Market Efficiency and Foreign Institutional Trading ... · 2. The data 2.1 Sample selection and market description Our transaction data is obtained from TAIFEX, which is one of the

Market Efficiency and Foreign Institutional Trading

Market Efficiency and Foreign Institutional Trading:

Evidence from Taiwan

Robin K. Chou Department of Finance, National Chengchi University, Taiwan

[email protected]

Keng-Yu Ho Department of Finance, National Taiwan University, Taiwan

[email protected]

Pei-Shih Weng Department of Finance, National Central University, Taiwan

[email protected]


Market Efficiency and Foreign Institutional Trading:

Evidence from Taiwan

Abstract Foreign institutional investors are frequently viewed as well-informed traders, and their information advantage would transform into informational efficiency on prices. Due to the relative importance of foreign institutions in emerging markets, this study investigates the information role of foreign institutional traders in Taiwan futures market (TAIFEX). TAIFEX has significant trading volume increase over the past few years, mainly because of the foreign institutions. However, it is puzzling that while trading volume has been increasing in TAIFEX, the market-wide liquidity costs tend to be much higher and the informational efficiency has worsen. We find that foreign institutions more frequently act as market makers, and such decrease in order aggressiveness has primarily resulted in the deterioration of market efficiency. Furthermore, extensive passive orders induce higher liquidity costs and impede possible arbitrage opportunity from mispricing. The evidence indicates that foreign institutions may earn profits by providing liquidity rather than trading with information advantage. JEL classification: G12; G14 Key words: Foreign institution; Market efficiency; Liquidity; Limit order; Order aggressiveness


3

1. Introduction In past few years, although it is evident that foreign institutions have more

frequently participated in emerging markets; the question whether increasing trading

activity of foreign institutions enhances market quality has not been widely-explored yet.

Conceptually, the policy makers of financial sectors in emerging markets claim that the

markets should be transitioning to more liquid and efficient markets by more foreign

institutional participation. In practice, however; such a statement lacks the solid

investigation. Though, some previous studies (e.g. Grinblatt and Keloharju (2000),

Dvorak (2005), and Huang and Shiu (2009)) show that foreign institutions are better able

to select winners in the investment markets than the domestic investors, implying that

foreign institutional traders could enjoy their information advantage in the local market;

these studies do not involve the comprehensive influence (if any) that comes from

foreign institutional trading. By the compact analyses for the futures transaction in an

important emerging market, Taiwan, the aim of this paper was to empirically explore the

upswing trading activity of foreign institutional investors and the accompanying change

in the market.

Recently, there is a marked uptrend in trading activity in the derivatives markets in

emerging countries, mainly because of foreign institutional investors. Taiwan futures

market also has the same phenomenon. Since foreign institutions are viewed as

important participants in emerging markets, their trading activity just creates a natural

opportunity to examine our question. Moreover, as well-known that markets are

incomplete in real world, lower transaction costs and higher leverage for derivatives

should attract more informed investors. In such a case, foreign institutional trading may

play a crucial part in the price discovery process in term of the presence of informed

trading.

Until the end of 2008, there are 52 derivatives exchanges worldwide reported to the

FIA (Futures Industries Association). Among top ten derivatives exchanges, there are five

exchanges listed in emerging countries.1 Moreover, the five fastest growing derivatives

exchanges are all from emerging markets.2 Since the financial derivatives products have

1 In top twenty derivatives exchanges, there eleven exchanges set up in emerging countries. 2 See the statistics from Futures Industries Association (http://www.futuresindustry.org/volume-.asp).


4

been regarded as an advanced tool for use in hedging and arbitraging by more

sophisticated traders, e.g., institutional investors, foreign institutions are hence viewed as

key contributors to the trend of increasing volume. Frequent foreign institutional

participant could be due to few reasons. First, these merging countries are eager to

improve, deepen and internationalize their financial markets by attracting more

international investors into their financial markets.3 Next, foreign institutional traders

also expect to conduct rent-seeking trading by entering these high-growth markets.4 In a

recent study, Barber, Lee, Liu, and Odean (2009) presented a clear portrait of who

benefits from trade in Taiwan stock market: individuals lose, institutions win. In

particular, they found that foreign institutions garner earn nearly half of institutional

profits, which indicated that foreign institutions would gain from trade by their

information advantage.

There have been previous studies of trading activity, many of which have

documented trading costs have declined and this decline has contributed to the trading

volume advances. For example, French (2008) and Chakravarty, Panchapagesan, and

Wood (2005) have argued that institutional commissions had declined over time. Further,

the advent of technology has made it easier for institutions to execute automated

algorithm trading (see Hendershott, Jones, and Menkveld (2008)). Since foreign

institutions may trade more frequently on their own account due to these mechanism

factors such as decreased commissions and improvement in trading technology; this

change would encourage them to trade on information advantage more efficiently

because decreased trading friction might increase returns via information-based trading.

For instance, Boehmer and Kelly (2009) established that institutions are important

traders in the markets and contribute to the informational efficiency of prices. In

addition, Albuquerque, Bauer, and Schneider (2009) suggested that foreign investors are

likely to have an advantage in processing global information and, therefore, contribute to

3 In the international finance literature, there has been a growing body of empirical evidence suggesting that opening a market to foreign investors is beneficial. This evidence suggests that stock market liberalizations lower cost of capital (Henry (2000a); Bekaert and Harvey (2000)), increase the real investment (Henry (2000b); Mitton (2006); Chari and Henry (2008); Bae and Goyal (2008)), and spur productivity and growth (Bekaert, Harvey, and Lundblad (2005, 2009)). 4 Griffin, Kelly, and Nardari (2010) mention that “the conventional wisdom is that emerging markets are less efficient than developed markets. Highly profitable trading strategies and prices that deviate from a random walk are often what people have in mind when describing the evidence.”


5

the incorporation of such information into security prices. Overall above the literature

expects that growing foreign institutional trading would lead to greater information

production and a more efficient market with reduced short-run fluctuations. Alternatively,

we might not ignore the likelihood that foreign institutions would merely pass along

domestic traders’ investment allocation decisions to the financial market. In such case,

one would not expect much change in information-based trading.

All the empirical issues in this paper have been examined in following stages. First,

we established some basic empirical features of the recent trading trend for market-wide

volume and for each trader’s class, including domestic individual investors, foreign

institutions, and domestic institutions. Our available data contain index futures contracts

in the Taiwan Futures Exchange (TAIFEX) from January 2003 through December 2008.

The detailed transactions data allow us more deeply compare the trading behavior among

three trade classes over time. In this stage, we documented that there is distinct uptrend

in the market-wide trading volume of TAIFEX, as well as in each trade class’s trading

volume. For the relative growth and decline in certain class, foreign institutional traders

have dramatic rises in trading percentages compared to their counterpart traders.

Second, we calculated bid-ask spread and posted depth in limit order book over our

sample period. We documented that there is declining pattern for liquidity in recent years.

Evidently, it seems that more frequent trading activity has no positive impact on market

quality in aspects of trading cost or price impact. Further, we directly examined the

change in information-based trading using variance ratios. While French and Roll (1986)

related these ratios to the amount of information incorporated into price, they showed

that hourly open-to-close return variances are greater than close-to-open variances and

conclude this finding is due to incorporation of private information during trading hours.

We implemented their approach to test the change in market efficiency over time.

Surprisingly, variance ratios of trading hours also tend to become lower in recent years,

which present that information-based trading has been relatively shrinking even that

trading activity is more intense lately, particularly for foreign institutional traders. This

finding suggested that increases in trading percentage of foreign institutions have not

translated into the improvement of market efficiency; instead, it seems that the market is

experiencing the deterioration of market quality.


6

We then explored the mechanism in several dimensions and discovered why

increases in foreign institutional trading do not lead to informational efficiency. As found,

the most relevant determinate to channelize the linkage between increased trading and

decreased market quality is the aggressiveness of order submission. At market level, the

percentage of non-marketable limit orders keeps rising gradually through time;

meanwhile, foreign institutions has the highest usage frequency for non-marketable limit

orders. Since Glosten (1994) suggested that market making activity would tend to involve

non-marketable limit orders while informed trades tend to be conducted by submitting

marketable limit orders, the relative proportion of non-marketable and marketable limit

orders submitted by certain trader class is a proxy to present the informativeness of

trading activity.

Finally, we indeed found that foreign institutional traders use more non-marketable

limit orders relative to marketable limit orders; they also tend to submit even more

non-marketable limit orders in recent years. In addition, variations in the proportion of

non-marketable limit orders of foreign institutional trading are significantly associated

with market efficiency; furthermore, more passive limit orders appear to have extra

influence on efficiency. On the other side, passive limit orders damage market liquidity,

and informational efficiency is related to liquidity since greater trading costs make

arbitrage trades more expensive, and presumably reduce efficiency. As a result, the

extensive usage of passive limit orders by foreign institutions produces a two-pronged

influence on informational efficiency, and has trapped the market in recent years.

Our study complements to current literature on the institutions’ information roles

in financial market since earlier studies document that institutions are well-informed and

their activities lead to more informationally efficient pricing.5 Evidently, we piece a new

picture for institutional trading activity, in particular when they trade as foreigners in

specific local market. That is, even if foreign institutions are better informed, it does not

trivially follow that their informational advantages translate into transaction prices.

Moreover, locking the market in inefficiency may create more opportunities for foreign

institutional traders to gain from mispricing. Given the prevailing tendency of attracting

5 The related empirical findings can be also seen in Grinblatt and Titman (1989, 1993), Nofsinger and Sias (1999), Wermers (1999, 2000), Chen, Jegadeesh, and Wermers (2000), and Bennett, Sias, and Starks (2003).


7

foreign institutional investment in an emerging market, the associated policy-makers

should not ignore the possibility of institutional trading inefficiency.

The remainder of this paper is organized as follows. Section 2 describes the sample

and TAIFEX. Section 3 presents the uptrend in trading activity and the changes in

liquidity patterns. Section 4 presents the changes of informational efficiency in the

market. Section 5 analyzes the association between market efficiency and trading

proportion among different trader classes. Section 6 analyzes the order submission and

order aggressiveness for each trader class, and Section 7 concludes.

2. The data 2.1 Sample selection and market description

Our transaction data is obtained from TAIFEX, which is one of the major

exchanges and the only one futures exchange in Taiwan. Taiwan, as an emerging market,

has impressively active financial markets. Until the end of 2009, Taiwan has been the

fourth largest emerging market in Asia-Pacific area. Such features would enable us to

interpret representative findings in Taiwan to other emerging markets.

The detailed transaction information in the dataset contains: date, the time of order

submission and order execution, the indicator for buy- or sell-initiation, the indicator for

contract opening or closing, ordered and traded quantity, the code for identification of

traders, and limit order book. The information above makes it feasible to construct all

necessary variables in this study, particularly for our purpose to classify traders as

individual investors, domestic institution investors, or foreign institution investors. Our

sample covers the period from 2003 though 2008 since both of the repealing of QFII

and the launching of TAIFEX continuous auction system are happened in 2003.

In recent years, the fast volume growing has make TAIFEX one of top derivatives

exchanges in the world. Until the end of 2008, TAIFEX is ranking as No. 17 among 52

derivatives exchanges reported to the FIA. In emerging markets, TAIFEX is the 8th

largest derivatives exchanges among 6 countries, including South Korean, South Africa,

China, Russia, and Taiwan. TAIFEX was established in 1998, and now offers futures and

options contracts on major Taiwan stock indices, government bond, equity, interest rate,

and commodity. Trading hours in TAIFEX is 08:45A M-1:45PM Taiwan time Monday


8

through Friday of the regular business days for the Taiwan Stock Exchange. TAIFEX is

an order-driven market and operated an automated auction system. Since there are no

designated market makers, investors submit their orders to the automatic trading system.

During trading hours, the system will continuously match unexecuted buy and sell orders.

TAIFEX used to launch the batch-call trading system until August 2002, and then

transformed to the continuous auction system. The active-growing period in TAIFEX

starts from 2003; in that year, the total trading volume reaches to 32 million contracts

while the annual trading volume is merely around 8 million in the end of 2002.

Among more than 20 products in TAIFEX, we select TXF (TAIEX futures) as our

sample. TXF is the first product in TAIFEX; its underlying index, Taiwan Stock

Exchange Capitalization Weighted Index (TAIEX), is the representative stock market

index in Taiwan. TXF is the most mutual and active product in TAIFEX and its

variations can reflect investors’ perspectives at market level without concerning

idiosyncratic information and cross-sectional correlation. The deliver months for TXF

are: spot month, the next calendar month, and the next three quarterly months. Since

TXF contacts delivered in three quarterly months are not as liquid as spot and the next

calendar month, our sample only incorporate spot and next-month futures. Because

summed trading percentage of spot futures and next-month futures is more than 90% of

daily trading volume, such exclusion does not alleviate the representativeness of our

sample.

2.2 Main variables

The trading information in our transaction data are recorded based on event time

within the trading day, to obtain a set of equal-interval observations, we reconstruct our

data by computing five-minute horizon observations for all variables. A five-minute

variable, such as return, was computed from the price closest to the end of time intervals

within the trading day. We computed each variable by picking up the values closest to the

end of five-minute interval. For each trading day, there are sixty five-minute interval

between 8:45 AM and 13:45 PM. Our intraday variables are defined as follows:

RET: the returns, computed from the midpoint of bid-ask spread in the end

of an intraday time interval.


9

QSPR: the quoted bid-ask spread, calculated as the difference between the best

bid and the best ask in the limit order book in the same interval as the return.

PSPR: the percentage bid-ask spread, calculated as dividing QSPR by the

midpoint of the bid-ask spread in the same interval as the return.

DEP1: the posted depth for the best bid and for the best ask in number of

order quantity in the same interval as the return.

DEP5: the posted depth for entire limit order book in number of order

quantity in the same interval as the return.

For a given trader class, we incorporate all executed orders for relevant calculation,

while canceled orders and unexecuted orders are excluded in calculations.6 In line with

the previous studies, we merely use the trades initiated by buyers or sellers to open new

futures contracts. Considering that closing contacts may be initiated by forced liquidation,

but not by investors themselves, new contracts therefore should contain more

meaningful information for future market and be better measure for trading activity (see,

for example, Pan and Poteshman (2006); Chang, Hsieh, and Lai (2009)).

To measure the tendency as a de facto market maker for a certain trader class, we

want to compute the proportion market order to limit orders (see Lee, Liu, Roll, and

Subrahmanyam (2004)). However, in an order-driven market like TAIFEX, there are no

designed dealers or specialists, and traders can merely submit limit orders. To overcome

the unavailability of market orders in limit order market, previous studies have suggested

that a marketable limit order is essentially equal to a market order (see Glosten (1994)).

When a transaction is just completing, the highest unexecuted bid price and the lowest

unexecuted ask price become prevailing quotes, so that we can use them to define a

marketable order for the next transaction. Any subsequent order to buy at or below the

prevailing ask or sell at or above the prevailing bid, moreover, the quantity needed in this

order can be cleared, that such order is deemed “marketable”. By doing so, we can

classify trades into non-marketable limit orders and marketable limit order.

3. Recent uptrend in trading activity 6 In addition to domestic individuals, domestic institutions, and foreign institutions, actually there are some trades placed by foreign individuals. However, they are too few to construct a meaningful group, that they are not included in our sample.


10

[Insert Figure 1 Here]

Figure 1 presents the monthly aggregate trading volume (in quantity) of TXF in

TAIFEX from 2003 through 2008 inclusive. Trading volume for individuals, domestic

institutions, and foreign institutions are also plotted separately. As can be seen, aggregate

trading volume in TAIFEX has increased over time, particularly; the uptrend is more

distinct from 2006. A potential cause for this phenomenon could be that TAIFEX

reduced the trading tax rate from 0.025% to 0.01% for all equity index futures, including

TXF in 2006.7 This 60% fee reduction is impressively high. Empirically, the lower

trading costs tend to stimulate the market participants and induce more trading activity.

This pattern is consistent with Chordia, Roll, Subrahmanyam (2011), who find that

declined trading cost has contributed significantly to the trading volume uptrend in

NYSE during the period from 1993 through 2008. Trading volume for either trader class

reveals similar pattern. However, while the variations of individual trading volume are

similar to that of the aggregate market, the growth rate of institutional trading volume is

markedly higher. Naively comparing the change between January 2003 and December

2008, domestic institutions appear to have twelve-fold increases in trading volume, while

foreign institutions appear to have more than twenty-five-fold increases. Such trends are

impressive, and also suggest that foreign institutions have recently acted as key

contributors to the advanced trading activity. Table 1 presents summary statistics

associated with trading volume for each trader class in each year. Except for the year of

2005, there is evidence that the trading volume at whole market level continuously

increases with time. For the exceptional decline in 2005, it could be due to the influence

from the U.S. market. As known, the real economy growth of U.S. in 2005 is much

slower relative to 2004. Since the economic cointegration between emerging markets and

the U.S. is quite strong, the economic conditions about the U.S. market tend to affect the

investors’ confidence in emerging markets, especially for retail investors. Such

phenomenon frequently happens in Taiwan.

[Insert Table 1 Here]

7 In TAIFEX, trading tax is charged for every executed transaction regardless of opening or closing position. The dollar amount is computed based on executed price. For instance, an executed price for TXF is (concurrent index points × NTD200).


11

Consistent with we have shown in Figure. 1, individual group is the largest trader

class in TAIFEX, therefore, its trading pattern closely co-moves with the whole market.

Interestingly, it seems that the trading decline in over all market in 2005 should not be

attributed to institutional traders. Evidently, both foreign and domestic institutional

traders do not trade less in that year; furthermore, the trading volume of foreign

institutions continuously increases while trading volume of domestic institutions stays in

the same level. In addition, given the increased trading volume for either trader class, the

trading proportion of retail investors has been decreasing over time. On the contrary,

even though both institutional investors only trade less than 20% of the whole market

volume in 2003, their trading proportion has grown to almost 40% in 2008. Especially,

increases in foreign institutional trading are even more dramatic, their trading proportion

is merely around 4% of total volume in 2003, but has been rising to 14.24% in 2008.

Foreign institutions used to be negligible traders in TAIFEX; however, they turn out to

be more important players in recent years. Overall evidence indicates that there is

significant uptrend in trading volume of recent years, and all trader classes contribute to

this trend. Among either trader class, foreign institutional traders appear to have the most

marked increases in their trading proportion. As we mention earlier, a reasonable

explanation for this phenomenon is that institutions may have traded to take advantage

of lower liquidity costs in the presence of reduced depths, as shown in Chakravarty,

Panchapagesan, and Wood (2005) as well as Jones and Lipson (2001) and Chordia, Roll,

Subrahmanyam (2011). We further explore this possibility.

In Table 2, we present summary statistics on several microstructure features

associated with trading activity, including bid-ask spread and posted depth in limit order

book. Statistics for all measures are calculated based on daily average value using

five-minute interval series within a trading day. Consistent with Table 1, we present

statistics for each year separately.


Not as expected, based on the results of Table 2, we find no evidence that trading

activity trends reasonably mirror the patterns in liquidity, via regardless of spreads or

depths. Furthermore, liquidity patterns seem to associate with trading volume trends


12

conversely. Except for PSPR, other liquidity measures (QSPR, DEP1, and DEP5) reveal

that market has continuously become more illiquid at least from 2005. In recent years,

the market-wide quoted spreads tend to widen, and market depths also become narrower.

All these findings may suggest that worse market quality would come up with increased

trading activity while such relationship seems strictly monotonic. For formal comparisons,

we test the liquidity measures between different years. Consistent with our expectation,

quoted spreads (posted depth) in 2008 are significantly greater (smaller) than those in

2003. Since the significant changes in 2008 could be extensively caused by global

financial crisis in that year, to exclude the illusion comes from the exogenous influence,

we also conduct a test for the difference between 2003 and 2007. Again, the alternative

comparison shows same results. Among the comparisons, PSPR is the only measure does

not appear distinct pattern. A reasonable explanation for the flat changes is the bull

market in Taiwan from late 2004 to early 2008. In that period, the price level in futures

market also had been increased over time. PSPR, as a relative spread measure (inverse to

the price level), hence has fewer variations between 2005 and 2007.

Since we have not seen the contemporaneous bid-ask spread decreases or price

depth increases, the increased proportion for domestic or foreign institutional trading is

not driven by liquidity costs. Interestingly, individual trading activities appear to have

much influence on the market liquidity. Among years, 2005 is the most liquid year,

because it shows up with the lowest bid-ask spread and the highest price depth.

Meanwhile, according to the uptrend in trading volume, only the year of 2005 have

decreased trading activity compared to 2004. Based on the findings of Table 1, the

declined volume should primarily attribute to the retail trading while there are no distinct

changes for institutional trading. These phenomena conclusively suggest that fewer

individual trades may have positive impact on the market quality, while individual traders

frequently act as noise traders (see, for example, Delong, Shleifer, Summers, and

Waldmann (1991); Statman (1999)). In addition, the results imply that a dramatic rise in

the trading proportion of foreign institutions does not lead to the improvement in

liquidity at the market level. However, such preliminary evidence does not necessarily

recognize that the increased institutional trading proportion, particularly for foreign

institutions, deteriorates the market liquidity in recent years. In practical, it could be many


13

reasons would contribute to liquidity variations. Since informational influence from

foreign institution is still unclear, it would be worthy to further ascertain the information

role of foreign institutions, and see variations of market efficiency in the same period.

4. Information-based trading and market efficiency Our findings by now confirm that institutions may trade more frequently on their

own account due to exogenous factors such decreased commissions or improvement in

trading technology. However, we do not find evidence that institutional trading plays a

beneficial role to market liquidity. As another dimension of market quality, we further

examine potential influence of institutional traders to market efficiency.

Recently, Chordia, Roll, and Subrahmanyam (2011) related trading activity trends to

price formation and suggested that more information-based trades from institutions may

benefit market price for incorporation of more information during the trading hours. A

seminal work to ascertain information-based trading and price formation can be seen in

French and Roll (1986). They showed that hourly open-to-close return variances are

greater than hourly close-to-open return variances and offered three potential

explanations for this finding: First, incorporation of private information during trading

hours; second, mispricing caused by investor misreaction or market frictions and

microstructure noise induced by bid-ask bounce; and third, greater incorporation of

public information into prices during trading hours. They rejected the last explanation

because the variance ratios were not significantly different on business days when the

stock market was closed. They concluded that the other two components would help

explain the higher ratio during market trading hours, with the first one being the

dominant factor. Chordia, Roll, and Subrahmanyam (2008) and Boehmer and Kelley

(2009) have also argued that variance ratios could reveal the degree of private

information produced by trading process. Relying on their conclusions, the variance ratio

is suitably linked to the informational efficiency of the pricing system in the sense of

Kyle (1985). Since foreign institutional investors are usually classified as informed traders,

given the increased foreign institutional trading proportion in recent years, we seek to

discern whether the market is more informatively efficient by comparing variances ratios

over time. Panel A of Table 3 first presents the hourly (open-to-close) to (close-to-open)


14

variance ratios for each year. To obtain per hour variances, open-to-close and

close-to-open returns are first used separately to compute raw variances.8 Then, each

computed variance is divided by the total number of trading (non-trading) hours.


As reported in Panel A, although all open-to-close variances far exceed

close-to-open variances, which is consistent with the argument of French and Roll (1986).

However, the variance ratios in more recent years are much lower than those in previous

years. In particular, the variance ratio in 2008 is merely the half of 2003. These findings

suggest, given dramatic increases in foreign institutional trading proportion recently,

overall market does not become more informative while the information-based

fluctuations within trading hours tend to be shrank. More specifically, while we expect

foreign institutional traders should more frequently trade with their private information,

and market prices should be determined by incorporating more information, the realized

outcome turns out to be opposite to the expectation.

For further comparisons, we alternatively consider the variance ratio computed

from five-minute interval returns and from open-to-close returns. In computing this

variance ratio, the five-minute return variance is multiplied by 60, which is the number of

five-minute intervals in a trading day. Early studies of market efficiency frequently use

variance ratios to test whether prices follow a random walk (see Barnea (1974)). A

random walk implies that the ratio of long-term to short-term return variances,

measured per unit time, equals 1. More deviation from random walk would be revealed

to be a variance ratio much greater than 1. Because we are interested in the gap between

actual and efficient prices in either direction, we compute the variance ratio deviation as

|1－VR(t, s)|, where VR(t, s) is the ratio of the return variance over t periods to the

returns over s periods, both divided by the length of the period. We report five-minute

variance ratio deviations in Panel B of Table 3. As can be seen in Panel B, the variance

ratios in more recent years are further away from the efficient level. Again, this finding

indicates that the extent of informational efficiency induced by trading is weaker lately,

8 To avoid contamination of return serial correlations by bid-ask bounce, we compute returns from quote midpoints.


15

especially in 2007 and 2008. Interestingly, according to Panel B of Table 3, we find that

the smallest efficiency deviations appear in 2005, which is consistent with our finding of

Table 2 that the market is the most liquid in 2005. This phenomenon may reveal the

close relationship between liquidity and market efficiency; we will address this issue

further in Section 5.

In addition to two variance ratios, we also conduct alternative efficiency measures

based on returns autocorrelations, which should be zero at all horizons if prices follow a

random walk. Since either positive or negative autocorrelations represent the departures

from efficient prices; we compute the absolute value of 30-minute mid-quote returns

autocorrelations and report the results in Panel C of Table 3. As reported, all values in

Panel C show similar patterns as that in Panel A or B.

5. The Relationship between Efficiency and Foreign Institutional Trading

Insofar as it can be ascertained, foreign institutional trading does not produce

information more frequently over time, meanwhile, the market is less informationally

efficient in more recent period. Indeed, foreign institutions trade more portions of

overall market in recent years, but the increases do not translate into the improvement of

market quality. Instead of that, the market quality has been worse than before. Since the

market efficiency should be more closely related to arbitrageurs’ trading, could it be that

the increasing institutional trading exploits market efficiency in some way? As noted

earlier, the extent of market inefficiency is modest when the market is the most liquid,

and vice versa. A possibility is that the increased liquidity cost in recent years (see Table 2)

may prohibit institutional investors to trade efficiently even though they have been

trading more. If that is the case, the market inefficiency is essentially driven by the

market illiquidity, but not the institutional trading itself. To allow for a nonlinear

relationship between efficiency and both liquidity and trading activity, we use a double

sort procedure. For whole period, we dividend the sample into three groups (low,

medium, and high) based on their posted depth (DEP5) and further divide these groups

into three subgroups (low, medium, and high) based on trading proportion of each

investor type (foreign institutions, individuals, and domestic institutions). Table 4 reports


16

the average values of efficiency measures within 9 sub-groups for each investor class. As

expected, efficiency increase monotonically with liquidity.


But more importantly, in Panel A we find that efficiency also increase monotonically

with foreign institutional trading proportion within each liquidity group whether

efficiency is measured by autocorrelations or variance ratio deviations. Among 6 liquidity

groups (3 groups for each of two efficiency measures), 5 groups show that inefficiency in

the subgroup with high foreign institutional trading proportion is significantly greater

than that with low foreign institutional trading proportion. This result is not specific to

posted depth as liquidity measure; we obtain similar results using the other measures in

Table 2 (not reported).9 The findings in Panel A of Table 4 suggest that foreign

institutional trading have an effect on efficiency that is orthogonal to the liquidity effect.

On the contrary, individual trading (Panel B) does not appear to have such effect. As for

domestic institutions (Panel C), although 4 subgroups with high trading proportion have

greater inefficiency level than their low trading proportion counterparts within the same

liquidity group; however, only one liquidity group appears to have significant difference

in inefficiency between high trading proportion subgroup and low trading proportion

subgroup.

5.1 Regression results for efficiency, trading proportion, and liquidity

Apparently the market inefficiency in recent years would be due to the market

illiquidity; however, we are surprised that some influence come from foreign institutional

traders can not be explained by liquidity effect, which implies that foreign institutions

deteriorate market quality extra. Given that possibility, since the foreign institutional

trading proportion has been increasing with time, the associated deterioration to market

efficiency should be more severe lately, and it just happens to what we have investigated

in Table 3. Nevertheless, an alternative but not mutually exclusive explanation, based on

9 The results using PSPR is relatively weak, only 4 groups show greater inefficiency in high foreign institutional trading proportion groups, and two of them have significant difference. For using the rest two measures (QSPR, DEP1) and an added measure QSPR/DEP1 (relative quoted spread), we obtain similar results.


17

our existing findings, is that an inefficient market will also attract more foreign

institutional arbitrageurs as long as the latent arbitrage profits is high. Thus, the pattern

in Panel A of Table 4 could be driven by a form of reverse causality. This possibility

warrants careful controls for the endogenous relationship between foreign institutional

trading proportion and market efficiency. To study the relation between efficiency and

trading activity in multivariate framework, we estimate the following time-series

regression:

tkt

K

kkittt XTPMIEMIE

1110 . (1)

MIEt is the change in market efficiency based on returns autocorrelations or

variance ratios in month t. TPit is the change in trading proportion for investor i (foreign

institutions, individuals, or domestic institutions) in month t. Xk is a set of control

variables, including average price level (in log) in month t, the change in monthly posted

depth, and total trading volume (in log) in month t. We also include the lagged dependent

variable in Equation (1) because the time series of efficiency measures are relatively

persistent and we want to make sure that the attendant autocorrelation does not affect

the estimates. As noted earlier, TPt may not solely work as exogenous variable in

Equation (1) but as endogenous variable with MIEt. To deal with the possible

endogeneity, we first conducted Hausman (1978) test to detect if TPt is endogenous

variable. If the results of Hausman test indicate existence of endogeneity, and thus the

two-stage least squares method (2SLS) is employed to estimate the Equation (1). We also

conducted Hausman test to posted depth since liquidity and efficiency could be likely

jointly determined. For Hausman test and two-stage least squares, we need to include

additional variables as instruments for the trading proportion and the posted depth. We

used monthly risk-free rate and the open interest, which was applied to the trading

proportion only.

In addition to two efficiency measures have used in Table 4, we employed three

extra autocorrelations-based or variance ratios-based measures using different sampling

frequency in the regression analyses. We considered another intraday measure based on

variance ratio of (5, 60) minutes, and two daily horizon measures based on daily returns


18

autocorrelations and variance ratios of (1, 5) days. By doing so, we presume make sure

the regression estimates does not depend on specific intraday pattern.

[Insert Table 5 here]

Panel A of Table 5 presents Hausman test results.10 Except for two models using

autocorrelations measure (model 1 and 2), the estimated residuals of posted depth in all

other models (model 3 to 5) are insignificant at 10% level, which indicates that liquidity

and efficiency are simultaneously determined only when efficiency is measured by

autocorrelations. Furthermore, for all models, the Hausman test for trading proportion

of each investor class does not detect that trading proportion is endogenous variable

since all the tests fail to reject the hypothesis that trading proportion is exogenous in

Equation (1). Based on the results of Hausman test, we only apply 2SLS for model 1 and

2.11 The regression estimates of Equation (1) for foreign institutions, individuals, and

domestic institutions are reported in Panel B, C, and D, respectively.

Panel B of Table 5 shows that, controlling for liquidity, market price level, and total

trading volume, increases in foreign institutional trading proportion are always associated

with decreases in efficiency (note that all efficiency measures we use are inversely related

to the degree of efficiency) and are statistically significant when we use different

efficiency measures. The only one exception does not perform significant estimate is

model 2, which uses daily returns autocorrelations as efficiency measure. As expected,

liquidity is significantly negatively related to efficiency. The results of Panel B are

consistent with the patterns in Table 4, and hence confirm foreign institutional trading

has extra effect on market efficiency that is orthogonal to the liquidity effect.

The results of Panel C and D also support the findings of Table 4. Interestingly, for

individuals, although none liquidity group in Table 4 has significant pattern for efficiency,

Panel C of Table 5 shows significant regression estimates when we use model 3 and 5.

Seemingly, increases in individual trading proportion are also associated with increases in

10 Based on the procedure of Hausman test, we should first regress the examined variable (trading proportion or liquidity) on a set of control variables and the instrument variable(s), and then estimate the residuals. Add the estimated residuals into original regression model as an additional regressor, and run the regression. If the coefficient of estimated residuals is statistically significant, we confirm the endogeneity. We report the coefficients of estimated residuals and t-value (in brackets) in Panel of Table 5. 11 To be concise, only results of the efficiency equation are reported for model 1 and 2, as the particular interest here is on the changes in market efficiency induced by the changes in trading proportion.


19

efficiency. Since the correlation coefficient between the change in individual trading

proportion and the change in foreign institutional trading proportion is only -0.391

(based on our unreported calculation), the significant effect from individual trading is not

primarily due to the substitute trading shifts between foreign institutions and individuals.

However, such individual trading effect on market efficiency is less robust among

different efficiency measures.

5.2 Vector autoregression analyses

Section 5.1 has recognized the contemporaneous relationship between foreign

institutional trading and market efficiency, our further goal is to explore intertemporal

associations between market efficiency and trading activity, either for institutional

investors or individuals. In this section we adopt a 5-equation vector autocorrelation that

incorporates 5 variables we have used in the multivariate regression of Table 4. Thus,

consider the following system:

K

ktktikt uXX

1. (2)

where X is a vector that represents the changes in efficiency, the changes in trading

proportion for investor i, the change in posted depth, average price level, and total

trading volume. In the empirical estimation, we choose K=2; the number of lags in

Equation (2) on the basis of Akaike information criterion (AIC) for each analysis. Table

6 reports our VAR estimation results. To be concise, only the results of first two

equations (efficiency and proportion) are reported in Table 6, as our particular interest

here is the intertemporal association between the changes in market efficiency and the

changes in trading proportion for each investor class.


As reported, both market efficiency and trading proportion appear to have negative

autocorrelation at least up to two-period lag and the coefficients are statistically

significant except for domestic institutions. More importantly, the changes in foreign

institutional trading proportion lead to efficiency deterioration, but not vice versa. The

effect is significant even for two-period lagged changes in trading proportion. The


20

finding indicates that foreign institutional trading influences market efficiency

contemporaneously and intertemprally. Because the effect is consistently negative, it

means that increases in relative foreign institutional trading would worsen current and

subsequent market quality. Compared to foreign institutions, lagged individual trading

has inverse influence on efficiency, which is similar to the results of Table 4. The

intertemporal effect from individuals, however, is relatively weaker than that from foreign

institutions because only one-period lagged changes in trading proportion is statistically

significant. As for domestic institutions, although their trading proportion changes tend

to lead efficiency decreases, but the intertemporal effect is not significant. Consistent

with the finding for foreign institutions, efficiency changes does not lead to changes in

both individual and domestic institutional trading proportion.

In summary, overall the preceding findings indicate that the informational

inefficiency in recent years can be attributed to increasing foreign institutional trading in

the market. Even though, the concurrent decreasing liquidity enhances the market

inefficiency, but liquidity changes are not the only driver for the changes in efficiency.

Foreign institutions appear to have additional influence for the efficiency deterioration.

Given the evidence, if the liquidity is not the only reason for the market inefficiency, why

and how would foreign institutional trading harm the market? As mentioned in the

introduction, although foreign institutions are frequently viewed as informed traders and

their trading could help incorporating more information into prices; however, foreign

institutions could alternatively play as uninformed roles that merely pass along retail

traders’ asset allocation decisions to the financial market and act as liquidity providers. If

that is the case, such trading strategy indeed benefit nothing to pricing efficiency of the

market. We analyze this possibility in the following section.

6. Order Submission and Aggressiveness 6.1 Non-marketable limit order

In this section, we analyzed the possible changes in order submission behavior of

either trader class to ascertain whether recent market inefficiency could be related to

order type used by specific investors based in our preceding findings. Glosten (1994)

suggested that in a limit order market, market marking activity tend to involve


21

non-marketable limit orders designed to capture the bid-ask spread, while traders who

demand liquidity immediately would more likely submit marketable limit orders. In

addition, Lee, Liu, Roll, and Subrahmanyam (2004) showed that the proportion of

non-marketable limit orders could proxy for the frequency of demand for immediacy. In

line with these studies, we classified the orders of each trader class based on all executed

order submission, and computed the submission proportion for either order type. Table

7 reports the proportion of non-marketable limit orders for either trader class over

time.12


As can be seen, though the changes are gradual, total proportion of limit orders in

overall market has been increasing from 54.91% in 2003 to 57.97% in 2008. Considering

the concurrent growth in trading volume, the 3% increases in limit order proportion are

not neglectable. When we look further into each investor group, the variation of limit

order proportion is quite different for different trader classes. Compared to total market,

individuals’ limit order proportion is much more stable; it is 51.65% in 2003, rises to the

highest level 54.3% in 2005 and slightly recesses to 52.02% in 2008. For domestic

institutions’ limit order proportion, there is a minor downtrend before 2008. It is around

60% in 2007 while the percentage in 2003 is 66.49%, however; the proportion rises again

to 65.07% in 2008. Different from individuals and domestic institutions, foreign

institutions’ limit order proportion shows a distinct uptrend. While the percentage is

53.76% in 2003, quite similar to the aggregate market level in the same year, the

proportion keeps rising with time, and has been as high as 71.76% in 2008. Such marked

increases in limit order proportion simply indicate that foreign institutional traders use

more and more limit orders in recent years, and also implies that they tend to have less

frequent demand for immediacy.

The patterns in Table 7 reveal a picture that increases in aggregate market’s limit

order proportion are driven by foreign institutional traders. Apparently, foreign

institutions are only traders who complete transactions by using more and more limit

12 For brevity, we use “market order” and “limit order” to indicate marketable limit order and non-marketable limit order, respectively.


22

orders. Considering their increasing trading proportion and concurrent increasing limit

orders usage, foreign institutions show the market a distinct trading tendency; that is,

their trading behavior becomes less aggressive over time and affects the whole market

more widely in recent years. During the same period, individuals consistently trade as

what they used to trade, and the changes to domestic institutions are less distinct and

inclusive.

To distinguish whether the extensive usage of limit orders by foreign institutions is

the main reason why increases in their trading proportion deteriorate the market

efficiency, we conduct similar analyses to those in Table 4 and 5 to address the

association between foreign institution’s limit order usage and market efficiency. Again, a

double sort procedure is applied. For foreign institutions, we first divide the sample

months into three groups based on their monthly average limit orders percentage and

further divide these groups into three subgroups based on foreign institutional trading

proportion. Panel A of Table 8 reports the average efficiency level within each subgroup.


While efficiency increases almost monotonically with the usage percent of limit

orders in foreign institutional trading; within each limit order group, the trading

proportion loses it significant association with efficiency no matter which efficiency

measure is used in the analysis. These results recognize our expectation based on

preceding findings, the influence of foreign institutional trading to market efficiency

should be primarily attributed to the frequent and extensive usage of limit orders by

foreign institutions.

For more specific analyses, we reestimated the regressions of model 1 and model 4

in Table 5 with replacing the change in trading proportion with the change in limit order

proportion. Meanwhile, we incorporated the residuals estimated from the regression of

changes in trading proportion on changes in limit order proportion as an additional

independent variable. If the effect of trading proportion can be fully explained by limit

order usage, the residuals would have no significant impact on market efficiency. Panel B

of Table 8 presents the results of regressions.

Apparently, increases in limit order percentage are associated with decreases in


23

market efficiency and the impact is statistically significant, these results are consistent

with those regression estimates using trading proportion. Furthermore, compared to

Table 5, the regressions we reestimate here perform higher explanatory power in the light

of R2. Finally, the residuals of trading proportion changes present no significant

influence to market efficiency. Overall findings by now indicate that it is more extensive

usage of limit orders rather than the foreign institutional trading itself to cause the

market efficiency deterioration. However, since foreign institutions place more and more

limit orders in recent years, their increasing trading proportion further enhances the

harmful influence.

6.2 Order aggressiveness

Thus far we have studied the determinants of market efficiency deterioration in

recent period, and find the main factor to the deterioration is that foreign institutions

trade with more limit orders. Viewing the whole market as a pricing system,

less-aggressive foreign institutions in the system would trade with uninformative strategy

and obstruct inefficient market prices to revert to efficiency. However, since foreign

institutions tend to act as liquidity providers more frequently, they are supposed to help

market liquidity and therefore alleviate the extent of inefficiency. Unfortunately, the fact

turns out to be the other way. This unexplained phenomenon tends to be a puzzle and

could dust our preceding findings. The primary goal of this subsection is to shed more

light to the information extent of foreign institutional limit orders and also to explore

whether there are essential differences for limit orders across different trader classes.

Each year, we classified each trader class’s submitted limit orders into four

categories based on corresponding submission aggressiveness. As for the definition of

submission aggressiveness, following a classification scheme similar to Griffiths, Smith,

Turnbull and White (2000), we used the price of limit order to define appropriate

categories. The first category is for any subsequent limit order to buy at or above the

prevailing ask or sell at or below the prevailing bid.13 The second category is for any

13 Note that the limit orders here still indicate “non-marketable” limit orders. Although the definition for category one seems marketable for orders; however, any order with marketable price could still be non-marketable when it has unexecuted quantity (more than available quoted depth). We only deem an order submission as marketable whenever its order price and quantity can fulfilled immediately during the


24

subsequent limit order to buy below the prevailing ask and above the prevailing bid or

sell above the prevailing bid and below the prevailing ask. The third category is for any

subsequent limit order to buy at the prevailing bid or sell at the prevailing ask. The fourth

category is for any subsequent limit order to buy below the prevailing bid or sell above

the prevailing ask. Although each preceding category is non-marketable category;

however, according to the corresponding marketability, we can deem the first category as

the least passive (relatively aggressive) group and the fourth category as the most passive

group, and etc.

For each trader class, Table 9 reports the frequencies of the limit order classification

scheme across each year.


Panel A presents the results of foreign institutions. Except for two extreme groups (the

first and the fourth category), the order submission frequencies in two medium groups

are relatively stable throughout the years. An interesting observation is that decreases in

the frequencies of the least-passive group are roughly close to increases in the

frequencies of the most-passive group. The variations present a frequency shift between

order submission aggressiveness, which indicates that the order placing strategy of

foreign institutions have been less and less aggressive. A complemented observation is

that the aggregate frequencies in first two categories are more than 50% in 2003 and

2004, but this proportion keeps falling to less than 40% in 2008.

For individual investors, based on the order submission scheme, their order placing

strategy seems quite stable over time. There is no distinct changing pattern for each

category. In addition, the order submission frequencies are primarily concentrated in the

last two groups, which indicate that the limit order placement of individuals is extremely

passive. Next, for domestic institutions, it seems that there is changing pattern in each

group. Both the least-passive and the most-passive group have lower proportion in recent

years, while the frequencies in two medium groups tend to be higher at the same time.

Though, compared to foreign institutions, the variations of order submission frequencies

for domestic institutions are sometimes ambiguous. To more precisely compare the

transaction.


25

timely variations of the order aggressiveness for each trader class, the following

quantitative measure of limit order aggressiveness is applied to allow us to conduct

comparisons based on a single measure. The aggressiveness index is defined as:

.)(

,)(

sellorderifQ

QPP

buyorderifQ

QPP

nessAggressive

TSi

SMS

TBi

BBM

(3)

where PM is the prevailing mid-quote price for specific order i in specific trading day; PB

(PS) is buy (sell) order price for specific order i in specific trading day; QB (QS) is buy

(sell) order quantity for specific order i in specific trading day; QTB (QTS) is daily total buy

(sell) order quantity. 14 The higher aggressiveness index, the more passive order

submission.

Each day, we calculated aggressiveness index for each trader class, and reported the

median value of each year in Panel A of Table 10.


As can be seen in each year, the aggressiveness index of individuals is the highest, the

next is domestic institutions, and the last is foreign institution. The sequence is consistent

with the frequencies distribution among categories in the order submission scheme,

which points out that foreign institutions are the most aggressive traders in the market

while individuals are the most passive participants. However, the reported increasing

aggressiveness index of foreign institutions represents their changes of being less

aggressive for trading. The pattern is consistent with our inference based on Table 9. In

addition, the stable aggressiveness index of individuals implies that retail traders behave

in their usual manner. As for domestic institutions, they become slightly more passive

compared to what they used to be; however, their changes in the aggressiveness index in

last three years are minor at best.

14 Our definition is similar to that of Chou and Wang (2009), while they used transaction price instead of mid-price in the formula.


26

When preceding evidence has shown that foreign institutions more frequently trade

as liquidity providers by using more limit orders than market orders. Further evidence

here explains why there is no simultaneous improvement for market liquidity. Given that

unchanged retail traders’ order aggressiveness, institutional order submissions are

relatively passive in recent years, in particular foreign institutions have markedly

monotonic changes. Passive order prices tend to widen quoted bid-ask spread, and it just

so happens to what we have reported in Table 2. In the unreported result, we also find

that the average order size in each transaction for foreign institutions becomes smaller

from 2005 through 2008. This phenomenon can further explain shrank market depth

pattern.

Finally, as a robustness check, we performed VAR estimates similar to Table 6, but

replaced trading proportion with limit order proportion and added aggressiveness index

in the regressors. Our aim is to investigate whether the extent of order aggressiveness

has extra influence which is orthogonal to limit order itself on market efficiency. For

brevity and our particular interest on aggressiveness index, we only report partial results

in Panel B of Table 10. Similar to Table 8, the intertemporal effect of changes in limit

order proportion to changes in efficiency is significant for foreign institutions.

Interestingly, we find that domestic institutions’ limit order proportion also has

intertemporal effect on the efficiency, although the two-period lag is a bit surprising.

More importantly, both lagged domestic and foreign institutions’ order aggressiveness

index have significant impact on the market efficiency.

In sum, the above estimates suggest that the passiveness of limit orders could be

the key determinant to realize the mechanism how increased limit order usage percentage

from institutional trading can harm the market efficiency. On the other side, extensive

passive limit orders also deteriorate liquidity and hence damage the efficiency; such

two-pronged influence has unavoidably trapped the market in recent years.

7. Conclusion

Aggregate trading volume in TAIFEX has increased significantly over the past few

years. Meanwhile, foreign institutions, rather than individual trades or domestic

institutional traders, are key contributors for the volume uptrend, since only foreign


27

institutional trading proportion has been increasing throughout the years. However, over

the same period when trading volume has been increasing, the market-wide spread is

widened and the concurrent price depth becomes narrower; moreover, the market

informational efficiency has deteriorated in recent years. Since foreign institutions are

generally viewed as informed traders and their trading would benefit to incorporating

information, such phenomenon seems puzzling.

This paper explores the anatomy of the increased foreign institutional trading and

accompanying shifts in market quality to channelize the causality. Evidently, two major

empirical findings would rationalize the puzzle. First, foreign institutions use relatively

more non-marketable limit orders than other traders, and the tendency is even more

intense with time. This finding indicates foreign institutions more frequently act as

liquidity providers. This change leads to the infrequent information-based trading and

induces the market inefficiency.

Second, there is further downward shift for the order aggressiveness of foreign

institutions’ non-marketable limit orders. The frequencies of their most passive limit

orders has increased over time, while the overall passiveness level of non-marketable

limit orders has also risen. Shifting to passiveness widens the bid-ask spread and directly

increases liquidity costs for trading. Since market illiquidity could abate the arbitrageurs’

willingness to trade for miss-pricing. Passive order submissions would finally affect the

market efficiency. Given that increasing foreign institutional trading proportion in recent

years, we wonder that the market could be locked in an inefficiency trap with spiral

influence from foreign institutions.

To some extent, this study sheds light into the discussion about opening market to

foreign institutions in emerging countries. Nowadays, it is a prevailing trend for emerging

countries to attract more foreign institutional traders to invest in their financial markets.

Indeed, increased trading activity from foreign institutions would contribute to volume

growth in the market, and attract more international investors’ interest. From the aspect

of market development, such change would have positive impact. However, as can be

seen in the case of TAIFEX, foreign institutions may not necessarily benefit to the local

market. Whenever foreign institutions trade passively or inefficiently, their overwhelming

and unavoidable influence would harm the market. This paper has provided evidence for


28

the possibility above. For policy-making purpose, it may be really worth to pay more

attention to the association between foreign institutions and the local market they have

participated.

Finally, our research also raises few questions. First, why would foreign institutions

more frequently act as passive traders? Next, just how much could foreign institutions

gain or lose from their passive trading strategies? These issues are left for future research.


29

Reference Admati, Anat R., and Paul Pfleiderer, 1988a, A theory of intraday patterns: Volume and price variability,

The Review of Financial Studies 1, 3-40. Admati, Anat R., and Paul Pfleiderer, 1988b, Selling and trading on information in financial markets, The

American Economic Review 78, 96-103. Albuquerque, Rui, Gregory H. Bauer, and Martin Schneider, 2009, Global private information in

international equity markets, Journal of Financial Economics 94, 18-46. Bae, Kee-Hong, Arzu Ozoguz, Hongping A. U. Tan, and Wirjanto Tony S., 2011, Do foreigners facilitate

information transmission in emerging markets?, Journal of Financial Economics Forthcoming. Barber, Brad M., and Terrance Odean, 2000, Trading is hazardous to your wealth: The common stock

investment performance of individual investors, Journal of Finance 55, 773-806. Bekaert, Geert, and Campbell R. Harvey, 2000, Foreign speculators and emerging equity markets, The

Journal of Finance 55, 565-613. Bekaert, Geert, Campbell R. Harvey, and Christian Lundblad, 2005, Does financial liberalization spur

growth?, Journal of Financial Economics 77, 3-55. Bloomfield, Robert, Maureen O’Hara, and Gideon Saar, 2009, How noise trading affects markets: An

experimental analysis, Review of Financial Studies 22, 2275-2302. Boehmer, Ekkehart, and Eric K. Kelley, 2009, Institutional investors and the informational efficiency of

prices, Review of Financial Studies 22, 3563-3594. Bozcuk, Aslihan, and M. Ameziane Lasfer, 2005, The information content of institutional trades on the

London stock exchange, Journal of Financial and Quantitative Analysis 40, 621-644. Chakravarty, Sugato, 2001, Stealth-trading: Which traders' trades move stock prices?, Journal of Financial

Economics 61, 289-307. Chakravarty, Sugato, Venkatesh Panchapagesan, and Robert A. Wood, 2005, Did decimalization hurt

institutional investors?, Journal of Financial Markets 8, 400-420. Chang, Chuang-Chang, Pei-Fang Hsieh, and Hung-Neng Lai, 2009, Do informed option investors predict

stock returns? Evidence from the Taiwan stock exchange, Journal of Banking & Finance 33, 757-764.

Chari, Anusha, and Peter Blair Henry, 2008, Firm-specific information and the efficiency of investment, Journal of Financial Economics 87, 636-655.

Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam, 2005, Evidence on the speed of convergence to market efficiency, Journal of Financial Economics 76, 271-292.

Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam, 2008, Liquidity and market efficiency, Journal of Financial Economics 87, 249-268.

Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam, 2011, Recent trends in trading activity and market quality, Journal of Financial Economics 101, 243-263.

Chou, Robin K., and Yun-Yi Wang, 2009, Strategic order splitting, order choice and aggressiveness: Evidence from the Taiwan futures exchange, Journal of Futures Market 29, 1102-1129.

Chung, Dennis, and Karel Hrazdil, 2010, Liquidity and market efficiency: A large sample study, Journal of Banking & Finance 34, 2346-2357.

Chung, Kee H., Bonnie F. Van Ness, and Robert A. Van Ness, 1999, Limit orders and the bid-ask spread, Journal of Financial Economics 53, 255-287.

De Long, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, 1991, The survival of noise traders in financial markets, Journal of Business 64, 1.

Dvorak, Tomas, 2005, Do domestic investors have an information advantage? Evidence from Indonesia, Journal of Finance 60, 817-839.

Foster, F. Douglas, and S. Viswanathan, 1990, A theory of the interday variations in volume, variance, and trading costs in securities markets, The Review of Financial Studies 3, 593-624.

French, Kenneth R., 2008, Presidential address: The cost of active investing, Journal of Finance 63, 1537-1573.

French, Kenneth R., and Richard Roll, 1986, Stock return variances : The arrival of information and the reaction of traders, Journal of Financial Economics 17, 5-26.

Froot, Kenneth A., David S. Scharfstein, and Jeremy C. Stein, 1992, Herd on the street: Informational inefficiencies in a market with short-term speculation, The Journal of Finance 47, 1461-1484.

Glosten, Lawrence R., 1994, Is the electronic open limit order book inevitable?, Journal of Finance 49, 1127-1161.

Griffin, John M., Patrick J. Kelly, and Federico Nardari, 2010, Do market efficiency measures yield correct inferences? A comparison of developed and emerging markets, Review of Financial Studies 23, 3225-3277.

Grinblatt, Mark, and Matti Keloharju, 2000, The investment behavior and performance of various investor types: A study of Finland’s unique data set, Journal of Financial Economics 55, 43-67.


30

Hendershott, Terrence, Charles M. Jones, and Albert J. Menkveld, 2011, Does algorithmic trading improve liquidity?, The Journal of Finance 66, 1-33.

Henry, Peter Blair, 2000a, Stock market liberalization, economic reform, and emerging market equity prices, The Journal of Finance 55, 529-564.

Henry, Peter Blair, 2000b, Do stock market liberalizations cause investment booms?, Journal of Financial Economics 58, 301-334.

Huang, Roger D., and Cheng-Yi Shiu, 2009, Local effects of foreign ownership in an emerging financial market: Evidence from qualified foreign institutional investors in Taiwan, Financial Management 38, 567-602

Jame, Russell E., and Qing Tong, 2009, Retail investor industry herding, SSRN eLibrary (Available at SSRN: http://ssrn.com/paper=1468952).

Johnson, Timothy C., 2008, Volume, liquidity, and liquidity risk, Journal of Financial Economics 87, 388-417. Jones, Charles M., Gautam Kaul, and Marc L. Lipson, 1994, Information, trading, and volatility, Journal of

Financial Economics 36, 127-154. Jones, Charles M., and Marc L. Lipson, 2001, Sixteenths: Direct evidence on institutional execution costs,

Journal of Financial Economics 59, 253-278. Kerr Christoffersen, Susan E., and Ya Tang, 2010, Institutional herding and information cascades:

Evidence from daily trades, SSRN eLibrary (Available at http://ssrn.com/paper=1572726). Kyle, Albert S., 1985, Continuous auctions and insider trading, Econometrica 53, 1315-1335. Lee, Yi-Tsung, Yu-Jane Liu, Richard Roll, and Avanidhar Subrahmanyam, 2004, Order imbalances and

market efficiency: Evidence from the Taiwan stock exchange, Journal of Financial & Quantitative Analysis 39, 327-341.

Mitton, Todd, 2006, Stock market liberalization and operating performance at the firm level, Journal of Financial Economics 81, 625-647.

Nicolosi, Gina, Liang Peng, and Ning Zhu, 2009, Do individual investors learn from their trading experience?, Journal of Financial Markets 12, 317-336.

Pan, Jun, and Allen M. Poteshman, 2006, The information in option volume for future stock prices, The Review of Financial Studies 19, 871-908.

Scharfstein, David S., and Jeremy C. Stein, 1990, Herd behavior and investment, The American Economic Review 80, 465-479.

Scharfstein, David S., and Jeremy C. Stein, 2000, Herd behavior and investment: Reply, The American Economic Review 90, 705-706.

Statman, Meir, 1999, Behaviorial finance: Past battles and future engagements, Financial Analysts Journal 55, 18-27.


31

Figures 1: Trading volume in quantity over time This figure consists of four plots. Each plot presents the monthly trading volume (in quantity)

from 2003 through 2008 for total market, individuals, foreign institutions, and domestic

institutions, respectively. Trading volume is counted based on TAIEX futures; TAIEX is the major

spot market index in Taiwan. The trading data are obtained from Taiwan futures exchange.

Total Market

0

500000

1000000

1500000

2000000

25000002003

2004

2005

2006

2007

2008

Year

Volum

eIndividuals

0

200000

400000

600000

800000

1000000

1200000

1400000

1600000

1800000

2003

2004

2005

2006

2007

2008

Year

Vol

ume

Foreign Institutions

0

50000

100000

150000

200000

250000

300000

2003

2004

2005

2006

2007

2008

Year

Volu

me

Domestic Institutions

0

100000

200000

300000

400000

500000

600000200

3

2004

2005

2006

2007

2008

Year

Vol

ume


32

Table 1: Trading volume summary statistics for each trader class This table presents the statistics of trading volume for each trader class in each year. The number in parentheses under the sum for certain trader class is the trading percentage in total aggregate volume in the corresponding year. Data are from 2003-2008 inclusive, 1,484 daily observations are used.

(Unit: 100 contracts) Year

Class 2003 2004 2005 2006 2007 2008

All Sum 130,040 176,725 136,932 197,288 234,581 389,335 Mean 542 736 571 822 977 1,622 Median 552 690 583 818 887 1,586 Std. dev. 176 174 78 141 301 432 Individuals Sum 106,035 130,264 88,403 128,947 145,558 244,269 (81.54%) (73.71%) (64.56%) (65.36%) (62.05%) (62.74%) Mean 442 543 368 537 606 1,018 Median 432 514 393 521 595 1,027 Std. dev. 153 152 60 110 204 291 Foreign Sum 4,629 11,399 13,460 16,868 32,325 55,441 institutions (3.56%) (6.45%) (9.83%) (8.55%) (13.78%) (14.24%) Mean 19 47 56 70 135 211 Median 18 41 57 70 129 224 Std. dev. 11 18 8 11 49 44 Domestic Sum 19,363 35,062 35,055 51,472 56,698 89,625 institutions (14.89%) (19.84%) (25.60%) (26.09%) (24.17%) (23.02%) Mean 81 146 146 214 236 373 Median 83 142 148 208 226 376 Std. dev. 23 21 27 32 68 115


33

Table 2: Bid-ask spread and quoted depth in each year This table reports mean and median of four liquidity measures in each year. Data are from 2003-2008 inclusive, 1,484 trading days are incorporated. QSPR is the quoted bid-ask spread, calculated as the difference between the best bid and the best ask in the limit order book. PSPR is the percentage bid-ask spread, calculated as dividing QSPR by the midpoint of the bid-ask spread. DEP1 is the posted depth for the best bid and for the best ask in number of order quantity. DEP5 is the posted depth for entire limit order book in number of order quantity. All measures are first recorded in the end of five-minute interval within a trading day, and a daily observation is produced by the average value of 60 five-minute observations. We conduct t-test/Wilcoxon test to compare the sample in different years and report significance. The asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels.

T-test / Wilcoxon test Year Measures

2003 2004 2005 2006 2007 2008 2003=2007 2003=2008

QSPR Mean 1.526 1.631 1.287 1.382 1.630 1.831 *** *** Median 1.475 1.492 1.262 1.344 1.537 1.754 * ***

PSPR Mean 0.030 0.027 0.021 0.020 0.020 0.027 *** *** Median 0.031 0.025 0.021 0.020 0.019 0.026 *** *

DEP1 Mean 27.562 27.901 33.807 27.534 21.416 18.182 *** *** Median 26.683 27.327 33.443 27.300 20.525 17.098 *** ***

DEP5 Mean 221.125 214.282 285.053 211.127 168.455 150.314 *** *** Median 216.776 210.639 275.502 208.051 169.655 144.975 *** ***


34

Table 3: Variances ratios and returns autocorrelations This table reports two variance ratios and the first-order return autocorrelations in each year. Data are from 2003-2008 inclusive, 1,484 trading days are incorporated. Panel A presents (open-to-close) ÷ (close-to-open) per hour return variance ratios, based on mid-quote returns. Panel B presents the ratios of five-minute return variance to open-to-close variance (scaled by the number of five-minute intervals in a day). Panel C presents 30-minute first-order return autocorrelations. All measures are calculated monthly by using daily or intraday observations. We conduct t-test/Wilcoxon test to compare the sample in different years and report significance. The asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels.

T-test / Wilcoxon test: 2003 2004 2005 2006 2007 2008 2003=2007 2003=2008

Panel A: Hourly open-to-close / close-to-open variance ratios Mean 11.711 15.333 11.433 9.005 7.394 6.436 *** *** Median 10.394 13.135 11.322 10.046 6.218 5.139 *** ***

Panel B: Deviations of 5-minute / open-to-close variance ratios Mean 0.304 0.391 0.199 0.244 0.325 0.437 - *** Median 0.297 0.351 0.219 0.254 0.304 0.418 - ***

Panel C: Absolute values of first order autocorrelations of 30-minute returns Mean 0.063 0.095 0.069 0.070 0.085 0.118 ** *** Median 0.050 0.071 0.049 0.056 0.065 0.092 * ***


35

Table 4: Liquidity and institutional/individual trading effect on market efficiency This table presents the levels of market efficiency within 9 subgroups using double sort. We first sort the sample (72 monthly observations, 2003-2008) by liquidity measure (DEP5) and form three liquidity groups. High liquidity group means the group consists of the most liquid months. We further divide each liquidity group into three subgroups based on trading proportion (foreign institutions, individuals, or domestic institutions) and report mean of efficiency measures (|AR| or |1-VR|) for each subgroup. |AR| and |1-VR| indicate “absolute values of first order autocorrelations of 30-minute returns” and “deviations of 5-minute / open-to-close variance ratios”, respectively. We conduct t-test to compare the means of high-subgroup and low-subgroup with each liquidity group and report significance. The asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels. The capital alphabet ‘F’ indicates the fail to reject the null hypothesis.

Market efficiency measures

Trading proportion |AR| |1-VR|

Liquidity Low Medium High T-Test:

Low=High

Low Medium High T-Test:

Low=High

Panel A: Foreign institution

Low 0.082 0.090 0.127 ** 0.304 0.405 0.458 **

Medium 0.076 0.086 0.084 F 0.319 0.273 0.438 **

High 0.061 0.068 0.076 ** 0.169 0.195 0.289 *

Panel B: Individuals

Low 0.112 0.089 0.097 F 0.439 0.357 0.374 F

Medium 0.083 0.084 0.079 F 0.346 0.301 0.373 F

High 0.071 0.072 0.063 F 0.251 0.184 0.228 F

Panel C: Domestic institution

Low 0.093 0.100 0.105 F 0.351 0.384 0.432 F

Medium 0.080 0.086 0.080 F 0.369 0.300 0.351 F

High 0.062 0.074 0.070 F 0.196 0.220 0.247 *


36

Table 5: The relationship between trading proportion and market efficiency This table presents the results of the estimates based on following time-series regression:

tkt

K

kkittt XTPMIEMIE

1110 .

MIEt is the change in market efficiency based on returns autocorrelations or variance ratios in month t. TPit is the change in trading proportion for investor i (foreign institutions, individuals, or domestic institutions) in month t. Xk is a set of control variables, including average price level (in log) in month t, the change in monthly posted depth, and total trading volume (in log) in month t. Estimates are based on 72 monthly observations, 2003-2008. We first conduct Hausman (1978) test to detect if TPt or Depth (DEP5) is endogenous variable. Panel A presents the results of Hauman test. We report the coefficients of estimated residuals for TPt or Depth and corresponding t-statistics (in brackets). If the results of Hausman test indicate existence of endogeneity, and thus the two-stage least squares method is employed for correction. The regression results for foreign institutions, individuals, and domestic institutions are reported in Panel B, C, and D, respectively. Five efficiency measures used as dependent variables: 30-minute returns autocorrelations, daily returns autocorrelations, and the deviations based on variance ratio of (5, 60) minutes, variance ratio of 5-minute to open-to-close horizon (300 minutes), or variance ratio of (1, 5) days. The asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels.

(1) (2) (3) (4) (5) Efficiency variables | AR(30 min) | | AR( daily) | | 1－VR(5m, 60m) | | 1－VR(5m, 300m) | | 1－VR(1d, 5d) |

Panel A: Hausman test results

Foreign institution 0.378 [0.123] -1.087 [0.352] 5.644 [0.429] 10.159 [0.517] -4.460 [0.745] Individuals 0.166 [0.458] 2.470 [0.450] -1.104 [0.807] -2.909 [0.892] 6.804 [0.868] Domestic institution 0.163 [0.735] -1.579 [0.351] 1.474 [0.232] 19.694 [0.734] -2.336 [0.854] Depth 0.418 [0.087] -1.423 [0.092] 0.003 [0.235] 0.002 [0.523] -0.023 [0.215]


37

Panel B: Foreign institutional trading proportion

Intercept -0.081 0.952 4.672 2.339 15.111* Lag_Efficiency -0.522*** -0.327*** -0.345* -0.443*** -0.532*** Proportion 0.061* 2.827 3.526** 10.153** 20.971* Depth -0.087** -1.043* -0.007* -0.004* -4.197** Log_Price 0.059 0.073* 0.522* 0.416* 0.382 Log_Volume 0.074 -0.067 0.318* 0.158 -1.057* R2 0.351 0.245 0.408 0.436 0.372

Panel C: Individual trading proportion

Intercept -0.122 0.668 2.276 -0.253 9.999 Lag_Efficiency -0.537*** -0.415*** -0.512* -0.515*** -0.549*** Proportion -0.201 0.094 -1.481* -2.209 -14.707* Depth -0.093** -2.934** -0.003 -0.003* -4.466** Log_Price -0.030 0.034 0.352* 0.213 -0.099 Log_Volume -0.045 -0.046 -0.153 0.054 -0.699 R2 0.289 0.239 0.371 0.338 0.388

Panel D: Domestic institutional trading proportion

Intercept -0.308 1.242 3.338 -0.478 16.345** Lag_Efficiency -0.522*** -0.321** -0.384* -0.528** -0.524*** Proportion 0.067 -1.141 1.416 4.964* 7.503 Depth -0.110** -1.288* -0.004 -0.004* -2.156 Log_Price -0.022 0.038 0.400* 0.386 -0.053 Log_Volume -0.021 -0.086 -0.226 0.099 -1.141* R2 0.268 0.185 0.324 0.359 0.313 Observation 71 71 71 71 71


38

Table 6: Vector autoregression estimates This table presents the results of a five-equation vector autocorrelation (VAR) that incorporates

five variables we have used in the multivariate regression of Table 5. Estimates are based on 72 monthly observations, 2003-2008. The following system is estimated:

K

ktktikt uXX

1.

where X is a vector that represents the changes in efficiency, the changes in trading proportion for investor i, the change in posted depth, average price level, and total trading volume. In the empirical estimation, we choose K=2; the number of lags in the equation, on the basis of Akaike information criterion (AIC) for each analysis. The results of first two equations (efficiency and proportion) are reported for brevity. We report the VAR coefficients and corresponding t-statistics (in brackets).

Foreign institutions Individuals Domestic institutions

Efficiency Proportion Efficiency Proportion Efficiency Proportion

Efficiency (-1) -0.870 -0.005 -0.750 0.013 -0.721 -0.014

[-6.432] [-0.429] [-5.727] [ 1.033] [-5.562] [-1.344]

Efficiency (-2) -0.483 -0.010 -0.364 0.030 -0.390 -0.017

[-3.286] [-0.800] [-2.499] [ 1.699] [-2.701] [-1.447]

Proportion (-1) 3.736 -0.475 -1.839 -0.472 2.123 -0.133

[ 2.405] [-3.652] [-1.756] [-3.681] [1.624] [-1.037]

Proportion (-2) 3.139 -0.270 -0.876 -0.251 1.913 -0.023

[ 1.868] [-2.019] [-0.899] [-2.105] [ 1.303] [-0.191]

Depth (-1) 0.0003 0.0000 0.0002 -0.0001 0.0002 0.0001

[ 0.519] [ 1.042] [ 0.391] [-2.312] [ 0.373] [ 2.578]

Depth (-2) -0.001 0.0000 -0.001 -0.0001 -0.001 0.0000

[-1.571] [ 0.762] [-1.726] [-0.882] [-1.253] [ 0.428]

Log_Price (-1) -0.582 0.019 -0.420 -0.038 -0.535 0.012

[-2.154] [ 0.851] [-1.487] [-1.098] [-1.896] [ 0.509]

Log_Price (-2) 0.538 -0.033 0.345 0.039 0.496 -0.016

[ 1.785] [-1.306] [ 1.071] [ 0.993] [ 1.578] [-0.622]

Log_Volume (-1) 0.078 0.001 0.083 0.004 0.049 -0.005

[ 1.295] [ 0.129] [ 1.277] [ 0.489] [ 0.791] [-1.072]

Log_Volume (-2) -0.066 0.005 -0.0645 -0.010 -0.046 0.007

[-1.139] [ 1.129] [-1.083] [-1.391] [-0.772] [ 1.385]

R2 0.453 0.293 0.420 0.331 0.406 0.212


39

Table 7: The proportion of non-marketable limit order for each trader class This table presents the mean/median proportion of non-marketable limit orders for each trader class based on the daily observations in each year. Data are from 2003-2008 inclusive, 1,484 trading days are incorporated. Marketable orders are defined as limit orders placed at prevailing inside quotes; that is, sell orders placed at or below the highest prevailing bid or buy orders placed at or above the lowest prevailing offer; non-marketable orders are counterpart of marketable orders. The numbers reported are mean and median (in parentheses) percentage of non-marketable limit orders based on daily observations for certain trader class in the specific year.

Year Class

2003 2004 2005 2006 2007 2008

Total Non-Marketable 54.91% 55.33% 56.44% 56.16% 56.65% 57.97% (55.01) (55.57) (56.67) (56.64) (55.97) (57.59) Individuals Non-Marketable 51.65% 53.40% 54.30% 53.98% 52.63% 52.02% (51.23) (53.51) (54.39) (54.26) (52.66) (51.90) Foreign Non-Marketable 53.76% 56.69% 60.85% 64.89% 68.31% 71.64% institutions (53.21) (56.28) (59.97) (65.49) (66.41) (73.51) Domestic Non-Marketable 66.49% 62.91% 62.24% 59.79% 60.95% 65.07% institutions (66.00) (62.76) (62.11) (60.28) (60.36) (63.58)


40

Table 8: Foreign institutional non-marketable limit orders and market efficiency Panel A presents the levels of market efficiency within 9 subgroups using double sort. We first sort the sample (72 monthly observations) by the proportion of non-marketable limit orders of foreign institutions and form three limit orders proportion groups. We further divide each limit orders proportion group into three subgroups based on foreign institutional trading proportion and report mean of efficiency measures (|AR| or |1-VR|) for each subgroup. |AR| and |1-VR| indicate “absolute values of first order autocorrelations of 30-minute returns” and “deviations of 5-minute / open-to-close variance ratios”, respectively. We conduct t-test to compare the means of high-subgroup and low-subgroup with each liquidity group and report significance. Panel B presents the results of reestimation for model (1) and (4) in Table 5 with replacing the change in trading proportion with the change in limit order proportion. Meanwhile, we incorporate the residuals estimated from the regression of changes in trading proportion on changes in limit order proportion as an additional independent variable. The asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels. The capital alphabet ‘F’ indicates the fail to reject the null hypothesis.

Panel A: Trading proportion effect

Foreign institutional trading proportion

Limit orders

proportion

Low Medium High

T-Test:

Low=High

% of limit

orders

Low | AR | 0.070 0.079 0.080 F 54.64

| 1-VR | 0.228 0.234 0.303 F

Medium | AR | 0.072 0.071 0.080 F 63.66

| 1-VR | 0.213 0.237 0.309 F

High | AR | 0.093 0.110 0.095 F 69.63

| 1-VR | 0.441 0.447 0.438 F

Panel B: Foreign institutional limit orders

| AR | | 1-VR |

Intercept -0.079 2.747

Lag_Efficiency -0.519*** -0.442***

Limit_order 0.105** 12.010**

Resd_proportion 0.005 0.013

Log_Price 0.059 0.429*

Depth -0.075* -0.006*

Log_Volume 0.087 0.209*

R2 0.375 0.454


41

Table 9: Order aggressiveness cross years, by order categories This table reports the frequencies of the limit order classification scheme in each year for each trader class. Data are from 2003-2008 inclusive, 1,484 trading days are incorporated. The numbers reported are mean and median (in parentheses) percentage of non-marketable limit orders based on daily observations. The first category (High) in the order submission scheme means the submissions are the least passive (relatively aggressive) while the last category (Low) incorporates the most passive order submissions.

Year 2003 2004 2005 2006 2007 2008

Panel A: Foreign institutions Aggressiveness High 37.54 40.47 41.98 31.97 29.07 24.99

(34.33) (38.80) (41.60) (31.51) (28.94) (24.05)

2 15.56 16.59 13.95 15.52 15.63 14.23 (13.56) (17.84) (13.92) (14.80) (15.17) (15.10)

3 34.70 28.54 28.45 28.38 29.33 34.18 (32.96) (25.23) (27.56) (29.96) (28.84) (34.19)

Low 12.20 14.40 15.61 22.02 25.95 27.59 (10.55) (12.76) (13.98) (21.42) (25.26) (27.03)

Panel B: Individuals Aggressiveness High 11.43 12.62 14.76 12.39 13.26 9.86

(11.34) (12.43) (14.85) (15.36) (13.04) (9.68)

2 11.78 14.07 10.04 13.22 14.40 13.62 (11.85) (13.61) (10.09) (13.05) (14.04) (13.17)

3 29.19 25.16 28.26 28.38 27.10 30.99 (29.19) (25.35) (28.17) (28.26) (27.48) (31.54)

Low 47.58 48.15 46.94 45.01 45.24 45.53 (47.20) (48.23) (47.06) (43.24) (45.18) (46.02)

Panel C: Domestic institutions Aggressiveness High 21.47 23.12 23.79 22.76 18.55 14.96

(20.71) (22.99) (23.55) (22.32) (18.34) (14.64)

2 13.04 17.32 12.21 16.99 17.92 17.21 (13.13) (17.13) (12.03) (16.74) (17.92) (16.31)

3 28.55 26.60 30.47 28.62 33.10 36.33 (28.25) (27.72) (30.20) (28.24) (32.67) (36.98)

Low 36.93 32.96 33.52 31.63 30.43 31.49 (37.27) (32.25) (33.04) (31.29) (29.87) (31.47)


42

Table 10: Order aggressiveness and market efficiency Panel A presents median aggressiveness index based on daily calculation in each year for each trader class. Data are from 2003-2008 inclusive, 1,484 trading days are incorporated. The aggressiveness index is defined as:

.)(

;)(

sellorderifQ

QPPorbuyorderif

Q

QPPnessAggressive TS

i

SMS

TBi

BBM

where PM is the prevailing mid-quote price for specific order i in specific trading day; PB (PS) is buy (sell) order price for specific order i in specific trading day; QB (QS) is buy (sell) order quantity for specific order i in specific trading day; QTB (QTS) is daily total buy (sell) order quantity. The higher aggressiveness index, the more passive order submission. Panel B reports the results of VAR estimates similar to Table 6, but we replace trading proportion with limit order proportion and add aggressiveness index in regressors. For brevity and our particular interest on aggressiveness index, we only report the results of first two equations with the first three variables. The numbers reported are the VAR coefficients and corresponding t-statistics (in brackets).

Panel A: Order aggressiveness index

2003 2004 2005 2006 2007 2008

Foreign institutions 0.276 0.174 0.325 0.733 1.128 1.676

Individuals 7.0428 9.691 7.016 6.582 7.344 6.892

Domestic institutions 2.807 4.122 2.267 3.746 3.798 3.996

Panel B: Vector autoregression estimates for order aggressiveness index and market efficiency

Foreign institutions Individuals Domestic institutions

Efficiency Aggsn. Efficiency Aggsn. Efficiency Aggsn.

Efficiency (-1) -0.734 0.261 -0.867 -2.631 -0.733 1.549

[-5.443] [ 0.194] [-6.277] [-0.725] [-5.450] [0.181]

Efficiency (-2) -0.277 0.854 -0.475 1.345 -0.412 4.761

[-2.137] [1.668] [-3.102] [ 1.334] [-2.904] [0.535]

Aggressiveness (-1) 0.086 0.462 0.002 0.584 0.015 0.247

[2.452] [ 3.504] [ 1.042] [ 4.349] [2.216] [1.699]

Aggressiveness (-2) 0.013 -0.113 -0.004 0.112 0.004 -0.013

[ 1.025] [-0.878] [-1.390] [ 0.876] [0.661] [-0.009]

Limit_order (-1) 2.458 2.534 1.971 -14.611 0.741 4.743

[ 1.825] [1.557] [1.477] [-1.082] [0.469] [1.251]

Limit_order (-2) 2.012 8.658 -0.509 -39.334 2.518 -7.312

[ 1.522] [ 0.528] [-0.818] [-1.583] [ 1.965] [-0.759]

Market Efficiency and Foreign Institutional Trading ... · 2. The data 2.1 Sample selection and market description Our transaction data is obtained from TAIFEX, which is one of the

Documents