Correlation Pattern among “ Asian Paper Tigers ” Currencies : A Dynamic Conditional Correlation Approach

This study attempts to investigate the Dynamic Conditional Correlation (DCC) for eight currencies in the East Asia region, known as Asian Paper Tigers from the period of July 2002 to July 2012. The estimation results generated from DCC model verify that each tested exchange rate's volatility is determined by its own previous volatility shock, however failed to find any evidence with its own residual shock. While for correlation estimation results, we support the evidence that the conditional correlations for all tested pairs currencies are highly affected by their previous correlation. Most of the Asian Paper Tigers currencies recorded a low conditional correlation over the tested sampling period except for CNYJPY, MYRCNY, MYRIDR, MYRTHB, JPYTHB and PHPKRW. The findings further verify that mixing the currencies within different monetary regime plays a significant role in enhancing the currency portfolio diversification results. Although in unstable period, both JPYTHB and MYRJPY are the most promising combinations to be included in the optimal currency investment basket where both pairs have small and stable correlations either during the global recession period or European liquidity crisis period.


INTRODUCTION
In recent years, interest in correlation has been growing rapidly since the parameter is crucial in explaining the co-movement behaviour within financial and economic application investigations (Li, 2011).He postulates that the results of currencies co-movements have implications in managing portfolio, for example, for purposes of predicting the changing directions of currencies co-movements in optimising the currency portfolios performances.In general, how the currencies move together is also matters towards the real economy activities.As for central banks, they are interested to know whether they could achieve the desired domestic appreciation or depreciation against other foreign currencies (Benediktsdóttir and Scotti, 2009).In practical perspective, currencies co-movements are vital in portfolio construction, where fund managers can use the correlation coefficients to determine the right international currencies combination or other financial assets combination in their investment portfolio basket.
There are two approaches used to find the currencies co-movements; correlation and tail dependence.Correlation looks at how pairs of currencies move together across distribution.This is relevant to investors and fund managers who are constructing portfolios with global minimum variance.Previously, many researchers estimate the unconditional correlation between assets over a certain period of study (Solnik, 1974;Solnik et al., 1996).However, new information flow has a tendency to influence the correlation pattern among asset classes (Makridakis and Wheelwright, 1974;Bennett and Kelleher, 1988).They are among the earliest to discover the co-movements of international stock markets are unstable over the time.As such, conditional or non-monotonic correlation estimation is much more relevant for portfolio managers.Thus, recently a large number of correlation investigation studies apply Dynamic Conditional Correlation (DCC henceforth) to estimate the conditional correlation pattern among tested currencies (Li, 2011;Zhang, 2011;Li et al., 2012;Tamakoshi and Hamori, 2013).When these countries' currencies are strong, they tend to display higher positive correlation among these currencies especially when they are tied with US dollar, Euro and Japanesse Yen.Bong-Han et al. (2011) compares the movement of Japanese yen with five other emerging Asian currencies.Their findings display downward correlations trend between tested currencies.
The other approach is tail dependence, which focuses on the level of dependence of two distribution tails (either lower-left quadrant or upper-right quadrant).Lately, many have explored the financial data distribution pattern within this paradigm (Hauksson et al., 2001;Fortin and Kuzmics, 2002;Beine et al., 2010) in equity market, Patton (2006) in currency market inter alia).Benediktsdóttir and Scotti (2009) however, combined the correlation and tail dependence techniques in their co-movements investigation.They portrayed the effect of recession in both correlation and tail dependence estimation.The finding confirms that currencies tend to move differently between high interest rate differential currencies.Using European stock indices, Fortin and Kuzmics (2002) find that these returns are highly dependent in its lower tail due to the downward movement of European exchange rate volatility but fail to find any asymmetric effect.Similar pattern is displayed in Asian stock markets, where the crisis affects the co-movements of currencies within Asian region (Caillault and Guégan, 2007).Additionally the introduction of the Euro currency have pushed up the co-movement between entire euro stock return distributions, hence minimize the portfolio diversification benefits within this European region (Beine et al., 2010).
Non monotonic correlation estimation is much more relevant since the market volatility vary over the time for any financial and economic series.Further, Li (2011) finds that correlation among five inflationtargeted currencies 1 tend to move in a time varying manner.Hence, in this research we intend to focus on correlations estimation between pairs of currencies in the East Asian region or also known as Asian Paper Tigers using a time varying Dynamic Conditional Correlation (DCC henceforth) on the currencies returns.Within the GARCH volatility framework, we estimate the conditional correlation using DCC model for these 28 pairs of currencies as time varying volatility is a pervasive phenomena that occurs to a greater or lesser degree with most time series of financial returns.We choose the Asian Paper Tigers currencies as these currencies are likely to move together with the other currencies within their geographically close countries (Mizuno et al., 2006).
Economic climate constantly changes over the time therefore; these changes may affect the movements of these pair's currencies.Economic turmoil's have affected Asian Tigers currencies in many ways including:  Changes in their liquidity climate  Upward movement of the level of risk aversion among Asian paper tigers  Loosen of carry-trade among these countries (Kohler, 2010) Apart from that, de-leveraging also plays a significant role in influencing the currencies co-movement in this region (Melvin and Taylor, 2009).Any liquidity instability in market condition may increase the correlation between Asian countries and US (Zhang, 2011).Tamakoshi and Hamori (2013) infer the effect of European liquidity bubble on the conditional correlation between European cross currencies swap.The liquidity turbulence have resulted an extreme incremental in the co-movement within cross currency swap market.Other example whose demonstrated the effect of economic climate on correlation movement between various currencies are King and Wadhwani (1990), Baig and Goldfajn (1999), Caillault and Guégan (2007)

METHODOLOGY
Within the GARCH volatility framework, using the computed returns, we employ Dynamic Conditional Correlation to estimate the conditional correlation between these eight Asian Paper Tigers currencies.We estimate the conditional correlations for the 28 pairs of currencies.Further, we also compute the descriptive statistics for each estimated conditional correlation series.

Dynamic
conditional correlation model specification: Bollerslev (1990) develops a simple conditional correlation (Constant Conditional Correlation Model) that encompasses the univariate GARCH framework.The model's basic assumption is correlation moves in a monotonic fashion.Economic and financial climate tend to change over time and markets incorporate these changes into their price movements.Hence, it is less accurate for the researcher to posit monotonic conditional correlations for many economic and financial variables.Consequently, Tse and Tsui (2000) and Engle (2002) introduce the time varying conditional correlation GARCH models, namely, DCC.The model is able to capture time varying correlation between two random currency returns.For the purpose of this research, Engle ( 2002)'s DCC model (Appendix A) is used; where the model is outlined according to the following requirement: where, D t is the diag and P follows the dynamic process (contrary to Constant Correlation model where p is set to constant): where the N N is a symmetric positive definite matrix Q t = (q ij,t ) and Q t formulates as follows: 1 where u t represents the standardized residual ( / ), is the N×N unconditional correlation matrix of standardize u t , α and β are the non-negative scalar parameters, which are restricted to be α + β<1.
The Q t matrix is written similar to the GARCH process and transformed into a matrix.However, a constant conditional correlation can be tested by restricting α = β = 0 towards the DCC model.This model, however, tends to drive the conditional correlation to similar dynamics because α and β are scalars.On the other hand, when N is large, the model becomes easy and flexible to use because the model can be estimated through the two-step process consisting; dynamic volatility clustering structure and dynamic correlations parameters.
DCC estimation results: Table 1 represents the estimation results for selected exchange rate generated from the DCC model.The results exhibit strong evidence that each exchange rate's volatility is determined by its own volatility shock (refer to A 11 and A 22 parameters) rather than its own shock (refer to B 11 and B 22 parameters).Overall, the correlation results support the evidence that the conditional correlations for all tested pairs currencies (except for PHPTHB) are highly affected by its previous period correlation (refer to B).However, lower magnitudes are displayed from their residual terms (refer to A). Contrary results are displayed for JPYSGD, MYRKRW, CNYPHP and PHPTHB correlations where each of the pair's correlation is strongly influenced by its own residual shocks and the results fail to find any significant influence from its previous period correlation.Unique by itself, there is no evidence to infer that KRWTHB current correlation is affected either by its own residual or its previous term correlation.

Conditional correlations results:
Full sample: Based on the mean equality test results 2 , we find that the conditional mean between the 28 pairs of currencies are significantly different with one and another.Table 2 summarises the statistical descriptive for the selected pairs of currencies' correlations.
For instance the conditional correlation for CNYJPY, MYRCNY, MYRIDR, MYRJPY, MYRTHB, JPYTHB, PHPKRW in Panel 1 are volatile over the full sampling period.Among the unstable correlation pairs, MYRCNY display an extreme correlation movement ranging from -0.125 to 0.979 over the 10 years tested period.While, MYRIDR conditional correlations swing quite widely from 0.033 to 0.87.Further the average conditional correlation for JPYTHB, CNYJPY, MYRCNY, MYRIDR, MYRTHB and PHPKRW are positively correlated to each pairs.And PHPKRW exhibits the highest average conditional correlation which is at 0.47.In contrast, MYRJPY are inversely correlated where the pair's average conditional correlation is -0.03.
Intuitively, we can suggest that investor should consider investing for all 19 pairs currencies in Panel 2 without the Indonesian Rupiah and Thai Baht combination (IDRTHB) and Malaysian Ringgit and Singapore Dollar (MYRSGD).These 19 pairs currencies have a weak and stable conditional correlation which are best for investment diversification results.However, investor should avoid investing in CNYJPY, MYRCNY, MYRIDR, MYRTHB, JPYTHB and PHPKRW since these pairs displayed a volatile conditional correlation pattern.Such unstable movement may not give the best diversification results for investors.On the other hand, relating the findings with the Paper Tigers exchange rate regime context, we find that good diversification synergy can be achieved when the investors combine currencies between countries that apply dirty floating and free floating exchange rate regime (see Panel 2).Dirty float countries adjust their domestic currency value via central bank intervention to maintain a desired domestic currency value accordingly, such regime will complement the movement of the country with a free floating regime.This does not work when the conditional correlations are in volatile state (refer to Panel 1).Based on Panel 1, JPYTHB and CNYJPY are under dirty and free floating regime but they have considerable among high pairs correlation.
Nevertheless, investors try to avoid combining two currencies that under the same regime (MYRSGD and IDRTHB for Dirty float regime and PHPKRW for free float regime).Within the same regime, both countries apply a very much similar exchange rate system for their currencies, hence no diversification effect could be generated.However, this does not apply when you tend to combine a stronger Big Paper Tigers with the smaller one (such as JPYKRW, JPYPHP and CNYIDR).
Three sub periods: Table 3 describes the descriptive statistic characteristics for selected pairs of currencies.
The findings in Panel 1 (Table 3) describes the mean and standard deviation conditional correlation in unstable situation according to three sub periods.We found that CNYJPY is more volatile in Sub Period 1 compared to the other two sub periods.Japan and China was a big rival in dominating the Asian market in mid 2000s (Bong-Han et al., 2011), as such it is not surprise that these two currencies correlation have a volatile movement during that period.Although CNYJPY correlation is highly volatile in sub period 1, but the average correlation is merely small (0.07), hence we suggest that investor should consider to combine these two currencies into their investment basket when the global financial market is quite stable.Despite the fact that both countries adopt different exchange rate regime which may produce good diversification results, however, this could not be manifested when the global market was not in a good shape.Hence, investor should avoid combining Chinese Yuan and Japanese Yen when the global financial climate is turbulence state (sub period 2).Contrary to CNYJPY, MYRCNY and JPYTHB are a bad investment combination in stable market (sub period 1).However, JPYTHB have a lower range of correlation during the recent European liquidity crisis (Sub Period 3), hence investor can combine these two currencies in their investment basket during the unstable market.While, we can see an accelerating pattern in conditional correlation from sub period 1 to sub period 3 for MYRIDR, MYRTHB and PHPKRW.These three pairs were not well diversified currencies since they have a strong positive correlation in these three sub period.Such strong correlation displayed may be due to these three pairs are within the same exchange rate regime where MYRIDR and MYRTHB are imposing the dirty float regime while, PHPKRW is adopting the free float system.A similar regime between two countries imply that both countries have a very much similar exchange rate climate and such similarity may led to a less promising diversification results.Therefore either in stable or less stable market condition, investor should not consider these three pairs in their currency investment portfolio.
The results in Panel 1, further showed that MYRJPY have a moderate correlation movement ranging from -0.6 to 0.3 in the recent European Liquidity crisis period (Sub Periods 3).Prior to the Global Financial crisis, we found that Malaysian Ringgit and Japanese Yen conditional correlation mean was merely small (Sub Period 1), while the other two sub periods displayed an opposite direction.During the year end of 2008, the other European currencies were weakening, Yen displayed a strong position parallel with dollar and Swiss France.Such strong currency have resulted an influx of investors shifted into YEN rather than European currencies during that European liquidity crisis.Due to this, combination of Malaysian Ringgit with Japanese Yen gives a good choice of investment.Further, such weak and negative correlations are the best criteria in getting the best diversification result in portfolio management.
Next in Panel 2 presents the summarization of the mean and standard deviation selected pairs currencies in a stable correlation movement in the three sub periods.Most of the standard deviation for all 9 pairs were at the lowest (below than 0.09).We can conclude that these 9 selected pairs have the most stable correlation fluctuation within the tested three sub periods.Stable does not always a good indication, since MYRSGD, SGDTHB, IDRTHB and IDRJPY are among the ineffective combination to be included in your currency investment portfolio either in stable or unstable global market condition.Moreover some of the good examples of investment are JPYKRW, CNYIDR, MYRKRW and JPYPHP.These 4 pairs of currencies recorded an extremely small correlation between each other (conditional correlation <0.08).As such at any market condition, investors may acquire a good investment result in these four pairs currency.Although JPYKRW, CNYIDR and JPYPHP are within the same exchange rate regime, however these pairs displayed a good diversification result due to strong Asian tigers paired with the smaller ones.
Overall, when we segregated the full sampling period into three sub periods, the results infer that JPYTHB and MYRJPY are a good choices to put in your currency investment basket when market are in crisis period (in Global recession and European liquidity crisis period).A strong free float currency such as Yen is able to minimize the effects of unstable dirty float currency such as Malaysian Ringgit and Thailand Bath.Although, both Malaysia's and Thailand's central banks play a significant role in buying and selling their currencies in open market to protect the currency value during that crisis period, a strong Yen sufficient to shield the crisis effect toward their currencies.On the other side of coin, CNYJPY turned out to be the best investment option when the market is stable (sub period 1).Either in crisis or noncrisis scenario, a weak, stable correlation movement, strong/small paper tigers combination either within similar regime or different regime are some of the investment criterions to be considered in achieving the best diversification result (such as JPYKRW, CNYIDR, MYRKRW and JPYPHP).

CONCLUDING REMARKS
This study explores the co-movements between eight Asian Paper tigers consisting Malaysian Ringgit (MYR), Singapore Dollar (SGD), China Yuan Renminbi (CNY), Japanese Yen (JPY), Indonesian Rupiah (IDR), Thailand Baht (THB), Philippine Peso (PHP), Korean Won (KRW) currencies for the period between 24/7/2002 and 24/07/2012.We employ the DCC model to estimate the dynamic condition correlations between 28 pairs currencies.Empirical evidence infers that co-movements between assets classes tend to change accordingly to the economic climate.Hence to identify the implication of crisis towards Asian Paper Tigers currencies co-movements, we further divide the sampling period into three sub periods consisting Pre Global financial crisis, Global financial crisis and recent European Liquidity Crisis.The main research findings are as follows:  Eighteen pairs of Asian paper tigers currencies recorded a low conditional correlation over the full sampling period except CNYJPY, MYRCNY, MYRIDR, MYRTHB, JPYTHB and PHPKRW. Mixing differ exchange rate regimes between currencies play a significant role in enhancing the currency portfolio diversification results. Similar exchange rate regime might not creating a diversification synergy, however it might be materialize when investors combining the stronger and weaker currencies. During the global recession and European liquidity crisis period, both JPYTHB and MYRJPY are the most promising combination to be included in the currency investment basket.
In the mist of slowdown Japan's economic recovery, Japanese Yen is able to maintain its strong position in this East Asian territory.This safe haven currency is able to neutralise the unstable currencies like Thailand Bath and Malaysian Ringgit.Although Thailand and Malaysia adopt the dirty float system, yet both central banks interventions may be insufficient to stabilize both currencies values at all time.Hence, by combining the safe haven currency into their foreign reserve basket it may create a good synergy.While for investor perspective, these combinations will create a better diversification synergy.
In a nutshell, investors should aim for Asian Paper Tigers currencies that fall under stable and weak correlations.Next condition, investors should combine the strong currencies with the weak currencies and differ exchange rate regime for an excellent diversification results.These suggestions are solely been made based on the average conditional correlation and volatility perspective.Future researcher should explore the portfolio diversification performance via construction the actual currencies basket and measure the constructed portfolio return throughout the stable and unstable period.

Table 1 :
Estimation for DCC model Panel 1 -

Table 2 :
Descriptive statistic for tested pairs conditional correlation (full sample period)

Table 3 :
pairs currencies with high correlation volatility (high standard deviation or S.D.) and panel 2 shows pairs currencies with stable correlation volatility (low standard deviation or S.D.Mean and standard deviation for selected pairs currencies conditional correlation according to three sub periods Panel 1 (high S.D.) ); DF: Dirty float exchange rate regime; FF: Free floating exchange rate regime; Min.: Minimum; Max.: Maximum; S.D.: Standard deviation , sub period 2 and sub period 3 show the pre global recession, global recession and European liquidity crisis, respectively; DF: Dirty float exchange rate regime; FF: Free floating exchange rate regime; S.D.: Standard deviation