Sie sind auf Seite 1von 22

Pacific-Basin Finance Journal 11 (2003) 1 22 www.elsevier.

com/locate/econbase

Investment patterns and performance of investor groups in Japan


Akiko Kamesaka a, John R. Nofsinger b,*, Hidetaka Kawakita c
b a Faculty of Economics, Ryukoku University, Kyoto 612-8577, Japan Department of Finance, Insurance and Real Estate, College of Business and Economics, Washington State University, Pullman, WA 99164-4746, USA c Nippon Life Insurance Company, 1-2-2, Yurakucho, Chiyoda-ku, Tokyo 100-8444, Japan

Received 29 November 2001; accepted 26 August 2002

Abstract Using weekly aggregate investment flow from Japan, we study the investment patterns and performance of foreign investors, individual investors, and five types of institutional investors. Securities firms, banks, and foreign investors perform well over the sample period. Individual investors perform poorly. We also find that foreign investor trading is associated with positive feedback market timing and that this trading earns high returns. Alternatively, individual investors use positive feedback trading in their market timing but earn low returns. Consequently, we document evidence consistent with information-based models (foreign investors) and behavioralbased models (individual investors). It is a particularly new and interesting finding that evidence of both information-based trading and behavioral-based trading occurs in the same market. D 2003 Elsevier Science B.V. All rights reserved.
JEL classification: G10; G15 Keywords: Investor behavior; Feedback trading; Herding; Foreign investors; Institutional investors

There is an ongoing debate whether investor trading decisions are influenced more by information about value or by psychological biases. Two categories of theoretical trading models have been developed to explain the two potential influences of behavior. The information-based category of models posits that trading is based on informational advantages. These models suggest that informed investor trading would exhibit a positive

* Corresponding author. Tel.: +1-509-335-7200; fax: +1-509-335-3857. E-mail address: john_nofsinger@wsu.edu. (J.R. Nofsinger). 0927-538X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 7 - 5 3 8 X ( 0 2 ) 0 0 0 9 5 - 1

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

feedback, or momentum, pattern of trading. That is, high (low) returns in one period will be associated with a high degree of investor buying (selling) in the next period. This herding pattern is the result of a group of investors trading on the same (or correlated) information signals (see Froot et al.,1992; Bikhchandani et al., 1992; Hirshleifer et al., 1994). The behavioral-based models posit that investor decisions are influenced by cognitive errors such as overconfidence and disposition effect. These behavioral models (see Daniel et al., 1998; Gervais and Odean, 2001) also suggest that a positive feedback trading pattern can be indicative of investor overconfidence. Therefore, both informationbased and behavioral-based theories predict that investors may engage in positive feedback trading. Indeed, empirical evidence suggests that some investors trading exhibits a positive feedback pattern. Grinblatt et al. (1995) argue that a trade imbalance by one investor type that is correlated with past returns is considered feedback trading. Nofsinger and Sias (1999) provide empirical evidence that U.S. institutional investors positive feedback trade. Grinblatt and Keloharju (2000) also find that foreign investors in Finland exhibit positive feedback trading patterns. Froot et al. (2001) use daily international portfolio flow data and find strong evidence of positive feedback trading in international flows. Lastly, Bange (2000) reports evidence of positive feedback trading by U.S. individual investors. While these institutional investors, foreign investors, and individual investors may exhibit positive feedback trading patterns, this pattern does not distinguish between the influence of information-based decisions and behavioral-based decisions. Nofsinger and Sias (1999) suggest that the source of positive feedback trading can be inferred by investment outcome. That is, positive feedback trading that earns high returns indicates that the trading was motivated by information. Positive feedback trading that results in a low return indicates a behavioral-based motivation. For example, an empirical test by Barber and Odean (2001) shows that one cognitive error, overconfidence, is associated with poor investment performance. Therefore, the existence of a positive feedback trading pattern could indicate an informed trader that will earn high returns (information models) or an investor affected by psychological biases that will earn low returns (behavioral models). As the two groups of models predict different outcomes, which group best describes investor behavior? Our study provides evidence that each explanation can describe an investor type. Additionally, we find that both information-based and behavioral-based positive feedback trading can occur in the same marketplace. Of course, investors may use investing strategies that do not manifest themselves as positive feedback trading. A contrarian strategy, or value investing, would manifest itself as a negative feedback pattern. That is, after stock prices decline and become cheap relative to value, value investors buy. Therefore, a negative return is followed by investor buying, which is called negative feedback trading. Lastly, investors may trade using strategies that are not associated with past market returns, such as indexing or trades based on liquidity needs. While theory suggests that investor trading may be characterized by specific trading patterns, like positive feedback trading, empirical studies can identify the actual trading patterns of investor groups. The purpose of this study is to empirically characterize the trading style of seven different investor groups in Japan. The groups are individuals,

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

foreign investors, and five types of institutional investors. To be consistent with theoretical models, we first look for the positive and negative feedback trading patterns. Where positive feedback trading exists, we attempt to identify its motivation (information-based or behavioral-based) using the post trading returns. While theory and existing empirical studies can lead to ex ante hypotheses for some investor groups, we find little direction for other groups. This discussion occurs in the next section. Our unique data includes aggregate weekly investment flow data from the Tokyo Stock Exchange. The data aggregates both the weekly purchases and sales by investor type during 1980 through October 1997. This data gives us the opportunity to examine something different. Whereas most studies of investor behavior study trading patterns associated with individual stocks, we investigate behavior associated with aggregate market timing. The data set also has several other advantages over prior studies. Specific advantages of this data set are the long time-series (nearly 18 years), the short intervals (weekly), and the division of flow data into seven investor types. We find that both foreign investors and Japanese individual investors exhibit positive feedback trading characteristics. Yet foreign investors appear to time the Japanese equity market better than individual investors. Our conjecture that foreign investors are informed traders is consistent with the findings of Froot et al. (2001) but contradicts the conclusions of Bohn and Tesar (1996). Additionally, our conclusions that Japanese individual investors are affected by behavioral biases support the findings of Bange (2000) and Odean (1998) for U.S. individual investors and Kim and Nofsinger (2002a) for Japanese individual investors. Some institutional investors in Japan exhibit negative feedback trading characteristics. Specifically, the weekly flow of banks, insurance firms, and investment trusts are negatively associated with prior market returns. This is contrary to the U.S. evidence, which suggests that institutions positive feedback trade.1 Over the period, the banks appear to have superior market timing ability while the insurance firms and investment trusts do not. Additionally, securities firms do not exhibit any type of feedback trading. Yet this group is one of the better performers. In summary, banks, securities firms, and foreign investors are the winners on the Tokyo Stock Exchange while the Japanese individual investors are the losers. The rest of this paper is organized as follows. The next section reviews existing literature on different investor groups. Section 2 describes the characteristics of the data and describes the initial test methods. Investor group behavior is analyzed in Section 3. Investment performance is evaluated in Section 4. Lastly, concluding remarks appear in Section 5.

1. Related investor literature This paper builds on three areas of investor behavior literature. Our data allows for the study of foreign investors, different types of Japanese institutional investors, and individual investors. The sections below briefly review the work in these three areas.
1

See Grinblatt et al. (1995), Nofsinger and Sias (1999), and Wermers (1999).

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

1.1. Foreign investors The Asian financial crisis of 1997 created considerable interest in the behavior of international investment flows and how they affect local markets. In the study of this crisis, Choe et al. (1999) use transactions data from Korea to examine its impact on investor trading. Their evidence suggests that foreign investors herd together. Using a similar data set to ours, Karolyi (2002) studies weekly investment flow from 1995 to 2001 to study investor reaction to the crisis in Japan. Both studies conclude that although foreign investors appear to exhibit positive feedback trading characteristics, their trading did not destabilize either market during the crisis. Also, Bae et al. (2001) find that since foreign investors are not encumbered by Japanese corporate accounting rules, they can exploit the tendencies for Japanese institutions to sell large amounts of shares at fiscal year-end to realize capital gains. A small number of studies have conducted more general tests of international investment flow. Using quarterly data from 1982 to 1992, Brennan and Cao (1997) find mixed evidence of feedback trading by international investors. However, the quarterly nature of their data may be too coarse for detecting herding and feedback trading when it is conducted within the 3-month intervals. Froot et al. (2001) overcome this issue by using daily international portfolio flow data, but have a shorter time period, 1994 to 1998. They find strong evidence of positive feedback trading in international flows. Additionally, the international flows have positive predictability for equity returns. That is, these international flows have success in timing foreign marketsespecially emerging markets. In a study of Finland, Grinblatt and Keloharju (2000) also find that foreign investors exhibit positive feedback trading and earn positive abnormal returns. This contradicts Bohn and Tesar (1996), who use monthly data and find that U.S. investors do poorly in timing foreign markets. Specifically, they conclude that they earn 15 basis points less than holding an international market portfolio. Focusing on Japan, Kang and Stulz (1997) use annual firm-holdings data (1975 to 1991) to analyze the type of firms held by foreign investors. Their evidence suggests that foreign investors are biased towards large firms since they face high information cost in Japanese markets. Additionally, Hamao and Mei (2001) use monthly portfolio positions of foreigners in Japan from 1974 to 1992. They find that foreign investors demonstrate some degree of market timing ability but the performance is not statistically significant. While the evidence is mixed, it appears that the preponderance of the empirical evidence suggests that foreign investors are positive feedback traders and that this trading is information based. We test this hypothesis. 1.2. Institutional investors On average, institutional investors are better trained and have better resources than individual investors. Institutional investors are subject to the same cognitive biases as individual investors. However, better information and analysis skills may allow institutions to overcome these biases. Indeed, Nofsinger and Sias (1999) find that U.S. institutional investors are positive feedback traders. They study annual firm level ownership and find that the stocks institutions purchase outperform stocks they

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

sell. Therefore, they attribute feedback trading to the information-based trading models. However, Prowse (1990, 1992) argues that the institutional environment in Japan is different than in the U.S. Institutions have more latitude to exert control over firms and often form a teamwork relationship with businesses known as a keiretsu. Kim and Nofsinger (2002b) suggest that this relationship has two important outcomes. First, this closer relationship allows institutional investors access to better information. Second, the teamwork relationship reduces their ability to use the information in trading. Using annual firm level data, they find that Japanese institutions are negative feedback traders and have a high degree of return predictability. However, these papers examine aggregate institutional behavior. Alternatively, this study investigates the trading behavior of five categories of institutional investors: securities firms, insurance firms, banks, industrial companies, and investment trusts. Not all categories of institutional investors may trade in the same manner. For example, each keiretsu is centered on a bank. These banks are likely to be well informed about the prospects of the firms in the keiretsu (Hoshi et al., 1991). While the keiretsu relationship may dictate how and when investment is made in the firms, we expect the trading of banks to reflect this inside knowledge. The industrial companies within the keiretsu are also long-term shareholders of the other companies in the group. But they may not have the same level of information as the banks. Insurance firms are also long-term investors. But they do not benefit from the close relationships of the keiretsu. Alternatively, investment trusts (similar to U.S. mutual funds) tend to have a shorterterm focus in seeking returns. Therefore, their trading behavior is likely to be different than the long-term focused institutions. U.S. equity mutual funds exhibit positive feedback trading characteristics on a quarterly basis (Grinblatt et al., 1995; Wermers, 1999). Whether this behavior is information based or behavioral based is unclear. The stocks mutual funds pick tend to beat the market, but the abnormal return is not large enough to compensate investors for fees and trading costs. Lastly, we investigate securities firms. Little research has been conducted on the behavior and performance of these investors. However, securities firms may have good firm-specific information through their dealings with the companies and may have good supply and demand information through their dealings with investors. Therefore, we expect securities firms behavior to be indicative of information-based trading. The theories and empirical studies mentioned in this section are silent on the behavior of institutions at the aggregate market level. That is, institutions may buy and sell specific firms because of informational advantages. But do they have market wide informational advantages? This study investigates trading activity at the market level, rather than at the firm level. 1.3. Individual investors The recent studies of individual investor behavior suggest that they frequently succumb to their cognitive biases. In letting behavioral problems such as disposition effect and overconfidence affect their investment decisions, U.S. individual investors tend to trade too much (Barber and Odean, 2000, 2001; Odean, 1999), are reluctant to realize losses

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

(Odean, 1998), and positive feedback trade (Bange, 2000). Consequently, their investment performance is poor. For example, the stock investors sell outperforms the stocks they purchase by nearly 6% the following year (Odean, 1999). However, the evidence of whether individual investors are positive or negative feedback traders is mixed. Bange (2000) and Kim and Nofsinger (2002a) report evidence of positive feedback trading while Grinblatt and Keloharju (2000), Jackson (2002), Murase (2001), and Odean (1998) find evidence of negative feedback, or contrarian, trading. The Bange study is particularly applicable to our study because it investigates investor sentiment about the aggregate market.

2. Investment flow data 2.1. Investor groups and trading The Tokyo Stock Exchange (TSE) collects and reports on the yen value of both aggregate buying and selling each week. Thus, our data uses weekly flow rather than monthly, quarterly, or even annual flows. Additionally, these buying and selling investment flows are classified by seven investor groups, including security firms, insurance firms, banks, individuals, industrial companies, foreigners, and investment trusts (mutual
Table 1 Summary statistics of weekly equity sales and purchases (Million yen) Foreign investors Selling Buying Securities firms Selling Buying Banks Selling Buying Insurance firms Selling Buying Investment trusts Selling Buying Companies Selling Buying Individual investors Selling Buying Mean (weekly) 256,628 255,492 471,995 499,920 282,502 304,865 21,212 20,353 133,339 133,356 183,789 176,030 499,697 471,732 Standard deviation 173,864 173,009 376,574 409,434 348,389 356,274 22,566 17,667 135,466 144,571 231,450 228,644 442,252 428,791 Minimum 5092 3023 12,462 13,537 1672 4552 52 431 1147 3018 2327 3841 18,711 16,120 Maximum 1,226,617 931,796 2,244,457 2,376,596 1,916,819 2,016,594 229,282 125,569 764,966 857,569 1,422,126 1,337,747 2,477,824 2,464,171

This table reports the descriptive statistics for the weekly buying and selling of equities on the Tokyo Stock Exchange in millions of yen. The trades are aggregated by investor type. The sample period is January 1980 to October 1997 and represents 926 weeks of trading.

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

funds). Lastly, the available data encompasses a longer time period than prior studies. The analysis uses this weekly investment flow data during January 1980 to October 1997 for the seven investor groups. The weekly investment flow of foreign investors begins to be somewhat contaminated in the middle of the 1990s by some domestic investors starting to trade through foreign accounts. The largest contamination is by Japanese pension funds, which started trading using foreign accounts in the mid-1990s. Our long sample period is an advantage because this problem only affects the last three out of the 18 years of our sample period. The problem may be much more pronounced in Karolyis (2002) sample of 1995 to 2001 or Bae et al.s (2001) sample of 1991 to 1999. Table 1 summarizes the buying and selling amounts of each investor group. Foreign investors sold an average 257 billion yen of stock per week on the TSE during the sample period. The minimum sales week, valued at 5 billion yen, was during the first week in January 1982. The maximum sales week was valued at 1227 billion yen and was during the fourth week in October 1987. Foreign investors made average weekly purchases of 255 billion yen. The minimum and maximum purchases occurred during the first week of January 1982 and the second week of March 1996, respectively. The data does not distinguish between foreign institutional and foreign individual investors. However, it is likely that most of the trading by foreign investors is done by foreign mutual funds, hedge funds, pension funds, and investment banks. Of the five types of domestic institutional investors in Japan, securities firms conduct the most trading.2 Securities firms were net purchasers during the period, averaging 471 billion yen in sales and 500 billion yen in purchases. It should be noted that the investment flow values for securities firms are for their trading on their own accounts only. It does not include commission trades of other investors. Banks are the next largest trading institution. They averaged 283 billion yen in weekly sales and 305 billion yen in purchases. Also note that banks have a very high standard deviation of trading relative to the level of their mean trading. Industrial companies and investments trusts have only a modest role in the trading during our period and insurance firms conducted very little trading, with only 21 and 20 billion yen in weekly sales and purchases, respectively. Japanese individual investors were one of the largest trading groups on the TSE during our sample. On average, individuals sold 500 billion yen of equities per week and purchased 472 billion yen. Of all seven investor types, individuals had the highest trading week, selling 2.48 trillion yen and purchasing 2.46 trillion yen during the third week in February, 1989. Although the mean trading of individuals in Japan is very high, the relative level of trading decreased during the period. For example, in the early 1980s, individuals traded equities at a level 75% higher then securities firms. By the late 1980s, individuals were trading about the same amount as securities firms. By the mid-1990s, individuals were trading less then half of the amount traded by securities firms.

2 We discuss the behavior and performance of investment trusts as an institutional investor. However, as these investment trusts are similar to U.S. mutual funds, it may be more appropriate to think of them as na ve individual investors. Although portfolio managers make the buy and sell decisions, the aggregate investment flow into (or out of) the market is probably more a function of the individual investors money flow into (or out of) the mutual funds.

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

2.2. Net investment flow In order to investigate investor behavior and performance, we calculate each weeks trade imbalance by investor group. We use weekly trading instead of aggregating to monthly or quarterly because Chan et al. (2000) find that stock momentum in Japan occurs on a weekly time period. As we are interested in whether investor group i was a net buyer or seller during week t, we calculate the following Net Investment Flow (NIF) measure: NIFit Purchasing Valueit Selling Valueit : Purchasing Valueit Selling Valueit 1

NIF will be positive (negative) when the investor group buys more (less) equities than sells during the week. We interpret large trade imbalances in either direction as an indication of market timing. Large net buying (selling) signals that the investor group thinks the TSE is under- (over-)valued relative to the alternatives. Table 2 summarizes the sample statistics of NIF. Foreign investors, with an average NIF of 0.004, were not net buyers or sellers during the period. However, they did experience large swings in investment flow, with a large net selling of 0.743 one week and a large net buying of 0.571 in another week. The standard deviation of 0.170 indicates that foreign investors ranked third among the seven investor groups in variation of net buying and selling. These estimates suggest that foreign investors are no more prone to market timing of the TSE than other major investor groups. For example, banks are a major trader on the TSE. Banks, a net buyer during our sample, experienced a variation of 0.167 and a minimum and maximum of 0.732 and 0.633, respectively. These estimates are very similar to those of foreign investors. With a standard deviation of 0.337, insurance firms had the largest variation between buying and selling. Their minimum and maximum NIFs were 0.946 and 0.890, respectively. However, insurance firms are under strict regulatory rules that cause them to do most of their trading near their fiscal year end, usually March (see Bae et al., 2001). Therefore,

Table 2 Summary statistics of weekly net investment flow Mean (weekly) Foreign investors Securities firms Banks Insurance firms Investment trusts Companies Individual investors 0.004 0.026 0.082 0.050 0.024 0.034 0.040 (0.72) (10.44)*** (14.93)*** (4.55)*** ( 4.24)*** ( 8.36)*** ( 16.96)*** Standard deviation 0.170 0.077 0.167 0.337 0.175 0.123 0.072 Minimum 0.743 0.368 0.732 0.946 0.557 0.556 0.340 Maximum 0.571 0.383 0.633 0.890 0.606 0.521 0.265

This table reports the descriptive statistics for each investor-groups weekly Net Investment Flow (NIF). The NIF is computed as NIFit Purchasing Valueit Selling Valueit =Purchasing Valueit Selling Valueit for each investor group i during week t. The sample period is January 1980 to October 1997 and represents 926 weeks of trading. *** Denotes significance at the 1% level.

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

their co-movement in trading is regulatory induced instead of information based. As they are the lowest traders on the TSE (in terms of yen amounts), their large variation in net investment flow is unlikely to affect the market. Securities firms, one of the largest trading groups on the TSE, were a net buyer of equities. However, their variation of net investment flow was one of the lowest of the groups. Investment trusts and companies were net sellers of equities during the period. The variation of NIF for both firms indicates that they were moderate market timers. Japanese individual investors were net sellers of equities in the TSE during the sample period. The standard deviation of 0.072 is the lowest among the seven investor groups. Additionally, measuring variation by using the minimum and maximum trading imbalance shows that 0.340 and 0.265 are also the smallest of the groups. In many studies, large trading imbalances are indicative of investor herding. Herding is defined as a group of investors buying or selling at the same time interval (Nofsinger and Sias, 1999). The length of the time interval is an empirical issue and could be as short as 1 day or as long as 1 year. Theory suggests that investors could herd for rational reasons such as they are following the same information signals (Froot et al., 1992; Hirshleifer et al., 1994). Or, investors could herd for irrational reasons like following fads (Bikhchandani et al., 1992). The large variation in NIF suggests that Japanese insurance firms engage in herding behavior the most. This suggests that either insurance firms tend to follow the same information signals or they are susceptible to investment fads. To determine whether the herding is rational or not, Nofsinger and Sias (1999) recommend examining the post herding returns. A high return after buy-herding (or low return after sell-herding) indicates rational herding. Investment trusts, foreign investors, and banks rank next in their level of herding. Lastly, securities firms and individual investors appear to herd to a lesser extent. Although the herding models are formed for herding on individual securities, our results should be interpreted as herding into and out of the stock market.

3. Investment flow patterns In this section we investigate the potential source of investor herding and feedback trading. Feedback trading is also known as either contrarian investing when the trade imbalance is negatively correlated with past return or momentum investing when the correlation is positive. By examining the returns in the week(s) prior to the trading week, we can examine the extent to which investors positive or negative feedback trade. Previous investigations of herding focus on the association between trading and individual stocks, whereas we investigate flows into and out of the aggregate stock market.3 We begin by reporting the Pearson correlation coefficients between the investor group NIF measures in Table 3. Foreign investor investment flow is uncorrelated with Japanese securities firms and significantly negatively correlated with the other five investor groups.

3 One notable exception is Karolyis (2002) study of the investor group reaction to the Asian financial crisis in 1997. Also, Murase (2001) investigates the correlation coefficients between investor groups and market returns.

10

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

Table 3 Correlation of weekly net investment flow of Japanese equity investors Foreign Foreign Securities Banks Insurance Trusts Companies Individuals Return (t = 0) Return (t = 1) Return (t = 2) Return (t = 3) Return (t = 4) 1 0.012 0.345*** 0.252*** 0.413*** 0.437*** 0.273*** 0.195*** 0.173*** 0.159*** 0.155*** 0.128*** Securities 1 0.344*** 0.064* 0.118*** 0.275*** 0.332*** 0.296*** 0.137*** 0.162*** 0.172*** 0.171*** Banks Insurance Trusts Companies Individuals

1 0.257*** 0.159*** 0.214*** 0.193*** 0.231*** 0.198*** 0.220*** 0.220*** 0.189***

1 0.221*** 0.344*** 0.035 0.153*** 0.116*** 0.083** 0.081** 0.056*

1 0.218*** 0.030 0.042 0.184*** 0.185*** 0.175*** 0.145***

1 0.458*** 1 0.352*** 0.159*** 0.096*** 0.004 0.093*** 0.011 0.077** 0.020 0.063* 0.020

Pearson correlation coefficients are reported between each investor groups NIF and market returns. Return is the weekly return of the TOPIX index for the week of the investment flow (t = 0) and the preceding four weeks (t = 1, 2, 3, and 4). The sample represents 926 weeks of investment flow. * Denotes significance at the 10% level. ** Denotes significance at the 5% level. *** Denotes significance at the 1% level.

Banks, insurance firms, investment trusts, and companies all have a significantly positive trade flow correlation with each other. Japanese individual investors have significantly negative trade flow correlation with foreign investors, securities firms, and banks. Individual investor trade flow is uncorrelated with insurance firms and investments trusts, and is positively correlated with companies trade flow. The table also reports the trade flow correlation with weekly returns. Specifically, the correlation between trade flow and TOPIX return during the week (t = 0) of the trading and during the previous 4 weeks (t = 1,. . ., 4) is shown. A positive (negative) correlation between trade flow and market return during the previous weeks indicates that the group is positive (negative) feedback trading. Foreign investor trade flow is positively correlated with the TOPIX return. The estimates for correlation on past returns are also significantly positive. This suggests that foreign investors are positive feedback, or momentum, traders. This herding by foreign investors is similar to the conclusions of Choe et al. (1999) and Grinblatt and Keloharju (2000) who find foreign investor herding in South Korea and Finland, respectively. This pattern also describes securities firms. The investment flow of securities firms is also positively correlated with the market return and the previous weeks returns. This is consistent with the findings of institutional investors in the U.S. That is, U.S. institutional investors tend to be momentum investors on individual stocks. The trading flow of banks, insurance firms, investment trusts, and companies is all negatively correlated with the current and past market returns. This suggests that these investor groups employ a negative feedback, or contrarian, trading strategy. This is different from the momentum trading of U.S. institutional investors (Nofsinger and Sias, 1999) and mutual funds (Wermers, 1999). However, Grinblatt and Keloharju (2000) find weak evidence of contrarian investing by Finnish institutions using daily data. Additionally, Kim and Nofsinger (2002b) study annual changes in stock ownership and find that

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

11

Japanese institutional investors engage in contrarian strategies. Although their evidence of this strategy was stronger during the bull market (through 1989) than during the bear market (post 1989). The high degree of correlation between these four groups and their common trading strategy suggests that they may herd into and out of the market together. Individual investor flow is negatively correlated with the TOPIX return during the week of the trading. However, individual investor flow is uncorrelated with past weekly market returns. That is, individual investors do not appear to be market timing feedback traders at least not on a weekly herding period. The evidence for individual investor feedback trading is mixed. For example, using firm level data, Odean (1998) and Grinblatt and Keloharju (2000) report evidence that U.S. and Finnish individual investors are negative feedback traders. Alternatively, Kim and Nofsinger (2002a) conclude that Japanese individual investors are positive feedback traders. Their position is also consistent with Bange (2000), who uses survey data to find that U.S. individual investors exhibit positive feedback trading characteristics. The Bange study is particularly applicable to our findings because her data deals with changes in aggregate equity holdings and thus reflects market timing into and out of the stock market. When an investor group herds into (or out of) the stock market, it may take a longer period than 1 week. If the herding period were longer than 1 week, the weekly NIFs would be positively correlated. However, since the investor groups NIF is correlated with both past flow and past returns, we need to be cautious in the interpretation of the Table 3 results. That is, multicollinearity may cause erroneous conclusions. To account for this problem, we investigate trading patterns by estimating a simple bivariate VAR( p) model. Specifically, we propose the VAR model: Yt C
p X j1

0 j Y t j ut

where Yt is a 2 1 vector of NIF and TOPIX return, C is a 2 1 matrix of constants, 0j is a 2 2 matrix of parameters, Yt j is the 2 1 matrix of NIF and TOPIX return for week lag j, and ut is the 2 1 error matrix. We specify a VAR(4) model which has four lags, p = 4. The VAR model is estimated for each investor group using heteroscedasticityconsistent standard errors. One benefit of using the VAR model is the ability to test for Granger causality between investor group trading and market returns. We implement the causality test by computing a Wald test with the null hypothesis that all four lag coefficients are equal to zero. Actually, we compute one test for the four lagged NIF coefficients and one for the lagged TOPIX return. We use four weekly lags because Chan et al. (2000) show that most of the return momentum in Japan occurs during the first 4 weeks. Additionally, Bange (2000) shows that U.S. individual investors positive feedback trade on a monthly basis. The estimation results are reported in Table 4 along with the coefficient of determination (R2). We first discuss the NIF side of the bivariate VAR(4) model. All seven investor groups show positive autocorrelation with their trading for at least 1 week. That is, the coefficient on the previous weeks NIF is significantly positive in each regression. In fact, many of the coefficients for NIF of 2, 3, and 4 weeks prior to the trading week

12

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

Table 4 VAR model estimates of net investment flow and performance Foreign NIF Constant NIF 1 2 3 4 TOPIX return 1 2 3 4 0.001 Security firms TOPIX NIF 0.001 0.013*** Banks Insurance firms TOPIX NIF 0.001 0.006 TOPIX 0.001

TOPIX NIF 0.001 0.024***

0.573*** 0.002 0.417*** 0.001 0.136*** 0.008 0.048 0.000 0.042 0.005 0.102** 0.009 0.105*** 0.004 0.001 0.006 0.412** 0.013 0.560*** 0.049 0.130 0.057 0.509*** 0.007 1.71 3.75 0.01 0.031 0.156 0.051 0.039 160.45*** 2.84 0.19

0.456*** 0.009 0.045 0.005 0.114*** 0.008 0.101*** 0.008

0.499*** 0.002 0.064 0.003 0.079* 0.004 0.171*** 0.002

0.008 0.734*** 0.028 0.805** 0.004 0.057 0.128 0.061 0.833** 0.061 0.054 0.061 0.078 0.103 0.077* 0.015 0.690*** 0.007 1.155*** 0.002 1.33 3.55 0.01 394.72*** 29.43*** 0.38 4.94 6.38 0.01 874.76*** 17.72*** 0.50 3.61 6.21 0.01

NIF Wald test 1044.26*** TOPIX Wald test 33.93*** R2 0.61

This table reports the bivariate VAR(4) model estimates by investor group for the variables; weekly NIF and weekly market return, as proxied by the TOPIX index return. Specifically, the estimated model is Y t C P 4 1 vector of NIF and TOPIX return, C is a 2 1 matrix of constants, 0j is a j1 0j Yt j ut , where Yt is a 2 matrix of parameters, Yt j is the matrix of NIF and TOPIX return for week lag j, and ut is the error matrix. * Denotes significance at the 10% level. ** Denotes significance at the 5% level. *** Denotes significance at the 1% level.

have significantly positive coefficients. None of the lagged NIF coefficients is significantly negative. The Wald tests for the lagged NIF coefficients are significant at the 1% level for all investor groups. Our interpretation is that there is significant Granger causality between current investment flow and the flow of the past month. Additionally, these results suggest that the net direction of the trading (buying versus selling) occur for multiple consecutive weeks. That is, the true herding period for these investor groups is likely to be longer than 1 week. Therefore, the correlation coefficients in Table 3 may not fully represent the feedback trading strategy. To investigate feedback trading behavior, we examine the coefficients on past returns by the investor groups. For foreign investors, the coefficients for the first lagged return, week t = 1, are significantly positive at the 1% level. However, the coefficients for weeks t = 2, t = 3, and t = 4 are significantly negatively. This indicates that foreign investors are positive feedback traders on a short-term (weekly) period but may be negative feedback traders over a longer period. Since the Wald test for the lagged returns is significant at the 1% level, foreign investment flow is Granger-caused by past returns. An R2 of 0.61 indicates that the VAR model fits the foreign investor NIF reasonably well. For securities firms, the coefficients for lagged returns are positive, but not statistically significant. The Wald test for lagged returns is not significant either. Securities firms trading does not appear to be Granger caused by past market returns. The low R2 of 0.19 suggests that the model does not explain their trading flow very well.

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

13

Trusts NIF 0.005 0.509*** 0.009 0.041 0.174*** 1.258*** 0.106 0.073 0.430** 635.05*** 40.62*** 0.41 TOPIX 0.001 0.004 0.007 0.003 0.006 0.010 0.055 0.070 0.006 1.38 4.97 0.01

Companies NIF 0.010*** 0.445*** 0.181*** 0.009 0.111 0.265* 0.372** 0.120 0.246* 400.15*** 18.48*** 0.37 TOPIX 0.002* 0.006 0.010 0.011 0.003 0.022 0.041 0.079 0.004 3.37 4.84 0.01

Individuals NIF 0.011*** 0.525*** 0.087 0.076* 0.049 0.302*** 0.083 0.067 0.024 358.19*** 15.37*** 0.41 TOPIX 0.001 0.017 0.005 0.003 0.001 0.002 0.061 0.062 0.003 1.48 4.49 0.01

Banks, insurance firms, and investment trusts appear to be negative feedback traders on the short-term. All three have significantly negative TOPIX Return (t = 1) coefficients. The coefficients turn positive for greater return lags. This indicates that they may be positive feedback traders over a longer (monthly or more) period. The coefficients on the lagged returns for company investment flow are positive. This suggests companies are positive feedback traders. Note that this is the opposite of the company negative feedback trader results in Table 3. Whether it is positive or negative feedback trading, the Wald test indicates that the investment flow of banks, insurance firms, investment trusts, and companies all can be considered to be Granger-caused by past market weekly returns. Lastly, after accounting for the past investment flow, individual investors exhibit positive feedback characteristics over the short term. The coefficient for the previous weeks return is 0.302 and is significant at the 1% level. This positive feedback trading result, which is contrary to the result in Table 3, is consistent with the findings of Bange (2000). The significant Wald test statistic confirms that past returns are important for individual investor trading flow. Although the focus of this study is investor behavior, our bivariate model allows us to examine whether the TOPIX return exhibits Granger causality with investment flow. None of the seven model estimates shows that investment flow affects the market return. In fact, past market weekly returns do not Granger-cause current returns either. This is different

14

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

from the results of Chan et al. (2000) who find that the Japanese market exhibits momentum on a weekly basis. The difference between our analysis and theirs is that we control for investment flow. In summary, after taking into account investment flow autocorrelation using the bivariate VAR(4) model, foreign investors, companies, and Japanese individual investors appear to be short-term positive feedback traders. Alternatively, banks, insurance firms, and investment trusts exhibit short-term negative feedback market timing characteristics. Securities firms trading appears unrelated to feedback trading.

4. Investment performance of investor groups Our evidence shows that in market timing, some investor groups in Japan are positive feedback traders while others are negative feedback traders. Both information-based and behavioral-based models predict positive feedback trading. However, the two groups of models have different predictions about post herding performance. Therefore, we evaluate the investment performance of these investor groups in this section. The nature of the data precludes a direct estimate of investor group performance because we do not have portfolio holdings, only aggregate weekly purchases and sales in the stock market. Nevertheless, we can gauge the market timing ability of these investor groups by examining the market returns after each trading week. Each groups market timing performance is evaluated using two different methods. First, we examine the market performance after those weeks where investors conducted particularly heavy buying or selling. Second, we estimate the cumulative yen return due to the weekly changes in investment flow and the following market return for each investor group. 4.1. Investment performance To study the performance of our investor groups after heavy buying and selling weeks, we sort each groups NIF onto five equal sets. The quintile of the weeks with the highest positive NIF is designated the buying weeks. The quintile of the largest negative NIF is the selling weeks. The buy and sell weeks represent the weeks with the greatest trade imbalance by the group of investors. To determine the market timing ability of the investors, we compute the 1-week, 1-month, and 2-month total return following the trading week. The results of the analysis are reported in Table 5. Foreign firms have a NIF of 0.232 during the buying weeks and a 0.235 NIF during the sell weeks. One week after the buying, the market increases 0.16%, on average. One week after the selling, the market declines 0.04%. The difference between the post trading week return after buy weeks and sell weeks is tested using a difference in means t-statistic. The post trading week returns are not significantly different for foreign investors. The month following the buy and sell weeks experienced a market return of a significant 1.02% and an insignificant 0.28%, respectively. The difference is not significant. The 2-month return following the foreign investor trading is a significant 1.72% after the buy weeks and an insignificant 0.39% after the sell weeks. The difference is significant at the 10% level. Overall,

Table 5 Market performance after buying and selling by investor type Net investment flow Buy Sell Mean return Following week Buy Foreign Securities Banks Insurance Trusts Companies Individuals 0.232 0.117 0.323 0.489 0.221 0.122 0.049 0.235 0.087 0.143 0.459 0.272 0.218 0.145 0.16% (0.93) 0.14 (0.68) 0.09 (0.53) 0.23 (1.47) 0.18 (1.11) 0.09 ( 0.51) 0.22 ( 1.02) Sell 0.04% ( 0.22) 0.06 ( 0.36) 0.12 (0.61) 0.06 (0.30) 0.19 (1.18) 0.01 (0.06) 0.43 (0.21) Difference (t-statistic) 0.77 0.75 0.13 0.69 0.03 0.40 0.93 Mean return Following month Buy 1.02% (2.69)*** 1.06 (2.45)** 0.51 (1.56) 1.23 (3.82)*** 0.55 (1.47) 0.17 (0.48) 0.72 (1.75)* Sell 0.28% (0.70) 0.57 ( 1.57) 0.39 (0.93) 0.42 (1.14) 0.81 (2.28)** 0.21 (0.56) 0.43 (1.07) Difference (t-statistic) 1.34 2.88*** 0.23 1.68* 0.50 0.06 1.99** Mean return A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122 Following 2 months Buy 1.72% (3.59)*** 1.49 (2.47)** 0.75 (1.53) 2.32 (5.80)*** 1.85 (3.29)*** 0.68 (1.38) 1.12 ( 1.84)* Sell 0.39% (0.67) 0.55 ( 1.02) 0.28 (0.48) 1.00 (1.97)* 1.09 (1.98)** 0.44 (0.76) 0.62 (1.06) Difference (t-statistic) 1.75* 2.52** 0.61 2.04** 0.98 0.32 2.06**

We sort each investor-groups NIF onto five equal sets. The quintile of the weeks with the highest positive NIF is designated the buying weeks for that investor group. The quintile of the largest negative NIF is the selling weeks. The buy and sell weeks represent the weeks with the greatest trade imbalance by the group of investors. Each quintile of buy or sell weeks has 185 observations. We compute the 1-week, 1-month, and 2-month total return following the buy or sell trading week. The mean NIF and returns are reported in the table. The difference between the post trading week return after buy weeks and sell weeks is tested using a difference in means t-statistic. * Denotes significance at the 10% level. ** Denotes significance at the 5% level. *** Denotes significance at the 1% level.

15

16

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

foreign investors appear to have some success in timing the Japanese stock market.4 Hamao and Mei (2001) do not find statistical significance for foreign investor market timing in Japan. We attribute our stronger results to two major differences between our methods and their methods. Their timing measure only incorporates whether the investors are net entering or exiting the stock market and whether the future excess return was positive or negative. Investor decisions and future outcomes are treated as binary variables. Our analysis also includes the magnitude of the trading decision and outcome. That is, we examine only large net purchases or sales and then examine the magnitude of the future return. By adding magnitude, we increase the power of our tests. The trading imbalance of the buy and sell weeks is low for securities firms. The NIFs for the buy and sell weeks are 0.117 and 0.087, respectively. The difference in trade imbalance (0.204) is less than half of that of foreign investors, banks, insurance firms, and investment trusts. The market experiences a 0.14% return the week following the buying and 0.06% return following the selling. The difference is not significant. A market return of 1.06% (significant at the 5% level) occurs 1 month after the buying. This is significantly different than the 1-month return of 0.57% following selling. The 2-month market return following the buying is also significantly larger than the return following the selling weeks. Securities firms appear to have some success in market timing. The stock market performance after the buying weeks for banks, investment trusts, and companies is not significantly different from the weeks after the selling. This indicates that these investor groups are not successfully timing the market. Alternatively, insurance firms have extreme trade imbalances of 0.489 and 0.459 during the buying and selling weeks, respectively. The market return after the buy weeks significantly outperforms the weeks after the selling weeks for 1- and 2-month periods. Individual investors experience the lowest levels of buying and selling trade imbalance. The buy and sell weeks NIFs are 0.049 and 0.145, respectively. Contrary to the performance of the other investor groups, individual investors exhibit poor market timing ability. The week after buying, individuals experience a market return of 0.22% while the return is 0.43% after selling. The 1- and 2-month returns after buying are a significant 0.72% and 1.12%, respectively. The 1- and 2-month returns after selling are 0.43% and 0.62%, respectively. The 1- and 2-month returns after buying are significantly different than after the selling weeks. The 1-month return after selling outperforms the month after buying by 1.15%. Compounded, the difference is 14.7% annually. Using U.S. individual investor brokerage account data, Odean (1999) finds that the stocks individual investors sell outperform the stocks they buy by 5.8% during the following year. While their study uses firm-level data, our findings confirm their results at the aggregate market level. That is, not only are individual investors poor stock traders, they are also poor market timers.

4 The investor flow data for the week is not publicly available until the following Thursday. We do not feel this is problematic because we are studying investor behavior, not advocating a trading strategy.

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

17

4.2. Cumulative performance Instead of examining the heavy buying and selling weeks, in this section we evaluate the relative market timing ability of the investor groups over the entire period. Grinblatt and Titman (1993) develop a performance measure based on the change in the portfolio holdings. Their Portfolio Change Measure captures the positive covariance between the return on asset j, Rj, and the change in proportional holdings of that asset, wj. The measure is COV
T X N X t 1 j1

wj;t wj;t1 Rj;t =T

where N is the number of the assets and T is the estimating sample period. In this paper, we estimate this portfolio change measure for each investor i, under slightly different conditions. We assume that there are only two assets. One is the stock market index and the other is the risk-free rate. The proxy for the weekly market return is the TOPIX index and the weekly risk-free rate is assumed to be zero. In this case, the Portfolio Change Measure simplifies to COV
T X wt wt1 Rt =T ; t 1

where Rt is the return on the market index during period t. Similar to Karolyi (2002), we estimate Eq. (4) by substituting the change in the portfolio weights with the net investment purchases in yen during the week. The following empirical specification estimates the cumulative yen return due to the weekly changes in investment flow and the following market return: Cumulative Yen Return
T X Purchasest1 Salest1 Rt : t 1

Eq. (5) is estimated for each investor group. The cumulative yen return is graphed in Fig. 1. The numbers on the horizontal axis of the figure represent time as denoted by year, month, and the week of the month (YYYYMMW). The y-axis shows the cumulate return in billion yen. Securities firms, banks, and foreign investors perform the best. These three investor groups earn around 100 billion yen each in market timing trading. The total yen return for insurance firms, investment trusts and companies is nearly zero over the 17-year period. Individual investors are clearly the market losers, losing 180 billion yen. Overall, market-timing performance does not appear to be related to feedback trading strategy. For example, foreign investors, securities firms, and banks all perform well. But foreign investors are positive feedback traders while banks are negative feedback traders. Securities firms trading is not related to feedback trading. Even though foreign investors follow a positive-feedback trading pattern and perform well, individual investors also

18

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

Fig. 1. Cumulative yen returns by investor type.

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

19

follow a positive feedback strategy but perform poorly. Apparently, not all positivefeedback trading is alike. Indeed, although both foreign and individual investors exhibit positive feedback trading patterns, their trading is negatively correlated (see Table 3). The figure also suggests that the difference between investor group performance is mostly due to the two periods 1986 and 1990. Although the Japanese stock market had steadily climbed in the few years prior, 1986 was the year in which the bull market accelerated. The TOPIX earned over 50% in 1986. Alternatively, 1990 was the year in which the Japanese stock market collapsed. The TOPIX lost 38% of its value in 1990. The index lost 18% just in the months of February and March. Table 6 shows the net trading activity of the investor groups in the two periods. The first column of results in Table 6 show that securities firms and banks were heavy net buyers in the market during 1986, the beginning of the great bull market. On average, securities firms bought 95.1 billion yen more in stocks than they sold each week. Alternatively, foreign investors and individual investors were heavy net sellers of stocks at the beginning of the bull market. In the first 2 months of the stock markets collapse (February and March of 1990), foreign investors and banks sold heavily. On average, they sold 293.6 and 233.5 billion yen more than they purchased each week, respectively. On the other hand, individual investors were heavy buyers in the market during these 2 months. Individuals bought 262.4 billion yen more than they sold each week. Further examination of the data (not shown in the table) indicates that the heaviest net buying by individual investors was during the first 2 weeks of March 1990. Table 6 and Fig. 1 show that much of the terrible performance by individual investors comes from trading in the wrong direction at the beginning of the bull market and at the beginning of the bear market. Alternatively, banks were able to successfully buy at the beginning of the bull market and sell at the beginning of the bear market. This accounts for much of their good investment performance over the sample period. While foreign investors traded in the wrong direction during the beginning of the bull market, they were successful in selling at the beginning of the bear market. This explains how the poor investment performance in the first half of the sample period is reversed and becomes a good performance by the end of the period.

Table 6 Mean weekly net investment at the beginning of bull and bear markets billion yen Beginning of bull market (1986) Foreign investors Securities firms Banks Insurance firms Investment trusts Companies Individual investors 81.7 95.1 50.1 1.5 16.7 3.8 58.5 Beginning of bear market (February and March 1990) 263.6 40.9 233.5 13.8 47.4 71.4 262.4

This table reports the mean weekly net investment for each investor-group during the two periods 1986 and February through March of 1990. The net investment is the total purchasing value less the total selling value for the week.

20

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

5. Summary We use a unique data set of weekly purchases and sales over an 18-year period on the Tokyo Stock Exchange (TSE). The data details the aggregate investment flow of seven investor groups: foreigner investors, Japanese individual investors, and five classifications of institutional investors. This data allows us to study the market timing investment behavior and performance for these groups of investors. Two categories of theoretical trading models have developed to explain investor positive feedback trading behavior: information-based and behavior-based motivations. The two groups of models predict different performance as a consequence of the trading. Our study provides evidence that each explanation can describe an investor type. Specifically, we find that foreign investor positive feedback trading appears to be associated with superior information as they experience high market-timing returns. For example, the TOPIX earns an average 1.72% in the 2 months after heavy buying by foreign investors but only 0.30% after heavy selling. We estimate that their market timing activities are associated with 100 billion yen in profits during the sample period. Interestingly, banks also have success in market timing but their trading is associated with negative feedback, or contrarian, trading. Their success may be attributed to their access to information through the keiretsu relationships. Securities firms have success but their trading is not associated with any type of feedback trading. Alternatively, the positive feedback trading of Japanese individual investors is likely driven by behavioral factors as they experience poor market timing returns. The TOPIX earns a 1.12% return during the 2 months after heavy individual investor buying and earns 0.62% after heavy selling. We estimate that market timing is responsible for aggregate individual investor losses of 180 billion yen over the 18-year sample period. We document that the same trading strategy can have different outcomes for different types of investors. This suggests that both information-based and behavioral-based positive feedback trading can occur in the same market. Or results show that foreign investors, banks, and securities firms are the clear market-timing winners of TSE. The market-timing performance of insurance firms, companies, and investment trusts is not notable. Japanese individual investors are the clear losers.

Acknowledgements The authors thank Syunji Tamiya of Nippon Life Insurance and Masahiko Kusumi of Takachiho University for providing the data. We also thank an anonymous reviewer, Masato Hirota, Akitoshi Ito, Andrew Karolyi, Takao Kobayashi, Allen Poteshman, Toyoharu Takahashi, Masashi Toshino, Yoshiro Tsutsui, Masahiro Watanabe, Takeshi Yamada, Katsunari Yamaguchi, Yasuhiro Yonezawa, Toshiki Yotsuzuka, and participants of the meetings of the Japanese Economic Association, the Midwest Finance Association, Finance Forum of Institute for Posts and Telecommunications Policy, Nippon Finance Association , Japan Society of Monetary Economics, and the APFA/PACAP/FMA Tokyo Finance Conference for helpful comments and suggestions. This paper was selected for the

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

21

Ibbotson Associates Japan Research Award, 2002, APFA/PACAP/FMA Finance Conference. All remaining errors are our own.

References
Bae, K., Ito, K., Yamada, T., 2001. How do foreigners profit from local investors? Evidence from trades in the Japanese stock market. Hong Kong University of Science and Technology working paper, August. Bange, M., 2000. Do the portfolios of small investors reflect positive feedback trading? Journal of Financial and Quantitative Analysis 35, 239 255. Barber, B., Odean, T., 2000. Trading is hazardous to your wealth: The common stock investment performance of individual investors. Journal of Finance 55, 773 806. Barber, B., Odean, T., 2001. Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics 116, 261 292. Bikhchandani, S., Hirshleifer, D., Welch, I., 1992. A theory of fads, fashion, custom, and cultural change as informational cascades. Journal of Political Economy 100, 992 1026. Bohn, H., Tesar, L., 1996. U.S. equity investment in foreign markets: portfolio rebalancing or return chasing. American Economic Review 86, 77 81. Brennan, M., Cao, H., 1997. International portfolio investment flows. Journal of Finance 52, 1851 1880. Chan, K., Hameed, A., Tong, T., 2000. Profitability of momentum strategies in the international equity markets. Journal of Financial and Quantitative Analysis 35, 153 172. Choe, H., Kho, B., Stulz, R., 1999. Do foreign investors destabilize stock markets? The Korean experience in 1997. Journal of Financial Economics 40, 31 62. Daniel, K., Hirshleifer, D., Subrahmanyam, A., 1998. Investor psychology and security market under- and overreactions. Journal of Finance 53, 1839 1885. Froot, K., Scharfstein, D., Stein, J., 1992. Herd on the street: Informational inefficiencies in a market with shortterm speculation. Journal of Finance 47, 1461 1484. Froot, K., OConnell, P., Seasholes, M., 2001. The portfolio flows of international investors. Journal of Financial Economics 59, 151 193. Gervais, S., Odean, T., 2001. Learning to be overconfident. Review of Financial Studies 14, 1 27. Grinblatt, M., Keloharju, M., 2000. The investment behavior and performance of various types: A study of Finlands unique data set. Journal of Financial Economics 55, 43 67. Grinblatt, M., Titman, S., 1993. Performance measurement without benchmarks: An examination of mutual fund returns. Journal of Business 66, 47 68. Grinblatt, M., Titman, S., Wermers, R., 1995. Momentum investment strategies, portfolio performance, and herding: A study of mutual fund behavior. American Economic Review 85, 1088 1105. Hamao, Y., Mei, J., 2001. Living with the enemy: An analysis of foreign investment in the Japanese equity market. Journal of International Money and Finance 20, 715 735. Hirshleifer, D., Subrahmanyam, A., Titman, S., 1994. Security analysis and trading patterns when some investors receive information before others. Journal of Finance 49, 1665 1698. Hoshi, T., Kahyap, A., Scharfstein, D., 1991. Corporate structure, liquidity, and investment: Evidence from Japanese industrial groups. Quarterly Journal of Economics 106, 33 60. Jackson, A., 2002. The aggregate behaviour of individual investors. London Business School working paper, April. Kang, J., Stulz, R., 1997. Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan. Journal of Financial Economics 46, 3 28. Karolyi, A., 2002. Did the Asian financial crisis scare foreign investors out of Japan? Pacific Basin Finance Journal 10 (4), 411 442. Kim, K., Nofsinger, J., 2002a. The behavior and performance of individual investors in Japan. Washington State University working paper, March. Kim, K., Nofsinger, J., 2002b. Institutional herding, business groups, and economic regimes: Evidence from Japan. Washington State University working paper, March.

22

A. Kamesaka et al. / Pacific-Basin Finance Journal 11 (2003) 122

Murase, A. (Kamesaka, A.), 2001. Stock investment performance of main investor groups in Japanese market. Review of Monetary and Financial Studies (in Japanese, March). Nofsinger, J., Sias, R., 1999. Herding and feedback trading by institutional and individual investors. Journal of Finance 54, 2263 2295. Odean, T., 1998. Are investors reluctant to realize their losses? Journal of Finance 53, 1775 1798. Odean, T., 1999. Do investors trade too much? American Economic Review 89, 1279 1298. Prowse, S., 1990. Institutional investment patterns and corporate financial behavior in the United States and Japan. Journal of Financial Economics 27, 43 66. Prowse, S., 1992. The structure of corporate ownership in Japan. Journal of Finance 47, 1121 1140. Wermers, R., 1999. Mutual fund herding and the impact on stock prices. Journal of Finance 54, 581 622.