2012asianfa the effect of foreign institutional investors · the effect of foreign institutional...

1

1

The Effect of Foreign Institutional Investors Before and After QFII Elimination:Evidence from Taiwan

Abstract

In this article, we found that Taiwan benefited from the elimination of the Qualified ForeignInstitutional Investor system (QFII) system and that the destabilizing effects of institutionalinvestors were diminished. Using 10 years of daily data, we concluded the following. (1) Foreigninvestors do have a stabilizing effect on the stock market, except during a crisis period. (2) There isno strong evidence to prove that foreign investors destabilized the currency market. (3) Foreigninvestors had a slight stronger impact than domestic investors on the stock market during thefinancial tsunami. (4) The role of foreign investors in Taiwan has changed following the eliminationof the QFII system.

The empirical results have policy implications for those countries that still impose regulations onforeign capital flows and shed light on the question of whether a small, flat, and export-orientedeconomy could benefit from financial deregulation.

Keywords: foreign institutional investors, domestic institutional investors, QFII, financial tsunami,VAR.

1. Introduction

1.1 Research motivationIn last few decades, financial liberalization has become the mainstream in many emerging

markets, including Taiwan. As an emerging country, the Taiwanese government has removedrestrictions on international capital flows step by step, loosening limitations on the degree offoreigners’participation in Taiwan’sstock market. However, before the Taiwanese governmentcompletely eliminated the Qualified Foreign Institutional Investor system (QFII) and prepared toenjoy the benefits of liberalization, we witnessed the damaging effects of powerful internationalcapital flows in emerging countries such as the Mexican peso crisis (1994/1995), the Asian financialcrisis (1997/1998), the Russian financial crisis (1998), and the Brazilian crisis (1999). ThoughTaiwan was not heavily affected by these crises, we still suffered moderately1. Radelet and Sachs(1998) concluded that large-scale movements of international capital flow were at the core of theAsian crisis. That is, the behavior of international capital flows may offset some of the advantagesof liberalization and internationalization. Hence, this study focused on the influence of foreign

1 Taiwan performed well compared with what other countries suffered during the Asian crisis period. Exchange rate:TWD dropped by 20%, Thai Baht by 55%, South Korean Won by 50%, Malaysian Ringgit by 42%, and IndonesianRupiah by 70%. Stock market: Taiwan rose by 18.1%, South Korea and Southeast Asia dropped by 30% to 55%.

2

2

capital flows. We examined the effect of foreign investors in the stock market and currency market.The objective of this paper is to determine the role of foreign investors before and after theelimination of the QFII system in Taiwan, including during the time of 2008 Financial Tsunami.

1.2 Research backgroundInternational capital flows from country to country. These cross-border transactions not only take

advantage of high-return markets but also reduce the risk to their portfolios by diversification. Atthe end of 2009, Taiwan’sstock market had accumulated approximately 1.768 trillion TWD capitalinflows from foreign investors since the elimination of the QFII system; 470 billion TWD net salesin 2008 were included.

Foreign investors held only 15.06% of the market capitalization of Taiwan’sstock market in 2000,but the number increased to 31.64% in 2009. Furthermore, this figure is approximately 15 timesgreater than that for domestic institutions, which held 1.79% for investment trusts and 0.31% forsecurity dealers, respectively. This market share suggests the increasing importance of foreigners toTaiwan’sstock market.

Because of the huge capital flow from foreign investors, when a crisis influenced the emergingmarket economies, foreign investors, as the providers of international capital flows, were blamedfor destabilizing stock prices and currency values. However, not all crises are bad. A crisis mightserve as a reflection of the flaws of the country’s existing policies. Policy flaws could be thereason why an economy was attacked by international speculators, which were blamed fortriggering financial disasters. For example, Mundell (2000) pointed out that the pegged rate mayserve as a useful measure under certain circumstances, but not as a perpetual measure, because itcauses policy conflicts and eventually leads to a currency crisis. The time bombs started countingdown when the government adopted pegged rate policy, as it did in Mexico, Thailand, andIndonesia.

As a shallow economy, Taiwan exhibited large fluctuations in the stock and exchange market,even in the noncrisis period (from 2000 to 2007). The foreign exchange rate (TWD/$) rangedfrom 30.302 to 35.168, and the stock index ranged from 3446.26 to 10202.2. As a free market, thecapital flows from foreign investors in Taiwan are considered one of the most influential marketpowers. By 2001, foreigners owned 15.06% of the stock market capitalization and 5.89% oftrading volume compared with 34.5% and 17.62% in 2007, respectively. Thus, there is evidencethat foreign investors had demonstrated their forceful growth in Taiwan before the crisis.

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Total(100m)

3

3

Table 1. Net purchases of institutional investors (in 100 million NTD)Source: Taiwan Economic Journal

Table 1 shows the 10-year net purchases of the three major institutional investors in the stockmarket. We can see the magnitude of foreign investors’capital flow in Taiwan and find they onlyact as net sellers during the crisis. Net purchases of domestic institutions displayed less volumethan foreigners most of time, and investment trusts seemed to be the opposite of foreign investors.Given the differences in magnitude between foreigners and locals, we would like to know whetherforeign investors’trading behavior, such as feedback trading2, exacerbates stock and exchangemarket volatility.

1.3 Research objectivesTaiwan is a good example of crisis effects when capital restrictions are removed. The difference

between the 1997 Asian crisis and the 2008 Financial Tsunami is telling, because the formeroccurred during a period of capital control regulation on foreign investors and the latter did not.Past studies could not provide evidence regarding whether free international flows stabilize ordestabilize the stock market and currency market during a catastrophe because they lacked suchsimilar events after deregulation. In this study, we aim to determine the effect of the elimination ofthe QFII system. Did the role of foreign investors in Taiwan change after financial liberalization?What role did foreign investors play during the crisis? Is the effect of foreign investors on stockand currency markets stabilizing or destabilizing before and after the elimination of QFII?

Our hypotheses and explanations are stated as follows: (1) foreign net purchases destabilize thestock market, which means market returns are affected by foreign net purchases; (2) foreign netpurchases destabilize the currency market; that is, currency returns are affected by foreign netpurchases. Further, we compare foreign institutional investors with domestic institutionalinvestors regardless of the existence of the QFII system and during the 2008 financial tsunami: (3)foreign net purchases destabilize the stock market and currency market more than domesticinstitutions. Then, we examine foreigners’role in Taiwan: (4) has the role of foreign investorschanged after the elimination of the QFII system? In other words, have foreign investorsdemonstrate different trading behaviors after deregulation? These hypotheses will be investigatedusing a vector autoregressive (VAR) model.

2 Feedback trading: Trading conduct based on historical data. There are two types of feedback trading: negative andpositive. Negative feedback trading means selling when prices rise and buying when they fall. Positive feedback tradingmeans buying when prices rise and selling when they fall.

ForeignInvestors 1,466.65 3,062.51 278.97 5,489.56 2,839.66 7,194.15 5,581.09 741.29 -4,700.02 4,801.40 26,755.26

InvestmentTrusts -160.25 -90.14 68.98 -56.73 -141.07 -863.28 -542.55 1,553.86 441.62 -289.26 -78.82

SecuritiesDealers -693.82 90.46 25.25 297.84 -143.41 169.13 69.01 159.54 437.77 100.08 511.85

4

4

The rest of this paper is organized as follows: Section 2 discusses the possible relationshipbetween foreign investors and the local market. This section also introduces the methodology thatwe used and explains why it applies to this study. Section 3 describes the data and themethodology. Section 4 presents the empirical results and analyzes the phenomena that we foundduring each period. Section 5 concludes the study and makes suggestions to further studies.

2. Literature review

The stock market in Taiwan is different from other countries; the majority is made up ofindividual investors. Thus far, individual investors own approximately 70% of daily tradingvolume. Three major institutional investors (securities dealers, investment trusts and foreigninvestors) only own approximately 30%. Institutional investors can be separated into two types:domestic and foreign. Since 2005, foreign institutions have exceeded domestic institutions indaily trading volume percentage, as Figure 1 shows. In 2008, foreign institutions reached a highof 22% of daily trading volume; their share has increased nearly 10 times in past 10 years.

We found the trading volume of foreign institutional investors is still much less than that ofindividuals (22% vs. 61%). However, their investment strategies and behavior are considered tobe indicators to individual investors. Unlike individual investors, foreign institutional investors haveprofessional backgrounds and deep pockets. Arbel et al. (1983) and Brous and Kini (1994)concluded that institutional investors use their own information and professionals, collecting andanalyzing data before devising strategies. Bacmann and Bolliger (2001) found that foreignanalysts produce more timely and accurate forecasts than local analysts. Thus, individualinvestors are always concerned with the actions and strategies of foreign institutional investors. Inaddition, media reports always focus on the movements of foreign institutional investors. Thus, itis even easier for individual investors to follow foreign institutionalinvestors’strategies.

This section discusses the effects of foreign investors on the stock market and foreign exchangemarket. Past studies did not reach any conclusion regarding whether foreign investors destabilizethese two markets. Thus, the role that foreign investors play, disturbing and enhancing thevolatility of local market, is still a controversial issue for academics. In the remaining sections, wediscuss the negative effects of foreign investors first, followed by their positive effects on themarkets. We also describe the reason that we chose to study Taiwan. Lastly, we explain why weuse a vector autoregressive (VAR) model as our methodology at the end of the section.

Figure 1. Trading volume of institutional investors in TSE (%)

5

5

0.00

5.00

10.00

15.00

20.00

25.00

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Year

%

Domestic Institution Foreign Institution

Source: Taiwan Stock Exchange

2.1 The negative sideThe Central Bank in Taiwan3 (CBC) held the opposite attitude toward foreign investors because

they were thought to cause instability and fluctuations in the economy. Taiwan is a shalloweconomy; both the stock market and currency market are easily affected by international and localevents. In addition, firms rely on importing and exporting so much that fluctuations in exchangerates can cause severe damages to them. Thus, the CBC preferred a step-by-step capitalderegulation process to open the market.

Because of the large capital flows from foreign investors, it is also said that foreign investors maycause market volatility and abnormal returns. Scholes (1972) indicated that the trading behaviorsof foreign investors often cause market instability and abnormal returns because of their largecapital flows. Some research has focused on the trading behaviors of foreign investors. Bekaertet al. (2002) studied 20 emerging countries and found that foreign investors were prone to “capitalflight”: whencapital leaves, it leaves faster than it entered most of these countries. The authorsalso suggested that foreign investors’trading behavior often causes stock market instability anddisturbs the exchange rate. Foreign investors’ strategieswere also examined. Dornbusch andPark (1994) argued that foreign investors’trading strategies allow stock prices overreact to changesin fundamentals and, thus, lead to higher volatility. Eswar et al. (2003) also indicated that becauseforeign investors employ herding4 and feedback trading strategies, which can move stock pricesaway from fundamentals, fickle capital flows could increase the volatility of equity returns anddestabilize equity markets.

3 The Central Bank in Taiwan (CBC) is the Central Bank of the Republic of China4 Herd behavior is the tendency for individuals to mimic the actions (rational or irrational) of a larger group.

%

6

6

There are also studies of the role of foreign investors during the 1997 East Asian Crisis.Krugman (1998) argued that speculative capital flows are responsible for the virulence of thesecrises and their contagion effects; he suggested that imposing temporary capital controls could be aneffective way to help stabilize the economy during a crisis. Nofsinger and Sias (1999) and Stiglitz(1998) indicated that positive feedback trading by foreign investors can partially explain an excessof volatility in stock prices, driving markets away from fundamentals and destabilizing them. Lin(2006) showed that the strategies of foreign investors, such as feedback trading strategies, did affectTaiwanese stock market volatility. The crisis led to a breakdown of international financialtransactions and resulted in costly adjustments for the economies of emerging markets. Previousstudies seem to support the CBC’s claim about the instability caused by foreign investors.

2.2 The positive sideOne of the reasons that the government opened the market was that such a policy helps stabilize,

enlarging the domestic stock market and enhancing market efficiency. The Taiwan Securities andExchange Commission5 (TSEC) emphasized the advantage of foreign investors through, forexample, stabilization and demonstration effects. The TSEC encouraged the government toaccelerate the speed with which it opened the stock market to foreign investors. Wang and Shen(1999) suggested that foreign investors have only mild influence on the volatility of stock returnsand had a positive influence on exchange rate volatility during the process of Taiwan’s financial liberalization. Bekaert and Harvey (1997) studied 21 emerging countries and found that capitalmarket liberalizations significantly decreased market volatility in emerging countries, especially inTaiwan, Mexico, Brazil, and Argentina. Bohn and Tesar (1996) pointed out that foreign investorswere not the main cause of stock market instability. Hamao and Mei (2001) found no evidence tosupport the claim that trading by foreign investors tends to increase market volatility more thandomestic investors. De Long et al. (1990) offered an analysis of foreign investors’ behavior,suggesting that, though positive feedback trading is detectable, it did not necessarily exacerbate thedestabilization of the market.

Some studies even suggest that foreign investors did not cause instability during the Asianfinancial crisis period. Choe et al. (1999) showed that foreign equity investors did practiceherding during the Korean crisis period, but they found no convincing evidence that foreigninvestors played a destabilizing role in the stock market. Kim and Wei (2002) also suggested thatthe offshore funds in Korea are the wrong group to blame for exacerbating the volatility in the

5 The Taiwan Securities and Exchange Commission (TSEC), which was renamed the Taiwan Securities and FuturesCommission in 1997 and the Taiwan Securities and Futures Bureau in 2004. The operation of the TSEC is overseen bythe Securities & Futures Bureau (SFB), which has the exclusive authority to inspect the securities market operations andprocesses of the TSEC.

7

7

emerging markets. Karolyi (2002) pointed out that, though foreigners became net sellers in theperiod that Japan was suffering the Asian financial crisis, there is no evidence to confirm thatforeign investors destabilized Japanese markets. Wang (2007) suggested that increasing theinvestor-base as a result of liberalization actually decreased total return volatility. Because eachinvestor has only limited information and a set of available securities, foreign investors make theinvestor base larger, which fosters liquid information. Khanna and Palepu (1999) showed thatforeign investors are better monitors of corporate management than domestic institutions. Thesestudies all concluded that foreign investors were not the crucial factor in the instability of the stockmarket, which means that TSEC’s claims were also supported by scholars.

2.3 Why study Taiwan?From the above literature review, we can say the role of foreign investors varies from country to

country. As Johnson and Mitton (2003) stated, the benefits of free capital flows are still debatedamong policymakers and scholars. Corsetti et al (2001) pointed out that large institutionalinvestors exacerbated market pressure on smaller- and medium-sized emerging economies in thecrisis period, but large-sized markets were not affected. Thus, the effect of foreign investors onthe stock market and currency market is also controversial in different periods. Furthermore, theproportion of foreign investors in the stock market and the magnitude of foreign investors in theeconomic system are not the same in each country. Taiwan is a worthy country to study for thefollowing three reasons. First, thus far, it has been the severest financial crisis Taiwan hasencountered after the elimination of QFII system. Thus, it addresses a key question: does financialliberalization help stabilize the stock market and currency market? Second, the proportion of dailytrading volume of foreign investors (Figure 1) has doubled since the government abandoned theQFII system. Does this cause damage along with the increase of foreign holding? Finally, asTaiwan is a shallow economy, how influential could foreign investors be on the Taiwan stockmarket and exchange market now that their capital can flow without limitation?

We focus on the role of foreign investors through daily data from 2000 to 2009. Do ourempirical results reveal what foreign investors’role was during these years, and are there anydifferent effects between domestic institutions and foreign investors after the elimination of theQFII system or during the Financial Tsunami? Answering this question can give policymakers incountries with their own QFII system, such as China, a reference with which to decide whether tospeed up the capital market liberalization process.

2.4 Why a vector autoregressive model?According to data from the Taiwan Stock Exchange (TSE), in our sample period, we found that

foreign investors were net purchasers, with an average of 1,552 million purchases per day beforecrisis. After the crisis, they became net sellers, with an average of 1,888 million sales per day in

8

8

2008. However, foreign investors turned back into net purchasers, averaging 1,913 millionpurchases per day as the stock market boomed. We used a VAR model to examine the jointrelationship of stock market returns and currency market returns and foreign investors’ net flows. Sims (1980) invented the VAR model, which allows every included variable to be endogenous inorder to avoid setting too many inappropriate limitations on variables. Karolyi (2002) employed atri-variable VAR model, allowing market returns and currency returns to depend not only on pastreturns but also on past net flows of foreigners. The “net purchase”variable in this study washighly serially correlated. Thus, it also depends on past net purchases and past returns in order totest for foreign investors’feedback trading behavior. As Lin (2006) suggested, feedback trading offoreign investors can be detected in Taiwan by employing market returns and net purchase offoreign investors in the VAR model. Because of the characteristics of the VAR model, we adoptedit as the methodology in this study. Moreover, we use Granger causality to find the relationshipamong market returns, currency returns, and net purchases. We used impulse response function toanalyze the effect of a random shock on our variables. To be certain, variance decomposition isalso adopted to show the magnitude of the effect of other variables.

3. Data and methodology

Our main data source is the Taiwan Economic Journal (TEJ). The daily data comprise returns ofthe stock market and currency market and net purchases of foreign investors, investment trusts, andsecurity dealers. In calculating the stock market return, we use ln(Indext / Indext-1), which wasalso provided by the TEJ database. Currency return was calculated as -ln(ERt / ERt-1); we add aminus to adjust the value. Because of differences in European terms (TWD/$), when the value ofthe TWD appreciates, the exchange rate decreases. This calculation provides a more intuitive wayfor us to distinguish the change in value of the TWD. Net purchase was computed by purchasesminus sales. In this study, we divided the entire sample into three sub-samples: sub 1 representsthe period before the elimination of the QFII system, from January 4, 2000 to October 1, 2003; sub2 represents the period without crisis after the elimination of the QFII system, from October 2,20036 to July 18, 2007; and sub 3 represents the period since the crisis, from July 19, 20077 toDecember 31, 20098. The entire sample covers the period between January 4, 2000 to December31, 2009. That is, our sample covers four years before the elimination of the QFII system and fouryears after its elimination, with 948 and 941 trading days, respectively. This large sample allows usto find differences in the relatively long run. The crisis after the elimination of the QFII systemlasted 2 years, with 615 trading days.

6 October 2, 2003 was the day that the Taiwanese government abandoned the QFII system.7 July 19, 2007 was the day that Bear Stearns Companies’two funds collapsed (Bear Stearns High-Grade StructuredCredit Fund and Bear Stearns High-Grade Structured Credit Enhanced Leveraged Fund). The same day, the Dow JonesIndustrial Average Index climbed to 14,000 points. From that point, the U.S. stock market started entering a bearmarket.8 December 31, 2009 was the end-point for the available data we could collect.

9

9

-2,000

0

2,000

4,000

6,000

8,000

10,000

12,000

1/2000

7/2000

1/2001

7/2001

1/2002

7/2002

1/2003

7/2003

1/2004

7/2004

1/2005

7/2005

1/2006

7/2006

1/2007

7/2007

1/2008

7/2008

1/2009

7/2009

-800

0

800

1,600

2,400

3,200

4,000

4,800

We provide Figure 2 to show Taiwan’s weighted stock index and currency values for the entire sample; net purchases of foreign investors were also included. The time series figure helps us toobserve the trend of our variables across the entire sample period. During the crisis, the stockindex reached 9,000 in July 2007 and achieved its peak of 9,859.65 in October 2007. However,the index experienced sustained decline to below 4,500 in late 2008. TSE volume presentedsignificant decline along with market index. TWD/$ also showed dramatic depreciation, from itspeak of 30.1 to its bottom of 35.174. However, local policy9 should be considered to partly explainthis decrease. Cumulativeforeign investors’net purchases demonstrate selling strategies, a tradingstrategy inverse to the one previously used. Moreover, we find that net purchases of foreigninvestors tend to be more volatile than before, possibly suggesting that these foreign flows are fickleand speculative.

Figure 2 Daily foreign investment in Taiwan stock market: 2000-2009

Left axis (-2,000~12,000): Market index: Taiwan weighted stock index.Right axis (-800~4,800): Exchange rate: (TWD/$)*100. Cumulativeforeign investors’net purchases (billion, TWD).

Foreign investors’net purchases (100 million, TWD). TSE volume (100 million, TWD)Source: Taiwan Economic Journal

9 Similar to other export-oriented countries, the Taiwanese government intended to depreciate the TWD in order toboost export volume in the early crisis.

Cumulate FI’snet purchases

Market indexExchange rate

FI’s net purchases

TSE volume

10

10

3.1 MethodologyTo analyze the destabilizing effect of foreign net purchases on stock or currency returns, and to

determine whether foreigners used feedback trading, we adopt the VAR model, which employsmarket returns, currency returns and net purchases. Because there are strong positive correlationsamong variables and autocorrelation within variables themselves, it is difficult to determine theirrelationship. Sims (1980) invented the VAR model to solve these problems by establishing adynamic structural model. The VAR model is a structural model that enables the variables in thisstudy be driven by the lags of other variables. Using Schwarz criterion (SC) and evaluating lagstructure in the dynamic system, we determined that four lags10 were enough to capture most of theeffects of institutional investors’net purchases. Let market returns, m; currency returns, x; netpurchases, n; then the tri-variate equations can be written as follows:

ztitiitititttt

xtitiitiititttt

ytitiitiititttt

enBxBmBnBxBmBCn

enBxBmBnBxBmBCx

enBxBmBnBxBmBCm

33324

311

3311

3211

31130

2322211

2311

2211

21120

1312111

1311

1211

11110

...

...

...

…(1)

To make it easier to read, Eq. (1) can be derived as follows:

)E(, 'L

1stttstst eeeYy

…(2)

Where yt is a 3×1 vector of daily observations of market returns, currency returns and netpurchases at time t; αis a 3×1 vector of constant parameters; βs is a 3×3 matrix of parameters attime s. L represents the lags for the model and μt is a 3×1 vector of errors of yt given all the pasty’s. The elements of βs measure the direct effect that a change in the return on the j-th variablewould have on the i-th variable in s days. Each investor’s net purchases series was used toestimate this model.

According to Sims (1980), the coefficients in the VAR model do not effectively reveal economicimplications. Three important applications were developed to interpret the meanings of the VARresults. Using a causality test, an impulse response analysis, and variance decomposition makesthe VAR model more readable and meaningful in its analysis. A Granger causality test was used todetermine whether the lags of past returns or purchases are jointly significant predictors of futurereturns or purchases. We tested market returns, currency returns, and net purchases of eachinstitutional investor. Eq. (1) can be written in the following form:

10 SC and AIC suggested one lag is the most parsimonious model. However, one lag is not enough for our purposes.We then tested our model from one lag to ten lags and found that four lags could reflect most of the information that wewant. Lags above four only needlessly extend the model with lags carrying less information.

11

11

zt

k

iiti

k

iiti

k

iitit

xt

k

iiti

k

iiti

k

iitit

yt

k

iiti

k

iiti

k

iitit

enGxGmGCn

enGxGmGCx

enGxGmGCm

1

33

1

32

1

3130

1

23

1

22

1

2120

1

13

1

12

1

1110

…(3)

Where G is the set of all coefficients of a certain variable. For example, x Granger causes m ifand only if the coefficients in G12 are not all equal to zero, which rejects the null hypothesis ofGranger causality that G = 0. There are two types of causality: unidirectional causality andbidirectional causality. In the former, one variable Granger causes the other variable. In the latter,two variables Granger cause each other.

To trace the cross-equation feedbacks through Eq. (2), an alternative method is to use themultiplier analysis or innovative accounting technique based on the moving average. The movingaverage form can be written as follows:

0sstst ey …(4)

Where yt is a linear combination of current and past one-step-ahead forecast innovations. The(i,j)-th element of matrix Φs measures the response of the i-th variable in s days to a unit randomshock in the j-th variable and none in the other variables. However, the vector of et is seriallycorrelated, a serious flaw because we would like to determine the distinct response patterns of theVAR model. Thus, we adopted Choleski factorization. The Σεcan be decomposed into VV’byusing Choleski factorization to remove the problem we faced. The orthogonalizing transformationproceeds as follows:

''')( 2/12/1' VVADADADAeeE tt …(5)

Where A is a square matrix of the sample covariance matrix Σε, D is a diagonal matrix, and V is anonsingular, lower triangular matrix with positive elements on the diagonal. Then, Eq. (4) can berewritten as follows:

tttst eVkkCy 1

0s

,

…(6)

12

12

Where Cs is a square matrix, the (i,j)-th elements of Cs stands for the impulse response of the i-thvariable in s periods to a shock of one standard error in the j-th variable. The series of kt is seriallyuncorrelated error. We plot the impulse responses and 95% confidence intervals using standarderrors from the Monte Carlo integration technique. Through Choleski factorization, the movingaverage form of Eq. (4) can be used to allocate the forecast variance of each element in y todifferent sources of shocks. The error of the optimal T+1 ahead forecast is:

1

0h

2,

T

hijt Ry…(7)

The element of forecast error variance in the T+1 step ahead forecast of yi, which is accountedfor by innovations in yj. That is, R2

ij,h represents the contribution by innovation in the j-th variableto variance of the i-th variable at time h. Variance decomposition analysis provides a measure ofthe overall relative importance of the markets in generating the fluctuations in daily net purchasesfrom past net flows, stock market returns and currency returns. We report the percentage of totalforecast error variance for stock market returns, TWD/dollar exchange rates returns and netpurchases up to 5, 10 and 15 days ahead accounted for by innovations of past market returns,exchange rates returns and net purchases.

4. Empirical results

4.1 Summary statisticsTable 2 presents summary statistics for market returns, currency returns, and net purchases by

each investor group during the QFII, QFII-eliminated, crisis, and entire sample. Over the entireperiod, foreign investors were the most aggressive net purchasers of Taiwanese equity, at 1,000million per day. Foreign investors’ netpurchases increased in sub 2 compared with sub 1,indicating that deregulation was successful at attracting international capital flows (from 9.57 to21.5). However, the shift to negative net purchases by foreign investors and negative marketreturns suggests that the expectation that foreign investors tend to sell in crisis is true. Domesticinstitutions did not show the same pattern as foreign investors. Investment trusts revealed oppositestrategies from foreign investors across the entire sample. Securities dealers apparently had theirown approach. However, both of them became net purchasers during the crisis. Comparing sub1 with sub 2, we found that, as the investor base broadened and information liquidation increased,the standard deviation of stock market returns decreased following liberalization. This result11

met the government’s goal for opening the stock market. Though the standard deviation ofcurrency returns slightly increased, the change is not strongly statistically significant (10% level).The result for the currency market suggested that the CBC was too concerned with the

11 The standard deviation of the stock index significantly decreased from sub 1 (52.97) to sub 2 (26.33). The currencymarket also benefited from liberalization, as the standard deviation of the exchange rate significantly decreased fromsub 1 (0.051) to sub 2 (0.026). However, both eventually rose to 62.99 and 0.044, respectively.

13

13

consequences of deregulation. It is also possible that the CBC still has a visible hand in the currencymarket.

However, when the U.S. subprime mortgage crunch tolled the bell for the crisis in late 2007, thebloom of deregulation started fading in Taiwan. The standard deviation of market returnsincreased to the same level as it had been before. Has deregulation helped stabilize these twomarkets? Or are there other reasons that offset the advantages of liberalization? The results of thisstudy could reveal the answers by using our analytical methods.

Table 2. Summary statistics

Entire sample (2000/01/04~2009/12/31) Mean S.D. Maximum Minimum Trading daysMarket returns -0.001252 1.620951 6.5246 -6.9123 2504TWD/$ returns -0.000664 0.24679 2.898567 -2.956315 2504Foreign Investors’ net purchases(100m) 10.68501 72.81381 1252.8812 -623.9813 2504Investment Trusts’ net purchases(100m) -0.031478 11.97509 64.53 -88.76 2504Securities Dealers’ net purchases(100m) 0.204413 11.6345 63.79 -111.77 2504

Sub 1 : QFII (2000/01/04~2003/10/01) Mean S.D. Maximum Minimum Trading daysMarket returns -0.043727 1.88326 6.1721 -6.7745 948TWD/$ returns -0.007458 0.227434 2.898567 -2.956315 948Foreign Investors’ net purchases(100m) 9.574473 33.79942 175.82 -154.6 948Investment Trusts’ net purchases(100m) -0.256983 13.09837 64.53 -60.13 948Securities Dealers’ net purchases(100m) -0.38192 9.079974 42.19 -38.31 948

Sub 2 : QFII-eliminated (2003/10/02~2007/07/18) Mean S.D. Maximum Minimum Trading daysMarket returns 0.056351 1.101084 5.4189 -6.9123 941TWD/$ returns 0.003155 0.241843 1.006908 -1.006487 941Foreign Investors’ net purchases(100m) 21.50359 79.07887 1252.88 -549.37 941Investment Trusts’ net purchases(100m) -1.102221 10.10109 52.72 -88.76 941Securities Dealers’ net purchases(100m) 0.349458 13.96242 63.79 -111.77 941

Sub 3 : In the crisis (2007/07/19~2009/12/31) Mean S.D. Maximum Minimum Trading daysMarket returns -0.023915 1.837494 6.5246 -6.7351 615TWD/$ returns 0.003967 0.280795 1.500921 -1.222136 615Foreign Investors’ net purchases(100m) -4.156455 99.37631 579.0914 -623.98 615Investment Trusts’ net purchases(100m) 1.954455 12.56904 50.01 -49.39 615Security Dealers’ net purchases(100m) 0.886293 11.18761 41.58 -47.8 615

Source: Taiwan Economic Journal

Table 3 shows the correlation coefficients of daily net purchases for each of the investmentsectors, with market returns and currency returns. Foreign investors shared the same investmentbehavior as domestic institutional investors; the correlation coefficients between net purchases offoreign investors and market returns, net purchases of investment trusts and market returns, and net

12 The largest net purchases of foreign investors, due to block trading when Philip Taiwan transferred all of its TSMCequity to its headquarters, Philip Deutschland. (December 28, 2005)13 The largest net sales of foreign investors caused by Dow Jones slump. (July 27, 2007)14 The virtually largest net purchases of foreign investors triggered by Ma’s victory in presidential election. (March 22,2008)

14

14

purchases of securities dealers and market returns are all strongly positive. The result is consistentwith the correlation table in Lin (2006), which uses daily data from 1994 to 2003. The significantpositive correlation coefficients among institutional investors suggest that herding behavior mayexist in the stock market, similar to Kim and Wei’s(1999) and Kaminsky and Schmukler’s(1999)findings. There is a strong positive relation between each institutional investors’trading activity.Nevertheless, foreign investors showed increasing magnitude in both stock and currency marketsduring the crisis while domestic institutional investors’magnitude diminished in the stock market.We also found that the correlation coefficients between market returns and currency returns arehigher since deregulation, which means the frequent inflow and outflow of international capitalflows enhanced the relationship between the stock market and the exchange market.

Table 3. Correlation coefficientsMarketreturns

TWD/$returns

ForeignInvestors

InvestmentTrust

Entire sampleMarket returns 1.000TWD/$ returns 0.276 *** 1.000Foreign Investors’ net 0.401 *** 0.270 *** 1.000Investment Trusts’ net 0.397 *** 0.101 *** 0.175 *** 1.000Securities Dealers’ net 0.565 *** 0.161 *** 0.354 *** 0.406 ***

Sub 1 : QFIIMarket returns 1.000TWD/$ returns 0.183 *** 1.000Foreign Investors’ net 0.362 *** 0.170 *** 1.000Investment Trusts’ net 0.465 *** 0.087 *** 0.280 *** 1.000Securities Dealers’ net 0.584 *** 0.150 *** 0.364 *** 0.496 ***

Sub 2 : QFII-eliminatedMarket returns 1.000TWD/$ returns 0.301 *** 1.000Foreign Investors’ net 0.395 *** 0.259 *** 1.000Investment Trusts’ net 0.373 *** 0.082 ** 0.162 *** 1.000Securities Dealers’ net 0.694 *** 0.202 *** 0.346 *** 0.396 ***

Sub 3 : In the crisisMarket returns 1.000TWD/$ returns 0.400 *** 1.000Foreign Investors’ net 0.565 *** 0.364 *** 1.000Investment Trusts’ net 0.325 *** 0.144 *** 0.224 *** 1.000Securities Dealers’ net 0.568 *** 0.111 *** 0.423 *** 0.374 ***

Note: When we adjusted the calculation of currency return to–ln(ERt/ERt-1), the correlation became positive.*** = 1% significance level; ** = 5% significance level; * = 10% significance levelSource: Taiwan Economic Journal

The correlation coefficients between market returns and currency returns reached their highest inthe crisis period compared to the noncrisis period. Similarly, the correlation coefficients between netpurchases of foreign investors and currency returns increased following deregulation. Domestic

15

15

institutions showed lower coefficients, which suggests lower demand in exchange than for foreigninvestors. We assume that domestic institutions would have less effect than foreign investors onthe exchange market, no matter the status of the QFII system.

4.2 VAR model resultsBefore we implement VAR in this study, we have to check whether the VAR model is suitable.

Unit root tests were used to ensure that all the series that we employed are stationary. Anaugmented Dickey-Fuller test and a Kwiatkowski-Phillips-Schmidt-Shin test were adopted. AsTable 4 shows, each variable series we employed is stationary. In this case, we do not have toimplement a Vector Error Correction Model (VECE). The VAR is good enough for us to analyzethe phenomena in Taiwan. Table 5 reports the estimation results by the VAR model for each offour different sectors in three periods. For each sector, the first four coefficients represent thelagged market returns, currency returns and net purchases. The Granger causality test shows thatone variable can be forecasted by the lags of the other variable. In this section, we focus on thedifference between sub 1 and sub 2 first. Then, we compare the changes in sub 2 with sub 3.Last, we will provide a summary of our overall empirical results.

Table 4. Unit root test

ADF KPSSVariable T-Stat. Critical value (1%) LM-Stat. Critical value (10%)

Market returns -47.369*** -3.433 0.179 0.347TWD/$ returns -46.859*** -3.433 0.115 0.347

Foreign Investors’ net purchases -20.195*** -3.433 0.185 0.347Investment Trusts’ net purchases -33.018*** -3.433 0.337 0.347Securities Dealers’ net purchases -34.467*** -3.433 0.462* 0.347

Series of securities dealers’net purchases did not reject the null hypothesis at a 5% significant level in the KPSS test.(critical value: 0.463).

4.2.1 Sub 1 vs. Sub 2The results show that there is a little evidence to support the idea that foreign investors’role in

the stock market has changed following the elimination of the QFII system (see Table 5.a and Table5.b). The first-lag coefficients of foreign investors’net purchases in the two periods for marketreturns, 0.0065 and 0.0017 respectively, are significant and positive. There is only a slight changebecause all the coefficients of foreign investors’purchases for market returns in sub 2 haddiminished. Foreign investors still adopted positive feedback trading strategies; most coefficientsof market returns for net purchases are positive and include statistically significant positivecoefficients (3.556 in sub 1 and 15.191, 5.576 in sub 2). A Granger causality test also showed thatnet purchases are affected by market returns; the chi-square values are 47.426 and 44.771 in the twoperiods. Information contributions were also detected for net purchases of foreign investors.Granger caused market returns (11.659 and 9.441, respectively) with bidirectional causality, which

16

16

indicates future net purchases have an impact on market returns and vice-versa. The results showonly a small difference after the QFII system was eliminated. We thought that the step-by-stepliberalization process effectively avoided drastic changes. However, net purchases of foreigninvestors Granger caused currency returns from an insignificant 6.623 to a significant 20.029,indicating that, as the volume of international capital flows increases, foreigners gained greaterimpact on currency market.

In contrast, domestic institutions became less influential on the stock market. Bothinvestment trusts and securities dealers were more powerful than foreigners in sub 1, as table 5.ashows. Domestic institutions’first-lag and third-lag coefficients were significant: 0.013 and -0.014for investment trusts 0.027 and -0.013 for securities dealers. However, their influence diminishedto only one significant coefficient: first-lag 0.009 for investment trusts and third-lag -0.006 forsecurities dealers (see table 5.b). Investment trusts revealed themselves to be negative feedbacktraders in two periods, as most of the coefficients are negative and were Granger caused by marketreturns (89.045 and 173.474). However, they had only a bidirectional causality with market returnsin sub 1. In other words, feedback trading and information contribution were both detected only insub 1. We found that investment trusts tended to sell stock when the TWD appreciated in twoperiods, though the coefficients are not statistically significant. Securities dealers turned intopositive feedback traders in sub 2; the first-lag coefficient (2.127) and third-lag coefficient (1.048)are significant and positive, with an inverse causality from 1.952 to 19.112. We also found thatcurrency returns had an increasing impact on securities dealers in sub 2 for the significant first-lagcoefficient (3.393) and third-lag coefficient (4.315) and Granger caused net purchases of securitiesdealers. The results suggest that security dealers tended to purchase stock when the TWDappreciated, unlike investment trusts.

The results show that all institutional investors were feedback traders in sub 2, consistent withFroot et al. (2001). We found no strong evidence that foreign institutional investors destabilized thestock market more than domestic institutions in sub 1 and sub 2. However, we did find that foreigninstitutional investors have had an increasing impact on the currency market since the elimination ofthe QFII system. Though domestic institutions once had greater influence, it disappeared afterderegulation. Liberalization was beneficial to Taiwan in the noncrisis period.

4.2.2 Sub 2 vs. Sub 3The role of foreign investors changed during the crisis; they demonstrated a destabilizing effect

on stock market, as table 5.c shows. The coefficients of foreign investors’net purchases formarket returns became stronger than ever, from a first-lag coefficient of only 0.002 in sub 2 to asecond-lag coefficient of -0.002 and a third coefficient of 0.003 in sub 3. Foreign investors’ positive feedback trading disappeared. The first-lag coefficient (3.505) of market returns to net

17

17

purchases of foreign investors became less significant, and this time only foreign investors Grangercaused market returns, a unidirectional causality (11.067). In other words, informationcontribution remains. This result shows that market returns were not able to predict net purchases offoreign investors but that net purchases of foreign investors still predicted market returns across theentire sample. The results were opposite those of Lin (2006), who found the existence of foreigninvestors’feedback trading behavior but no information contribution in Taiwan during the Asiancrisis. Apparently, the magnitude of shock and the degree of regulation on foreign investors maylead to different consequences in crisis.

The influence of currency returns on net purchases of foreign investors has risen in the crisis;the first-lag (21.663), third-lag (-31.881), and fourth-lag (29.523) coefficients of currency returns onthe net purchases of foreign investors are statistically significant. However, we cannot tellwhether foreign investors prefer buying stock when the TWD appreciates or depreciates because thesigns were both positive and negative. The causality between currency returns and foreigninvestors’net purchases has changed its direction. In sub 2, net purchases of foreign investorsGranger caused currency returns. In sub 3, currency returns Granger caused the net purchases offoreign investors. It is also worth mentioning that we found a bidirectional causality betweencurrency returns and market returns, which always happens alongside a financial crisis like theAsian crisis. This result is consistent with Granger et al. (2000), indicating that the impact ofcurrency in crisis in a flatter economy was exacerbated by outside shocks.

Domestic institutions also changed their behavior and showed asymmetry in the crisis:investment trusts and securities dealers embraced different strategies. Investment trusts still holdon to negative feedback trading strategies, with significant negative coefficients and net purchasesGranger caused by market returns (85.08). Still, we found no information contribution in sub 3.In other words, only a unidirectional causality is detectable since the elimination of the QFII system.Investment trusts still tend to sell their stock holdings when TWD appreciates. Securities dealersreversed their trading behavior again; feedback trading strategies disappeared in sub 3. For thecoefficients of market returns to net purchases of security dealers, only the second-lag coefficient,(-0.441) is significant, and net purchases were not Granger caused by market returns (3.157). Asfor investment trusts, information contribution was not found after deregulation—the only patternthey shared in sub 3. Securities dealers’preference for purchasing stock according to change inthe exchange rate was unclear. The coefficients showed no significance and Granger causalityalso vanished. One thing worth to noticing is that securities dealers had a more slight destabilizingeffect on market returns than foreign investors (for the third-lag, 0.018 and for the fourth-lag, -0.011;coefficients are significant.)

The results show the differences between foreign investors’and domestic institutions’

18

18

behavior and impact in the crisis. Foreign investors had a stronger impact than domesticinstitutions on the stock market and were affected by currency market more (see table 5.c). It isworth pointing out that securities dealers also had a slight effect on the stock market regardless ofthe small volume they trade in sub 3. Investment trusts are less likely to be destabilizing thanthose for foreign investors and securities dealers. Finally, compared with foreign investors,domestic institutions showed a weaker relationship with the currency market.

4.2.3 Summary of VAR analysisIn this section, we found that, despite the small scale (see table 1), securities dealers also

demonstrated some destabilizing effects similar to foreign investors. The asymmetry in scale ofthe two institution investors resulted in similar consequences, suggesting that the amount of capitalflows does not necessary stand for the influence on the stock market. Overall, the elimination ofthe QFII system does benefit Taiwan, as is evident in table 2 and from the VAR results. Thedestabilizing effects of institutional investors were also reduced, as shown in table 5.b. Foreigninvestors were feedback traders until the crisis hit Taiwan, but their net purchases constantly affectthe next day of stock market returns across the entire sample. These findings were different fromthose for domestic institutional investors: investment trusts were negative feedback traders andsecurities dealers frequently changed strategies. The results of positive feedback traders for foreigninvestors and negative feedback traders for investment trusts are similar to Karolyi’s(2002)findings in Japan. For domestic institutions, the information contribution effects have disappearedsince deregulation. The currency market was influenced only by foreign investors; domesticinstitutional investors showed no effect on it. In the VAR results, we could say that foreigninvestors still have greater influence, as their stock holding has increased and restrictions have beenremoved.

4.3 Impulse response function analysisIn this section, we discuss relations among market returns, currency returns and institutional

investors. Then the results will be compared with the VAR results and the Granger causality test toensure the robustness of the study. The impulse responses in each plot are the coefficients fromthe former moving average function in Eq. (3). A Monte Carlo integration procedure is used todisplay the 95% confidence intervals around the impulse response coefficients. From figure 3, wecan see that each variable exhibited strong autocorrelation patterns. In short, the VAR model isappropriate for this study. During the crisis (figure 3.c), relations between variables wereamplified and revealed different, discontinuous movement compared with previous periods (figures3.a, 3.b), showing the impact of the crisis shock.

First, we check the relation between market returns and net purchases of foreign investors (FI) inthree periods. The response of FI to market returns (bottom-left) is affected by market returns

19

19

from the beginning and the existence of positive feedback trading strategies. The response ofmarket returns to FI (top-right) was suppressed in sub 2, indicating that deregulation did reduceforeign investors’impact on the stock market. However, as we mentioned before, benefits ofderegulation

20

20

Table 5.a QFII (2000/01/04~2003/10/01)

Foreign Investors Investment Trusts Securities Dealers

Market

return

TWD/$

return

Net

purchase

Market

return

TWD/$

return

Net

purchase

Market

return

TWD/$

return

Net

purchase

Market -1 0.0143 0.0070 ** 3.5562 *** 0.0126 0.0092 ** 1.2439 *** -0.0182 0.0084 ** 0.1757

return -2 -0.0084 0.0020 0.1386 0.0179 0.0066 * -1.5846 *** -0.0028 0.0067 * -0.1566

-3 0.0519 * 0.0012 0.3975 0.1196 *** 0.0050 -0.6843 *** 0.0822 ** -0.0007 -0.0471

-4 -0.0657 ** -0.0037 -0.7841 * -0.0431 -0.0020 -0.6763 *** -0.0645 * -0.0044 -0.0484

TWD/$ -1 0.1793 0.0194 -5.6372 0.2532 0.0262 -0.6926 0.2614 0.0271 1.2959

return -2 0.0505 0.0409 4.6297 0.0565 0.0424 1.1745 0.0438 0.0414 1.0790

-3 0.2383 -0.0032 -2.3405 0.2026 -0.0022 -0.4433 0.2125 0.0008 -1.4668

-4 0.4355 * 0.0565 ** 1.1553 0.4344 * 0.0551 ** -0.8050 0.4307 * 0.0577 ** -0.0715

Net -1 0.0065 *** 0.0004 * 0.3800 *** 0.0132 ** 0.0002 0.2623 *** 0.0267 *** 0.0005 0.4349 ***

purchase -2 0.0021 0.0003 0.1173 *** -0.0014 -0.0003 0.0984 *** 0.0030 -0.0008 0.0266

-3 -0.0033 0.0000 -0.0225 -0.0140 *** -0.0000 -0.0427 -0.0132 * 0.0021 -0.0357

-4 0.0004 -0.0001 0.0947 *** 0.0054 -0.0003 0.0175 0.0066 -0.0002 0.0183

Constant -0.0961 * -0.0144 ** 4.3273 *** -0.0340 -0.0078 -0.2468 -0.0292 -0.0073 -0.2821

Adj. R-square 0.0287 0.0295 0.3260 0.0266 0.0232 0.2019 0.0273 0.0268 0.2130

Granger causality tests (Chi-sq)

Market return 4.3096 47.4261 *** 8.8498 ** 89.0449 *** 6.0309 1.9522

TWD/$ return 3.6549 3.0221 3.8571 0.6885 3.7996 2.9458

Net purchase 11.6588 ** 6.6525 9.6217 ** 0.5967 10.2849 ** 4.0444

*** = 1 % significance level; ** = 5 % significance level; * = 10 % significance level.

21

21

Table 5.b QFII eliminated (2003/10/02~2007/07/18)


Market

return

TWD/$

return

Net

purchase

Market

return

TWD/$

return

Net

purchase

Market

return

TWD/$

return

Net

purchase

Market -1 -0.0264 0.0001 15.1911 *** -0.0156 0.0112 * 1.8671 *** -0.0107 0.0196 ** 2.1265 ***

return -2 -0.1051 *** -0.0069 1.9243 -0.0970 *** 0.0082 -2.7302 *** -0.0755 * 0.0174 * -0.5593

-3 0.0533 -0.0040 5.5761 ** 0.0780 ** 0.0074 -1.1261 *** 0.1202 *** 0.0085 1.0480 **

-4 -0.0955 *** -0.0132 * -2.3418 -0.0585 * -0.0009 -1.1512 *** -0.0831 ** -0.0015 -0.4902

TWD/$ -1 0.3777 *** 0.0300 7.7828 0.4496 *** 0.0525 ** -0.3479 0.4342 *** 0.0516 * 3.3925 **

return -2 0.1070 -0.0159 -1.6031 0.1290 -0.0065 -0.0712 0.1065 -0.0054 4.3148 ***

-3 -0.0492 -0.0012 1.7061 -0.0324 0.0057 1.4687 -0.0592 0.0121 -0.3222

-4 -0.0415 0.0232 -0.5879 -0.0308 0.0324 -1.0286 -0.0195 0.0402 -1.0374

Net -1 0.0017 *** 0.0004 *** 0.2142 *** 0.0088 ** 0.0000 0.3728 *** 0.0031 -0.0010 0.2162 ***

purchase -2 -0.0002 0.0001 0.1037 *** -0.0016 -0.0009 0.1276 *** -0.0013 -0.0010 0.0434

-3 -0.0001 0.0001 0.0902 *** -0.0002 0.0000 0.1299 *** -0.0062 * 0.0001 -0.0492

-4 0.0000 0.0001 0.0516 * -0.0044 -0.0009 0.0544 ** 0.0031 -0.0001 0.0018

Constant 0.0287 -0.0110 * 10.2701 *** 0.0576 ** -0.0008 -0.1610 0.0531 * 0.0011 0.1138

Adj. R-square 0.0218 0.0180 0.2197 0.0287 0.0125 0.3628 0.0150 -0.0001 0.1471


Market return 3.1426 44.7707 *** 3.1311 173.4744 *** 5.0373 19.1121 ***

TWD/$ return 6.5057 0.6475 9.2759 ** 2.4499 8.5292 * 9.4145 **

Net purchase 9.4411 ** 20.0294 *** 4.0268 2.7562 3.0706 2.9291


22

22

Table 5.c In the crisis (2007/07/19~2009/12/31)


Market

return

TWD/$

return

Net

purchase

Market

return

TWD/$

return

Net

purchase

Market

return

TWD/$

return

Net

purchase

Market -1 -0.0269 -0.0191 *** 3.5025 * -0.0104 -0.0126 ** 1.5992 *** -0.0367 -0.0138 ** 0.2779

return -2 0.0692 * 0.0119 * 4.5356 ** 0.0176 0.0100 * -1.4606 *** 0.0465 0.0143 ** -0.4405 *

-3 -0.1083 ** -0.0038 -0.4330 -0.0629 -0.0033 -1.2806 *** -0.1008 ** -0.0061 -0.1730

-4 -0.0683 * -0.0037 -1.0219 -0.0518 0.0010 -0.3070 -0.0356 0.0014 0.2208

TWD/$ -1 0.8985 *** 0.2184 *** 21.6633 * 0.9260 *** 0.2301 *** -0.1739 0.9966 *** 0.2382 *** 0.1314

return -2 0.3722 -0.0068 -5.0908 0.3439 -0.0063 -0.6050 0.3084 -0.0069 1.1425

-3 -0.315 -0.0529 -31.8805 ** -0.1620 -0.0468 -0.5888 -0.0931 -0.0410 -0.2468

-4 0.451 * 0.0185 29.5232 ** 0.4500 ** 0.0172 1.4738 0.4289 * 0.0150 -1.2500

Net -1 0.0010 0.0003 ** 0.3483 *** 0.0008 0.0006 0.2470 0.0090 0.0008 0.2282 ***

purchase -2 -0.0021 ** -0.0001 -0.0137 0.0029 0.0013 0.2317 *** -0.0091 -0.0004 0.0779 *

-3 0.0032 *** 0.0001 0.0914 ** 0.0076 0.0003 0.0603 * 0.0183 ** 0.0017 * 0.1099 **

-4 -0.0007 0.0000 -0.0273 -0.0077 -0.0004 0.0745 ** -0.0109 * -0.0010 -0.0799 *

Constant -0.0298 0.0044 -2.6192 -0.0423 -0.0004 0.6101 * -0.0426 0.0021 0.5817 *

Adj. R-square 0.0293 0.0421 0.1874 0.0150 0.0397 0.2770 0.0228 0.0390 0.0717


Market return 8.7956 * 5.1293 5.8018 85.0797 *** 6.8182 3.1567

TWD/$ return 14.6874 *** 9.8266 ** 14.6900 *** 0.9764 15.4486 *** 0.9865

Net purchase 11.0673 ** 4.3822 2.2310 2.8486 6.9999 2.4556


23

23

faded in the crisis period; the response of market returns to FI became larger and fluctuated more asGranger causality turned to unidirectional and only FI Granger caused market returns.

Second, we examined market returns and currency returns. The response of market returns tocurrency returns (top-center) decreased in sub 2, and the impact shortened to the 3rd day. As theplot shows, it became smoother than in sub 1. But, the impact turned more volatile in sub 3. Theresponse of currency returns to market returns (middle-left) increased only in the first day yetbecame smoother in sub 2, compared with sub 1. However, the response of currency returns tomarket returns also fluctuated more and became discontinuous during the crisis. Both of theresults showed that the relationship between market returns and currency returns became volatile incrisis and are consistent with the bidirectional causality in the Granger causality test.

Currency returns and net purchases of foreign investors were checked last. The response of FI tocurrency returns (bottom-center) gradually grew larger over time, especially in crisis-time, ascurrency returns Granger caused FI in sub 3. The response of currency returns to FI (middle-tight)had become larger since deregulation, as FI Granger caused currency returns and began tofluctuated more in sub 3.

For domestic institutions, net purchases of investment trusts (IT) and net purchases of securitiesdealers (SD) were put into one figure and plotted for three periods as figures 4.a, 4.b, and 4.c. Wedid so because of the similar scope of the two sectors and because doing so made it easier tocompare the differences between them. The response of IT to market returns (bottom-left)exhibited negative feedback trading strategies for IT; the coefficients went below 0 in two days andstayed negative for several days. The response of SD to market returns (bottom-left) had reachedits peak in sub 2 (figure 4.b), only became feedback traders supported SD in this period. We foundthat the magnitude of the response of market returns to IT (top left) had decreased after deregulation,as for SD. Though their magnitude to market returns increased in the crisis (figure 4.c), the impactwas smaller than in sub 1.

The response of IT to currency returns (bottom-left) showed negative coefficients, just as wefound with VAR that IT tended to sell stock when the exchange rate rose. The response of SD tocurrency returns, however, presented different patterns. The coefficients were larger than IT’s andchanged through time. The magnitude was around zero in sub 1, became positive in sub 2, andwent negative in sub 3. This result supports the claim that SD were Granger caused by currencyreturns in sub 2, as in the VAR results. The response of currency returns to IT and SD (top-right)showed little impact in sub 3 (figure 4.c), also similar to the VAR results.

To summarize the results of the impulse response analysis, the elimination of the QFII system

24

24

reduced not only foreign investors’but also domestic institutions’influence on the stock market.Yet, it also increased the

25

25

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14

Response of MARKET to MARKET

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14

Response of MARKET to CURRENCY

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14

Response of MARKET to FI

-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14

Response of CURRENCY to MARKET

-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14

Response of CURRENCY to CURRENCY

-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14

Response of CURRENCY to FI

-20

0

20

40

60

80

2 4 6 8 10 12 14

Response of FI to MARKET

-20

0

20

40

60

80

2 4 6 8 10 12 14

Response of FI to CURRENCY

-20

0

20

40

60

80

2 4 6 8 10 12 14

Response of FI to FI

Figure 3.a Sub 1, QFIIinfluence of the currency market, especially when Taiwan was hit by a crisis. Foreign investorshad a greater impact than locals, but were not that powerful compared with their scope. Most ofthe results of the impulse response analysis are consistent with the VAR model, suggesting that theresults of this study are robust.

4.4 Variance decomposition analysisVariance decomposition analysis reveals how much of the forecast error variance of one variable

can be explained by exogenous shocks to the other variables. In other words, it reveals the amountof information that one variable contributes to the other in a VAR model. Table 6 shows the resultsof variance decomposition of the three periods in entire sample period. Because there are fivetrading days per week in Taiwan, we adopted 5 days (1 week), 10 days and 15 days ahead for eachvariable. On average, 30% of variance for institutional investors can be associated with innovationin the market returns. In detail, the proportion of net purchases of foreign investors by market

26

26

returns

27

27

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-20

0

20

40

60

80

2 4 6 8 10 12 14


-20

0

20

40

60

80

2 4 6 8 10 12 14


-20

0

20

40

60

80

2 4 6 8 10 12 14


Figure 3.b Sub 2, QFII eliminatedincreased gradually and soared to approximately 40% during the crisis (table 6.c). However, theproportion of net purchases of investment trusts by market returns decreased gradually from 30% to20%. The proportion of net purchases of security dealers, showed a distinct track: an inverseU-shape. In sub 2, the proportion reached 60%, but it was 30% and 35% in sub 1 and sub 2,respectively. This inverse U-shape track can be explained in the previous VAR analysis, whichshows that security dealers turned into feedback traders in sub 2. On the contrary, all institutionalinvestors’effects on market returns had shrank in sub 2, but rose again in sub 3, showing that therewere benefits of deregulation. The proportions of institutional investors by currency returns alsorose following the elimination of the QFII system, especially for foreign investors and securitydealers, with 3-fold and 5-fold increases in sub 2. Moreover, net purchases of foreign investorsand currency returns presented a growth-and-decline relation, which was similar to the directionalchange in Granger causality. In the crisis period, the proportion of currency returns by marketreturns had increased to its peak of 14%. The proportion of market returns by currency returns also

28

28

increased, doubling compared to sub 2.

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-20

0

20

40

60

80

2 4 6 8 10 12 14


-20

0

20

40

60

80

2 4 6 8 10 12 14


-20

0

20

40

60

80

2 4 6 8 10 12 14


Figure 3.c Sub 3, In the crisisThe results of the innovation of market returns on net purchases of foreign investors and currency

returns are similar to Karolyi (2002). Karolyi (2002) found that the proportions of net purchasesof foreign investors and currency returns by market returns went up in Japan during the Asian crisis.Investment trusts and securities dealers exhibited different patterns in this analysis as well. Theresults of this analysis coincide with the VAR model and impulse response analysis, indicating therobustness of this study.

5. Conclusion

In this study, we investigated market returns, currency returns and net purchases of institutionalinvestors before QFII was eliminated, after QFII was eliminated and in the crisis. We used a VARmodel and then impulse response analysis and variance decomposition to ensure robustness. Ourhypotheses were tested and stated as follows: (1) Foreign investors do have stabilizing effect on

29

29

Taiwanese stock market, except during the crisis period. (2) There is no strong evidence to prove

30

30

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14

Response of MARKET to CholeskyOne Standard Deviation Innovations

-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14

Response of CURRENCY to CholeskyOne Standard Deviation Innovations

-2

0

2

4

6

8

10

2 4 6 8 10 12 14

Response of IT to CholeskyOne Standard Deviation Innovations

-2

0

2

4

6

8

10

2 4 6 8 10 12 14

MARKET CURRENCYIT SD

Response of SD to CholeskyOne Standard Deviation Innovations

Figure 4.a Sub 1, QFII (domestic institutions)that foreign investors destabilized the currency market, but their impact on currency market didincrease. (3) Compared with domestic institutions, foreign investors were slightly stronger thansecurity dealers and investment trusts on the Taiwanese stock market during the financial tsunami.(4) The role of foreign investors has changed twice after the elimination of the QFII system. Onechange occurred following deregulation, when the impact on currency market increased; the otherchange occurred during the financial tsunami, when a destabilizing effect was detected on the stockmarket. However, we should note that domestic institutions also had the ability to destabilize thestock market before deregulation. Securities dealers even cast a destabilizing effect during the crisisperiod, similar to foreign investors. The destabilizing effect of foreign investors only becamestronger in the crisis. Foreign investors adopted feedback trading strategies until the crisis hit

31

31

Taiwan, but they still have a greater impact than domestic institutions on

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

2 4 6 8 10 12 14


-2

0

2

4

6

8

10

2 4 6 8 10 12 14


-2

0

2

4

6

8

10

2 4 6 8 10 12 14



Figure 4.b Sub 2, QFII-eliminated (domestic institutions)the stock market because information contribution exists in entire sample. For all institutionalinvestors, trading behavior did not stand for the potential to destabilize the stock or currencymarkets.

Though foreign investors had demonstrated their destabilizing effect on the stock market duringthe crisis, we should understand the reason why they came to Taiwan. International capital flowsseek profit all around the world; it is their nature to escape from a downturned economy.Moreover, the flows returned after 2009, indicating that the flight was a temporary move. Ourfindings are similar to Bohn and Tesar (1996) and Hamao and Mei (2001). Foreign investors are

32

32

not the main factor because we find that domestic institutions such as securities dealers also had adestabilizing effect during the crisis, even with smaller scale of daily trading volume. We thoughtthat the CBC’s exchange rate management diminished the

-0.5

0.0

0.5

1.0

1.5

2 4 6 8 10 12 14


-.05

.00

.05

.10

.15

.20

.25

.30

2 4 6 8 10 12 14


-2

0

2

4

6

8

10

2 4 6 8 10 12 14


-2

0

2

4

6

8

10

2 4 6 8 10 12 14



Figure 4.c Sub 3, In the crisis (domestic institutions)foreign investors’impact on the currency market, resulting in an increase in international flows, butthe data showed a slight influence.

We think that once we know the nature of international flows and the ways to deal them withproper financial supervisory mechanisms, international capital flows can have a positive effect on ashallow economy country such as Taiwan. If we do have a good financial environment with astrong economic foundation, the flows are not too difficult to manage. The government has

33

33

currency policy, money policy, and fiscal policy to protect the economic system from“hot”money.What we should worry about is the rising fiscal deficit, which could swell to cripple the flexibilityof these economic balancing policies. The case of Taiwan could provide policy implications tothose countries that still impose regulations on foreign capital flows and shed light on an importantquestion: could a small, flat, and export-oriented economy benefit from financial deregulation? Wethink the answer is a resounding“Yes.”

Table 6.a Variance decomposition by N-days ahead forecasts (%). Sub 1, QFII

Table 6.b Variance decomposition by N-days ahead forecasts (%). Sub 2, QFII eliminated


Marketreturns

TWD/$returns

Netpurchases

Marketreturns

TWD/$returns

NetPurchases

Marketreturns

TWD/$returns

Netpurchases

Market 5 98.33 3.76 22.75 98.65 4.18 33.26 98.54 4.11 38.17

returns 10 98.29 3.77 22.83 98.62 4.20 33.70 98.53 4.13 38.16

15 98.28 3.77 22.84 98.62 4.20 33.70 98.53 4.13 38.16

TWD/$ 5 0.39 95.35 0.36 0.39 95.76 0.11 0.35 95.38 0.24

returns 10 0.39 95.28 0.39 0.40 95.7 0.16 0.36 95.35 0.24

15 0.39 95.28 0.39 0.40 95.74 0.17 0.36 95.35 0.24

Net 5 1.26 0.88 76.90 0.93 0.045 66.64 1.097 0.50 61.59

purchases 10 1.29 0.93 76.78 0.94 0.054 66.14 1.10 0.51 61.60

15 1.29 0.93 76.77 0.94 0.054 66.13 1.10 0.51 61.60


Marketreturns

TWD/$returns

Netpurchases

Marketreturns

TWD/$returns

Netpurchases

Marketreturns

TWD/$returns

Netpurchases

Market 5 98.07 8.71 27.08 98.65 9.48 26.50 98.69 9.49 58.04

returns 10 98.06 8.85 27.20 98.64 9.49 26.17 98.69 9.49 58.07

15 98.06 8.85 27.20 98.64 9.49 26.07 98.69 9.49 58.07

TWD/$ 5 0.90 89.25 1.29 0.96 90.33 0.24 0.90 90.17 1.35

returns 10 0.90 88.97 1.30 0.96 90.21 0.36 0.90 90.17 1.35

15 0.90 88.96 1.30 0.96 90.19 0.370 0.90 90.17 1.35

Net 5 1.02 2.03 71.62 0.37 0.17 73.24 0.39 0.32 40.60

purchases 10 1.02 2.17 71.49 0.38 0.28 73.46 0.40 0.32 40.57

15 1.02 2.18 71.49 0.38 0.30 73.55 0.40 0.32 40.57

34

34

Table 6.c Variance decomposition by N-days ahead forecasts (%). Sub 3, In the crisisForeign Investors Investment Trusts Securities Dealers

Marketreturns

TWD/$returns

Netpurchases

Marketreturns

TWD/$returns

Netpurchases

Marketreturns

TWD/$returns

Netpurchases

Market 5 95.89 14.39 38.66 97.22 14.40 21.03 96.44 14.19 34.90

returns 10 95.85 14.39 38.59 97.21 14.42 20.72 96.43 14.20 34.90

15 95.85 14.39 38.59 97.21 14.42 20.69 96.43 14.20 34.90

TWD/$ 5 2.35 84.83 2.81 2.40 85.07 0.10 2.48 85.38 1.63

returns 10 2.36 84.80 2.94 2.40 84.99 0.19 2.48 85.37 1.65

15 2.36 84.80 2.94 2.40 84.99 0.20 2.48 85.37 1.65

Net 5 1.75 0.77 58.52 0.36 0.51 78.86 1.07 0.41 63.45

purchases 10 1.78 0.80 58.46 0.38 0.57 79.07 1.07 0.42 63.44

15 1.78 0.80 58.46 0.38 0.57 79.10 1.07 0.42 63.44

35

35

6. Reference1. Bacmann, J.C., & G, Bolliger. (2001). Who are the best? Local versus foreign analysts on the Latin American stock

markets. University of Neuchâtle, working paper,2. Bekaert,G., & C,R, Harvey. (1997). Emerging equity market volatility. Journal of Financial Economics, 43, 29-773. Bekaert,G., & C,R, Harvey.. (2000). Foreign Speculators and Emerging Equity Markets, Journal of Finance, Vol. 55,

No. 2, 565-6134. Bekaert, G., C, R, Harvey, & Lumsdaine, R. (2002). The dynamics of emerging market equity flows. Journal of

International Money and Finance, 21, 295–350.5. Bohn, H., & L, Tesar. (1996). U.S. equity investment in foreign markets: Portfolio rebalancing or return chasing?

American Economic Review, 86, 77–81.6. Brous, P.A., & O, Kini.. (1994). The Valuation Effects of Equity Issues and the Level of Institutional Ownership:

Evidence from Analysts' Earnings Forecasts, Financial Management, Vol. 23, No. 1, 33-467. Choe, H., B, Kho., & R. M, Stulz. (1999). Do foreign investors destabilize stock markets? The Korean experience in

1997. Journal of Financial Economics, 54, 227–264.8. Corsetti, G., P, Pesenti., & N, Roubini. (2001). The role of large players in currency crises. NBER working paper, No

w83039. DeLong, B., A, Shleifer., L, Summers., & R, Walkmann. (1990). Positive feedback investment strategies and

destabilizing rational speculation, Journal of Finance, 45, 379–395.10. Dornbusch, R., & Y.C, Park.(1994). Financial integration in a second-best world: are we still sure about our

classical prejudices? Korea Institute of Finance, Policy issue series, 94-1, February11. Eswar, P., R, Kenneth., S. J, Wei., & M. A, Kose. (2003). Effects of financial globalizagion on developing countries:

some empirical evidence. International monetary fund.12.Froot, K., P, O’Connell, & Seasholes, M. (2001). The portfolio flows of international investors. Journal of

Financial Economics, 59, 151–193.13. Granger, C.W.J., B.N, Huang, & C.Y, Yang. (2000). A Bivariate Causality between Stock Prices and Exchange

Rates: Evidence from Recent Asian Flu. Quarterly Journal of Economics and Finance, 40, 337–354.14. Hamao, Y., & J, Mei. (2001). Living with the enemy: an analysis of foreign investment in Japanese equity market.

Journal of International Money and Finance, 20, 715–735.15. Johnson, S., & T, Mitton. (2003). Cronyism and capital controls: Evidence from Malaysia. Journal of Financial

Economics, 67,351–382.16. Kaminsky, G., R , Lyons., & Schmukler, S. (2001). Mutual fund investment in emerging markets: an overview.

World Bank Economic Review,15, 315–340.17. Kaminsky, G. L., & S. L, Schmukler. (1999). What triggers market jitters? A chronicle of the Asian crisis. Journal

of International Money and Finance, 18, 537–560.18. Karolyi, G. A. (2002). Did the Asian financial crisis scare foreign investors out of Japan? Pacific-Basin Finance

Journal, 10, 411–442.19. Khanna,T. & K, Palepu.. (1999) Emerging Market Business Groups, Foreign Investors, and Corporate Governance.

National Bureau of Economic Research, working paper 6955.20. Kim, W., & S, Wei. (2002). Foreign portfolio investors before and during a crisis. Journal of International

Economics, 56, 77–96.21. Krugman, P. (1998). Saving Asia: it is time to get radical. Fortune, September22. Lin, A.Y. (2006). Has the Asian crisis changed the role of foreign investors in emerging equity markets: Taiwan’s

experience. International Review of Economics and Finance, 15,364-382.23. Mundell, R. (2000). Currency areas, exchange rate systems and international monetary reform. Journal of Applied

Economic, Vol. III, No. 2, 217-25624. Nofsinger, J. R., & R. W, Sias. (1999). Herding and feedback trading by institutional and individual investors.

Journal of Finance, 54, 2263–2295.25. Scholes, M.S. (1972). The market for securities: Substitution versus price pressure and the effects of information on

sharpe prices. The Journal of Business, 45(2), 170.26. Sims, C. (1980). Macroeconomics and reality. Econometrica, 48, 1–48.27. Stiglitz, J(1998). Boat, blain and capital flows. Financial Times, March 25.28. Wang, L.R, & C.H, Shen. (1999). Do foreign investments affect foreign exchange and stock markets—the case of

Taiwan. Applied Economics, v31, i11, p1303.29. Wang, J. (2007) Foreign equity trading and emerging market volatility: Evidence from Indonesia and Thailand,

Journal of Development Economics, 84, 798-811.

2012asianfa the effect of foreign institutional investors · the effect of foreign institutional...

Documents