Академический Документы
Профессиональный Документы
Культура Документы
12137
Robin K. Chou
Department of Finance, National Chengchi University
Keng-Yu Ho
Department of Finance, National Taiwan University
Pei-Shih Weng**
Department of Finance, National Dong Hwa University
Abstract
Using a large data set with detailed classifications of different trader types, this study sheds
further light on the trading activity of various trader types on the Taiwan futures exchange
(TAIFEX). Compared with domestic traders on the TAIFEX, we find that foreign institutional
traders act as contrarians. In addition, when the market becomes illiquid, foreign institutional
traders are net buyers and individual traders are net sellers. The result implies that both for-
eign institutional traders and individual traders may contribute to the recovery of liquidity
dry-ups together. Among all traders, only the order imbalance of foreign institutional
*The suggestions and advice of two anonymous referees are gratefully acknowledged. Ying
Hao gratefully acknowledges the financial support provided by the National Natural Science
Foundation of China (Grant Nos. 71372137 and 71232004) and the Fundamental Research
Funds of the Central Universities of China (Grant Nos. CD JSK11002 and
106112015CDJXY020010). Robin K. Chou gratefully acknowledges financial support from the
Ministry of Science and Technology of Taiwan (MOST 101-2410-H-004-067-MY3) and from
the National Natural Science Foundation of China (No: 71232004 and 71373296). Pei-Shih
Weng would like to express his gratitude to the Ministry of Science and Technology of Tai-
wan for its financial support (MOST 102-2410-H-259-075-MY2 and MOST 104-2410-H-259-
008). Any remaining errors are the authors’ own.
**Corresponding author: Pei-Shih Weng, Department of Finance, College of Manage-
ment, National Dong Hwa University, Hualien 97401, Taiwan. Tel: +886-3-8633148,
Fax: +886-3-8633148, email: psweng@mail.ndhu.edu.tw.
investors presents return predictability when the market is illiquid, which suggests they have
an information advantage in such a market.
Keywords Domestic institutional traders; Foreign institutional traders; Individual traders;
Liquidity; Order imbalance
JEL Classification: G12, G14
1. Introduction
Although previous studies have explored the associations between trading activity,
liquidity, and market returns (Hasbrouck and Seppi, 2001; Chordia et al., 2002,
2005, 2008), these studies use aggregate market data with no analysis of the trading
activities of different trader types. Although many studies examine the differences
between institutional and individual investors (De Long et al., 1990; Cohen et al.,
2002; Griffin et al., 2003), they focus solely on local traders in developed markets
and ignore foreign institutional traders. Indeed, the role of foreign institutional tra-
ders, particularly in emerging markets, is less explored. Among these types of tra-
ders, foreign institutional traders in emerging markets draw much greater attention
than their counterparts in developed markets. Although the number of studies on
foreign institutional traders is growing, findings remain inconclusive. Foreign insti-
tutional investors may be less familiar with local markets even though they are
often sophisticated traders from developed markets. Alternatively, foreign institu-
tional traders may be better informed than the local traders due to their better
sense of the global financial market. Thus, recent empirical findings have not settled
the debate.1 By investigating the association between trading activity, market liquid-
ity, and market returns for various trader types, with a special interest in the role of
foreign institutional investors, our study adds to the discussion and provides an
investment reference for investors in the emerging market.
We use a large and longer data set with a detailed classification of different tra-
der types on the Taiwan futures exchange (TAIFEX). The detailed trade data make
it possible to investigate trading activities for domestic individual traders, foreign
institutional traders, and domestic institutional traders, respectively.
To capture better the information content of the trading activity of various tra-
der types, we measure trading activity by order imbalance. Although past research
commonly uses trading volume to measure trading activity, numerous recent stud-
ies use order imbalance (Chordia et al., 2008; Easley et al., 2008; Subrahmanyam,
2008; Barber et al., 2009a; O’Hara et al., 2011). Following these studies, we reason
1
For instance, Grinblatt and Keloharju (2000), Froot et al. (2001), and Froot and Ramadorai
(2008) find evidence that foreign investors’ trades lead to price movements, implying that
foreign institutions are well-informed. Conversely, Choe et al. (2005) find no such evidence
in Korea. Similarly, at the market level, Griffin et al. (2004) show that after controlling for
the contemporaneous relation between flows and returns, foreign investors are generally not
able to time the market at a daily frequency.
that reported volume alone may conceal the trading direction and may be a noisy
proxy for order flow.2 Related studies using data from the TAIFEX also widely con-
sider order imbalance to examine trading behavior. For instance, Huang and Chou
(2007) provide a comprehensive analysis of the TAIFEX and the Singapore deriva-
tives exchange and compare the impact of the behavior arising from different trad-
ing mechanisms on market efficiency.
In the literature, Chordia et al. (2002) and Lee et al. (2004) are closely related to
this study. Our empirical setting follows the models of Chordia et al. (2002), but we
further extend the discussion regarding order imbalances by various trader types,
while their focus is on aggregated market trading. In addition, because transaction
data with detailed order information were not available to Chordia et al., they must
estimate whether an order is buyer-initiated or seller-initiated by the specific algo-
rithm, which is a task unavoidably induces estimation errors. By contrast, our data set
discloses precise indicators that whether an order is buyer-initiated or seller-initiated
and can avoid such estimation errors. Lee et al. (2004) study order imbalance on the
Taiwan Stock Exchange (TWSE) from September 1996 through April 1999. With
detailed transaction data similar to ours, they identify trader types and investigate the
association between order imbalance and market returns. However, their study does
not address the role of market liquidity. As mentioned, Huang and Chou (2007) also
conduct a comprehensive analysis of the relationship between order imbalances,
liquidity, and returns for the TAIFEX, using an approach similar to that of Chordia
et al. (2002). However, their focus is on aggregated market trading, while we discern
the difference in trading activities between various types of traders.
We perform several stages of analysis in this study. The first stage of our analy-
sis, which is similar in spirit to the analysis of Chordia et al. (2002), presents how
order imbalance emanating from various trader types would react to different states
of previous market return and liquidity. The second stage of our analysis focuses
on return predictability of order imbalance, which allows us to discern the informa-
tion content (if any) of trading behavior by each trader type. The third stage of our
analysis extends the second stage to further investigate return predictability of order
imbalance in an illiquid market.
To summarize the major findings on foreign institutional traders, we show that
foreign institutional traders are essentially different from domestic traders, as for-
eign institutional traders are contrarians. In addition, when the market becomes
illiquid, foreign institutional traders are net buyers and individual traders are net
sellers. Our empirical findings also suggest that the order imbalance of foreign
2
Chordia et al. (2002, p. 112) describe a simple and clear case: “Consider, for example, a
reported volume of one million shares. At one extreme, this might be a million shares sold
to the market maker while at the other extreme it could be a million shares purchased. Per-
haps more typically, it would be roughly split, about 500,000 shares sold to and 500,000
shares bought from the market maker. Each scenario has it own specific implications for
price movement or liquidity changes.”
institutional investors is positively associated with market returns – their trades pre-
sent return predictability when the market is illiquid. Overall the results suggest
that, compared with domestic traders on the TAIFEX, foreign institutional traders
are net buyers and relatively better informed in illiquid markets. They also help to
stabilize prices. These characteristics are quite different from what we observe from
domestic traders. Our study hence sheds further light on the behavior of foreign
institutional investors in emerging markets, particularly when the market is illiquid.
Some recent studies have also examined the trading behavior of various types of
investors using data from the TAIFEX. Our findings are comparable to the results
of these studies.3 For example, Li et al. (2013) examine the disposition effect for
different types of traders on the TAIFEX. They find that foreign institutional traders
are less likely to show the disposition effect, and they also show that foreign institu-
tional traders are better informed than other types of traders. Similarly, Wang
(2014) studies the order-splitting of various types of traders, and finds that foreign
institutional traders have a high tendency to split their orders, implying that foreign
institutional traders are well informed and try to protect their information when
they trade. Further, Kuo et al. (2015) study the association between trading imbal-
ances and future returns for domestic individual traders, foreign institutional tra-
ders, and domestic institutional traders on the TAIFEX. They show that among
these different types of traders, foreign institutional traders are the best informed,
which is consistent with the studies cited above. While our study finds similar
results regarding the information content of foreign institutional trading, it also
focuses on a particular issue that is ignored in these other studies: how the trading
activities of various types of traders change with previous market price movements
and different conditions of market liquidity.
Chiu et al. (2014) study the order submission behavior of different types of tra-
ders. They find that foreign and domestic institutional traders, but particularly for-
eign institutional traders, tend to submit more limit orders on the TAIFEX. Hao
et al. (2015) find similar results. These studies imply that foreign institutional tra-
ders may act as liquidity providers on the market, and their findings are similar to
our findings that foreign institutional traders are net buyers when the market is
illiquid. However, Chiu et al. (2014) and Hao et al. (2015) do not address the
information role of different traders. Further, for trading patterns, Chou et al.
(2015) study the intraday trading activities of day traders and non-day traders. They
report that individual day traders behave as contrarians, while individual non-day
traders act as momentum traders. Chou et al. (2015) examine intraday trades, but
our study analyzes trading activities on a daily basis. We also find that individual
traders tend to buy when the market is advanced, acting partly as momentum tra-
ders. Our study and that of Chou et al. (2015) uncover several similar patterns,
although Chou et al. do not study the information content of the various types of
trader on the TAIFEX.
3
We thank a referee for suggesting the comparison.
While previous studies use different approaches and different measures to study
the information content and trading activities of various types of traders on the
TAIFEX, by using different sample periods, our study attempts to look at both
issues by using a single measure—order imbalance—to obtain a comprehensive
understanding of the trading activities of different types of traders on the TAIFEX.
Furthermore, though in an emerging financial market like the TAIFEX, market liq-
uidity could suddenly dry up, previous studies have not considered the conse-
quences of such an event. Given this lack of research, our analysis addresses how
the trading activities of each type of trader change in an illiquid market. In sum,
we believe that our study can make a useful contribution to the field as we address
topics that have seldom been explored in previous studies.
The remainder of this paper is organized as follows. Section 2 develops our
hypotheses. Section 3 describes the data. Section 4 presents the main empirical
results, and Section 5 reports the findings of additional analyses. Finally, Section 6
concludes the paper.
4
We do not make this presumption, because the literature is not conclusive. For example,
Grinblatt and Keloharju (2000) present firm evidence that foreign institutional investors and
domestic institutional investors in Finland trade with significantly different patterns. While
Lee et al. (2004) and Barber et al. (2009c) also report a difference between foreign and
domestic institutional investors in the Taiwan stock market, both Wang (2014) and Chou
et al. (2015) show that domestic institutional traders and foreign institutional traders behave
similarly in their order submissions and order-splitting. The empirical evidence on whether
foreign or domestic investors are more informed is also mixed (Froot et al., 2001; Choe
et al., 2005; Dvorak, 2005; Froot and Ramadorai, 2008; Li et al., 2013; Kuo et al., 2015).
5
Barber and Odean (2011) also state that buying is forward-looking and selling backward-
looking. Individuals buy stocks because of what they hope will happen and sell stocks because
of what has already happened. If so, individuals are not contrarians, but institutions, who in
aggregate are trading against individuals, would act as contrarians.
market liquidity is relevant when examining the trading activities of various types
of traders on the TAIFEX.
Anand et al. (2013) suggest that some institutional investors are able to insulate
themselves and earn a premium by providing liquidity when market liquidity
declines. Further, undervalued prices may be more likely in illiquid markets because
of sentimental-selling activities by irrational investors. According to the findings of
Anand et al. (2013), we can expect that institutional traders, who are usually classi-
fied as rational investors, may be engaged in long positions when the market
becomes illiquid. By contrast, since individual traders are very frequently character-
ized in the previous literature as irrational traders, an illiquid market should trigger
individual traders to sell more shares. Our expectation is similar to that of Baker
and Stein (2004), who argue that when irrational investors observe the trading deci-
sions of rational investors, they tend to underreact to the information contained in
those decisions. This means that overconfidence encourages individual traders to
trade against institutional traders in an illiquid market. Grinblatt and Keloharju
(2009) analyze both sensation and overconfidence as mechanisms that lead to trad-
ing, while Odean (1998) and Kyle and Wang (1997) apply a similar argument to tie
liquidity to overconfidence.
In sum, for the relation between lagged market liquidity and subsequent trading
activity, we hypothesize:
H2: When the market is illiquid, individual investors tend to continue to sell,
whereas institutional investors trade in the opposite position.
6
Griffin et al. (2003) also examine such a relation for institutional trading imbalances. They
find a positive relation between institutional trading imbalance and returns, which is consis-
tent with most of the literature.
More recently, Kuo et al. (2015) present evidence that trades by institutional
investors positively predicted returns over short horizons on the TAIFEX during the
period from 2002 to 2005, while individual trading had no significant impact on
future returns. Most related studies find that institutional traders are informed traders
and individual traders are noisy traders. For instance, Li et al. (2013) show that both
domestic and foreign institutional traders have a lower tendency to show the disposi-
tion effect and have a better trading performance because of their information advan-
tage. Wang (2014) also shows that institutional traders, when compared to individual
traders, are more likely to split orders for the purpose of hiding their information
when they trade. Both these studies suggest that individual traders are uninformed
and institutional traders are informed. In this sense, we can expect that institutional
trading will be positively associated with returns and that individual trading will have
a negative or no effect on returns. In sum, we formulate our hypotheses regarding the
relationship between returns and order imbalance as follows:
H3a: Market returns are positively related to order imbalances of institutional tra-
ders.
H3b: Market returns are negatively related (or unrelated) to order imbalances of
individual traders.
Because short-horizon return predictability from order flows can be an inverse
indicator of market efficiency, widespread agreement exists in the literature that
increased liquidity tends to increase market efficiency and to make returns more
unpredictable, although there is an ongoing debate about return predictability for
each investor class. For instance, Chordia et al. (2008) compare the return pre-
dictability from order imbalance across different liquidity regimes in the market,
and argue that such predictability diminishes as the market becomes more liquid.
In addition, Chung and Hrazdil (2010) find that less information is incorporated
into the formulation of prices when the market is illiquid. Because the predictability
of returns can be much higher in an illiquid market, we also examine Hypotheses
3a and 3b under an illiquid scenario.
3. Data
3.1. Description of the TAIFEX and Sample Selection
The rapid growth in trading volume has made the TAIFEX one of the major emerg-
ing derivatives exchanges in the world. At the end of 2008, TAIFEX was ranked
17th among 52 derivatives exchanges reported to the Futures Industries Associa-
tion.7 Taiwan is also the fifth largest emerging economy in the Asia-Pacific area.8
7
The TAIFEX is ranked eighth among emerging markets.
8
International Financial Statistics of the International Monetary Fund places Taiwan fifth on
its list of emerging Asia-Pacific countries in 2008.
9
Our sample period ends in 2008 because in 2008 the TAIFEX stopped releasing data sets
that identified the type of trader and included some other relevant items, such as the trading
direction, for researchers. Studies using similar data, such as Chou et al. (2015), Wang
(2014), and Chiu et al. (2014), also choose a sample period ending in 2008.
10
In addition to domestic individuals, domestic institutions, and foreign institutions, some
trades are placed by foreign individuals. However, the trades are too few to construct a
meaningful group and so we do not include them in our sample.
11
We also conduct empirical examinations using number of transactions and number of con-
tracts size. All three measures yielded qualitatively similar results.
tests, we calculate order imbalance using only market orders or limit orders.12
To make the coefficients of regressions easier to compare across different agents,
we scale the order imbalance for certain trader types by averaging the absolute
order imbalance of that type.
3.2.2. Market Index.
Because we study trading activity on the futures market, we use the TXF index as the
market index. Instead of using the index directly, we calculate the volume-weighted
TXF index using the spot futures index and the next-month futures index.13 Because
both contracts are the most actively traded futures products on the TAIFEX, our
market index can better reflect market-wide information for most trades.14
3.2.3. Market Liquidity.
Liquidity is not a simple concept that can be directly observed. However, it is gener-
ally associated with the price impact induced by trades. Our daily liquidity measure
is based on Amihud’s (2002) measure of security illiquidity, which is calculated as
the ratio of the absolute value of daily return over the dollar volume. The Amihud
measure corresponds to the notion of price impact. Previous studies (Korajczyk and
Sadka, 2008) report that many measures of liquidity, especially the Amihud measure,
are highly correlated and are driven by a common systematic component. Owing to
the illiquidity expression of the Amihud measure, for ease of disposition, we add a
negative sign to the Amihud measure (in log) as our liquidity measure:
jrm;t j
LIQ ¼ log ;
DVOLt
where Rm,t is market return based on TXF index at date t, and DVOLt is total trad-
ing volume in term of contract dollar value at date t. Usually the unit of trading
volume on the futures market is the quantity of contracts because traders pay a
margin for per futures contract they purchase, instead of paying full contract dollar
value. Because the denominator in the original Amihud measure is the dollar trad-
ing volume, we calculate and sum the dollar value of all TXFs traded during a trad-
ing day as daily total trading volume.15
12
We classify the order into market order and limit order based on the following algorithm.
When a transaction is just being completed, the highest unexecuted bid price and the lowest
unexecuted ask price become prevailing quotes. Thus, we can use them to define a market
order for the next transaction. Any subsequent order to buy at or below the prevailing ask or
to sell at or above the prevailing bid and the quantity needed in this order can be cleared;
such an order is deemed a market order. Otherwise, the order is categorized as a limit order.
13
The weights are proportional to daily trading volume of each futures contract.
14
We also use spot TXF index and next-month TXF index as the market index, respectively.
The results are mainly consistent with those reported based on the volume-weighted TXF
index. The results are available upon request.
15
In unreported results, we measure liquidity using total trading quantity and produce similar
results.
Table 1 Summary statistics: trading volume and proportion of limit and market order
This table presents the trading volume and trading proportion of different trader types. Market orders
are defined as 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. Limit orders are counterparts
of market orders. The numbers in brackets under the total volume for certain trader types are the trading
percentage in total aggregate volume. The numbers in parentheses under number of limit (market) orders
for certain trader types present the percentage of limit (market) orders in that trader type over total vol-
ume. Data are from 2003–2008 inclusive with 1484 daily observations.
Foreign Domestic
All classes institutions Individuals institutions
Number of transactions
Total volume 18 734.88 1160.47 14 690.35 2884.06
[6.20%] [78.41%] [15.39%]
# of limit orders 11 110.84 825.50 8216.75 2068.59
% of limit orders (59.38) (70.03) (56.38) (71.33)
# of market orders 7624.04 334.97 6473.60 815.47
% of market orders (40.62) (29.97) (43.62) (28.61)
Number of quantities
Total volume 35 414.68 2892.76 25 519.75 7002.18
[8.17%] [72.06%] [19.77%]
# of limit orders 19 574.59 1818.37 13 333.05 4423.17
% of limit orders (55.20) (61.69) (52.48) (63.24)
# of market orders 15 840.09 1074.39 12 186.70 2579.01
% of market orders (44.80) (38.31) (47.52) (36.76)
Number of dollars (trillions)
Total volume 47.96 4.23 34.21 9.52
[8.82%] [71.33%] [19.85%]
# of limit orders 26.56 2.69 17.85 6.02
% of limit orders (55.20) (61.69) (52.48) (63.24)
# of market orders 21.40 1.55 16.36 3.49
% of market orders (44.80) (38.31) (47.52) (36.76)
4. Empirical Results
4.1. Order Imbalance Emanating from Various Trader Types
To test Hypotheses 1 and 2, we respectively regress the order imbalances emanating
from foreign institutional traders, individual traders, and domestic institutional tra-
ders on lagged market returns and lagged market liquidity after controlling for past
lagged order imbalance.16 To ensure that the correlations between the explanatory
16
Previous studies recognize the persistence in order imbalance as a stylized feature (Chordia
et al., 2002; Lee et al., 2004). We add lagged order imbalance up to five lags in the regressions to
make sure the attendant autocorrelations do not affect the estimates. However, our results show
that the results of the estimates remain similar without controlling lagged order imbalance.
variables do not affect our regression results and to alleviate possible collinearity
among predictors, we perform regressions by incorporating different sets of
explanatory variables successively.17 Table 3 reports the empirical results.
Panel A of Table 3 reports the regression results for the order imbalance of foreign
institutional investors. The findings suggest that foreign institutional traders act as
contrarians. Except for Model 7, contrarian trading persists for up to at least 2 days.
The trading activity of foreign institutional investors tends to reduce the market
volatility because foreign institutional investors buy after the market declines and sell
after the market advances. More importantly, the coefficients of lagged market
illiquidity on order imbalance are negative among all model specifications, suggesting
that foreign institutional traders are net buyers after market liquidity shrinks. Panel B
reports regression results for the order imbalance of individual traders. The trading
patterns reacting to lagged market returns are mixed. Trend-chasing trading occurs
right after up-market moves but reverse trading occurs right after down-market
moves. The finding is consistent with our argument in Section 2 that the literature has
not reached a conclusion on whether to classify individual traders as momentum tra-
ders or contrarians. These findings are also consistent with the statistics reported in
Table 2 showing that individual traders tend to be net buyers on the TAIFEX.
As for the association with lagged market liquidity, we find that individual tra-
ders are eager to sell when the market becomes illiquid, but find no distinct trading
patterns for individual traders when the market is liquid. Panel C of Table 3 reports
regression results for the order imbalance of domestic institutional traders. We find
no significant reaction for domestic institutional traders when the market increases;
however, domestic institutional traders tend to be net sellers when the market
declines. This finding suggests that domestic institutions may trade against individ-
uals and foreign institutions when the market goes down. As for the reaction of
domestic institutional trading to market liquidity, we find no significant evidence to
reach a solid conclusion.
Overall, the main results in Table 3 are consistent with Hypothesis 1 that for-
eign institutions act as contrarians on the TAIFEX, although the same cannot be
seen with domestic institutions. Individuals act according to different market states,
and such an asymmetry is quite intriguing. The trading pattern may reveal the irra-
tional feature of individual investors: they tend to be overoptimistic about the
future and thus continue to buy regardless of up- or down-market conditions.
The findings in Table 3 are also consistent with Hypothesis 2 that after the mar-
ket liquidity shrinks foreign institutions and individuals are net buyers and net sell-
ers, respectively. Interestingly, given the possibility that negative sentiment may
trigger individuals to sell, and cause the market prices in the illiquid state to be
undervalued, the trading behavior of foreign institutional traders and individual tra-
ders may provide sensible explanations for why individual traders lose to foreign
17
We also examine collinearity by using variance inflation factors (VIF), and all cases have a
VIF of less than five, meaning that our results do not induce serious collinearity problems.
Foreign Domestic
All classes institutions Individuals institutions
Table 2 (Continued)
Foreign Domestic
All classes institutions Individuals institutions
institutional traders on the local market. While our findings about foreign institu-
tional trading can support our expectation regarding Hypothesis 2, again we cannot
find the same evidence for domestic institutions. Testing Hypotheses 1 and 2 indi-
cates that domestic institutional traders and foreign institutional traders behave
quite differently on the TAIFEX.
In sum, the regression results using order imbalance emanating from various
trader types show that foreign institutional traders help to stabilize the market via
contrarian trading. In addition, domestic traders and foreign institutional traders
on the TAIFEX trade in opposite directions when the market becomes illiquid.
level. The order imbalance (OIB) is in trading volume (quantity). All OIB variables are scaled by the average absolute imbalance level of the trader type. R denotes the
daily returns based on a mixed futures index, which is quantity-weighted from the futures index of the current contract month and the futures index next to the cur-
rent contract month. The logarithm market liquidity level (LIQm) is –1 multiplied by the logarithmic illiquidity measure of Amihud (2002). HLIQ and LLIQ present
the high and low market liquidity, which are defined as Max[LIQm – mean of LIQm , 0] and Min[LIQm – mean of LIQm , 0], respectively. The Yule–Walker procedure
is applied to adjust for the series dependence in the residuals. R2 is the transformed R2, which only incorporates the structural part the model but not the lagged resid-
uals. Data are from 2003–2008 inclusive with 1484 daily observations. The standard errors are in parentheses. ***, **, and *represent the 1%, 5%, and 10% significance
levels, respectively.
Explanatory Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat)
variables (1) (2) (3) (4) (5) (6) (7)
Explanatory Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat)
variables (1) (2) (3) (4) (5) (6) (7)
515
Table 3 (Continued)
516
Explanatory Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat)
Y. Hao et al.
Max(0, Rt–3) 0.073 (0.04)* 0.035 (0.03) 0.256 (0.06)*** 0.049 (0.06)
Max(0, Rt–4) 0.046 (0.04) 0.017 (0.03) 0.181 (0.06)*** 0.009 (0.06)
Max(0, Rt–5) 0.034 (0.03) 0.027 (0.03) 0.053 (0.06) 0.084 (0.06)
OIBt–1 0.379 (0.03)*** 0.372 (0.03)*** 0.375 (0.03)*** 0.341 (0.03)***
OIBt–2 0.182 (0.03)*** 0.193 (0.03)*** 0.184 (0.03)*** 0.182 (0.03)***
OIBt–3 0.120 (0.03)*** 0.103 (0.03)*** 0.118 (0.03)*** 0.103 (0.03)***
OIBt–4 0.039 (0.03) 0.04 (0.03) 0.042 (0.03) 0.04 (0.03)
OIBt–5 0.05 (0.03)* 0.055 (0.03)* 0.049 (0.03)* 0.057 (0.03)*
LLIQt1 0.141 (0.06)*** 0.112 (0.06)* 0.299 (0.11)*** 0.331 (0.11)***
LLIQt2 0.095 (0.06) 0.062 (0.06) 0.404 (0.11)*** 0.124 (0.11)
LLIQt3 0.035 (0.06) 0.015 (0.05) 0.091 (0.10) 0.037 (0.10)
LLIQt4 0.03 (0.05) 0.038 (0.05) 0.038 (0.10) 0.023 (0.11)
LLIQt5 0.015 (0.05) 0.041 (0.06) 0.087 (0.10) 0.106 (0.11)
HLIQt1 0.015 (0.04) 0.018 (0.04) 0.006 (0.04) 0.005 (0.04)
HLIQt2 0.038 (0.04) 0.035 (0.03) 0.019 (0.04) 0.037 (0.05)
HLIQt3 0.039 (0.03) 0.027 (0.04) 0.026 (0.04) 0.036 (0.04)
HLIQt4 0.058 (0.04) 0.036 (0.04) 0.043 (0.04) 0.035 (0.04)
HLIQt5 0.049 (0.04) 0.024 (0.03) 0.052 (0.04) 0.034 (0.04)
Week day Included but not reported
R2 (%) 7.67 35.29 2.68 39.44 39.99 18.99 45.29
Panel C: Domestic institutional investors’ order imbalance
Intercept 0.073 (0.12) 0.087 (0.08) 0.089 (0.15) 0.127 (0.1) 0.173 (0.11) 0.139 (0.15) 0.204 (0.13)
Min(0, Rt–1) 0.167 (0.04)*** 0.07 (0.04)* 0.153 (0.08)* 0.137 (0.08)*
Explanatory Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat) Coeff. (t-stat)
variables (1) (2) (3) (4) (5) (6) (7)
Min(0, Rt–2) 0.10 (0.04)** 0.002 (0.04) 0.009 (0.08) 0.124 (0.08)
Min(0, Rt–3) 0.243 (0.04)*** 0.193 (0.04)*** 0.332 (0.08)*** 0.283 (0.08)***
Min(0, Rt–4) 0.016 (0.04) 0.118 (0.08) 0.031 (0.08) 0.127 (0.08)
Min(0, Rt–5) 0.013 (0.04) 0.063 (0.05) 0.023 (0.08) 0.054 (0.08)
517
R2 (%) 8.26 18.14 1.34 23.82 20.60 8.54 24.25
518
Y. Hao et al.
Intercept 0.196 (0.07)*** 0.207 (0.07)*** 0.155 (0.06)** 0.152 (0.06)** 0.118 (0.07)* 0.056 (0.07)
Excess buy orders, 0.143 (0.06)** 0.095 (0.05)* 0.399 (0.05)***
Max[0, OIBt]
Excess sell orders, 0.228 (0.04)*** 0.714 (0.19)*** 0.611 (0.04)***
Min[0, OIBt]
Lagged excess buy orders, 0.029 (0.06) 0.009 (0.05) 0.037 (0.06) 0.012 (0.06) 0.021 (0.04) 0.140 (0.06)**
Max[0, OIBt1]
Lagged excess sell orders, 0.048 (0.04) 0.431 (0.19)** 0.305 (0.05)*** 0.062 (0.04)* 0.228 (0.18) 0.048 (0.05)
Min[0, OIBt1]
Lagged positive return, 0.078 (0.05) 0.075 (0.05) 0.103 (0.05)** 0.094 (0.05)* 0.086 (0.05)* 0.125 (0.05)**
Max[0, Rt1]
Lagged negative return, 0.020 (0.05) 0.008 (0.05) 0.036 (0.04) 0.006 (0.05) 0.022 (0.05) 0.034 (0.05)
Min[0, Rt1]
Adjusted R2 0.064 0.017 0.229 0.0054 0.0047 0.0071
Rt in the months with average monthly mar- Rt in the months with average monthly market liq-
ket liquidity in the top 10% group uidity in the bottom 10% group
519
Y. Hao et al.
18
Aggregate trading volume measured in number of quantity or transaction yields results sim-
ilar to those reported in Panel B of Table 6.
521
522
Y. Hao et al.
Table 6 (Continued)
5. Additional Analyses
5.1. Generalized VAR Analysis and Impulse Responses
Since we analyzed separately the relation between lagged market returns and liquid-
ity and order imbalances of the various trader types, and between lagged order
imbalances of the various trader types and future returns, one might wonder
whether and how the associations between all these variables would affect our find-
ings in Section 4. Therefore, we re-estimated the results in Tables 3 and 4 using the
generalized VAR (GVAR) method proposed by Koop et al. (1996) and Pesaran and
Shin (1998). From our use of the GVAR, we can better understand the dynamic
relationship between market returns, market liquidity, and order imbalances of the
various trader types, using the information about the generalized impulse response
functions.19 In estimating GVAR, a suitable number of lags for the model has first
to be defined. We consider AIC and SIC as the selection criteria and set the lag
number at seven to estimate GVAR.20 In addition, since we consider seven variables
in the GVAR (two excess market returns [positive and negative], market liquidity
[high and low], and order imbalances of foreign institutional traders, domestic
institutional traders, and individual traders), and lag numbers up to seven, we only
report the coefficients for the first three lags for all the variables, to shorten our
table.21 We first report our estimate for GVAR in Table 6, and then report the esti-
mates for the generalized impulse responses in Table 7.
Panel A of Table 6 presents the coefficients for the order imbalances of different
types of trader to lagged market returns. As can be seen, the reactions of foreign
institutional investor order imbalances to previous market returns imply that for-
eign institutional traders are contrarians. By the same token, the reactions of indi-
vidual investor order imbalances to previous market returns indicate that individual
traders are net buyers, regardless of whether the market conditions are going up or
down. Both findings are consistent with our results in Table 3. Interestingly, when
19
The advantage of generalized impulse response functions is their invariance to the ordering
of variables, enabling them to be used to compare the magnitude of standard deviation
shocks for the independent variables and the dependent variables. In this way, we can com-
pare the impact of order imbalance on returns for the various types of investors simultane-
ously. The authors appreciate the suggestion from an anonymous referee that we follow this
approach in our further analysis.
20
In the pre-estimate test, we find that AIC suggests that a suitable number of lags is three,
and that SIC suggests six. We repeat the estimate using between three and six lags, and find
the results of the estimate to be qualitatively similar. In addition, we find that all variables
are stationary without differentiating in advance.
21
Most of the coefficients for more than three lags are also not statistically significant.
Table 7 Impulse response to a generalized one S.D. innovation for various variables
This table reports the results of the impulse responses of order imbalance for different types of traders
and market returns to a generalized one S.D. innovation for various variables that have been tested by
GVAR analysis in Table 6. OIB_F, OIB_I, OIB_D, MAX_R, MIN_R, HLIQ, and LLIQ are defined as in
Table 6. The estimated period is 10 trading days for each relation, and the sum of the coefficients in all
the periods is reported in the cells.
Table 3 shows only that domestic institutional traders sell more after the market
declines, Table 6 shows that they also buy more after the market advances. This
result implies that domestic institutional traders act as momentum traders, the
opposite of foreign institutional traders.
Panel B of Table 6 further reports the coefficients of order imbalances of the
different types of trader to lagged market liquidity. Consistent with our findings
in Table 3, after market liquidity decreases, foreign institutions are still the only
traders who act as net buyers. Individual traders tend to act as net sellers at the
same time, as is reported in the results in Table 3, but the coefficients here are
not significant across all values of lagged market liquidity. Instead, domestic insti-
tutional traders have a stronger tendency to trade against foreign institutional tra-
ders when the market is illiquid, since the coefficient of OIB_Dt to LLIQ t-2 is
positive and statistically significant. In addition, Panel B also shows that foreign
institutional traders behave as net sellers when the market is relatively liquid,
because the coefficient of OIB_Ft to HLIQt2 is negative and statistically signifi-
cant. In general, the overall findings shown in Panel A and Panel B of Table 6
are similar to our previous findings and support the argument that foreign insti-
tutional traders behave as contrarians and tend to help with the recovery of liq-
uidity when the market is illiquid, and that domestic traders (institutions or
individuals) still increase the selling pressure.
Finally, we move our focus to consider whether lagged order imbalances are
related to market returns. Panel C shows that order imbalances for both domestic
and foreign institutions are positively related to future market returns, whereas for-
eign institutional trading has a more persistent influence on market returns than do
domestic institutions. By contrast, individual trading is negatively related to future
returns. In sum, Panel C of Table 6 provides more evidence to support Hypotheses
3a and 3b: institutional traders, particularly foreign ones, are more likely to be
informed, and individuals are noisy traders.22
As mentioned previously, in Table 7 we provide the estimates of the general-
ized impulse response functions to the major relations of GVAR, including the
responses of order imbalance to a generalized one standard deviation innovation
in market returns and liquidity, and the responses of market returns to a general-
ized one standard deviation innovation in order imbalances for each trader type.
The coefficients of the impulse responses are up to 10 trading days. To shorten
the table, we report the sum of the coefficients for all the periods. We highlight,
in particular, the estimates that are comparable to the findings in previous sec-
tions. First, for either an up- or a down-market shock, foreign institutions
respond negatively to market returns. At the same time, domestic institutions
have the opposite response to the market, and the responses of individual traders
are mixed.
Next, a one standard deviation innovation in market illiquidity (LLIQ) has an
impact of around 0.4 on foreign institutional trading, whereas the impulses are
positive for domestic institutions and individual traders, but with lower strength.
This finding indicates that the influence of market illiquidity on whether foreign
institutions act as net buyers is greater than it is on whether other traders do.
Finally, among all the traders, a one standard deviation innovation in order
imbalances for foreign institutional investors has the greatest positive impact
(0.166) on positive market returns, whereas domestic institutions have the second
greatest positive impact (0.149) and the impact of individual traders is very small
(0.018). For negative market returns, a one standard deviation innovation in the
order imbalance of foreign institutions also has the greatest positive impact
(0.677), whereas the impact of domestic institutions is much smaller (0.114) than
that of foreign ones. Conversely, the coefficient for a one standard deviation inno-
vation in the order imbalance of individual traders to negative market returns is
0.531.
22
Panel C also confirms the persistence of order imbalance autocorrelation for each trader
type.
23
We also thank the anonymous referee who suggested this analysis.
Table 8 Market returns and lagged order imbalance in extreme market return states
This table examines the relation between market returns and lagged order imbalance in high and low
return states for each trader type by quantile regression. The dependent variable is daily market returns.
The seventh quantile and the third quantile are used to define extreme market returns. Explanatory vari-
ables include the contemporaneous and lagged daily excess buy orders (Max[0, OIBt]) and excess sell
orders (Min[0, OIBt]) by trader type, the lagged total dollar market volume and the lagged market
returns. Data are from 2003–2008 inclusive with 1484 daily observations. The standard errors are in
parentheses. ***, **, and * represent the 1%, 5%, and 10% significance levels, respectively.
extreme market returns scenarios, we use the seventh quantile against the third
quantile to conduct the estimates.24 The results are reported in Table 8.
As reported, in a state of high market returns, neither trading by individuals
nor trading by foreign institutions has an impact on returns. Although the coeffi-
cient of lagged excess buy orders for domestic institutions is positively significant
(the lagged excess sell is insignificant), the explanatory power of the model is weak,
at 0.006, which is essentially equal to zero. To sum up the findings, the return pre-
dictability in a period of high market returns is similar to that in a period of high
market liquidity.
Interestingly, in a state of low market returns, return predictability of order
imbalance can be found for each trader type. While the results in Table 5 show that
only order imbalance for foreign institutional traders is positively related to next-
day market returns in illiquid months, the results in Table 8 show that in a state of
24
The results using the ninth quantile against the first quantile and the eighth quantile against
the second quantile are similar to the reported results that use the seventh quantile against
the third quantile.
low market returns, order imbalances for both foreign institutions and domestic
institutions are positively related to next-day market returns. Moreover, we find
that order imbalance of individual traders is negatively related to next-day market
returns. The overall results in the low market returns scenario of Table 8 suggest a
similar conclusion to that reported in Table 5. Further, the findings in Table 8 pro-
vide indirect support for Hypotheses 3a and 3b.
25
We thank an anonymous referee for suggesting this analysis.
26
For brevity, the coefficients for foreign institutions and individual traders are not reported
here, since the results are qualitatively similar to those in Table 6.
528
This table performs additional analysis for two different types of domestic institutional trader: mutual funds and proprietary firms. Generalized VAR analysis is per-
formed as for Table 6, by replacing OIB_D with OIB for domestic mutual funds (OIB_M) and OIB for domestic proprietary firms (OIB_P) in the variable vectors. For
Y. Hao et al.
brevity, the coefficients of OIB_F and OIB_I are not reported in the table. All other variables are defined as in Table 6. Data are from 2003–2008 inclusive with 1484
daily observations. The standard errors are in parentheses. ***, **, and *represent the 1%, 5%, and 10% significance levels, respectively.
27
For each paired group, the trading status on the stock market has been controlled. The dif-
ferences for the two variables, if any, should be related to the different conditions of the
futures trading.
28
For brevity, we do not report the results of mean or median tests in the table.
This table reports the scheme to classify net selling (order imbalances) of foreign institutions into eight groups that represent hedging and non-hedging with different
trading levels. To measure different trading levels, if the order imbalance in a specific day exceeds the median level, we classify it as a large trade. If the order imbalance
does not exceed the median level we classify it as a small trade. We therefore define four hedging groups and four non-hedging groups with different trading sizes. The
classification can be seen in Table 6. Groups 5 to 8 are the hedge groups, while Groups 5 and 6 (Groups 7 and 8) is a comparable pair with a different level of net sell-
ing on the futures market and the same level of net buying on the stock market. The other four groups (Groups 1 to 4) are non-hedging groups, while Groups 1 and 2
(Groups 3 and 4) is also a comparable pair with a different level of net selling on the futures market and the same level of net selling on the stock market. Two vari-
ables, 1-day lagged market returns and liquidity, are applied for comparisons in each pair. All differences for the variables in each pair are statistically significant at the
10% level, except for the test for the Groups 1 and 2 pair on 1-day lagged market liquidity.
1 2 3 4 5 6 7 8
Large spot net Large spot net Small spot net Small spot Small spot Small spot Large spot Large spot
selling & large selling & small selling & net selling & net buying & net buying & net buying & net buying &
futures net futures net large futures small futures small futures large futures small futures large futures
Variables selling selling net selling net selling net selling net selling net selling net selling
6. Conclusion
By using a large data set with detailed classification of different trader types, this
study sheds further light on the order imbalance of various traders on the futures
market. We also make comparisons between foreign institutional traders and their
local competitors, which helps us to understand differences among traders. Our
major findings can be summarized as follows.
First, we find that, unlike other trader types, foreign institutional traders act as
contrarian traders to lagged market returns. This finding indicates that foreign insti-
tutions frequently trade against local traders on the TAIFEX. Second, individual tra-
ders and foreign institutional traders on the TAIFEX trade in opposite directions
when the market becomes illiquid. The finding suggests that both foreign institu-
tional traders and individual traders may contribute to the recovery of liquidity
dry-ups together.
Third, we find that market returns are positively associated with the order imbal-
ance of foreign institutional investors when the market is illiquid. At the same time,
other investors underreact to the information content in foreign institutional trading.
In sum, compared with domestic traders on the TAIFEX, foreign institutional
traders are only net buyers and relatively better informed in illiquid markets; more-
over, they help to stabilize prices. These characteristics are different from what we
observe in other domestic traders.
References
Amihud, Y., 2002, Illiquidity and stock returns: Cross-section and time series effects, Journal
of Financial Markets 5, pp. 31–56.
29
For brevity, we do not report the results using limit orders and market orders separately.
All the results are available upon request.
Anand, A., A. Puckett, P. J. Irvine, and K. Venkataraman, 2013, Institutional trading and
stock resiliency: Evidence from the 2007–2009 financial crisis, Journal of Financial
Economics 108, pp. 773–797.
Avramov, D., T. Chordia, and A. Goyal, 2006, Liquidity and autocorrelations in individual
stock returns, The Journal of Finance 61, pp. 2365–2394.
Baker, M., and J. C. Stein, 2004, Market liquidity as a sentiment indicator, Journal of
Financial Markets 7, pp. 271–299.
Barber, B. M., and T. Odean, 2011, The behavior of individual investors, SSRN eLibrary
(Available at SSRN: http://ssrn.com/abstract=1872211).
Barber, B. M., T. Odean, and N. Zhu, 2009a, Do retail trades move markets? Review of
Financial Studies 22, pp. 151–186.
Barber, B. M., T. Odean, and N. Zhu, 2009b, Systematic noise, Journal of Financial Markets
12, pp. 547–569.
Barber, B. M., Y.-T. Lee, Y.-J. Liu, and T. Odean, 2009c, Just how much do individual
investors lose by trading?, Review of Financial Studies 22, pp. 609–632.
Bloomfield, R., M. O’Hara, and G. Saar, 2005, The “make or take” decision in an electronic
market: Evidence on the evolution of liquidity, Journal of Financial Economics 75, pp.
165–199.
Brennan, M. J., and H. H. Cao, 1996, Information, trade, and derivative securities, Review of
Financial Studies 9, pp. 163–208.
Chang, C.-C., P.-F. Hsieh, and H.-N. Lai, 2009, Do informed option investors predict stock
returns? Evidence from the Taiwan Stock Exchange, Journal of Banking and Finance 33,
pp. 757–764.
Chang, S. S., L. V. Chang, and F. A. Wang, 2014, A dynamic intraday measure of the
probability of informed trading and firm-specific return variation, Journal of Empirical
Finance 29, pp. 80–94.
Chiu, J., H. Chung, and G. H. K. Wang, 2014, Intraday liquidity provision by trader type in
a limit order market: Evidence from Taiwan index future, Journal of Futures Markets 34,
pp. 145–172.
Choe, H., B.-C. Kho, and R. M. Stulz, 1999, Do foreign investors destabilize stock markets?
The Korean experience in 1997, Journal of Financial Economics 54, pp. 227–264.
Choe, H., B.-C. Kho, and R. M. Stulz, 2005, Do domestic investors have an edge? The
trading experience of foreign investors in Korea, Review of Financial Studies 18, pp. 795–
829.
Chordia, T., R. Roll, and A. Subrahmanyam, 2002, Order imbalance, liquidity, and market
returns, Journal of Financial Economics 65, pp. 111–130.
Chordia, T., R. Roll, and A. Subrahmanyam, 2005, Evidence on the speed of convergence to
market efficiency, Journal of Financial Economics 76, pp. 271–292.
Chordia, T., R. Roll, and A. Subrahmanyam, 2008, Liquidity and market efficiency, Journal of
Financial Economics 87, pp. 249–268.
Chou, R. K., G. H. K. Wang, and Y.-Y. Wang, 2015, The impacts of individual day trading
strategies on market liquidity and volatility: Evidence from the Taiwan index futures
market, Journal of Futures Markets 35, pp. 399–425.
Chung, D., and K. Hrazdil, 2010, Liquidity and market efficiency: A large sample study,
Journal of Banking and Finance 34, pp. 2346–2357.
Kuo, W.-H., S.-L. Chung, and C.-Y. Chang, 2015, The impacts of individual and institutional
trading on futures returns and volatility: Evidence from emerging index futures markets,
Journal of Futures Markets 35, pp. 222–244.
Kyle, A. S., and F. A. Wang, 1997, Speculation duopoly with agreement to disagree: Can
overconfidence survive the market test? Journal of Finance 52, pp. 2073–2090.
Lee, Y.-T., Y.-J. Liu, R. Roll, and A. Subrahmanyam, 2004, Order imbalances and market
efficiency: Evidence from the Taiwan Stock Exchange, Journal of Financial and
Quantitative Analysis 39, pp. 327–341.
Li, H. C., C. H. Lin, T. Y. Cheng, and S. Lai, 2013, How different types of traders behave in
the Taiwan futures market, Journal of Futures Markets 33, pp. 1097–1117.
Lin, M. C., 2011, Information content for investor groups in TAIEX futures trading, Asia-
Pacific Journal of Financial Studies 40, pp. 433–466.
Odean, T., 1998, Volume, volatility, price, and profit when all traders are above average,
Journal of Finance 53, pp. 1887–1934.
O’Hara, M., C. Yao, and M. Ye, 2011, What’s not there: The odd-lot bias in TAQ data,
SSRN eLibrary (Available at SSRN: http://ssrn.com/abstract=1892972).
Pan, J., and A. M. Poteshman, 2006, The information in option volume for future stock
prices, Review of Financial Studies 19, pp. 871–908.
Pesaran, H. H., and Y. Shin, 1998, Generalized impulse response analysis in linear
multivariate models, Economics Letter 58, pp. 17–29.
Seasholes, M. S., and G. Wu, 2007, Predictable behavior, profits, and attention, Journal of
Empirical Finance 14, pp. 590–610.
Stoll, H. R., 2000, Friction, Journal of Finance 55, pp. 1479–1514.
Subrahmanyam, A., 2008, Lagged order flows and returns: A longer-term perspective,
Quarterly Review of Economics and Finance 48, pp. 623–640.
Wang, Y.-Y., 2014, Order splitting behavior by different types of traders in the Taiwan
index futures markets under diverse market conditions, Journal of Futures Markets 34,
pp. 883–910.