The effect of attention on buying behavior during a financial crisis: Evidence from the Taiwan stock exchange

(1)

Other uses, including reproduction and distribution, or selling or

licensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of the

article (e.g. in Word or Tex form) to their personal website or

institutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies are

encouraged to visit:

(2)

The effect of attention on buying behavior during a

ﬁnancial crisis: Evidence from the

Taiwan stock exchange

Hsin-Yi Yu

a,1

, Shu-Fan Hsieh

b,

⁎

a_{Department of Finance, National University of Kaohsiung, Taiwan}

b_{Department of Money and Banking, National Kaohsiung First University of Science and Technology, Taiwan}

a b s t r a c t

a r t i c l e i n f o

Article history: Received 22 March 2010

Received in revised form 7 August 2010 Accepted 17 August 2010

Available online 22 August 2010 Keywords:

Trading volume Buy–sell imbalance Financial crisis

We confirm that investors in different categories have different trading patterns caused by attention-grabbing factors. Stocks with extreme one-day returns catch the attention of both individual and institutional investors. Individual investors are net buyers of losers whereas institutional investors are net buyers of winners. Unlike institutional investors, individual investors also regard volume as a conditional attention-grabbing factor. We alsofind that attention-driven buying behavior is mitigated by the financial crisis of 2007, which indicates that the buying behavior of investors is less emotional during a period of financial crisis.

1. Introduction

This paper examines the attention-driven buying behavior of different types of investors. When there are many choices, options that attract attention are more likely to be considered and chosen. Investors are more likely to buy the stocks which grab their attention.

Barber and Odean (2008)find that individual investors are net buyers of stocks with extremely high volume and extreme returns, whether positive or negative. However, is there an interaction between attention-grabbing factors? Is the attention-driven buying behavior strengthened or mitigated by the market situation? Wefirst examine the attention-grabbing behavior of different types of investors by observing the relationship between buy–sell imbalances and the prior day's volumes and returns and further investigate whether the financial crisis has influenced such behavior.

The term ‘attention-grabbing’ was first coined by Barber and Odean (2008). Attention-grabbing stocks refer to those in the news, those experiencing high abnormal trading volume, and those with extreme one-day returns. Previous research tends to view individuals and institutions differently. Institutions are commonly viewed as informed investors, and individuals are believed to have psychological biases and are often thought of as proverbial noise traders (Black, 1986; Kyle, 1985).Barber and Odean (2008)confirm that individual investors are net buyers of attention-grabbing stocks, while institu-tional investors are least influenced by attention.

However, it is interesting to note thatBarber and Odean (2008)

ﬁnd that both extreme positive and extreme negative returns lead to signiﬁcant buying behavior. Based on the work ofBarber and Odean (2008), we analyze a complete dataset, which contains all trading records of all investors on the Taiwan Stock Exchange (TSE). Our sample covers trading from 3rd January 2005 to 31st December 2009, which includes the pre-crisis and in-crisis periods. The dataset contains the entire transaction data and the identity of each trader in the Taiwan stock market, which means that our data allow us to identify trading of individuals and different kinds of institution. Compared to the US, the stock market in Taiwan possesses four characteristics. First, the TSE operates in a consolidated limit-order book environment in which only limit orders are accepted. Orders are executed according to strict price and time priority. Second, the turnover rate in the TSE is very high _{— averaging 184% annually} during our sample period2_{. In contrast, annual turnover on the New}

York Stock Exchange (NYSE) averaged 128% annually from 2005 through 2009. Additionally, the majority of trades are made by individual investors, which account for over 70% capitalization. Third, day trading3_{is prevalent in Taiwan (}_{Barber, Lee, Liu, & Odean, 2008}_).

Virtually all day trading (97.5%) can be traced to individual investors in Taiwan (Barber et al., 2008). Lastly, compared to the US stock market, in which institutional trades account for the majority of total trades, individual investors are very active in the Taiwanese stock market. High turnover and prevalent individual trading provide us International Review of Financial Analysis 19 (2010) 270–280

2_{We calculate turnover as half the sum of buys and sells in each year divided by the}

average daily market capitalization for the year.

3

Day trading is deﬁned as the purchase and sale of the same stock on the same day by an investor.

⁎ Corresponding author. Tel.: +886 7 6011000x3110; fax: +886 7 6011039. E-mail addresses:hyyu@nuk.edu.tw(H.-Y. Yu),shufan@nkfust.edu.tw(S.-F. Hsieh).

1

Tel.: + 886 7 5919709; fax: + 886 7 5919329.

Contents lists available atScienceDirect

(3)

with a good opportunity to observe the attention-driven buying behavior of individual investors.

Wefind that investors are contrarians in Taiwan. Since there is no officially designated market maker on the TSE, individual investors are more likely to indulge in de facto market making activity and act as liquidity providers of practitioners (Chui, Titman, & Wei, 2010; Kaniel, Saar, & Titman, 2008; Lee, Liu, and Subrahmanyam, 2004). In addition, we find that unusual trading volume only encourages individual investors to be net buyers when the stock returns are negative.

We further explore whether the behavior of different investor clientele is influenced by the market situation (Bateman, Islam, & Louviere, 2010; Chiang & Zheng, forthcoming; Karolyi, 2002; Lim, Brooks, & Kim, 2008). The impact of the globalfinancial crisis from 2007 to 2009 on the stock markets has been severe. The literature on thefinancial crisis, especially the massive falling of stock prices in several Asian countries during late 1997, is extensive; however, one facet of thefinancial crisis that did not receive much attention is its impact on investor behaviorChiang and Zheng (forthcoming)). Given this, we empirically investigate the effects of the_{financial crisis from} 2007 to 2009 on attention-driven buying behavior.

To test this, we separate the whole period into two sub-periods— pre-crisis and in-crisis. The results demonstrate that compared to the pre-crisis period, attention-driven buying behavior and the interac-tion between returns and volume are all weaker during the in-crisis period. Thisfinding implies that investors are less emotional during a financial crisis. Specifically, we also find evidence that foreign institutional investors sold off their holdings during the crisis, which indicates that they are less likely to be long-term investors.

This paper differs from previous research in three respects. First, the dataset used byBarber and Odean (2008)in their investigation of attention-grabbing behavior is restricted to a few brokerages. Our study contains all the trading records of individual investors and institutional investors of different types. Second, we document that there is an interaction between different attention-grabbing factors. Individual investors are more attracted by returns than volume. Third, we identify the role of theﬁnancial crisis in testing attention-grabbing behavior. The evidence shows that both individual and institutional investors are less attracted by extreme returns and volume during a ﬁnancial crisis. Investors are less emotional when the market falls.

The reminder of the paper is organized as follows. Section 2

describes the data and sorting methodology.Section 3presents the main empirical results of the attention-grabbing behavior of investors of different types. Section 4 tests whether the ﬁnancial crisis has inﬂuenced attention-grabbing behavior.Section 5concludes. 2. Data and variable construction

2.1. The trading environment in Taiwan

The TSE is managed by a consolidated limit-order book environ-ment in which only limit orders are accepted. There is no designated market maker in Taiwan. During the regular trading time, from 9.00 to 13.30, buy and sell orders interact to determine the executed price subject to applicable auto-matching rules. In TSE, trades can be matched once or twice times every 90 s throughout the trading day. Orders are executed according to strict price and time priority. Although market orders are not permitted, traders can submit an aggressive price-limit order to obtain matching priority. There is a daily price limit of 7% in each direction. The TSE charges commissions at 0.1425% of the value of a trade. Some brokers offer lower commissions for larger traders. Taiwan also imposes a transaction tax on stock sales of 0.3%. Capital gains (both realized and unrealized) are not taxed, while cash dividends are taxed at ordinary income tax rates for domestic investors and at 20% for foreign investors. Corporate income is taxed at a maximum rate of 25%, while personal income is taxed at a maximum rate of 40%.

2.2. Return and volume sorts

Our data include all orders submitted to the Taiwan Stock Exchange (TSE) from January 2005 through December 2009. The available data were collected from the Taiwan Economic Journal Database (TEJ). Based on the data provided by TEJ, we categorize traders to four groups–individuals, dealers, foreign investors, and mutual funds. The latter three groups are institutional investors.

Investors are more likely to notice when stocks have extreme one-day returns, because such returns are often reported in the news and subsequently drive the attention-grabbing behavior of some investors (Barber & Odean, 2008). For example, the media and the trading system constructed by brokerage firms routinely sift the previous day's big gainers and losers out of the market. Therefore, we can expect that those investors whose trading behavior is influenced by attention will tend to purchase or sell in response to price changes. To test the extent to which each of our two investor groups– individual and institution– are net purchasers of stocks in response to large price moves, we sort stocks based on lagged day returns and then calculate the buy–sell imbalances (BSI) for the following day. LikeBarber and Odean (2008), we calculate the buy–sell imbalances for the day following the extreme returns rather than the same day as extreme returns. For each time point (t−1), we sort all stocks for which returns are reported in the TEJ database daily returnsfile into 10 deciles and calculate the BSI in each partition for each investor group as: BSIidvpt = ∑ npt i = 1 Buyidv it−Sell idv it

The number of shares tradedit

npt ð1Þ BSIintpt = ∑ npt i−1 Buyint it−Sell idv it

The number of shares tradedit

npt ð2Þ

where nptis the number of stocks in partition p on day t, Buyitis the

number of shares purchased of stock i on day t, and Sellitis the number

of shares sold of stock i on day t. BSIptidvis the buy–sell imbalance for

individual investors and BSIptintis for institutional investors. For the

days that we have trading data, we calculate the time series mean of the daily BSI for each partition on the day following the return sort for each investor type.

In addition to returns, volume is another attention-grabbing factor argued in prior studies (Barber & Odean, 2008; Chordia, Roll, & Subrahmanyam, 2002). On the days when a stock experiences high volume, it is likely that investors pay more attention to it than other stocks. Therefore, we also sort stocks into 10 deciles on each day on the basis of the prior day's volume. To consider the inﬂuence of ﬁrm size on the trading volume, we adjust the volume by dividing the number of shares traded for each stock on each trading day by the total number of outstanding shares for that stock to be the sorting basis. For each day, (t−1), we sort all stocks into 10 deciles based on the prior day's adjusted volumes. We then calculate the time series mean of the daily BSI for each partition on the day following the volume sort. This calculation is analogous to that for our sorts based on returns.

3. Attention-driven buying behavior 3.1. Returns sorts

The sample covers all transaction data from 3rd January 2005 to 31st December 2009. Our data allow us to identify trading of individuals and different kinds of institution. Table 1 reports the descriptive statistics and serial correlation of buy–sell imbalances

(4)

(BSI). The individual investors and foreign institutional investors are net sellers buyers during the sample period. In Panel B, five lags comprise a calendar week, since the TSE operate Monday through Friday. Most autocorrelations of BSI are positive and signi_{ficant. Such} finding is consistent withLee, Liu, Roll, and Subrahmanyam (2004).

According toBarber and Odean (2008), investors are likely to pay attention to the stocks which exhibit extreme price moves, whether positive or negative. That is, they ﬁnd that attention-grabbing behavior forms a U-shape pattern for investors at the large discount brokerage.Table 2presents the numbers of the BSI of different kinds of investors for stocks sorted on the previous day's returns.Fig. 1plots the buy–sell imbalances (BSI) sorted on the previous day's return. We can observe that the BSI for individual investors decreases monoton-ically with the previous day's returns inFig. 1(a). In Panel A ofTable 2, the BSI of individual investors is 4.139% for stocks in the lowest return decile and−3.029% for stocks in the highest return decile. The result implies that individual investors in Taiwan are contrarians.

Mean-while,Fig. 1(b), (c), and (d) plot the BSI for foreign institutional investors, mutual fund, and dealers respectively. Unlike the behavior of individual investors, the three kinds of institutional investor are all net buyers of stocks with extreme positive returns on the previous day. For example, the BSI of mutual funds is 3.715% in the highest return decile and−5.716% in the lowest return decile. Institutional investors in Taiwan follow the momentum strategy. It is interesting to notice that institutional investors exhibit the opposite tendency to individual investors.

Ourfindings are different from those in the work ofBarber and Odean (2008)but similar to that ofChordia et al. (2002), Chui et al. (2010), Goetzmann and Massa (2002), Kaniel et al. (2008) and Richards (2005). As has been found by Linnainmaa (2010), limit orders significantly alters inferences about individuals' trading intentions and investment abilities. The most aggressive limit order traders are very contrarian with respect to returns. Since individual investors in Taiwan are heavy users of limit orders (Lee et al., 2004), ourfinding provides support for the argument raised in Linnainmaa (2003;2010) and Richards (2005). The inconsistency between the result ofBarber and Odean (2008)and ours may rely on the limited data used byBarber and Odean (2008)from brokerage and the heavy trading by individual traders in Taiwan, which account for over 70% capitalization. However, whether the relatively greater use of limit orders by individual investors is the only motive for the different trading behavior between individual and institutional investors requires further examination.

3.2. Volume sorts

Trading volume is another indicator of the attention a stock is receiving.Table 3presents the BSI for stocks sorted on the previous day's trading volume. Based on the numbers inTable 3,Fig. 2(a)–(d) plots the BSI conditional on trading volume for four kinds of investors. From Panel A,Table 3andFig. 2(a) we can observe that individual investors do not exhibit signiﬁcant attention-driven buying and selling for stocks with high volume. However, foreign institutional investors are aggressive net buyers of stocks in the bottom decile and mutual funds are the net buyers of stocks in the top decile. The complementing role of foreign institutional investors and mutual funds implies that foreign institutional investors are the liquidity providers for mutual funds. Unlike the results of the returns in the previous section, there is no obvious monotonic pattern between volume and BSI for individual investors, which means that individual Table 1

Summary statistics and serial correlation of returns and buy–sell imbalance. Daily buy– sell imbalances (BSI) are computed from 3rd January 2005 to 31st December 2009. BSI is defined as buy shares less sell shares during the day (using limit orders) and tabulated separately for individuals and institutional investors. Panel A reports the means, medians, and standard deviations, computed over the entire sample. Panel B presents the serial correlation of buy–sell imbalance. TheLjung, and Box (1978)Q test is for the null hypothesis that all five coefficients are zero. *, **, and *** indicate significance levels of 10% 5% and 1% respectively.

Panel A. buy–sell imbalance summary statistics

Individual investors Institutional investors

Foreign Mutual funds Dealers Mean 0.006 0.002 −0.016 −0.003 Median 0.000 0.000 −0.000 0.000 Std. Dev. 0.190 0.197 0.150 0.067 Panel B. serial correlation of returns and buy–sell imbalance

Lag (days) Individual investors Institutional investors

Foreign Mutual funds Dealers 1 0.073*** 0.097*** 0.04*** 0.051*** 2 0.065*** 0.09*** 0.031*** 0.044*** 3 0.07*** 0.089*** 0.031*** 0.046*** 4 0.065*** 0.087*** 0.028*** 0.043*** 5 0.062*** 0.084*** 0.022*** 0.043*** Q test 16276*** 20109*** 1209*** 5229*** Table 2

Buy–sell imbalances by investor type for stocks sorted on the previous day's return – the whole period. This table presents the buy–sell imbalance in percentage for different investor types. The table reports the mean for each time series of daily imbalances for a particular investor group and partition. Stocks are sorted daily into deciles on the basis of the previous day's return as reported in the TEJ daily stock returnﬁles for all TSE stocks. The buy–sell imbalances (BSI) are reported for the trades of four groups of investors: individual investors, foreign institutional investors, mutual funds, and dealers. For each day/partition/investor group, we calculate the buy–sell imbalance for each stock on each day as the number of purchased shares minus the number of sold shares divided by the total number of traded shares. The whole period covers 3rd January 2005 to 31st December 2009. Standard errors, calculated using a Newey-West correction for serial dependence, appear in parentheses.

Deciles 1 (extremely negative) 2 3 4 5 6 7 8 9 10 Panel A: individual investors

The whole period 4.139 3.123 2.391 1.639 1.128 0.324 −0.165 −1.122 −2.137 −3.029 (0.126) (0.135) (0.143) (0.146) (0.154) (0.154) (0.154) (0.155) (0.156) (0.122) Panel B: foreign institutional investors

The whole period −1.892 −1.107 −0.603 −0.051 0.489 1.096 1.617 2.199 2.457 2.212 (0.135) (0.153) (0.166) (0.171) (0.179) (0.184) (0.186) (0.181) (0.182) (0.156) Panel C: mutual fund

The whole period −5.716 −4.569 −3.730 −3.156 −2.567 −1.900 −1.033 0.034 1.763 3.715 (0.120) (0.115) (0.109) (0.116) (0.114) (0.109) (0.109) (0.098) (0.097) (0.089) Panel D: dealers

The whole period −0.765 −0.661 −0.525 −0.490 −0.430 −0.267 −0.248 −0.098 0.050 0.206 (0.038) (0.042) (0.043) (0.043) (0.046) (0.046) (0.045) (0.045) (0.048) (0.037)

(5)

investors are more easily attracted by returns than volume. However, volume is a good indicator of the trading behavior of some institutional investors.

3.3. The interaction between returns and volume

Prior results indicate that returns, rather than volume, lead to the attention-driven buying and selling behavior of individual investors.

Note that previous tests assume implicitly that individual investors will treat the two attention-grabbing signals equally, i.e. individual investors are equally attracted by stocks with the greatest volume and stocks with the greatest returns. However, individual investors may regard one factor as a priority and the other factor as secondary. Although the BSI is not inﬂuenced by the prior days' volume, this does not mean that volume has no inﬂuence on the BSI. We group our observations into two sub-samples_{– stocks with a positive previous}

The whole period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10

Partitions sorted on prior day's return Buy-Sell Imbalance _{The whole period}

The whole peiod

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000 6.000

Partitions sorted on prior day's return Buy-Sell Imbalance

The whole peiod

(a)

Individual investors

(b)

Foreign institutional investors

The whole period

-7.000 -6.000 -5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

The whole period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Group 1 Group 2 Group 3 Goup 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10

(c)

Mutual funds

(d)

Dealers

Fig. 1. Buy–sell imbalances for stocks sorted on the prior day's return for investors of different types — the whole period.

Table 3

Buy–sell imbalances by investor type for stocks sorted on the previous day's trading volume – the whole period. This table presents the buy–sell imbalance in percentage for different investor types. Stocks are sorted daily into deciles on the basis of the previous day's trading volume, which is calculated as the number of traded shares divided by the total number of outstanding shares for stock i on day t. The buy–sell imbalances (BSI) are reported for the trades of four groups of investors: individual investors, foreign institutional investors, mutual funds, and dealers. The table reports the mean for each time series of daily imbalances for a particular investor group and partition. For each day/partition/investor group, we calculate the buy–sell imbalance as the number of purchased shares minus the number of sold shares divided by the total number of traded shares. The whole period covers 3rd January 2005 to 31st December 2009. Standard errors, calculated using a Newey-West correction for serial dependence, appear in parentheses.

Deciles 1 (extremely small) 2 3 4 5 6 7 8 9 10 Panel A: individual investors

The whole period −0.052 0.120 0.392 0.642 0.991 0.893 1.209 1.102 0.836 −0.026 (0.242) (0.208) (0.181) (0.158) (0.146) (0.131) (0.120) (0.109) (0.101) (0.087) Panel B: foreign institutional investors

The whole period 2.625 1.475 1.073 0.664 0.417 0.324 0.154 0.042 −0.003 −0.122 (0.362) (0.255) (0.210) (0.185) (0.157) (0.135) (0.123) (0.106) (0.087) (0.063) Panel C: mutual fund

The whole period −3.443 −2.529 −2.258 −2.178 −2.155 −1.849 −1.602 −0.839 −0.427 0.735 (0.172) (0.120) (0.118) (0.111) (0.106) (0.105) (0.092) (0.092) (0.088) (0.086) Panel D: dealers

The whole period −0.569 −0.373 −0.328 −0.405 −0.341 −0.343 −0.352 −0.232 −0.188 −0.036 (0.100) (0.054) (0.046) (0.038) (0.037) (0.036) (0.036) (0.037) (0.036) (0.031)

(6)

day's return and stocks with a negative previous day's return– and then sort stocks in each sub-sample by the previous day's volume.

Table 4 separately reports the return-partition BSI of individual investors for stocks sorted by the previous day's volume.

Fig. 3 (a)–(b) plot the BSI sorted on the prior day's volume for individual investors conditional on the previous day's returns. We can observe that depending on the sign of the return, the BSI of stocks with a negative prior day's return increase monotonically with the previous day's volume for individual investors. Meanwhile, the prior day's volume does not signiﬁcantly inﬂuence the BSI of individual investors for stocks with a positive prior day's return.

In unreported analysis, we also repeat the analysis for institutional investors. However, the phenomenon documented inFig. 3(a) and (b) cannot be observed from institutional investors. This ﬁnding reveals that individual investors regard volume as the secondary attention-grabbing factor, where as institutional investors do not.

4. The inﬂuences of the ﬁnancial crisis on the buy–sell imbalance In the end of February, the biggest US house builder DR Horton warns huge losses from sub-prime fall-out4_{. In March 2007, late}

mortgage payments and home repossessions in the US have hit their highest level since records began5_{. If attention-grabbing behavior is}

initiated by search problems and emotion, it is possible that uncertainty brought by a financial crisis affects the emotion of investors and investor buying behavior. In this section, we repeat prior tests but separate the research periods into two sub-periods– before and after 1st March 2007– to examine whether financial crises influence attention-driven trading behavior.

Table 4

Buy–sell imbalances of individual investors for stocks classified by the previous day's return and then sorted on the previous day's trading volume — the whole period. This table presents the buy–sell imbalance in percentage for different investor types. Stocks are first classified into two groups. Panel A includes stocks with a positive previous day's return and Panel B includes stocks with a negative previous day's return. In the two groups, stocks are sorted daily into deciles on the basis of the previous day's trading volume, which is calculated as the number of traded shares divided by the total number of outstanding shares for stock i on day t. The buy–sell imbalances (BSI) are reported for the trades of four groups of investors: individual investors, foreign institutional investors, mutual funds, and dealers. The table reports the mean for each time series of daily imbalances for a particular investor group and partition. For each day/partition/investor group, we calculate the buy–sell imbalance as the number of purchased shares minus the number of sold shares divided by the total number of traded shares. The whole period covers 3 rd January 2005 to 31st December 2009. Standard errors, calculated using a Newey-West correction for serial dependence, appear in parentheses.

Deciles 1 (extremely small) 2 3 4 5 6 7 8 9 10 Panel A: the previous day's return is positive

The whole period −1.268 −1.576 −1.770 −1.736 −1.804 −1.662 −1.672 −1.769 −2.004 −2.089 (0.268) (0.244) (0.245) (0.231) (0.194) (0.196) (0.172) (0.158) (0.134) (0.123) Panel B: the previous day's return is negative

The whole period 0.823 1.493 1.900 2.339 3.049 3.182 3.294 3.632 3.823 3.517 (0.260) (0.235) (0.203) (0.180) (0.180) (0.169) (0.165) (0.162) (0.144) (0.139)

4

BBC news, 22 February 2007.

5

BBC news, 13 March 2007. The timeline of sub-prime losses can be traced from the webpage: http://news.bbc.co.uk/2/hi/business/7096845.stm.

The whole peiod

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance _{The whole peiod}

The whole period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance

The whole period

(a)

Individual investors

(b)

Foreign institutional investors

The whole period

-7.000 -6.000 -5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

The whole period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance _{The whole period}

(c)

Mutual funds

(d)

Dealers

Fig. 2. Buy–sell imbalances for stocks sorted on the prior day's volume for investors of different types — the whole period. H.-Y. Yu, S.-F. Hsieh / International Review of Financial Analysis 19 (2010) 270–280

(7)

4.1. Return and volume sorts

Table 5presents the numbers of the BSI sorted on the prior day's returns of different types of investors in two sub-periods— the pre-crisis and in-pre-crisis periods. Based on the numbers inTable 5,Fig. 4

plots the tendency of attention-grabbing behavior of different types of investors in two sub-periods. The slopes of the BSI lines are less steep during the financial crisis, especially for individual investors and foreign institutional investors. That is, the attention-grabbing behav-ior is less significant in the financial crisis period for the two types of investors. The mitigation of attention-grabbing trading behavior implies that investors, whether individual or institutional, are less influenced by emotion after the financial crisis occurs. Our results are qualitatively similar when we sort on same-day returns.

The numbers in Table 5also indicate that foreign institutional investors withdrew money from the Taiwanese market during the in-crisis period. Although foreign investors still possess net buying behavior of top-performing stocks sorted by the previous day's return,

they become the net sellers of stocks in other deciles. In Panel B, foreign institutional investors display attention-driven buying for stocks in deciles 3–10 during the pre-crisis period. However, this attention-driven buying is reversed to selling during the in-crisis period in deciles 1–7. For example, the BSI of foreign institutional investors is 3.940% for stocks in the seventh decile during the pre-crisis period. However, it becomes −0.411% during the in-crisis period. The attention-driven buying behavior in deciles 9–10 is also reduced during the in-crisis period. This result is similar to that found byKarolyi (2002), which demonstrates that Asianﬁnancial crisis in 1997 scared foreign investors out of Japan.

Fig. 5plots the BSI sorted on the prior day's volume of different types of investors during the pre-crisis and in-crisis periods. Similar to theﬁnding inSection 3.2, volume is not the determinant of the BSI for individual investors, whether during the pre-crisis or in-crisis periods. For foreign institutional investors, compared to the pre-crisis period, the slope of the BSI line is less steep in the in-crisis period. The BSI of mutual funds also exhibits a similar pattern. For example, the BSI of

The whole period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance The whole period

The whole period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance The whole period

(a)

Positive return

(b)

Negative return

Fig. 3. Buy–sell imbalances of individual investors for stocks sorted on the previous day's volume conditional on the previous day's returns — the whole period.

Table 5

Buy–sell imbalances by investor type for stocks sorted on the previous day's return — pre-crisis and in-crisis periods. This table presents the buy–sell imbalance in percentage for different investor types. The table reports the mean for each time series of daily imbalances for a particular investor group and partition. Stocks are sorted daily into deciles on the basis of the previous day's return as reported in the TEJ daily stock returnﬁles for all TSE stocks. The buy–sell imbalances (BSI) are reported for the trades of four groups of investors: individual investors, foreign institutional investors, mutual funds, and dealers. For each day/partition/investor group, we calculate the buy–sell imbalance for each stock on each day as the number of purchased shares minus the number of sold shares divided by the total number of traded shares. The pre-crisis period is from 3 rd January 2005 to 27th February 2007. The in-crisis period covers 1st March 2007 to 31st December 2009. Standard errors, calculated using a Newey-West correction for serial dependence, appear in parentheses.

Deciles 1 (extremely negative) 2 3 4 5 6 7 8 9 10 Panel A: individual investors

Pre-crisis period 3.656 3.042 1.894 1.044 0.245 −0.525 −1.147 −2.146 −3.352 −3.960 (0.136) (0.174) (0.188) (0.192) (0.206) (0.202) (0.195) (0.197) (0.195) (0.140) In-crisis period 4.671 3.254 2.922 2.268 2.036 1.210 0.859 −0.073 −0.934 −2.139

(0.215) (0.213) (0.221) (0.226) (0.230) (0.234) (0.237) (0.239) (0.240) (0.201) Panel B: foreign institutional investors

Pre-crisis period −2.254 −0.980 0.196 1.007 2.174 3.083 3.940 4.723 5.203 4.862 (0.217) (0.267) (0.291) (0.301) (0.316) (0.321) (0.313) (0.300) (0.296) (0.249) In-crisis period −2.197 −1.627 −1.754 −1.236 −1.156 −0.737 −0.411 0.102 0.223 0.367

(0.214) (0.225) (0.235) (0.241) (0.247) (0.250) (0.255) (0.251) (0.253) (0.219) Panel C: mutual fund

Pre-crisis period −6.105 −5.301 −4.248 −3.804 −2.994 −2.477 −1.657 −0.412 1.359 3.417 (0.190) (0.179) (0.169) (0.181) (0.182) (0.167) (0.167) (0.149) (0.157) (0.130) In-crisis period −5.956 −4.203 −3.434 −2.621 −2.323 −1.423 −0.564 0.479 2.244 4.005 (0.198) (0.193) (0.181) (0.186) (0.184) (0.181) (0.181) (0.166) (0.157) (0.155) Panel D: dealers Pre-crisis period −0.989 −0.871 −0.672 −0.608 −0.541 −0.320 −0.395 −0.190 0.042 0.184 (0.072) (0.082) (0.084) (0.086) (0.091) (0.094) (0.091) (0.091) (0.098) (0.074) In-crisis period −0.691 −0.595 −0.506 −0.516 −0.399 −0.277 −0.178 −0.037 0.019 0.235 (0.047) (0.051) (0.050) (0.050) (0.054) (0.050) (0.051) (0.051) (0.051) (0.044)

(8)

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000 6.000

Partitions sorted on prior day's return Buy-Sell Imbalance _{Pre-crisis period} _{In-crisis period}

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000 6.000

(a)

Individual investors

(b)

Foreign institutional investors

-7.000 -6.000 -5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's return Buy-Sell Imbalance Pre-crisis period In-crisis period

(c)

Mutual funds

(d)

Dealers

Fig. 4. Buy–sell imbalances for stocks sorted on the prior day's return for investors of different types — pre-crisis and in-crisis periods.

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance _{Pre-crisis period} _{In-crisis period}

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000 6.000 7.000

(a)

Individual investors

(b)

Foreign institutional investors

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance Pre-crisis period In-crisis period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

(c)

Mutual funds

(d)

Dealers

Fig. 5. Buy–sell imbalances for stocks sorted on the prior day's volume for investors of different types — pre-crisis and in-crisis periods. H.-Y. Yu, S.-F. Hsieh / International Review of Financial Analysis 19 (2010) 270–280

(9)

mutual funds is−4.562% for stocks in the bottom decile during the pre-crisis period but becomes _{−2.729% during the in-crisis period} (Table 6).

We also use the samples in the two sub-periods to examine the BSI sorted on the prior day's volume for individual investors conditional on the previous day's returns.Table 7presents the numbers of the BSI of individual investors andFig. 6illustrates this. Compared toFig. 3, the results of using sub-periods are similar to those when using the whole period as the sample. The BSI of stocks with a negative prior day's return increase monotonically with the previous day's volume for individual investors in both pre-crisis and in-crisis periods. However, for stocks with a positive prior day's return, there is no monotonic pattern for the BSI. Overall, whether during the pre-crisis

or in-crisis periods, individual investors pay more attention to the prior day's return rather than the prior day's volume.

4.2. Regression analysis

We further use multiple regressions to examine measure the level of influence of the financial crisis on attention-driven buying. Given that BSI could conceivably be caused by many factors, we control for past daily returns, weekly regularities, and past lagged values of BSI. Similar to the findings ofChordia et al. (2002) and Gibbons and Hess (1981), the daily BSI in the number of shares (BSINUM) is regressed on day-of-the-week dummies, variables designed to capture past up-market and down-market moves, past values of BSI, and thefinancial crisis dummies. Table 6

Buy–sell imbalances by investor type for stocks sorted on the previous day's trading volume — pre-crisis and in-crisis periods. This table presents the buy–sell imbalance in percentage for different investor types. Stocks are sorted daily into deciles on the basis of the previous day's trading volume, which is calculated as the number of traded shares divided by the total number of outstanding shares for stock i on day t. The buy–sell imbalances (BSI) are reported for the trades of four groups of investors: individual investors, foreign institutional investors, mutual funds, and dealers. The table reports the mean for each time series of daily imbalances for a particular investor group and partition. For each day/partition/investor group, we calculate the buy–sell imbalance as the number of purchased shares minus the number of sold shares divided by the total number of traded shares. The pre-crisis period is from 3 rd January 2005 to 27th February 2007. The in-crisis period covers 1st March 2007 to 31st December 2009. Standard errors, calculated using a Newey-West correction for serial dependence, appear in parentheses.

Deciles 1 (extremely small) 2 3 4 5 6 7 8 9 10 Panel A: individual investors

Pre-crisis period −0.884 −0.799 −0.662 −0.318 0.149 0.260 0.594 0.530 0.309 −0.582 (0.300) (0.270) (0.242) (0.202) (0.189) (0.167) (0.150) (0.139) (0.129) (0.118) In-crisis period 0.855 1.024 1.449 1.566 1.845 1.570 1.845 1.738 1.407 0.558

(0.392) (0.322) (0.270) (0.241) (0.225) (0.204) (0.190) (0.168) (0.157) (0.128) Panel B: foreign institutional investors

Pre-crisis period 6.067 4.263 3.251 2.421 1.698 1.571 1.057 0.863 0.740 0.466 (0.595) (0.451) (0.365) (0.323) (0.265) (0.227) (0.211) (0.182) (0.150) (0.106) In-crisis period −0.024 −0.933 −0.989 −1.083 −0.986 −1.019 −0.956 −0.977 −0.798 −0.585

(0.505) (0.341) (0.292) (0.258) (0.233) (0.202) (0.185) (0.158) (0.130) (0.096) Panel C: mutual fund

Pre-Crisis Period −4.562 −3.115 −2.771 −2.792 −2.846 −2.463 −1.973 −1.061 −0.723 0.669 (0.254) (0.180) (0.181) (0.176) (0.171) (0.166) (0.139) (0.147) (0.142) (0.132) In-Crisis Period −2.729 −2.029 −1.687 −1.575 −1.499 −1.437 −1.282 −0.821 −0.463 0.536 (0.300) (0.205) (0.198) (0.175) (0.171) (0.169) (0.156) (0.147) (0.142) (0.142) Panel D: Dealers Pre-crisis period −0.745 −0.529 −0.494 −0.592 −0.487 −0.524 −0.488 −0.276 −0.225 0.044 (0.200) (0.113) (0.094) (0.076) (0.072) (0.070) (0.072) (0.076) (0.075) (0.063) In-crisis period −0.550 −0.297 −0.264 −0.308 −0.277 −0.243 −0.333 −0.276 −0.216 −0.116 (0.112) (0.055) (0.050) (0.046) (0.044) (0.045) (0.042) (0.042) (0.038) (0.034) Table 7

Buy–sell imbalances of individual investors for stocks classified by the previous day's return and then sorted on the previous day's trading volume — pre-crisis and in-crisis periods. This table presents the buy–sell imbalance in percentage for different investor types. Stocks are first classified into two groups. Panel A includes stocks with a positive previous day's return and Panel B includes stocks with a negative previous day's return. In the two groups, stocks are sorted daily into deciles on the basis of the previous day's trading volume, which is calculated as the number of traded shares divided by the total number of outstanding shares for stock i on day t. The buy–sell imbalances (BSI) are reported for the trades of four groups of investors: individual investors, foreign institutional investors, mutual funds, and dealers. The table reports the mean for each time series of daily imbalances for a particular investor group and partition. For each day/partition/investor group, we calculate the buy–sell imbalance as the number of purchased shares minus the number of sold shares divided by the total number of traded shares. The pre-crisis period is from 3 rd January 2005 to 27th February 2007. The in-crisis period covers 1st March 2007 to 31st December 2009. Standard errors, calculated using a Newey-West correction for serial dependence, appear in parentheses.

Deciles 1 (extremely small) 2 3 4 5 6 7 8 9 10 Panel A: the previous day's return is positive

Pre-crisis period −1.939 −2.680 −3.359 −2.962 −3.125 −2.727 −2.594 −2.509 −2.778 −2.801 (0.316) (0.321) (0.297) (0.262) (0.250) (0.264) (0.209) (0.207) (0.166) (0.126) In-crisis period −0.543 −0.577 −0.204 −0.554 −0.503 −0.604 −0.742 −1.030 −1.242 −1.400

(0.452) (0.375) (0.393) (0.391) (0.298) (0.293) (0.280) (0.245) (0.214) (0.217) Panel B: the previous day's return is negative

Pre-crisis period 0.095 0.966 1.308 1.892 2.432 2.847 3.254 3.225 3.537 3.219 (0.318) (0.313) (0.268) (0.239) (0.226) (0.206) (0.210) (0.181) (0.193) (0.177) In-crisis period 1.653 2.082 2.428 2.877 3.649 3.469 3.359 4.107 4.118 3.878

(10)

Table 8reports the regression results for four kinds of investors respectively. The Durbin–Watson statistics express that there is no serious autocorrelation. The negative coefﬁcients of Max(Rt− 1, 0)and

Min(Rt− 1, 0)indicate that the daily return on day t−1 is negatively

related to the BSI on day t. That is to say, individual investors in aggregate act as contrarians— buy after market declines and sell after market advances. However, contrarian trading can only be observed immediately on day t−1. The daily returns lagged for 2, 3, 4, and 5 days are positively related to the BSI on day t, which means that the contrarian strategy of individual investors caused by attention-grabbing behavior is short-lived and subsequently reverses to momentum trading. In other words, in the long term, individual investors still naively chase the winners.

Unlike individual investors, institutional investors are momen-tum-style investors in the short term and contrarians in the long term. The signs of the return coefﬁcients reverse from positive to negative

after two days' trading. Due to the extensive use of limit orders, individual investors are more likely to be contrarians and the contrarian tendency of individuals leads them to act as liquidity providers to institutions that require immediacy (Campbell, Gross-man, & Wang, 1993; Grossman & Miller, 1988; Kaniel et al., 2008; Richards, 2005; Stoll, 1978). Institutional investors who require immediacy must offer price concessions to induce risk-averse individuals to take the other side of their trades.

Moreover, the attention-driven buying behavior of individual investors driven by the prior day's return is significantly mitigated by the_{financial crisis. Given the positive coefficients of the interaction} terms between thefinancial crisis dummy variable and the previous day's return (0.125 and 0.419), attention-driven buying can still be observed but at a lower level. Similarly, there still remains evidence of momentum trading on day t−1 for institutional investors, but the numbers of the BSI are smaller in thefinancial crisis. This finding

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance Pre-crisis period In-crisis period

-5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 5.000

Partitions sorted on prior day's volume Buy-Sell Imbalance _{Pre-crisis period} In-crisis period

(a)

Positive return

(b)

Negative return

Fig. 6. Buy–sell imbalances of individual investors for stocks sorted on the previous day's volume conditional on the previous day's returns — pre-crisis and in-crisis periods.

Table 8

The influence of the financial crisis on buy–sell imbalance. This table presents a regression analysis of the relationship between the buy–sell imbalance caused by attention-grabbing behavior and thefinancial crisis. The dependent variable BSINUM is buy–sell imbalances measured by the number of shares purchased minus the number of shares sold and then divided by the total number of shares traded in the market on day t. Rtdenotes the stock daily return on day t. The dummy variables Mon, Tue, Wed, and Thu control the weekly

regularities in buy–sell imbalance. The variable Crisis is a dummy variable which is equal to 1 during 1st March 2007 to 31st December 2009 and 0 otherwise. The p values are in parentheses. ***, ** and * denote signiﬁcance at the 1%, 5% and 10% levels respectively.

Individual investors Foreign investors Mutual funds Dealers

Intercept −0.344 *** 2.161 *** −2.571 *** −0.230 *** Max(Rt− 1, 0) −0.666 *** 0.269 *** 1.120 *** 0.120 *** Max(Rt− 2, 0) 0.053 *** −0.026 * −0.047 *** 0.000 Max(Rt− 3, 0) 0.062 *** −0.038 ** −0.032 ** 0.005 Max(Rt− 4, 0) 0.039 *** −0.034 ** −0.020 0.002 Max(Rt− 5, 0) 0.053 *** −0.062 *** −0.062 *** −0.008 Min(Rt− 1, 0) −1.203 *** 1.422 *** 0.986 *** 0.176 *** Min(Rt− 2, 0) 0.082 *** −0.039 * −0.200 *** 0.001 Min(Rt− 3, 0) 0.067 *** −0.030 −0.185 *** −0.009 Min(Rt− 4, 0) 0.077 *** −0.027 −0.195 *** −0.018 ** Min(Rt− 5, 0) 0.098 *** −0.014 −0.254 *** 0.000 BSIt− 1 0.057 *** 0.070 *** 0.035 *** 0.044 *** BSIt− 2 0.048 *** 0.062 *** 0.027 *** 0.035 *** BSIt− 3 0.054 *** 0.061 *** 0.028 *** 0.038 *** BSIt− 4 0.048 *** 0.059 *** 0.026 *** 0.035 *** BSIt− 5 0.045 *** 0.056 *** 0.016 *** 0.035 *** Mon −0.088 0.056 0.061 −0.005 Tue −0.106 0.226 *** −0.073 −0.033 Wed −0.098 0.097 0.190 ** −0.075 *** Thu 0.296 *** −0.290 *** 0.096 −0.197 *** Crisis 1.216 *** −2.496 *** 1.113 *** 0.048 * Crisis*Max(Rt− 1, 0) 0.125 *** −0.079 ** −0.377 *** −0.030 *** Crisis*Min(Rt− 1, 0) 0.419 *** −1.028 *** 0.053 −0.067 ***

The number of observations 723,866 501,367 252,666 504,213

Durbin–Watson 2.004 2.006 2.001 2.002

Adjusted R2 _0.029 _0.037 _0.033 _0.011

(11)

indicates that theﬁnancial crisis of 2007 enabled both individual and institutional investors to be less attention-driven.

Note that the negative coefficient of the financial crisis dummy variable on the regression of foreign institutional investors also reveals that foreign institutional investors in aggregate reverse to net sellers during thefinancial crisis period, which is consistent with the finding of Karolyi (2002)thatfinancial crises scare foreign institu-tional investors out of the market.

We next attempt to discern whether the influence of the interaction between return and volume on BSI is affected by the financial crisis. We include three interaction terms here to measure whether volume has a significant influence on investors for stocks with a negative prior day's return and whether such influences are mitigated by thefinancial crisis of 2007.

The empirical results are presented inTable 9. Wefind that high volume does not lead to significant attention-driven buying for individual investors. However, the significant positive coefficient of the interaction term Volt− 1* DRt− 1b 0reveals that individual investors

are attracted by stocks with high volume when the prior day's returns are negative. However, given the negative coefﬁcient of the interaction term DRt− 1b 0* Crisis * Volt− 1, the attraction of high volume

stocks with a negative prior day's return declines during thefinancial crisis period. Consistent with the finding of the return sorts, the financial crisis in 2007 mitigates attention-driven buying for individ-ual investors. Conversely, high volume encourages institutional investors to trade, whether buying or selling. The influence of volume, however, is also alleviated by thefinancial crisis.

Some may argue that there is no a definite date to define when turmoil begins. The effects of_{financial crisis on the Taiwanese stock} market may be postponed. Therefore, we do the sensitivity analysis toward the date offinancial crisis. Cooper, Gutierrez, and Hameed (2004)define the market state as up when the market's three-year return is non-negative; otherwise as down. We use the market's six-month return to judge whether the market is down. When the market's six-month return is transferred from positive to negative, we define that the crisis occurs. We also consider a one-year market return. Under this method, the dates of separating sub-periods are transferred to 13th December 2007 and 22nd January 2008 for the six-month and one-year returns respectively. We repeat all analysis by using the two dates to separate sub-periods. All majorfindings remain unchanged.

5. Conclusion

Attention-driven buying results from the difﬁculty that investors have in searching the thousands of stocks they can potentially buy. By using the trading data across different investor types that trade on the Taiwan Stock Exchange (TSE), this paper empirically investigates the attention-grabbing behavior of investors and the effect of theﬁnancial crisis of 2007 on this behavior.

We test for attention-driven buying behavior by sorting stocks on extreme one-day returns and trading volume. We ﬁnd that the attention-driven buying behavior of individual investors only focuses on the stocks which performed poorly on the previous day. Contrarily,

Table 9

The interaction between return and volume during thefinancial crisis. This table presents a regression analysis to examine whether the influence of volume on the buy–sell imbalances (BSI) is affected by thefinancial crisis. The dependent variable BSINUM is the buy–sell imbalances measured by the number of shares purchased minus the number of shares sold and then divided by the total number of shares traded in the market on day t. Voltis calculated by the number of shares traded for each stock on each trading day divided

by the total number of outstanding shares for that stock. Rtdenotes the stock daily return on day t. The dummy variables Mon, Tue, Wed, and Thu control the weekly regularities in BSI.

The variable Crisis is a dummy variable which is equal to 1 during 1st March 2007 to 31st December 2009 and 0 otherwise. DRt− 1b 0is also a dummy variable which equals 1 when the

previous day's return is negative and 0 otherwise. The p values are in parentheses. ***, ** and * denote signiﬁcance at the 1%, 5% and 10% levels respectively. Individual investors Foreign investors Mutual funds Dealers

Intercept −0.194 *** 1.835 *** −2.572 *** −0.291 *** Volt− 1 −0.034 −0.530 *** 0.579 *** 0.079 *** Volt− 2 −0.009 0.054 ** −0.067 *** 0.009 Volt− 3 −0.136 *** 0.130 *** −0.007 −0.014 ** Volt− 4 −0.023 0.111 *** −0.025 −0.006 Volt− 5 −0.107 *** 0.157 *** −0.090 *** −0.009 BSIt− 1 0.057 *** 0.069 *** 0.035 *** 0.045 *** BSIt− 2 0.050 *** 0.062 *** 0.028 *** 0.035 *** BSIt− 3 0.051 *** 0.061 *** 0.027 *** 0.038 *** BSIt− 4 0.046 *** 0.058 *** 0.023 *** 0.034 *** BSIt− 5 0.045 *** 0.056 *** 0.016 *** 0.031 *** Max(Rt− 1, 0) −0.560 *** 0.330 *** 0.734 *** 0.087 *** Max(Rt− 2, 0) 0.057 *** −0.038 ** −0.033 ** −0.004 Max(Rt− 3, 0) 0.089 *** −0.067 *** −0.019 0.006 Max(Rt− 4, 0) 0.039 *** −0.056 *** −0.014 0.009 Max(Rt− 5, 0) 0.073 *** −0.099 *** −0.056 *** −0.007 Min(Rt− 1, 0) −0.893 *** 0.705 *** 1.038 *** 0.147 *** Min(Rt− 2, 0) 0.090 *** −0.074 *** −0.204 *** 0.005 Min(Rt− 3, 0) 0.075 *** −0.048 ** −0.177 *** −0.018 ** Min(Rt− 4, 0) 0.095 *** −0.036 −0.226 *** −0.021 *** Min(Rt− 5, 0) 0.066 *** 0.008 −0.241 *** −0.007 Mon −0.100 0.056 0.105 0.001 Tue −0.082 0.157 * −0.036 −0.035 Wed −0.094 0.097 0.187 ** −0.071 ** Thu 0.288 *** −0.308 *** 0.102 −0.198 *** Crisis 1.173 *** −2.275 *** 0.928 *** 0.143 *** Crisis*Volt− 1 −0.055 0.287 *** −0.183 *** −0.076 *** DRt− 1b 0* Volt− 1 0.640 *** −0.315 *** −0.293 *** −0.015 DRt− 1b 0* Crisis * Volt− 1 −0.593 *** 0.952 *** 0.057 0.047 **

The number of observations 656,965 461,310 233,628 463,253

Durbin–Watson 2.004 2.007 2.002 2.002

(12)

institutional investors are net buyers of stock which performed well on the previous day. In brief, individual investors are contrarians and institutional investors are momentum-style investors.

Furthermore, this paper demonstrates that high volume is another attention-grabbing factor for institutional investors. Institutional investors, especially foreign institutional investors and mutual funds, are more prone to indulge in stocks with high volume. Conversely, volume is not a signiﬁcant determinant on the buy–sell imbalance (BSI) of individual investors. However, within the subset of stocks that do attract their attention through negative returns, individual investors are prone to buy stocks with high volume. This ﬁnding also indicates that individual investors do not treat the two attention-grabbing factors equivalently but regard returns as the priority.

Our analysis provides extra information for the attention-driven buying behavior during the financial crisis period. We observe attention-driven buying behavior in the pre-crisis period and the in-crisis period separately. Our tests exhibit strong evidence that buying behavior is less attention-driven during the _{financial crisis period} from 2007 to 2009 for both individual investors and institutional investors. The influences of financial crises on other trading behaviors are interesting topics left for future research.

References

Barber, B. M., Lee, Y. T., Liu, Y. J., & Odean, T. (2008). Just how much do individual investors lose by trading? Review of Financial Studies, 22, 609−632.

Barber, B. M., & Odean, T. (2008). All that glitters: the effect of attention and news on the buying behavior of individual and institutional investors. Review of Financial Studies, 21, 785−818.

Bateman, H., Islam, T., & Louviere, J. (2010). Retirement investor risk tolerance in tranquil and crisis periods: Experimental survey evidence. Working Paper.

Black, F. (1986). Noise. Journal of Finance, 41, 529−543.

Campbell, J. Y., Grossman, S. J., & Wang, J. (1993). Trading volume and serial correlation in stock returns. Quarterly Journal of Economics, 108, 905−939.

Chiang, T. C., Zheng, D., forthcoming. An empirical analysis of herd behavior in global stock markets. Journal of Banking and Finance.

Chordia, T., Roll, R., & Subrahmanyam, A. (2002). Buy–sell imbalance, liquidity, and market returns. Journal of Financial Economics, 65, 111−130.

Chui, A. C. W., Titman, S., & Wei, K. C. J. (2010). Individualism and momentum around the world. Journal of Finance, 65, 361−392.

Cooper, M. J., Gutierrez, R. C., & Hameed, A. (2004). Market states and momentum. Journal of Finance, 59, 1345−1365.

Gibbons, M., & Hess, P. (1981). Day of week effects and assets returns. Journal of Business, 54, 579−596.

Goetzmann, William N., & Massa, Massimo (2002). Daily momentum and contrarian behavior of index fund investors. Journal of Financial and Quantitative Analysis, 37, 375−390.

Grossman, S. J., & Miller, M. H. (1988). Liquidity and market structure. Journal of Finance, 43, 617−637.

Kaniel, R., Saar, G., & Titman, S. (2008). Individual investor trading and stock returns. Journal of Finance, 63, 273−310.

Karolyi, G. A. (2002). Did the Asianﬁnancial crisis scare foreign investors out of Japan? Paciﬁc-Basin Finance Journal, 10, 411−442.

Kyle, A. S. (1985). Continuous auctions and insider trading. Econometrica, 53, 1315−1335.

Lee, Y. T., Liu, Y. J., Roll, R., & Subrahmanyam, A. (2004). Buy–sell imbalances and market efﬁciency: evidence from the Taiwan Stock Exchange. Journal of Financial and Quantitative Analysis, 39, 327−341.

Lim, K. P., Brooks, R. D., & Kim, J. H. (2008). Financial crisis and stock market efﬁciency: empirical evidence from Asian countries. International Review of Financial Analysis, 17, 571−591.

Linnainmaa, J. (2003). Who Makes the Limit Order Book? Implications for Contrarian Strategies, Attention-Grabbing Hypothesis, and the Disposition. Working Paper, University of California, Los Angeles.

Linnainmaa, J. (2010). Do Limit Orders Alter Inferences about Investor Performance and Behavior? Journal of Finance, 65, 1473−1506.

Ljung, G., & Box, G. (1978). On a measure of lack ofﬁt in time-series models. Biometrika, 75, 355−361.

Richards, A. (2005). Bigﬁsh in small ponds: The trading behavior and price impact of foreign investors in Asian emerging equity markets. Journal of Financial and Quantitative Analysis, 40, 1−27.

Stoll, H. (1978). The supply of dealer services in securities markets. Journal of Finance, 33, 1133−1151.