

    not compute correlations of large returns, but instead measure the joint occurrences of large

    returns. (Bae, 2003, 719). In other words, by isolating extreme events on the marginal

    distribution of returns it is then possible to measure the joint occurrence of extreme market

    events across different markets without having to use any analysis of correlation coefficients

    which are wrought with statistical obstacles. A time series of returns data from various

    national stock markets can be used to analyse such joint occurrences between countries.

    Bae et al.s research was based on eight years of daily stock indices returns during the 1990s

    (1992-2000). This thesis uses a more recent data set from the same emerging markets which

    incorporates approximately a five and half year period both before and after the onset of the

    global financial crisis (January 2002-December 2012) with the date of June 15th2007 used as

    the cut off point for the pre-crisis and post-crisis analysis. The data and results are therefore

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    7/75

    7

    broken up into three segments. These segments are the total 11 year period, the pre-crisis

    period, and the post-crisis period. The results are summarised into tables representing their

    respective periods. The extreme positive and negative returns are measured separately and

    illustrated in tables for the differing time periods.

    A binary logistic regression model is used to further characterise the determinants of the

    extreme returns. It is similar to that employed by Boyson et al. (2006) in which the

    relationship between hedge funds and variables representing broad economic performance is

    assessed. In this thesis, the dependent variable used in the model is representative of extreme

    returns across a particular region while the explanatory variables are representative of the

    broad market indicators of interest rates, exchange rates, and market volatility. The model is

    useful for predicting extreme events based on contemporaneous movements in these broad

    market indicators. The model tests for the determinants of both positive and negative

    contagion in Asia and Latin America and is furthermore split into the three different time

    periods.

    This thesis is motivated largely by the fact that Bae et al.s approach to measuring financial

    contagion has not yet been applied to a more recent and turbulent time period. Whereas Bae

    et al.s work examined contagion in the 1990s; this work examines the last eleven years

    (January 2002- December 2012) of daily returns on stock indices from emerging markets and

    furthermore dissects the data in light of the global financial crisis. The binary logistic

    regression model further characterises contagion within Asia and Latin America. The

    following literature review will illustrate part of the existing research that has been carried

    out on the subject of measuring financial contagion and the different methodologies and

    conclusions which have been drawn.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    8/75

    8

    2. Literature review

    The predominant theme emanating from the existing literature on the subject of financial

    contagion indicates that the origins of the phenomenon are not well understood. More

    specifically the channels by which market shocks spread from one country to another and

    how such flows can be measured remain unclear. Both the methodologies used and the

    approaches taken have often focused on the analysis of correlation coefficients of stock

    returns which has yielded mixed results to date. Other alternative approaches which move

    away from a reliance on correlations have also led to greater ambiguity surrounding the

    subject. Below is an exploration of the literature on financial contagion which has attempted

    to define, measure, and explain a subject area which has proved to be a significant challenge

    for the academic community.

    Rigobon (2002) illustrated the difficulties that are widespread on the subject and discussed

    the most commonly used methodologies on the phenomenon which are linear regressions,

    logistic regressions and tests on returns correlations. He highlighted the three most common

    problems that have prevailed in the data which have been used in such research (data such as

    stock market returns, interest rates, exchange rates, or linear combinations of these). The

    three problems are simultaneous equations, omitted variable biases (as there is a lack of

    consistent and compatible data) and heteroskedasticity (caused by volatility increases during

    crisis periods thus making analysis correlation coefficients difficult). Much of the research

    outlined below has attempted to address the issue of measuring contagion and furthermore

    overcoming the hurdles highlighted by Rigobon.

    One such attempt was made by Forbes et al. (2001). They attempted to define and measure

    financial contagion. Their work highlighted the fact that although its existence is widely

    accepted, the definition and interpretation of contagion is unclear. They stated that any

    continuation of cross-linkages that are present during stable market periods cannot be

    classified as contagion. For contagion to occur, stock market shocks that occur in different

    markets at the same time need to be the result of some other unique link between them that is

    absent during normal market periods. Thusly contagion was defined as a significant increase

    in such cross-market linkages when shocks occur simultaneously across countries. They

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    9/75

    9

    classified increases in correlation as indicative of interdependence and increases in co-

    movements as increases in contagion. By establishing a workable definition for contagion,

    this work evaluated prior research carried out on the subject and concluded that there was no

    evidence of contagion and instead found evidence of interdependencies between markets.

    Similar findings were made by Candelon et al. (2005) who attempted to measure financial

    contagion by analysing the Hong Kong stock market crisis (1st Jan 1996 to 31st December

    1998) and the Mexican stock market (1994, Peso Crisis) during these turbulent market

    periods. Their methodology focused on an analysis of co-movement according to recurrent

    common economic cycles. Candelon et al. inferred that large cross-market shocks are not

    unique occurrences in their own right but rather a continuation of linkages that are already in

    existence during more stable market periods. This finding suggested that return shocks are

    spread through non-crisis related channels such as those associated with trade, policy co-

    ordination and random shocks. This conclusion tends to favour the interdependencies inferred

    above by Forbes et al. (2001) however the methodology used was criticised by Corsetti et al.

    (2005) who stated that such an approach creates a bias towards the null hypothesis of

    interdependence.

    Corsetti et al. (2005) examined the international effects of the 1997 Hong Kong stock market

    crisis on a sample of 17 countries. Their approach investigated previous research including

    that of Forbes et al. (2001) which suggested evidence of interdependence rather than

    contagion. Corsetti et al. outline a critique of the previous work by suggesting that much

    research had taken the variance of stock returns in the market where a crisis has originated as

    a proxy for the volatility affecting all other markets. This failure to distinguish between

    common and country-specific elements of returns data creates a bias towards there being no

    contagion. Corsetti et al. therefore used a model of interdependence that did not create an

    imposition on the variance of common factors relative to the variance of country-specific

    shocks.

    Corsetti et al.s (2005) findings suggested that the conclusion from much of the previousresearch on contagion, that interdependence between countries is evident rather than

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    10/75

    10

    contagion, is the result of arbitrary assumptions regarding the country-specific variance in the

    market where the crisis originates. These assumptions thusly bias tests towards the null

    hypothesis of interdependence. Corsetti et al. found evidence that for at least 5 of the 17

    countries analysed there was little evidence of interdependence. However their findings also

    suggested that country-specific noise cannot be disregarded when testing for transmission

    mechanisms of shocks. This work therefore created further questions over the measuring of

    contagion and the need for alternative methods to be developed. Karolyi (2003), for example,

    outlined an alternative approach which attempted to overcome some of the difficulties

    involved.

    Karolyi (2003) researched the various definitions of financial contagion across both the

    academic literature and the interpretations by the mass media on the subject. Karolyi then

    compared these definitions to the empirical evidence on international capital flows and asset

    prices. He found that the existence of financial contagion between markets was not as

    extensive as many researchers and commentators had inferred. Karolyi found that there have

    been limitations to much of the research on contagion as it had focused largely on the

    correlations between markets. The problem with simply analysing correlation coefficients is

    that they provide an equal weighting for small and large changes in returns. This excludes an

    evaluation on the uniqueness of large returns. Karolyi highlighted that there had been little

    research that attempted to solve this issue of correlation analysis and furthermore stated that

    even the research that had employed alternative statistical methods to measure financial

    contagion had not controlled for economic fundamentals. Karolyi claims that Bae et al.

    (2003) created a new measure of contagion that had addressed these problems.

    Bae et al. (2003) did this in three ways. Firstly they focused specifically on measuring

    extreme events which are considered to be returns that lie in the top or bottom 5% of the

    marginal distribution of returns. By isolating these extreme events, they could measure the

    occurrence of shared extreme events between markets referred to as co-exceedances

    (Karolyi, 2003, 194). Secondly they did not focus on correlations but on the conditional

    probabilities of these co-exceedances. Thirdly they used a multinomial logistic regression

    model to measure the probability of shared extreme returns or co-exceedances occurring. Thebenefits of this model were that it allowed for explanatory variables that characterise the

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    11/75

    11

    likelihood of extreme return shocks such as changes in interest rates, exchange rates and

    market volatility. They also used contagion in different regions as explanatory variables.

    They found that there was little evidence of contagion across regions as the model showed

    low significance figures where co-exceedances in one region were used to explain co-

    exceedances in another.

    Other approaches had been continuously developed such as the case with Iwatsubo et al.

    (2007) who focused their research on measuring contagion between the US and Asian

    markets. This was done by analysing the returns of 22 dually-traded stocks of Asian firms

    (Asian firms with stocks traded both on the NYSE and in their home countries). This

    methodology was useful in that it could distinguish between contagion and fundamentals-

    based(2007, p217) stock price co-movement for markets which traded non-synchronously

    in separate time zones. This approach could thusly control for the fundamental factors

    inherent in the stock prices and identify the role of other factors such as an individual

    countrys stock index.

    They found that there were significant bilateral contagion effects in both returns and returns

    volatility. Secondly, that contagion effects from the US to Asia were stronger than in the

    opposite direction. This suggested that the US may play a major role in the transmission of

    contagion in financial markets. Thirdly, they found that the intensity of the contagion was

    greater during the Asian crisis of 1998 than afterwards. These findings were somewhat

    contradictory to that of Diebold et al. (2009) which suggested that there was a significant

    disparity between the transmissions of returns compared to transmissions of volatility

    between equity markets.

    Diebold et al. (2009) provided a simple measure of linkages between equity markets by

    measuring interdependence of weekly asset returns and/or volatilities from 7 developed and

    12 emerging equity markets by constructing a Spill over Index. Similar to Bae et al. (2003),

    extreme events (in this case extreme returns and extreme volatility) were isolated and

    measured in order to illustrate the characteristics of such events. They measured returnspillovers as separate to volatility spillovers and included time periods which accommodated

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    12/75

    12

    periods of both market stability and market turmoil (the time-series of returns data ranged

    from January 1992 to November 2007). They furthermore identified trends and sudden bursts

    in the occurrence of these spillovers.

    They found that there existed an extreme divergence between the characteristics of returns

    and volatility spillovers. Returns spillovers displayed a gradually increasing trend without

    any bursts in the occurrence of such spillovers, which may perhaps be explained by the

    increased financial integration which took place over the last 15 years. Contrastingly,

    volatility spillovers displayed no trend whatsoever but rather clear bursts that can be readily

    associated with turbulent market periods. This finding raised the question of why there

    existed such a stark contrast between returns and volatility spillovers. Diebold et al.s work

    had therefore raised further questions surrounding interdependence, more specifically

    contagion, and how this can be identified, measured and understood.

    Further attempts to measure contagion include the work of Chiang et al. (2007) who used a

    conditional correlation model on 9 Asian stock market returns data from 1996 to 2003. This

    model was a multivariate GARCH model which was appropriate for measuring the time-

    varying conditional correlations between countries. This model enabled Chiang et al. to

    address the heteroskedasticity problem highlighted above by Rigobon without dividing the

    time-series data into two sample periods. Chiang et al. furthermore employed the same model

    on lagged US stock returns as an exogenous factor in order to further address the omitted

    variable problem also outlined by Rigobon (2002).

    Chiang et al. (2007) identified two phases in the Asian financial crisis; the first took place in

    the early weeks of the Asian crisis and showed an increase in correlation, which they

    classified as contagion, as increasing volatility spread from the earliest crisis-effected

    countries to other countries. The investor activity here was governed mainly by local (within

    the country) information. The second phase began at the end of 1997 through to 1998 as

    awareness of the crisis became more widespread internationally. This second phase showed a

    continued heightened correlation between stock returns and their volatility (which theyclassified as herding). The statistical analysis applied to correlation coefficients indicated that

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    13/75

    13

    there was a significant shift in variance throughout the crisis period which suggested that the

    benefits to international diversification may be somewhat limited. They also found evidence

    that changes sovereign credit-ratings had a significant effect on the dynamic correlations

    within the markets analysed. This work found evidence that Investor behaviour and credit-

    rating agencies therefore play a significant role in the transmission of contagion between

    countries.

    Other research on contagion that departed from the analysis of correlation coefficients was

    undertaken by Pritsker (2001). This work took a theoretical approach to studying the channels

    through which contagion may have spread and factors which may have made a country

    susceptible to contagion. These channels were sector, financial market, and financial

    institution linkages. The interaction between markets and financial institutions was seen as a

    possible originator for a tightening of liquidity and a flight by investors to safer assets.

    However Pritsker admited that the channels through which shocks spread from one country to

    another were not well understood and that this failure to identify such channels through

    which shock propagations flow illustrated the need for further research. Pritsker admited that

    his work had barely scratched the surface in termsof modelling propagation (2001, p20)

    and a need to develop theoretical models which can be tested was essential to developing a

    more in-depth understanding of contagion.

    Another alternative approach was undertaken by Goldstein et al. (2004) who examined the

    role that investors played in the contagion of market shocks between countries rather than

    attempting to measure contagion itself. This was done by developing a theoretical model of 2

    countries with independent fundamentals but with the same group of shared investors, the

    rationale being that each country may be susceptible to a self-fulfilling crisis as investors may

    withdraw their capital from a country fearing that other investors will do the same. In other

    words, a crash in one country will make an investor risk-averse and may thusly lead to that

    investor withdrawing their investments from the second country. When this occurs, there is a

    positive correlation between the returns of the two countries and thusly an increase in

    contagion. The mechanism that generated contagion in their model was based on a wealth

    effect. This model found that decreases in the wealth of investors in one country increased thelikelihood of a negative shock occurring in the second country. Although Goldstein et al.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    14/75

    14

    provided a possible explanation for contagion they provided little in terms of a framework by

    which contagion could be measured.

    It is clear from the literature outlined above that there remains a great difficulty in

    understanding the causes of contagion and furthermore measuring its occurrence. Both the

    focus and methodologies within the previous research on contagion have been varied with the

    analysis of correlation coefficients being the most popular area of focus. Despite the

    extensive work, it appears that a definitive framework is still an allusive target. Further

    research is this area is therefore likely to remain a contentious area of research for the

    foreseeable future. This thesis attempts to contribute to the existing literature by measuring

    contagion using the unique approach of Bae et al. (2003) whose alternative methodology

    moves away from an over-reliance on correlation analysis. The time-series of data used in the

    analysis will incorporate a more recent and lengthier time-period taking account of the global

    financial crisis.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    15/75

    15

    3. Data

    The purpose of this thesis is to try measure the occurrence and explain the determinants of

    financial contagion between emerging equity markets. A similar data set to that which was

    analysed by Bae et al. (2003) of daily stock market returns is used except that in this thesis,

    the data is more recent and extends over a greater time period. This time-series of data is

    furthermore split into two parts to enable an analysis of the pre and post financial crisis

    period as well as the total time period. This section will outline the data used and the rationale

    for its inclusion in this work.

    To try and accurately capture the characteristics of developing economies stock indices from

    17 different countries from across Latin America and Asia were chosen. These stock indices

    are representative of the largest publicly traded firms by market capitalisation within each

    economys equity market. They are also reflective of stocks that are accessible to foreign

    investors which is a key element required for measuring the spreading of cross-border

    extreme returns shocks as; a number of explanations of contagion are based on the actions

    by foreign investors (Bae, 2003, 721). Daily returns were used as these are the most

    sensitive to any sudden shocks that may occur.

    Roughly eleven years of daily returns from 2nd January 2002 to 28thDecember 2012 were

    used (2868 observations). In addition to the stock indices of the emerging markets chosen, the

    US (S&P 500) and Europe (Stoxx Europe 600) were added to provide a control for the extent

    to which contagion can also impact on developed markets. The daily closing prices of the

    selected stock indices for the chosen time period were downloaded from the Bloomberginterface and the returns were then calculated. The data set of returns was arbitrarily divided

    into two segments of approximately five and a half years each which were used for the pre

    and post financial crisis analysis (pre and post 15thJune 2007) as well as for the total 11 year

    period.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    16/75

    16

    3.1 Asian markets

    To capture the developing equity markets of Asia the following countries were included;

    China (CSI 300), Korea (KOSPI), Philippines (PCOMP), Taiwan (TWSE), India (S&P

    SENSEX), Indonesia (JCI), Malaysia (KLCI), Pakistan (KSE 100), Sri Lanka (CSEALL),

    and Thailand (SET). See Appendix 2 for detailed explanation of these markets.

    3.2 Latin American markets

    Similarly, to capture the developing equity markets of Latin America the following countries

    were included; Argentina (MERVAL), Brazil (IBOV), Chile (IPSA), Colombia (IGBC),

    Mexico (MEXBOL), Peru (IGBVL), and Venezuela (IBVC). See Appendix 3 for detailed

    explanation of these markets.

    3.3 Broad market indicators (3-month Treasury Bills, US Dollar, and the VIX)

    For the binary logistic regression model it was necessary to use explanatory variables that

    were representative of broad market performance. These variables were therefore

    representative of interest rates, exchange rates, and market volatility. The variable used for

    interest rates were 3-month Treasury Bills. For exchange rates a US Dollar index that values

    the US Dollar against a basket of major world currencies was used. Finally for market

    volatility, the VIX was used as a measure of implied future volatility. See Appendix 5 for

    detailed explanation of these variables.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    17/75

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    18/75

    18

    4.3 Exceedances (extreme returns)

    After the computation of the summary statistics, the occurrences of extreme positive and

    negative returns as well as the joint occurrences of such returns were measured. The 95 thand

    5thpercentiles of the marginal distribution of returns are chosen as the boundaries that define

    extreme positive and negative returns respectively. Positive returns were treated separately

    from negative returns and were therefore measured separately. The exceedances of

    individual countries as well as the coincidences of exceedances shared between countries

    were calculated over the three distinct time periods (pre-crisis, post-crisis, and total 11 year

    period). Where an exceedance occurred, a dummy variable of 1 wasattributed and for all

    other observations a 0 wasgiven where no extreme event had occurred.

    4.4 Binary logistic regression

    In attempting to characterise the determinants of contagion a binary logistic (logit) model was

    used. This model was similar to that used by Boyson et al. (2006) and examined the

    relationship between extreme returns within each region and on several broad market

    variables. These variables were reflective of market performance globally and therefore

    provide a useful measure by which contagion could be assessed. The total number of

    exceedances experienced within a region was used as the dependent variable while the broad

    market indicators of interest rates (3 month Treasury Bills), exchange rates (US Dollar

    exchange rate against a basket of major world currencies), and market volatility (VIX) were

    used as the independent or explanatory variables.

    The logit regression model was implemented over the three distinct time periods of the pre-

    crisis period, the post-crisis period, and the total 11 year period. The purpose of this was to

    attempt to characterise the determinants of contagion within the two emerging market regions

    analysed and to furthermore illustrate any changes in such determinants that may be

    evidenced in light of the global financial crisis. The positive and negative exceedances were

    analysed separately.

    In the logit model, the unobserved continuous variable, Z, represented the propensity towards

    an extreme return shock occurring and therefore larger values of Z indicated a higher

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    19/75

    19

    probability of this shock occurring. The following function describes the relationship

    between Z and the probability of an extreme return shock;

    Or

    Where;

    Is the probability the ithcase experiences an extreme return shock.

    is the value of the unobserved continuous variable for the ithcase.

    The logit model assumed that the unobserved continuous variable Z had a linear relationship

    with the explanatory variables of interest rates, exchange rates, and market volatility. This is

    illustrated as follows;

    = Where

    is the jth predictor for the ith case

    is the jthcoefficient

    P is the number of predictors

    Due to the fact that Z is an unobserved variable, a simple linear regression did not suffice as

    such a regression model would produce an output that was difficult to interpret as the

    dependent variable of exceedances within a region are categorical (exceedance or no

    exceedance i.e. 1 or 0). It was therefore necessary to relate the explanatory variables to the

    probability of an exceedance by substituting for Z;

    =

    The coefficients in the logit model were estimated through an iterative maximum likelihood.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    20/75

    20

    The Hosmer and Lemeshow test were used to test for the models predictive abilities. In this

    test, weaknesses in the predictive abilities of the model are sought. It was therefore preferable

    for this test to be incorrect. Due to this, it was preferred that the p-values were greater than

    .05 as p-values lower than this indicated that the model was a poor predictor.

    4.5 Summary of the methodology

    The summary statistics provide a preliminary outline of the relationship between all of the

    equity markets analysed. The analysis was then taken further in the calculation of bothpositive and negative extreme returns (exceedances). The occurrences as well as the joint

    occurrences of extreme returns between markets were measured providing an alternative

    illustration of contagion across Asia and Latin America. The total amount of exceedances to

    occur within Asia and Latin America were used as the dependent variables when the binary

    logistic model was run separately for both regions. This was done for positive and negative

    exceedances separately providing a characterisation of exceedances against the broad market

    indicators outlined above. All of the aforementioned methods were employed over the 3

    different time periods to illustrate the impact that the global financial crisis has had on

    contagion.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    21/75

    21

    5. Results

    5.1 Summary Statistics

    The summary statistics illustrated that there was an enormous increase in the correlations

    both within regions and across regions after the onset of the global financial crisis. In most

    cases the correlations more than doubled.

    5.2 Total period summary statistics

    The summary statistics for the total period (Table 5.1) show the highest average return out of

    all the markets observed was that of Venezuela with 0.16%. Argentina had the largeststandard deviation of 1.98%. Pakistan had the highest median with 0.06%. The maximum

    daily return experienced was by India with 17.34% while the lowest return was that of

    Venezuela with -18.66%.

    The correlations (Table 5.2) between countries were higher within regions than across

    regions. South Korea and Taiwan had the highest correlation in Asia of .64, while the lowest

    correlation in Asia was that of China and Sri Lanka with almost no relationship observed

    (.004). Brazil and Mexico had the highest correlation in Latin America with .68, while the

    weakest relationship in the same region was shared between Mexico and Venezuela (.02).

    The strongest relationship across the two regions was between India and Peru with a

    correlation of .29, while the weakest across the two regions was between Venezuela and

    Taiwan with a .034 correlation.

    Focusing solely on the correlations between developed markets and individual developing

    countries, the US had its weakest relationship with Sri Lanka (-0.17) and its strongest

    relationship with Brazil (.65). Europes strongest relationship was with Mexico (.565), while

    the lowest correlation it shared with an emerging market was .047 with Sri Lanka.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    22/75

    22

    Table 5.1: Summary statistics of daily returns for the total period;

    Mean

    Standard

    Deviation Median Minimum Maximum

    ARG 0.10% 1.98% 0.02% -12.15% 13.42%

    BRA 0.07% 1.82% 0.01% -11.39% 14.66%

    CHI 0.05% 1.04% 0.04% -6.92% 12.53%

    COL 0.10% 1.34% 0.06% -10.46% 15.82%

    MEX 0.08% 1.32% 0.08% -7.01% 11.01%

    PER 0.11% 1.53% 0.04% -12.45% 13.67%

    VEN 0.16% 1.39% 0.00% -18.66% 10.42%

    CHN 0.03% 1.73% 0.00% -13.17% 9.39%

    KOR 0.05% 1.51% 0.05% -10.57% 11.95%

    PHI 0.06% 1.28% 0.00% -12.27% 9.82%

    TAI 0.02% 1.36% 0.00% -6.68% 6.74%

    IND 0.07% 1.56% 0.05% -11.14% 17.34%

    INO 0.09% 1.43% 0.07% -10.38% 7.92%

    MAL 0.03% 0.77% 0.02% -9.50% 4.35%

    PAK 0.10% 1.40% 0.06% -7.45% 8.88%

    SRI 0.08% 1.21% 0.00% -12.97% 12.31%

    THA 0.06% 1.35% 0.00% -14.84% 11.16%

    US 0.02% 1.32% 0.03% -9.03% 11.58%

    EUR 0.01% 1.31% 0.04% -7.62% 9.87%

    LAT 0.09% 0.98% 0.14% -6.67% 7.53%

    ASA 0.06% 0.77% 0.11% -4.64% 4.44%

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    23/75

    23

    ARG BRA CHI COL MEX PER VEN CHN KOR PHI TAI IND INO MAL PAK SRI THA US EUR LAT ASA

    Pearson

    Correlati

    on

    1

    Pearson

    Correlati

    on

    .500** 1

    Pearson

    Correlati

    on

    .414**

    .533** 1

    Pearson

    Correlati

    on

    .334**

    .329**

    .338** 1

    Pearson

    Correlati

    on

    .475**

    .679**

    .550**

    .355** 1

    Pearson

    Correlati

    on

    .391**

    .434**

    .430**

    .341**

    .429** 1

    Pearson

    Correlati

    on

    .040*

    .045*

    .041*

    .083** .019 .050

    ** 1

    Pearson

    Correlati

    on

    .101**

    .152**

    .129**

    .112**

    .120**

    .146** .016 1

    Pearson

    Correlati

    on

    .189**

    .234**

    .258**

    .215**

    .262**

    .245** .016 .245

    ** 1

    Pearson

    Correlati

    on

    .127**

    .117**

    .157**

    .174**

    .105**

    .204** .033 .152

    **.350

    ** 1

    Pearson

    Correlati

    on

    .196**

    .175**

    .198**

    .178**

    .173**

    .203** .034 .219

    **.638

    **.365

    ** 1

    Pearson

    Correlati

    on

    .229**

    .271**

    .263**

    .258**

    .279**

    .293** .002 .199

    **.398

    **.244

    **.336

    ** 1

    Pearson

    Correlati

    on

    .216**

    .209**

    .262**

    .221**

    .209**

    .272** .025 .219

    **.463

    **.379

    **.456

    **.421

    ** 1

    Pearson

    Correlati

    on

    .174**

    .155**

    .231**

    .168**

    .175**

    .255** .025 .241

    **.449

    **.394

    **.435

    **.332

    **.480

    ** 1

    Pearson

    Correlati

    on

    .006 .037* .025 .083

    ** .031 .053** .019 .059

    **.092

    **.093

    **.108

    **.103

    **.094

    **.113

    ** 1

    Pearson

    Correlati

    on

    .053** .030 .080

    **.037

    * .027 .084** -.0 10 . 00 4 .058

    **.063

    **.052

    **.053

    **.054

    **.049

    ** .010 1

    Pearson

    Correlati

    on

    .235** .255** .268** .213** .254** .278** .050** .201** .419** .305** .388** .385** .457** .404** .092** .060** 1

    Pearson

    Correlati

    on

    .458**

    .650**

    .511**

    .270**

    .704**

    .387** .021 .060

    **.182

    ** .010 .131**

    .230**

    .115**

    .075** .003 -.017 .183

    ** 1

    Pearson

    Correlati

    on

    .426**

    .518**

    .525**

    .374**

    .565**

    .464**

    .047*

    .119**

    .345**

    .178**

    .281**

    .381**

    .310**

    .272**

    .052**

    .047*

    .322**

    .598** 1

    Pearson

    Correlati

    on

    .739**

    .794**

    .691**

    .594**

    .765**

    .678**

    .264**

    .171**

    .307**

    .199**

    .254**

    .349**

    .308**

    .256**

    .054**

    .065**

    .340**

    .663**

    .637** 1

    Pearson

    Correlati

    on

    .270**

    .295**

    .329**

    .296**

    .293**

    .357** .037 .487

    **.732

    **.572

    **.703

    **.630

    **.707

    **.639

    **.315

    **.227

    **.650

    **.179

    **.408

    **.409

    ** 1

    COL

    MEX

    PER

    ARG

    BRA

    CHI

    US

    VEN

    CHN

    KOR

    PHI

    TAI

    IND

    INO

    MAL

    PAK

    SRI

    THA

    EUR

    LAT

    ASA

    **. Correlation is s ignificant at the 0.01 level (2-tailed).

    *. Correlation is s ignificant at the 0.05 level (2-tailed).

    Table 5.2: Correlations for total period

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    24/75

    24

    5.3 The pre-crisis period summary statistics

    The highest average daily return observed during the pre-crisis period (Table 5.3) for an

    emerging market was that of Peru (0.21%), while the lowest average return was experienced

    by Taiwan with 0.04%. The highest standard deviation was that of Argentina with 1.98%,

    while the lowest was that of Malaysias 0.69%. Pakistan had the highest median return

    (0.18%) while the lowest medians were shared by Venezuela, China, Philippines, Taiwan,

    and Thailand (all with 0%). The maximum daily return was Colombias 15.82% and the

    minimum experienced was Thailands with -14.84%.

    The correlation figures (Table 5.4) suggest that in the pre-crisis period, the relationships

    within regions were weak, though they were still stronger within regions than across.Argentina and Brazil had the strongest correlation of .281 within Latin America (although it

    was Brazil and Mexico for total period), while Chile and Venezuela shared the weakest

    relationship in the same region (.014) (although it was Mexico and Venezuela for total

    period). Korea and Taiwan had the strongest correlation within Asia (same for total period)

    with .554, while India and Sri Lanka had the weakest correlation (.000) in the same region

    (China and Sri Lanka during the total period). The strongest relationship across the two

    regions was between Chile and Korea (India and Peru for the total period) with .192, while

    the weakest relationship across the regions was shared between Venezuela and Sri Lanka

    (Venezuela and Taiwan for total period) with a correlation of -.017.

    The correlations between the advanced markets (US and Europe) and the emerging markets

    show that the US had its strongest relationship with Mexico (it was Brazil for the total period)

    with a correlation of .531. The US had its weakest relationship with Sri Lanka (it was the

    same for the total period) with a correlation of -0.28. Europes strongest relationship with an

    emerging market was with Mexico (.459). Its weakest relationship was with Sri Lanka with a

    correlation of -.014 (both were the same as for the total period).

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    25/75

    25

    Table 5.3: Summary statistics for the pre-crisis period;

    Mean

    Standard

    Deviation Median Minimum Maximum

    ARG 0.15% 1.98% 0.09% -10.68% 13.42%

    BRA 0.11% 1.66% 0.05% -6.63% 6.34%

    CHI 0.08% 0.85% 0.06% -4.97% 3.00%

    COL 0.17% 1.45% 0.13% -10.46% 15.82%

    MEX 0.12% 1.15% 0.11% -5.80% 6.73%

    PER 0.21% 1.09% 0.12% -7.59% 8.55%

    VEN 0.14% 1.58% 0.00% -18.66% 10.42%

    CHN 0.08% 1.47% 0.00% -13.17% 9.39%

    KOR 0.07% 1.42% 0.08% -7.15% 7.64%

    PHI 0.09% 1.16% 0.00% -7.92% 4.89%

    TAI 0.04% 1.27% 0.00% -6.68% 5.64%

    IND 0.11% 1.32% 0.12% -11.14% 8.25%

    INO 0.13% 1.26% 0.09% -10.36% 5.47%

    MAL 0.05% 0.69% 0.03% -4.64% 3.14%

    PAK 0.17% 1.50% 0.18% -7.45% 8.88%

    SRI 0.11% 1.38% 0.02% -12.97% 12.31%

    THA 0.07% 1.26% 0.00% -14.84% 11.16%

    US 0.02% 0.98% 0.04% -4.15% 5.73%

    EUR 0.03% 1.10% 0.05% -5.03% 5.80%

    LAT 0.14% 0.78% 0.18% -4.54% 4.64%

    ASA 0.09% 0.62% 0.12% -4.03% 2.85%

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    26/75

    26

    Table 5.4: Correlations for pre-crisis period

    ARG BRA CHI COL MEX PER VEN CHN KOR PHI TAI IND INO MAL PAK SRI THA US EUR LAT ASA

    Pearson

    Correlatio

    n

    1 .281** .228** .184** .279** .148** .059* .023 .104** .043 .138** .104** .086** .090** -.0 21 .04 0 .116** .203** .195** .636** .142**

    Pearson

    Correlatio.281** 1 .385** .186** .540** .197** .079** .097** .153** .062* .134** .143** .118** .075** .050 .008 .166** .499** .325** .690** .206**

    Pearson

    Correlatio.228** .385** 1 .200** .403** .153** .014 .084** .192** .063* .183** .131** .187** .141** .000 .044 .161** .409** .364** .527** .237**

    Pearson

    Correlatio.184** .186** .200** 1 .262** .180** .096** .051 .114** .108** .107** .180** .118** .079** .088** -.005 .109** .129** .173** .538** .193**

    Pearson

    Correlatio.279** .540** .403** .262** 1 .202** .027 .080** .211** .064* .157** .165** .136** .125** .0 27 - .0 03 .189** .591** .459** .655** .232**

    PearsonCorrelatio .148**

    .197**

    .153**

    .180**

    .202**

    1 .069**

    .062*

    .117**

    .150**

    .104**

    .143**

    .110**

    .145**

    .080**

    .039 .110**

    .162**

    .213**

    .446**

    .209**

    Pearson

    Correlatio.059* .079** .014 .096** .027 .069** 1 -.011 -.006 .015 .016 -.005 -.001 .006 -.007 -.017 .061* .049 .051 .381** .008

    Pearson

    Correlatio

    .023 .097** .084** .051 .080** .062* -.011 1 .086** .006 .050 .043 .094** .129** .0 24 - .0 25 .078** .036 .021 .090** .326**

    Pearson

    Correlatio.104** .153** .192** .114** .211** .117** -.006 .086** 1 .247** .554** .328** .359** .341** .058* .012 .326** .111** .273** .210** .670**

    Pearson

    Correlatio

    .043 .062* .063* .108** .064* .150** .015 .006 .247** 1 .223** .202** .258** .275** .105** .028 .187** -.013 .083** .120** .484**

    Pearson

    Correlatio.138** .134** .183** .107** .157** .104** .016 .050 .554** .223** 1 .260** .341** .306** .097** .021 .301** .119** .223** .206** .630**

    Pearson

    Correlatio.104** .143** .131** .180** .165** .143** -.00 5 .0 43 .328** .202** .260** 1 .336** .215** .100** .000 .233** .064* .236** .211** .550**

    Pearson

    Correlatio.086** .118** .187** .118** .136** .110** -.001 .094** .359** .258** .341** .336** 1 .336** .078** .009 .319** .036 .199** .177** .617**

    Pearson

    Correlatio.090** .075** .141** .079** .125** .145** .006 .129** .341** .275** .306** .215** .336** 1 .088** -.014 .287** .004 .155** .155** .522**

    Pearson

    Correlatio

    -.021 .050 .000 .088** .027 .080** -.00 7 .0 24 .058* .105** .097** .100** .078** .088** 1 .013 .066* -.009 .047 .050 .362**

    Pearson

    Correlatio

    .040 .008 .044 -.005 -.003 .039 -.017 -.025 .012 .028 .021 .000 .009 -.014 .013 1 . 033 -.028 -.014 .025 .237**

    Pearson

    Correlatio.116** .166** .161** .109** .189** .110** .061* .078** .326** .187** .301** .233** .319** .287** .066* .033 1 .064* .200** .225** .560**

    Pearson

    Correlatio.203** .499** .409** .129** .591** .162** .049 .036 .111** -.013 .119** .064* .036 .004 -.009 -.028 .064* 1 .534** .493** .081**

    PearsonCorrelatio

    .195** .325** .364** .173** .459** .213** .051 .021 .273** .083** .223** .236** .199** .155** .0 47 - .0 14 .200** .534** 1 .425** .284**

    Pearson

    Correlatio.636** .690** .527** .538** .655** .446** .381** .090** .210** .120** .206** .211** .177** .155** .050 .025 .225** .493** .425** 1 .294**

    Pearson

    Correlatio.142** .206** .237** .193** .232** .209** .008 .326** .670** .484** .630** .550** .617** .522** .362** .237** .560** .081** .284** .294** 1

    COL

    ARG

    BRA

    CHI

    SRI

    MEX

    PER

    VEN

    CHN

    KOR

    PHI

    TAI

    IND

    INO

    MAL

    PAK

    *. Correlation is significant at the 0.05 level (2-tailed).

    THA

    US

    EUR

    LAT

    ASA

    **. Correlation is significant at the 0.01 level (2-tailed).

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    27/75

    27

    5.4 The post-crisis period summary statistics

    Average daily returns and median daily returns were noticeably lower in the post-crisis

    period (Table 5.5) compared to the pre-crisis period. The highest average daily return

    observed during the post-crisis period for the emerging markets was that of Venezuela

    (0.18%), while the lowest average return was experienced by China with -0.02%. The highest

    standard deviation was again Argentina with 1.98%, while the lowest was again Malaysias

    0.84%. Indonesia had the highest median return (0.05%) while 11 countries shared medians

    of 0%. The maximum daily return was Indonesias 17.34% and the minimum experienced

    was Venezuelas with-12.6%.

    The correlation figures (Table 5.6) suggest that in the since-crisis period, the relationshipswithin regions were strong, and that the relationships were stronger within regions than

    across regions. Argentina and Brazil had the strongest correlation of .683 (a large increase

    from the pre-crisis .281) within Latin America, while Brazil and Venezuela shared the

    weakest relationship (it was Chile and Venezuela for the total period) in the same region

    (.011). Korea and Taiwan again had the strongest correlation within Asia (the same for the

    total period) with .702, while Pakistan and Sri Lanka had the weakest correlation (.002) in the

    same region (this was India and Sri Lanka during the pre-crisis period). The strongest

    relationship across the two regions was between Peru and Thailand (this was Chile and Korea

    for the pre-crisis period) with .372, while the weakest relationship across the regions was

    once again shared between Venezuela and Sri Lanka as in the pre-crisis period with a

    correlation of .006.

    The correlations between the advanced markets (US and Europe) and the emerging markets

    show that the US had its strongest relationship with Brazil (despite being Mexico for the pre-

    crisis period) with a correlation of .738. The US had its weakest relationship yet again with

    Sri Lanka with a correlation of -0.1. Europes strongest relationship with an emerging market

    was with Brazil with .638 (and Mexico in the pre-crisis period). Its weakest relationship was

    with Pakistan with a correlation of .056 (and with Sri Lanka for the pre-crisis period).

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    28/75

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    29/75

    29

    Table 5.6: Correlations for post-crisis period

    ARG BRA CHI COL MEX PER VEN CHN KOR PHI TAI IND INO MAL PAK SRI THA US EUR LAT ASA

    Pearson

    Correlatio

    n

    1 .683** .553** .511** .631** .550** .018 .159** .264** .196** .247** .324** .320** .242** .033 .070** .339** .633** .601** .828** .363**

    Pearson

    Correlatio.683** 1 .622** .479** .771** .558** .011 .187** .294** .155** .204** .351** .268** .208** .024 .055* .319** .738** .638** .860** .348**

    Pearson

    Correlatio.553** .622** 1 .473** .631** .545** .072** .153** .301** .213** .208** .332** .303** .282** .045 .121** .337** .556** .609** .770** .374**

    Pearson

    Correlatio.511** .479** .473** 1 .453** .483** .066* .169** .326** .243** .256** .338** .324** .257** .070** .102** .326** .398** .568** .677** .397**

    Pearson

    Correlatio.631** .771** .631** .453** 1 .536** .014 .143** .297** .131** .184** .345** .253** .206** .033 .059* .298** .765** .626** .825** .326**

    PearsonCorrelatio

    .550** .558** .545** .483** .536** 1 .045 .182** .318** .233** .258** .359** .349** .311** .035 .129** .372** .467** .573** .771** .416**

    Pearson

    Correlatio

    .018 .011 .072** .066* .014 .045 1 .047 .044 .056* .056* .010 .055* .049 .063* .006 .041 .000 .047 .188** .068**

    Pearson

    Correlatio.159** .187** .153** .169** .143** .182** .047 1 .353** .243** .330** .286** .294** .309** .089** .033 .284** .071** .173** .211** .571**

    Pearson

    Correlatio.264** .294** .301** .326** .297** .318** .044 .353** 1 .426** .702** .445** .536** .527** .128** .115** .491** .225** .394** .370** .779**

    Pearson

    Correlatio.196** .155** .213** .243** .131** .233** .056* .243** .426** 1 .466** .271** .458** .474** .081** .105** .392** .022 .235** .244** .626**

    Pearson

    Correlatio.247** .204** .208** .256** .184** .258** .056* .330** .702** .466** 1 .386** .535** .525** .121** .089** .454** .140** .319** .285** .753**

    Pearson

    Correlatio.324** .351** .332** .338** .345** .359** .010 .286** .445** .271** .386** 1 .472** .403** .108** .109** .485** .306** .459** .419** .671**

    Pearson

    Correlatio.320** .268** .303** .324** .253** .349** .055* .294** .536** .458** .535** .472** 1 .572** .111** .103** .553** .154** .375** .378** .758**

    Pearson

    Correlatio.242** .208** .282** .257** .206** .311** .049 .309** .527** .474** .525** .403** .572** 1 .138** .120** .488** .111** .342** .312** .707**

    Pearson

    Correlatio

    .033 .024 .045 .070** .033 .035 .063* .089** .128** .081** .121** .108** .111** .138** 1 .002 .119** .011 .056* .055* .292**

    Pearson

    Correlatio.070** .055* .121** .102** .059* .129** .006 .033 .115** .105** .089** .109** .103** .120** .002 1 .093** -.010 .109** .107** .240**

    Pearson

    Correlatio.339** .319** .337** .326** .298** .372** .041 .284** .491** .392** .454** .485** .553** .488** .119** .093** 1 .250** .403** .413** .712**

    Pearson

    Correlatio

    .633** .738** .556** .398** .765** .467** .000 .071** .225** .022 .140** .306** .154** .111** .0 11 - .0 10 .250** 1 .629** .736** .220**

    Pearson

    Correlatio.601** .638** .609** .568** .626** .573** .047 .173** .394** .235** .319** .459** .375** .342** .056* .109** .403** .629** 1 .743** .471**

    Pearson

    Correlatio.828** .860** .770** .677** .825** .771** .188** .211** .370** .244** .285** .419** .378** .312** .055* .107** .413** .736** .743** 1 .461**

    Pearson

    Correlatio.363** .348** .374** .397** .326** .416** .068** .571** .779** .626** .753** .671** .758** .707** .292** .240** .712** .220** .471** .461** 1

    COL

    ARG

    BRA

    CHI

    SRI

    MEX

    PER

    VEN

    CHN

    KOR

    PHI

    TAI

    IND

    INO

    MAL

    PAK

    *. Correlation is significant at the 0.05 level (2-tailed).

    THA

    US

    EUR

    LAT

    ASA

    **. Correlation is significant at the 0.01 level (2-tailed).

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    30/75

    30

    5.5 Negative exceedances over the total 11 year period

    Table 5.7: The negative (co-) exceedances for the total period;

    Table 5.7 above can be interpreted as follows; for example in Asia, there were 134 days when

    only 2 countries experienced same day negative exceedances (negative return shocks) in the

    11 year period. There were 16 days when China shared negative exceedances with 6 or more

    countries in Asia. When interpreting returns, Thailand had a mean return of -3.18% on the

    days when 6 or more countries in Asia were experiencing a negative exceedance at the same

    time. The same country had a mean return of -3.87% during the days when Thailand was

    specifically one of the 6 or more countries experiencing exceedances on the same day (no

    single Asian country participated in all of the 34 days when 6 or more Asian countries shared

    negative exceedances). The Total returns were simply the average of the returns in each

    column.

    5.5.1 Asia:

    For the total 11 year period there were 812 days when extreme negative returns occurred

    within the Asian countries (2868 observations, 2056 days where no shocks occurred) South

    Korea was shown to be the most susceptible to extreme negative contagion in Asia as it

    shared 32 out of the total of 34 days where 6 or more Asian countries experienced negative

    shocks simultaneously. Koreas high susceptibility to extreme contagion was only slightlyhigher than that of Malaysia (31 days), Indonesia (30 days) and Taiwan (29 days). In contrast

    number of negative (co-)exceedances

    Mean when individually >=6 Mean when >=6 >=6 5 4 3 2 1 0

    China -5.08% -2.68% 16 9 5 19 35 60 2056

    Korea -4.79% -4.65% 32 14 18 21 25 34 2056

    Philipinnes -3.70% -3.14% 27 13 13 21 20 50 2056

    Taiwan -4.04% -3.60% 29 16 16 20 34 29 2056

    India -4.78% -3.83% 25 9 13 16 28 52 2056

    Indonesia -4.72% -4.20% 30 17 14 18 33 32 2056

    Malaysia -2.26% -2.09% 31 15 15 26 20 37 2056

    Pakistan -3.59% -1.05% 7 7 6 6 26 93 2056

    Sri Lanka -2.80% -0.76% 8 0 5 10 22 99 2056

    Thailand -3.87% -3.18% 26 10 15 20 25 47 2056

    Total -3.96% -2.92% 34 22 30 59 134 533 2056

    Argentina -6.80% -6.80% 15 15 19 23 18 54 2262

    Brazil -6.02% -6.02% 15 14 23 20 31 41 2262

    Chile -3.76% -3.76% 15 16 13 21 30 49 2262

    Colombia -3.65% -3.65% 15 11 13 13 25 67 2262

    Mexico -4.65% -4.46% 14 16 22 24 30 38 2262

    Peru -5.73% -5.39% 14 15 18 15 23 58 2262

    Venezuela -2.75% -0.18% 3 3 4 4 19 109 2262

    Total -4.77% -4.32% 15 18 28 40 88 416 2262

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    31/75

    31

    to this, Pakistan only endured 7 days of negative return shocks out of the potential 34 days in

    which 6 or more countries suffered extreme negative shocks. Sri Lanka also showed similar

    signs of immunity with only 8 days shared in the same category. Apart from Pakistan and Sri

    Lanka, China also showed signs of independence by only participating in 16 out of the 34

    days.

    The average return for all Asian countries during these 34 days was -2.92%. South Korea had

    the lowest average return (-4.65%) in Asia during the 34 days when 6 or more countries

    shared negative shocks. This was contrasted by Sri Lanka (-0.76%) with the highest return

    during that period. This is perhaps unsurprising since Sri Lanka was largely absent from the

    34 days of shared contagion (by 6 or more countries). It is therefore more illustrative to

    consider Sri Lankas average return for the 8 days that it did in fact share with 6 or more

    countries (-2.80%). From the countries that shared in the majority of the 34 days (i.e.

    excluding Pakistan and Sri Lanka), Malaysia shows the least negative average return during

    all 34 days of shared contagion (-2.09%) and also the least negative average return when it

    participated in the 34 days (-2.26%). This second figure is even lower than those of Pakistan

    (3.59%) and Sri Lanka (-2.80%) which avoided extreme negative contagion for the most part.

    5.5.2 Latin America

    There was a total of 606 days in which extreme negative returns were experienced by every

    country (2868 observations; 2262 where no shocks occurred). Latin America suffered fewer

    days of extreme contagion than Asia. For the 15 days where 6 or more countries experience

    extreme negative returns simultaneously, 4 countries participated in each of the 15 days

    (Argentina, Brazil, Chile and Colombia). A further 2 countries (Mexico and Peru) shared in

    14 of these 15 days. Venezuela was largely immune to the extreme contagion with only 3

    days shared amongst the 15, however Venezuela claimed the highest number of days in

    which negative shocks were shared with just one other country (109 days). Second to

    Venezuela in this category was Colombia with a comparatively low 67 days. This level of

    contagion sharply declined for Venezuela as the number of shared days of negative shocks

    increased across Latin America.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    32/75

    32

    Argentina had the lowest average return (-6.8%) during the 15 days of contagion with Brazil

    (-6.02%) and Peru (-5.39%) just above. Unsurprisingly, Venezuela had the least negative

    average return within this category (-0.18%). In the same period, the average return for all the

    Latin American countries was -4.32%. The countries below this, other than Venezuela, were

    Chile (-3.76%) and Colombia (-3.65%).

    5.6 Positive exceedances over the total 11 year period

    See table 5.8.

    Table 5.8: The positive (co-) exceedances for the total period;

    5.6.1 Asia

    There were fewer days where extreme positive returns were shared in Asia in comparison to

    the amount of days where extreme negatives were shared (18 days in total compared to 34

    days for negative). Taiwan had the highest level of positive contagion by participating in 17

    number of positive (co-)exceedances

    0 1 2 3 4 5 >=6 Mean when >=6 Mean when individually >=6

    China 1988 80 34 11 10 6 6 1.91% 3.89%

    Korea 1988 28 38 30 33 8 14 3.89% 4.52%

    Philipinnes 1988 65 33 17 10 8 10 2.66% 4.22%

    Taiwan 1988 35 39 27 20 4 17 3.48% 3.74%

    India 1988 41 46 24 12 6 13 3.98% 4.88%

    Indonesia 1988 42 32 27 16 9 16 4.59% 5.04%

    Malaysia 1988 53 32 22 16 4 16 2.21% 2.36%

    Pakistan 1988 80 41 12 6 1 3 0.86% 3.59%

    Sri Lanka 1988 101 21 12 3 3 4 0.79% 2.49%

    Thailand 1988 47 42 19 15 6 16 4.11% 4.38%

    Total 1988 572 179 68 32 11 18 2.85% 3.91%

    Argentina 2193 72 29 12 11 14 6 4.97% 4.97%

    Brazil 2193 55 33 26 13 12 6 4.82% 4.82%

    Chile 2193 54 34 23 14 13 6 3.46% 3.46%

    Colombia 2193 76 35 16 2 10 5 3.11% 3.44%

    Mexico 2193 51 29 29 15 14 6 4.30% 4.30%

    Peru 2193 71 32 14 9 12 6 4.57% 4.57%

    Venezuela 2193 104 30 9 0 0 1 0.56% 2.34%

    Total 2193 483 111 43 16 15 6 3.68% 3.99%

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    33/75

    33

    of the 18 days where 6 or more countries simultaneously experienced extreme positive

    returns. Taiwan was closely followed by Indonesia, Malaysia and Thailand which all shared

    16 days within the total of 18. Pakistan and Sri Lanka again showed signs of immunity with

    only 3 and 4 days shared respectively within the 18. Similarly, China again showed a lack of

    susceptibility to positive contagion with only 6 days shared in the same category.

    The average return for all the Asian countries within the >6category was 2.85%. This was

    surpassed by Indonesia (4.59%) with the largest return in the region. Thailand followed

    behind with an average return of 4.11%. Pakistan (0.86%), Sri Lanka (0.79%) and China

    (1.91%) had the lowest average returns in this category.

    5.6.2 Latin America

    Similarly to the extreme negative returns, Latin America had less extreme positive returns

    than Asia (675 days out of the 2868 observations, 2193 days with no extremes). There was a

    total of 6 days in which 6 or more countries shared extreme positive returns. The countries

    that participated in each of these 6 days were Argentina, Brazil, Chile, Mexico and Peru.

    Venezuela again avoided contagion for the most part by only experiencing 1 day of extreme

    positives in this category.

    3.68% was the average return across all countries when 6 or more countries shared extreme

    positive returns. Argentina experienced the highest average return in the >6category with

    4.97%. This was closely followed by Brazil with 4.82%. Unsurprisingly Venezuela had the

    lowest average return of 0.56%. Colombia had the second lowest with 3.11%.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    34/75

    34

    5.7 Negative exceedances during the pre-crisis period

    Table 5.9: The negative (co-) exceedances for the pre-cr isis per iod;

    5.7.1 AsiaIn the pre-crisis period (1422 days), there were 1017 days where no extreme returns occurred

    in Asia. There were only 6 days when 6 or more countries had extreme negative returns

    simultaneously. Indonesia shared in all of 6 of these days while China and Sri Lanka only

    shared 1 day each in this category. Although Pakistan and Sri Lanka had few days

    experienced in the greater than or equal to 6 category (2 and 1 respectively), they had the 2

    highest amount of days in which extreme negative returns were shared exclusively between 2

    countries in Asia (59 and 65 days respectively).

    The average return for Asian countries for the 6 days in which 6 or more shared extreme

    returns was -2.90%. Indonesia had the lowest average return in this category with a -4.73%

    return. Contrastingly, Sri Lanka had the least negative average return with -0.75%. Despite

    having only 1 day the 6 or more category, China had the most extreme negative return of -

    13.17%.

    number of negative (co-)exceedances

    Mean when individually >=6 Mean when >=6 >=6 5 4 3 2 1 0

    China -13.17% -1.70% 1 1 0 3 10 17 1017

    Korea -4.84% -4.39% 5 4 9 9 16 27 1017

    Philipinnes -3.10% -2.58% 5 3 5 4 9 32 1017

    Taiwan -4.16% -3.70% 5 2 7 8 17 15 1017

    India -5.82% -4.39% 4 3 6 3 5 26 1017

    Indonesia -4.73% -4.73% 6 6 4 5 15 19 1017

    Malaysia -2.65% -1.90% 4 3 5 9 11 18 1017

    Pakistan -3.88% -1.94% 2 5 2 2 11 59 1017

    Sri Lanka -2.35% -0.75% 1 0 3 3 12 65 1017

    Thailand -3.99% -2.97% 4 3 7 5 10 28 1017

    Total -4.87% -2.90% 6 6 12 17 58 306 1017

    Argentina -8.37% -8.37% 1 4 4 6 9 41 1107

    Brazil -5.46% -5.46% 1 3 5 6 14 31 1107

    Chile -2.20% -2.20% 1 4 1 3 13 25 1107

    Colombia -4.59% -4.59% 1 3 2 7 14 46 1107

    Mexico -3.58% -3.58% 1 4 4 6 13 21 1107

    Peru n/a -0.57% 0 1 2 0 8 20 1107

    Venezuela -1.77% -1.77% 1 1 2 2 15 68 1107

    Total -4.33% -4.33% 1 4 5 10 43 252 1107

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    35/75

    35

    5.7.2 Latin America

    Latin America experienced less contagion in this period than Asia with 1107 days where no

    extreme returns happened anywhere. There was only 1 day in which 6 or more countries

    shared extreme negative returns simultaneously. Every country except Peru participated in

    this category. Venezuela had the highest number of days with an extreme negative return

    shared exclusively with one other country in the region (68 days).

    The average return for all of the countries on the day with 6 or more shared negative

    exceedances was -4.33%. The least negative return was of course Peru (0.57%) as it avoided

    a return shock on this day. The most negative return was that of Argentina with -8.37%. This

    was followed by Brazils return of -5.46%.

    5.8 Positive exceedances during the pre-crisis period

    Table 5.10: The positive (co-) exceedances for the pre-cr isis period;

    5.8.1 Asia

    There were 965 days where no positive exceedances occurred in Asia and a total of only 3

    days where 6 or more countries had positive exceedances simultaneously. China and Sri

    Lanka did not participate in these days, whereas South Korea, Taiwan, Indonesia and

    number of positive (co-)exceedances

    0 1 2 3 4 5 >=6 Mean when >=6 Mean when individually >=6

    China 965 33 12 3 2 2 0 0.85% n/a

    Korea 965 18 22 16 7 3 3 4.16% 4.16%

    Philipinnes 965 29 21 8 4 4 1 1.66% 3.60%

    Taiwan 965 15 27 14 5 1 3 3.34% 3.34%

    India 965 19 18 5 3 2 2 2.60% 3.10%

    Indonesia 965 27 20 12 6 2 3 4.15% 4.15%

    Malaysia 965 30 22 10 4 0 2 1.15% 1.33%

    Pakistan 965 53 23 7 1 1 1 2.01% 4.68%

    Sri Lanka 965 60 12 6 2 2 0 0.79% n/a

    Thailand 965 24 27 9 6 3 3 3.64% 3.64%

    Total 965 308 102 30 10 4 3 2.43% 3.50%

    Argentina 1061 54 17 5 2 2 0 n/a n/a

    Brazil 1061 35 18 13 2 1 0 n/a n/aChile 1061 28 14 9 1 2 0 n/a n/a

    Colombia 1061 49 23 10 0 2 0 n/a n/a

    Mexico 1061 31 15 12 2 2 0 n/a n/a

    Peru 1061 29 13 4 1 1 0 n/a n/a

    Venezuela 1061 48 26 7 0 0 0 n/a n/a

    Total 1061 274 63 20 2 2 0 n/a n/a

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    36/75

    36

    Thailand were involved in all 3 days. Pakistan and Sri Lanka had the highest number of days

    where positive exceedances were shared exclusively with 1 other country (53 and 60 days

    respectively).

    The average return for the Asian countries during the 3 days of extreme positive contagion (6

    or more shared exceedances) was 2.43%. The highest were South Korea (4.16%) and

    Indonesia (4.15%) while the lowest were China and Sri Lanka which avoided participating in

    any of the 3 days with average returns of 0.85% and 0.79%

    5.8.2 Latin America

    There were no observed instances of extreme positive contagion in Latin America in the pre-

    crisis period. There were only 4 days in which 4 or more countries shared positive

    exceedances. In total there were 361 days where positive exceedances occurred (1061 days

    where no exceedances occurred anywhere out of the 1422 days observed in the pre-crisis

    period). No average return figures during periods of extreme contagion are available due to

    the fact that no extreme contagion occurred.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    37/75

    37

    5.9 Negative exceedances during the post-crisis period

    Table 5.11: The negative (co-) exceedances for the post-cr isis period;

    5.9.1 Asia

    There were 1039 days where there were no negative exceedances whatsoever (1446 days

    observed in post-crisis period). In contrast to the pre-crisis period, there were 28 days in the

    post crisis period in which negative exceedances were shared between 6 or more countries at

    the same time. Malaysia participated more than any other Asian country in this category (27

    days), while Pakistan and Sri Lanka (5 and 7 days respectively) participated the least.

    The average return across Asia during the most contagious days was -2.92%. The most

    negative average return was South Koreas with -4.70% while Pakistan (-0.86%) and Sri

    Lanka (-0.76%) were the least affected countries.

    5.9.2 Latin America

    There were 1155 days in which no negative exceedances occurred in Latin America in the

    post-crisis period. There was a large increase of extreme contagion in this period compared to

    the pre-crisis period with a total of 14 days of 6 or more countries experiencing exceedances

    number of negative (co-)exceedances

    Mean when individually >=6 Mean when >=6 >=6 5 4 3 2 1 0

    China -4.54% -2.89% 15 8 5 16 25 43 1039

    Korea -4.79% -4.70% 27 10 9 12 9 1 1039

    Philipinnes -3.82% -3.26% 22 10 8 17 11 18 1039

    Taiwan -4.02% -3.58% 24 14 9 12 17 14 1039

    India -4.58% -3.70% 21 6 7 13 23 26 1039

    Indonesia -4.72% -4.08% 24 11 10 13 18 13 1039

    Malaysia -2.13% -2.13% 27 12 10 17 9 19 1039

    Pakistan -3.47% -0.86% 5 2 4 4 15 34 1039

    Sri Lanka -2.86% -0.76% 7 0 2 7 10 34 1039

    Thailand -4.02% -3.23% 22 7 8 15 15 19 1039

    Total -3.89% -2.92% 28 16 18 42 76 227 1039

    Argentina -6.69% -6.69% 14 11 15 17 9 13 1155

    Brazil -6.06% -6.02% 14 11 18 14 17 10 1155

    Chile -3.87% -3.87% 14 12 12 18 17 24 1155

    Colombia -3.58% -3.58% 14 8 11 6 11 21 1155

    Mexico -4.73% -4.52% 13 12 18 18 17 17 1155

    Peru -5.73% -5.73% 14 14 16 15 15 38 1155

    Venezuela -1.49% -0.06% 2 2 2 2 4 41 1155

    Total -4.59% -4.35% 14 14 23 30 45 164 1155

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    38/75

    38

    simultaneously. Argentina, Brazil, Chile, Colombia and Peru participated in each of the 14

    days of extreme contagion, while Venezuela only participated for 2 days during this 14 day

    period.

    The average return across Latin America during the 14 days of extreme contagion was -

    4.35%. The lowest average return was suffered by Argentina (-6.69%) and the next lowest

    was that of Brazil (-6.02%). The least extreme was average return was that of Venezuela (-

    0.06%).

    5.10 Positive exceedances during the post-crisis period

    Table 5.12: The positive (co-) exceedances for the since-crisis period;

    5.10.1 Asia

    Extreme contagion also increased massively for positive exceedances in the since-crisis era

    with 15 days in total in which 6 or more countries simultaneously experienced positive

    exceedances. Indonesia and Thailand participated the most in this category with 13 days

    number of positive (co-)exceedances

    0 1 2 3 4 5 >=6 Mean when >=6 Mean when individually >=6

    China 1023 47 22 8 8 4 6 2.12% 5.02%

    Korea 1023 10 16 17 13 5 11 3.83% 4.62%

    Philipinnes 1023 36 12 9 6 4 9 2.86% 4.29%

    Taiwan 1023 20 12 13 15 3 14 3.51% 3.91%

    India 1023 22 28 19 9 4 11 4.26% 5.27%

    Indonesia 1023 15 12 15 10 7 13 4.67% 5.21%Malaysia 1023 23 10 12 12 4 14 2.42% 2.51%

    Pakistan 1023 27 18 5 5 0 2 0.63% 3.05%

    Sri Lanka 1023 41 9 6 1 1 4 0.79% 2.49%

    Thailand 1023 23 15 10 9 3 13 4.21% 4.55%

    Total 1023 264 77 38 22 7 15 2.93% 4.09%

    Argentina 1132 18 12 7 9 12 6 4.97% 4.97%

    Brazil 1132 12 15 13 11 11 6 4.82% 4.82%

    Chile 1132 26 20 14 13 11 6 3.46% 3.46%

    Colombia 1132 27 12 6 2 8 5 3.11% 3.44%

    Mexico 1132 20 14 17 13 12 6 4.30% 4.30%Peru 1132 42 19 10 8 11 6 4.57% 4.57%

    Venezuela 1132 56 4 2 0 0 1 0.56% 2.34%

    Total 1132 209 48 23 14 13 6 3.68% 3.99%

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    39/75

    39

    each. China, Pakistan and Sri Lanka were the least affected countries with 6, 2 and 4 days

    respectively involved in these 15 days.

    The average return across Asia during these 15 days was 2.93%. The countries that were

    below this average were China (2.12%), Philippines (2.86%), Malaysia (2.42%), Pakistan

    (0.63%) and Sri Lanka (0.79%). Indonesia had the highest average return of 4.67%. China

    experienced the highest average return over the 6 days in which it participated in the extreme

    contagion with 5.02%.

    5.10.2 Latin America

    There were 1132 days in which no negative exceedances occurred in Latin America in the

    since-crisis period. There was also a large increase of extreme contagion in this period

    compared to the pre-crisis period with a total of 6 days of 6 or more countries experiencing

    exceedances simultaneously. All of the Latin American countries participated in at least 5 of

    the days of extreme positive contagion with the exemption being Venezuela which only

    participated in 1 of these days. The average return across Latin America during the 6 days of

    extreme positive contagion was 3.68%. The highest average return was experienced by

    Argentina (4.97%) and the lowest was that of Venezuela (2.34%).

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    40/75

    40

    5.11 Regression results

    In attempting to further determine the characteristics of contagion a binary logistic regression

    model was used. The regression model was similar to that used by Boyson (2006) in which

    explanatory variables in the model were representative of broad market indicators which may

    help to characterize contagion. These explanatory variables were changes in interest rates,

    changes in the US Dollar exchange rate against a basket of other major currencies, and

    changes in market volatility over the same time period. Extreme positive and negative returns

    were analysed separately. Furthermore the regression analysis was implemented for each of

    the different time periods of pre-crisis, since-crisis and the total 11 year period. The response

    variable used was representative of the total number of same day exceedances across each of

    the 2 regions (Asia and Latin America). This enabled the logit model to measure the

    relationship between the broad market indicators and extreme returns across an entire region

    rather than within the individual countries analysed.

  • 8/13/2019 Measuring financial contagion between emerging equity markets before and after the onset of the global financial

    41/75

    41

    5.11.1 Asian Negative exceedances; Total period regression output (Table1)

    Negative exceedances in Asia was the first response variable used. The Hosmer-Lemeshow

    test indicated that the model was a good predictor of contagion. The likelihood of an extreme

    negative return in Asia was increased slightly with an increase in market volatility.

    Contrastingly, increases in the Dollar and Treasury Bills slightly reduced the probability of an

    extreme negative return in Asia. Both market Volatility and Treasury Bills had a significant

    relationship with negative exceedances in Asia as indicted by the low p-values. The Dollar

    however did not have a significant relationship.

    Table 1: Asian Negative exceedances; Total period regression output

    Hosmer and Lemeshow Test

    Step Chi-square df Sig.

    1 10.957 8 .204

    Variables in the Equation

    B S.E. Wald df Sig. Exp(B)

    Step 1a

    Doll