SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References

This study examines for evidence of peer effects in the trading decisions of individual investors from Mainland China, a country whose cultural and social structures are vastly different from those of Western countries. Cultural differences, as widely documented, play a significant role in social interactions and word-of-mouth behavior. In contrast to US studies, we find robust evidence that the trading decisions of Chinese investors are influenced, via word of mouth, by those of their peers who maintain brokerage accounts at the same branch, but not by those whose accounts are maintained at another branch located in the same city.

“People who interact with each other regularly tend to think and behave similarly.”

Robert Shiller (1995)

Peer effects have long been recognized in economic, psychological, and sociological literature as important determinants of individual outcomes. Theory suggests that in many economic and social settings, individuals can be influenced by the decisions of others, through “word of mouth” (Banerjee, 1992; Bikhchandani, Hirshleifer, and Welch, 1992; Ellison and Fudenberg, 1993, 1995; Cao and Hirshleifer, 2002; Banerjee and Fudenberg, 2004). Many financial economists believe that peer influences are important when explaining financial phenomena, such as banking panics and stock market crashes.1 Recent empirical evidence has demonstrated that peer interactions affect investor trading decisions, as well (Hong, Kubik, and Stein, 2005; Ivkovic and Weisbenner, 2006). In the psychological literature, peer effects are a crucial environmental factor affecting developmental consequences, personality, and behavioral problems (Valliant, 1995; Curran, Stice, and Chassin, 1997; Dishion, McCord, and Poulin, 1999). Sociological literature addressing neighborhood effects indicates that neighborhood peer influences may be driven by evolving social norms and social stigma (Jencks and Mayer, 1990; Glaeser, Sacerdote, and Scheinkman, 1996; Katz, Kling, and Liebman, 2001). All these studies on peer effects and social interactions are primarily based on developed Western countries, especially the United States; hence, their evidence does not necessarily extend to other countries, particularly Asia, where cultural and social structures are vastly different from those of Westerners. Cultural differences, as widely documented, play a significant role in social interactions and word-of-mouth behavior (Brown and Reingen, 1987; O’Keefe and O’Keefe, 1997; Money, Gilly, and Graham, 1998).

In this study, we search for evidence of peer effects in the trading decisions of investors from an emerging Asian market, Mainland China. This particular market offers a couple of appealing features for our study. First, China has a unique investment environment that allows us to examine several implications of social learning theories that cannot be directly addressed using US brokerage accounts (data widely employed in existing studies). In Mainland China, individual investors are only allowed to open one account at any brokerage branch, and all trades must be placed through this specific branch. Groups of individual investors are typically in the same trading room when they place their stock orders.2 Their close physical proximity naturally helps promote constant face-to-face interaction and word-of-mouth exchange of information. This unique setting offers a natural experiment on the impact of peer effects in financial markets.

Second, the Chinese have different cultural and social structures than those of Westerners. Lam and Lin (2003) study the effect of word of mouth in consumer purchases among Chinese and White Caucasians. They find that the Chinese group engaged in more word-of-mouth behavior than did their White Caucasian counterpart. They argue that this behavioral difference is associated with the Chinese cultural focus on Guanxi (meaning social relation in Chinese). Thus, examining the influence of peer relations in the Chinese market allows us to compare our results with those based on Western countries, such as the United States. Of particular interest to this study is that Hong et al. (2005) and Ivkovic and Weisbenner (2006) find evidence of peer effects in trading and investment decisions in US investors who live in the same city or the same community. Any evidence from our study would enable us to determine the extent to which social and cultural differences affect peer effects in investor trading behavior.

Our study employs Shanghai Stock Exchange (SHSE) data that contain daily detailed records of trades executed by individual brokerage accounts and their locations across Mainland China for the period April 2001-April 2002.3 Results indicate strong evidence of word-of-mouth effects in the trading decisions of these individual investors. These effects are more apparent in their purchase than sale of locally headquartered stocks. On average, a 1% increase in the likelihood of buying a particular stock by investors at a branch would induce an economically 3.5% difference in the likelihood of buying the stock by an investor from the same branch and by an investor from a different branch. Similarly, a 1% increase in the tendency to sell a stock by investors at a branch, on average, would lead to about 1.6% difference in the likelihood of selling the same stock by an investor from the branch and by an investor from another branch. Additionally, the results indicate that the trading decisions of individual investors have little influence on the trading decisions of an individual investor located at a different branch even within the same city. These findings are robust to several alternative interpretations associated with stock visibility and familiarity, within-branch bias, location, local bias, firm size, and calendar effects.

Furthermore, the results also demonstrate the importance of geographic proximity in word-of-mouth communication in Mainland China. Word-of-mouth effects exhibit a strong influence on the trading decisions of individual investors of close physical proximity, but the effects disapate as the physical distance between investors increases. Specifically, peer effects are evident only in the trades of individual investors from the same branch but not in the trades of individuals from nearby branches within a five-mile radius. Therefore, our results suggest no evidence of city or community effects, as previously shown in US studies. The difference in our findings suggests that given the Chinese social and cultural behavior, close proximity between individual investors is essential to foster face-to-face communication and strong relational ties, both of which facilitate the exchange of information and ideas and the influence of one another's trading decisions.

Our evidence of word-of-mouth effects on the correlated trading by individual investors may have an important implication about the widely documented “local bias” (strong investor preference for nearby firms). Coval and Moskowitz (1999) indicate that US professional money managers exhibit a strong bias toward locally headquartered firms. Zhu (2002) and Ivkovic and Weisbenner (2005) find a similar pattern in individual investors. These studies argue that the local bias is possibly due to information asymmetries or familiarity biases. Our study further suggests that this local bias may also be driven, in part, by peer effects. Whether peer effects contribute significantly to the phenomenon of local bias is an interesting issue that warrants future research.

Our work is somewhat related to that of Feng and Seasholes (2004) in that both studies test for correlated trading of Chinese individual investors. However, our study differs from theirs in many significant ways. One, we examine for evidence of peer-induced correlated trading at the branch level, while they focus on location-induced correlated trading at the cross-branch level. Two, using a different methodology from ours, Feng and Seasholes (2004) document that public or market-wide information can explain about 32% of the correlated net trades between branches, thereby concluding that public information is a major determinant to correlated trading. In contrast, we find strong peer effects at the branch level while controlling for cross-branch effects. This finding suggests that word-of-mouth or private information is a source of within-branch correlated trading. Finally, Feng and Seasholes’ (2004) study employs brokerage accounts data from a single brokerage firm in Mainland China to examine correlated trading in Shenzhen Stock Exchange (SZSE) stocks. Our study, however, uses brokerage account data from various branches of different brokerage firms across Mainland China. A detailed discussion of the data differences is presented in subsequent sections of the paper.

This paper is organized as follows. The next section reviews related literature and develops the hypotheses that we test in this study. Section II describes the data. Section III contains the baseline results, while Section IV performs various robustness tests. The final section concludes and discusses some implications of our findings.

I. Related Literature and Hypotheses Development

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References

Peer or social interactions provide opportunities for information exchanges via word of mouth, in which individuals use information about the experiences of others to guide their own decisions. Existing studies have argued that word of mouth is more evident in a high-context culture such as China than in her Western counterparts. This is, in part, due to the Chinese cultural focus on guanxi, which refers to the existence of direct particularistic ties between an individual and others. Guanxi plays a crucial role in Chinese societies and is based on factors that promote shared social experience between and among individuals; information exchange is one of the key variables that foster guanxi (Chiao, 1982; Osland, 1989; Leung, Wong, and Tam, 1995; Farth et al., 1998). At the low end of the spectrum (e.g., the United States), however, communication depends less on long-standing personal relationships or other contextual factors. China's culture also differs from that of Western countries regarding the dimension of individualism/collectivism. In Western culture, people are viewed as self-contained, autonomous individuals, whereas in highly collectivist cultures (e.g., China), people are primarily identified as group members (Morris and Peng, 1994). Such diverse studies in different fields seem to suggest that countries with different cultures may have varying impacts on peer effects.

To the best of our knowledge, there are two US studies that are closely related to our current work. Hong et al. (2005) present the first attempt to study the diffusion of word-of-mouth information among US mutual fund managers located in the same city, using the quarterly holdings of these managers from 1991 to 1996. Their study assumes that fund managers who are located in the same city have many opportunities to meet in formal or informal gatherings, such as local investor conferences, to exchange ideas via word of mouth. They find evidence that word-of-mouth effects are more prevalent in managers located in the same city than by the decisions of those located in other areas.

Ivkovic and Weisbenner (2006) study word-of-mouth diffusion effects among US individual investors by using information on common stocks that 35,673 households traded through a discount brokerage firm from 1991 to 1996. Their study also assumes that individuals who live in the same city have more opportunities to communicate via word of mouth. They provide evidence of information diffusion effects in the quarterly investment choices of individuals whose decisions are related to those made by their neighbors located within 50 miles of them. These two studies infer the process of information exchange from aggregate data using individuals’ varying levels of sociability with friends and neighbors as a proxy for information exchanged through social networks. Their results, however, cannot rule out alternative interpretations such as common personality traits and preferences and common reaction to public news or announcements, among others.

Motivated by the US evidence, we examine peer effects in Mainland China, a country that is culturally and socially different from the United States. For a high-context culture such as China, knowledge is situational and/or relational. Therefore, decisions and activities of individuals focus around personal face-to-face communication and strong social ties. Hence, we propose two testable hypotheses. First, we predict that word-of-mouth effects in investor trading decisions would be mainly at the branch level as the close physical proximity of individual investors is important to facilitate interpersonal relationships and collectivism. That is, the trading decision of an individual investor at a given branch is affected more by the trading decisions of other individuals from the same branch than by those of individuals from different branches located in the same city. Second, we test whether greater proximity encourages greater flow of word-of- mouth information. Greater proximity between individual investors increases the intensity of social interactions and thus is more likely to influence each other's decision making.4 The existing social interaction literature provides no consensus on whether the effect of peer interaction is a local or a community effect, nor do the recent studies that examine the influence of word-of- mouth effects in the financial decisions of individuals. Our analysis contributes to this growing literature by presenting the first attempt to test the importance of geographic proximity in word-of-mouth communication.

Our tests of word-of-mouth effects also differ from those of existing studies in that we differentiate word-of-mouth effects in the buying and selling decisions of individual investors. Typically, individual investors have different motivations for buying and selling a stock. Investor buys are generally motivated by positive information and/or the investor's optimistic expectations of the future performance of the stock. On the other hand, investor sales could reflect negative information that causes investors to sell a stock and/or various motivations, such as liquidity needs, portfolio diversification, and the investor's pessimistic expectations of the stock's future performance. Furthermore, in Mainland China, short sales are not allowed. A recent study by Bris, Goetzmann, and Zhu (2007) demonstrates that such restrictions have an asymmetric impact on the information diffusion of investor buys and investor sales. Particularly, they find that in markets with short sale constraints, stock prices tend to incorporate negative information at a slower rate. Thus, we should expect stronger word-of-mouth effects in investor buys than investor sales. In contrast, prior studies that employ either net buys/sales or total trades (buys and sells) might not be able to accurately capture word-of-mouth effects as such trading information might distort individual buy and sell effects.

II. Data Description and Sample Selection

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References

The primary database contains daily detailed records of A-Shares trades executed by individual brokerage accounts across Mainland China for the period April 2001-April 2002.5 The data are compiled by the SHSE for the purpose of an audit trail between the stock exchange and member brokerage firms, and are similar to the New York Stock Exchange's consolidated equity audit trail data.

A. Brokerage Accounts and Base Sample

As of December 2001, there are 255 SHSE members (112 are securities firms and 143 are nonsecurities firms) in Mainland China. These firms, in total, have about 2,700 branches nationwide.6 Our study exploits one paradigmatic feature of these brokerage offices, the layout of a typical brokerage office, that facilitates a lot of interaction and open conversations between individuals (see Feng and Seasholes, 2004, for the layout of a typical trading floor of a brokerage branch office). Typically, each office has a single, large digital display from which individual investors can watch the constant updates of stock prices, as well as the brokerage firm's releases of any public news and in-house research information. This simple setting of a branch inadvertently promotes social interaction between individual investors who frequent a branch office regularly and offers us an attractive opportunity to study the effects of social interaction via word of mouth in investor trading decisions.

Individuals are allowed to open only one brokerage account at any one branch in order to trade SHSE stocks, and the account is opened using their National Identity Card. However, investors are allowed to open more than one account at a different branch for trading SZSE stocks. All trades must be placed through computer terminals, telephone service, online trading, or cashier counters located in the particular branch at which the account is opened. (This policy does not apply to stocks traded on the SZSE, where investors can open more than one brokerage account.) It is important to stress that during the time period of our study, online trading was rare and costly. For example, online accounts comprised approximately 4.8% of the overall investing population in 2001 and about 7.4% in 2002, with online trading turnover of about 4.4% and 9% of the total market turnover, respectively (Chinese Securities Regulatory Commission). A survey conducted by the research department of SZSE (Chen, Li, and Du, 2002) shows that based on 2,587 survey responses, about 8% use online trading, 70% use cashier counters, and 19% use in-branch telephones or their own phone. Feng and Seasholes (2004) document that on average about 64% of their sample of SZSE trades are physically placed in a branch office. Therefore, a majority of Chinese individual investors typically place their trades in person at their brokerage branch. While our data offer no such information, we assume that about 60% of our sample individuals are in their brokerage offices when they place their trades. If the number is smaller than what we have assumed, then this should work against our finding of strong peer effects in investor trading.

We select a base sample of brokerage branches that are all located in Shanghai, the home of the SHSE. Out of 2,700 branches nationwide, 463 are in Shanghai. Therefore, we rank all Shanghai branches in our data set according to the number of individual brokerage accounts and the number of trades generated during the entire sample period. The reasons for this selection criterion are to ensure that 1) each branch has a large enough number of brokerage accounts to facilitate social interaction among individuals, and 2) investors who reside in the same city would receive the same local or national news. We exclude branches that are from the same brokerage firm as some brokerage firms may tend to attract a certain group of investors who behave more similarly than others. For example, it is possible that group of investors who belong to a certain social club and, hence, share similar preferences, open their brokerage accounts only with branches of a specific well-known brokerage firm in their neighborhood. To rule out such biases, we choose different brokerage branches, and as a result, our base sample consists of 30 branches with the most number of brokerage accounts and the largest number of trades transacted.

B. Trade Records of Individual Investors

We look at how brokerage accounts trade the 30 highest volume SHSE stocks measured in terms of total market value traded during the sample period and whose corporate headquarters are in Shanghai. Selecting stocks whose firms are headquartered in the same city helps differentiate word-of-mouth effects from local stock preference effects, thereby controlling for any investor biases toward holding stocks of nearby firms (Coval and Moskowitz, 1999; Ivkovic and Weisbenner, 2005). Our sample, therefore, consists of 498,982 trade records of 30 A-Shares executed by 121,082 individual brokerage accounts from 30 different branches located in Shanghai. Each trade record contains in detail all key elements of a stock transaction, including the stock code, the number of shares purchased or sold, the execution price and date, and an account identifier.

Table I reports descriptive statistics associated with each of the 30 different brokerage branches located in Shanghai and for the selected 30 SHSE stocks that investors at these branches trade. For convenience, we label the branches Branch 1-Branch 30. The number of brokerage accounts varies from 993 (Branch 3) to 7,416 (Branch 12). The average number of accounts is about 4,036. During the entire sample period, these accounts generate almost half a million trades of the 30 stocks. The total numbers of buys and sells are 273,625 and 231,572, respectively. The sum of the buys and sells is greater than the total number of trades. This is because an individual may buy and sell the same stock on the same day. The two executions will be aggregated into one trade record in our data. The total market value is RMB 3.97 billion for buys and is RMB 3.84 billion for sells.7 Across the 30 branches, the average purchase size range is RMB 5,276-RMB 26,982, whereas the average sale size is RMB 4,940-RMB 30,236.

Table I.  Descriptive Statistics of Brokerage Accounts by Branch For each branch office, the table presents the number of brokerage accounts, the total number of trades, the number of buy trades, the number of sell trades, the total buy value in RMB millions, the total sell value in RMB millions, and their respective average values in RMB in the last two columns. All these statistics are associated with the base sample of 30 stocks that are headquartered in Shanghai. The sample period is from April 2001 to April 2002.
BranchNo. of AccountsNo of TradesNo. of BuysNo. of SellsTotal BuysTotal SellsAverage BuyAverage Sell
Branch 14,10516,2668,9367,48081.1476.656,8306,718
Branch 22,69611,4656.2875,289112.73111.2213,04413,466
Branch 39932,8331,5421,30953.0944.9926,86623,968
Branch 43,61615,2088,3607,006140.91138.9812,67913,188
Branch 55,38731,0551,679714,761120.81111.085,2764,940
Branch 63,57314,0177,4056,860148.41142.9115,72115,774
Branch 76,57537,64619,91218,415443.42454.5616,91317,760
Branch 86,72327,69015,18512,818142.61134.487,0756,895
Branch 94,65119,52510,8948,853146.28128.6910,1109,533
Branch 101,7737,4463,9983,53361.0359.1611,21311,086
Branch 115,09919,37210,7458,811150.25168.6410,47812,485
Branch 127,41627,52215,31712,534163.60150.738,0777,786
Branch 133,37813,4027,3516,228101.16118.8510,57312,992
Branch 143,94419,23910,3959,097150.68143.6210,96310,950
Branch 155,29420,43711,2859,377115.11104.417,6677,374
Branch 167,01824,22113,43810,948177.89163.789,9319,742
Branch 171,4715,2002,8692,398100.45104.0426,98230,236
Branch 182,25210,0925,5264,764125.72118.2817,29717,155
Branch 195,26219,43510,9878,664127.63115.148,7658,396
Branch 203,33314,2667,9356,580113.18104.2310,85210,722
Branch 212,63512,2146,5995,878135.34131.0815,49615,614
Branch 226,18021,84212,2149,769118.64105.627,3286,899
Branch 232,0368,2474,6393,68679.3991.4813,15216,260
Branch 242,99912,9217,0475,99290.2283.319,5679,272
Branch 252,89112,0036,7365,521129.29120.9414,44414,226
Branch 263,69813,7537,4406,51991.9491.2510,04010,575
Branch 273,33316,4688,7437,978103.46107.668,8749,382
Branch 283,20210,5785,8004,858138.61130.9318,54418,317
Branch 297,18025,97514,38811770161.90143.498,4148,096
Branch 302,3698,6444,8553,876140.04139.6422,05423,146
Total12,108249,898227,362523,15723,9653,840  

At this juncture, it is necessary to draw some comparisons between our sample and that of Feng and Seasholes (2004). The latter have accounts-level data of a single brokerage firm and focus only on individual investors’ trades of SZSE stocks executed through seven branches of this firm. In contrast, our use of the SHSE audit trail data enables us to select different branches of different brokerage firms with the largest number of brokerage accounts and trade records. For example, the average number of accounts across our sample of 30 branches is 4,036 as compared with 1,139 in Feng and Seasholes's (2004) sample (see their Table I on page 2124). However, the advantage of their data over ours is that they are able to identify the means in which the trades are placed (e.g., phone, terminals at a branch, cashier windows, or computer links).

In comparison with China's nationwide averages, our sample of 30 branches has an average of 9,821 brokerage accounts that trade in all SHSE stocks, with average monthly trading value of RMB 168.2 million. However, the country's average number of accounts maintained at a branch is 14,841, with average monthly trading value of RMB 119.3 million (SHSE Fact Book, 2001). The difference in the number of brokerage accounts is partly due to the fact that the SHSE audit trail data only contain executed trades during our sample period. The SHSE Fact Book, on the other hand, reports the total number of brokerage accounts opened during 2001 and includes those brokerage accounts that were opened but inactive. Anecdotal evidence has suggested that about 10 million of these accounts were opened simply for the purchase of initial public offerings and were generally inactive. The large number of inactive accounts contributes to a lower nationwide average trading value per brokerage branch, as compared with our sample of active accounts in each branch. Nevertheless, while our sample is larger than that employed by Feng and Seasholes (2004), our results still have to be interpreted within the context of the sample we use.

III. Word-of-Mouth Effects in Investor Trading Decisions

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References

A. Baseline Specifications

We introduce two simple specifications to test for word-of-mouth effects in investor trading decisions. While our reduced-form models are similar to that of Hong et al. (2005), we separate the word-of-mouth effects on investor buying and selling decisions. Our base specification for investor buys is given by

  • image(1)

and for investor sells is given by

  • image(2)

where inline image is the ratio of the value of stock j purchased (sold) by investor i at a given branch b to the total purchase (sale) value of the SHSE stocks, or the buy (sell) ratio in stock j at time t; inline image is the equal-weighted average of buy (sell) ratios in stock j for all investors, excluding investor i, whose trades are placed through branch b at time t; inline image is the equal-weighted average of buy (sell) ratios in stock j for all investors in brokerage branch b at time t; B(k = b) is an indicator that takes the value one if branch k and branch b are the same and zero otherwise; and B(k ≠ b) takes the value one if branch k and branch b are not equal and zero otherwise. Given that the interpretations of Specifications (1) and (2) are similar other than the former examines investor buys effects while the latter examines investor sells, our subsequent discussion shall focus only on Specification (1).

Specification (2) states that an individual's decision to buy stock j is potentially influenced by the decisions of their peers in the same branch as well as by the decisions of their peers in other branches.8 For each of the 30 branches, we estimate inline image and inline image parameters that capture, respectively, the “own branch” and “other branch” effects. For instance, inline image measures how the buying decisions of individual investors from Branch 1 affect the buying decisions of individual investor i from Branch 1, and inline image measures how the buying decisions of individual investors from the other 29 branches affect the buying decisions of investor i from Branch 1. Our first hypothesis on word-of-mouth information transmission implies that, for any branch b, the own branch effect should always be larger than the other branch effect (i.e., inline image). The implication would be consistent with our prediction that the buying (selling) decision of an individual is influenced more by the buying (selling) decisions of other individuals from the same branch than those of other individuals from other branches. If correlated trading by individual investors from a branch is driven by their common interpretation of some public information that the national or city media release to the public, then the own branch coefficient should be insignificantly different from the other branch coefficient (i.e., inline image). For example, if publicly known information is transmitted to the city where all of our sample branches are located, we assume that investors in any of the branches would be informed of, or would have received such information.

The difference between the own branch and other branch coefficients determines the relative importance of the own branch and other branch word-of-mouth effects in investor trading behavior. To draw inferences on the extent of their relative importance (e.g., word-of-mouth effects) in the trading decisions of individual investors, we provide two test statistics: one based on an equal-weighted average of the differences between own branch and other branch coefficients, and another based on an accounts-weighted average. The latter weighting scheme is primarily motivated by existing evidence that individuals are more likely to participate in the stock market if there is a greater participation rate among individuals in the local community (Hong, Kubik, and Stein, 2004; Brown et al., 2005).

In our particular setting, a brokerage branch office is the geographic location that facilitates our test of word-of-mouth effects in investor trading decisions. We select a monthly time interval to allow for the speed of word-of-mouth information to be exchanged among individuals at the same brokerage branch. Prior studies use varying time intervals from weekly (Feng and Seasholes, 2004), to quarterly (Hong et al., 2005; Ivkovic and Weisbenner, 2006), to yearly information (Brown et al., 2005).

B. Baseline Results

Table II contains regression estimates of Specifications (1) and (2), with t-statistics in parentheses computed using Newey and West (1987) heteroskedasticity- and autocorrelation-consistent standard errors. It also presents the equal- and accounts-weighted average differences between own and other branch coefficients.

Table II.  Own Branch versus Other Branch Word-of-Mouth Effects in Buying and Selling Decisions of Individual Investors The table reports results of own branch and other branch effects as determinants of investor trading decisions for both “Investor Buys” and “Investor Sells.” These effects are given by baseline specifications: 1) investor buys and 2) investor sells, as defined in the text. Their associated t-statistics in parentheses are computed using Newey-West (1987) heteroskedasticity- and autocorrelation-consistent standard errors. The two rows above the number of observations indicate the equal-weighted (EW) and accounts-weighted (AW) average differences between own branch coefficients and other branch coefficients across the 30 branches.
 Investor BuysInvestor Sells
Own Branchest-Stat.Other Branchest-Stat.Own Branchest-Stat.Other Branchest-Stat.
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Branch 18.617***(16.2)0.773**(2.22)6.973***(14.5)2.837***(8.63)
Branch 25.109***(8.33)−3.033***(−7.03)2.395***(4.39)−1.582***(−3.60)
Branch 35.975***(12.8)−0.603**(−2.09)1.984***(6.70)−0.966***(−4.61)
Branch 47.219***(11.7)3.511***(8.12)3.842***(8.14)1.342***(3.54)
Branch 54.559***(5.65)−1.507***(−6.87)4.155***(4.44)2.042***(8.13)
Branch 68.281***(14.6)−0.085(−0.25)5.544***(12.4)1.621***(4.95)
Branch 70.028(0.06)1.763***(5.82)−2.244***(−4.91)1.832***(6.53)
Branch 83.126***(5.70)−0.789***(−5.51)1.830***(3.39)0.322**(2.05)
Branch 98.140***(15.2)2.547***(9.07)2.591***(5.01)−0.276(−0.85)
Branch 106.757***(13.7)1.001***(4.11)1.415***(4.04)−1.274***(−5.45)
Branch 115.239***(12.0)1.009***(5.95)2.042***(12.0)0.419***(6.47)
Branch 120.499(0.87)1.482***(6.74)−0.307(−0.91)0.298(1.59)
Branch 136.629***(14.1)1.722***(8.36)4.574***(13.6)1.899***(10.2)
Branch 142.638***(5.90)−0.176(−0.93)4.511***(10.3)2.883***(17.3)
Branch 154.036***(6.39)−0.616*(−1.84)2.455***(7.13)−0.667***(−3.07)
Branch 162.296***(4.50)−0.574***(−4.13)2.348***(5.13)0.309*(1.87)
Branch 17−0.895(−1.12)−4.071***(−12.7)0.310(0.39)1.810***(5.36)
Branch 181.442***(2.58)−0.008(−0.05)3.601***(7.81)0.996***(8.10)
Branch 191.211**(2.17)0.106(0.54)0.315(0.82)−0.208(−1.07)
Branch 201.635***(2.70)−1.262***(−3.74)0.248(0.47)−0.050(−0.15)
Branch 210.735***(7.91)−0.129**(−2.45)0.341***(3.16)−0.579***(−11.9)
Branch 220.459(1.35)−0.119(−1.27)−0.513*(−1.88)−0.333***(−3.84)
Branch 239.145***(7.28)−0.014(−0.07)3.119*(1.82)0.016(0.07)
Branch 24−0.218(−0.40)−2.361***(−11.6)−0.619(−1.22)−1.787***(−6.86)
Branch 253.523***(3.54)−0.516***(−2.97)0.294(0.30)−0.877***(−6.63)
Branch 261.161***(4.07)−0.053(−0.51)2.185***(8.97)0.904***(7.82)
Branch 271.197**(1.97)−0.999***(−8.58)−1.470**(−2.15)−2.290***(−11.8)
Branch 282.636***(4.30)0.620***(2.75)4.470***(9.03)2.098***(7.61)
Branch 290.731*(1.93)−0.451***(−3.61)−0.754**(−2.16)−1.570***(−10.5)
Branch 300.628 *(1.85)−0.346***(−4.67)0.265(0.92)−0.636 ***(−7.64)
EW Diff.3.524***(6.96)  1.579***(5.02)  
AW Diff.3.875***(4.76)  1.706***(3.44)  
No. of obs.22,721   19,358   

The table reveals strong evidence of word-of-mouth effects in the trading decisions of individual investors who place their trades through the same branch. For investor buys, the equal- and accounts-weighted average differences between own and other branch coefficients are 3.52 (t-statistic = 7.0) and 3.88 (t-statistic = 4.8), respectively. To determine the economic magnitude of word-of-mouth effects, we provide an example using equal-weighted differential results: given a 1% increase in the tendency to buy a particular stock by investors at a branch, on average, the difference in the tendency to buy the same stock by an investor from the same branch and by another from a different branch is 3.5%. For individual branches, 28 of the 30 own branch coefficients are positive, and 26 of them are larger than their other branch counterparts. Out of the 28 positive coefficients, 23 are statistically significant at the 5% level and 2 at the 10% level. The two negative own branch coefficients are statistically insignificant at conventional levels. These findings strongly suggest that an individual's buying decision is more strongly influenced by peers from the same branch than by those from other branches.

We find weaker word-of-mouth effects at the cross-branch level. As Table II indicates, 20 of the 30 other branch coefficients are negative, and more than half are statistically significant at the 5% level. On the other hand, almost all of the remaining 10 positive other branch coefficients are statistically significant at conventional levels; the magnitude of these coefficients is generally smaller than their own branch counterparts. The weak evidence of cross branch peer effects suggests that constant face-to-face interactions seem necessary to facilitate word-of-mouth information transmission among individual investors. Alternatively, it also might suggest no community or city effects, an evidence that contradicts those previously reported using US data.

Furthermore, the findings also indicate weaker word-of-mouth effects in investor sells. The magnitude of the weighted average differentials is less than half that of their buys counterparts. The equal- and accounts-weighted average differentials for investor sells are 1.58 (t-statistic = 5.0) and 1.71 (t-statistic = 3.4), respectively. Across the branches, 24 of the 30 own branch coefficients for investor sells are positive, and 19 of these positive coefficients are statistically significant at conventional levels and larger than their other branch counterparts. The weaker word-of-mouth effects in the selling decisions of individual investors could reflect negative information causing investors to sell a stock and/or various other motivations, such as liquidity needs, portfolio diversification, and/or investors’ pessimistic expectations of the stock's future performance.

It is evident that informal information transmission, via word of mouth, induces correlated trading by investors who have brokerage accounts at the same branch. More importantly, our finding of a stronger within-branch correlation than cross-branch correlation suggests that the decision of an investor is more likely to be affected by the decisions of others from the same branch than by the decisions of others from different branches. However, one might argue that the stronger own branch effect relative to the other branch effect possibly subsumes the significant cross-branch correlation of individual trades shown in Feng and Seasholes (2004). Using the principal components analysis, the two authors find that public information accounts for about 32% of the correlated trading by investors located in the same city. The implication of their result is that the remaining 68% of the correlated trades must stem from nonpublic, possibly private or word-of-mouth, information. Using their methodology,9 we compute the average pairwise correlation coefficients to be 0.350 (z-statistic = 26.95) for investor buys and 0.516 (z-statistic = 43.47) for investor sells. Note that the z-statistics are calculated using the asymptotic approximation of standard errors of correlation coefficient described in Feng and Seasholes (2004). While our cross-branch correlation of investor trades of about 35%-52% is consistent with theirs, this evidence further substantiates our within-branch effect shown in Table II. Overall, it implies that the within-branch word-of-mouth effects cannot simply be driven by common local or national market shocks.

C. Does Physical Proximity Matter?

We hypothesize that physical proximity matters for the working of word of mouth in Chinese societies. We design the test of this hypothesis as follows. We first define word-of-mouth effects at the branch level as local peer effects and at the cross-branch level as community peer effects. We then determine the downtown area of Shanghai city by using the Shanghai City Council as the center. We categorize 15 of the 30 sample branches located within a five-kilometer radius from the Shanghai City Council as “close” branches and the remaining 15 branches outside the five-kilometer radius and scattered across the city as “far” branches.

We modify our baseline Specification (1) as follows:

  • image(3)

where Dt is a branch dummy that takes the value one if k ∈{close} and zero otherwise. In contrast to Specification (1), Specification (3) has an additional term with a parameter inline image that measures the incremental influence of the buying decisions of individual investors from a close branch on the buying decision of individual investor i in Branch b. If community peer effects exist, then the buying decision of an individual in Branch b is influenced not only by the decisions of individuals at their own branch but also by the decisions of those whose brokerage offices are located close by. Then the coefficient on inline image should be positive. The same interpretation can be applied to investor selling activity. For brevity, we do not present the specification for investor selling; the specification is similar to Specification (3) other than replacing investor buys with investor sells. Given that the magnitude and the statistical significance of the own branch effect inline image remain fairly constant, Table III only summarizes estimates of inline image and inline image of close branches associated with investor buys and estimates of their counterparts associated with investor sells.

Table III.  “Community” versus “Local” Peer Effects in Trading Decisions of Individual Investors The table summarizes the test results of “close branch” (“local” effects) or “community” word-of-mouth effects in the trading decisions of individual investors. For this analysis, we group the 30 branches in Shanghai into “close” and “far” branches. The near branches consist of 15 brokerage offices that are located within a five-kilometer radius, encompassing the downtown area of Shanghai City with the Shanghai City Council as the center. The 15 far branches are located outside of the five-kilometer radius and generally scattered across the large city. To examine community peer effects, we run the following regression model: inline imagewhere Dt is a branch dummy that takes the value one if k ∈{near} and zero otherwise. The parameter inline image measures how the buying decisions of individual investors at nearby branches influence the buying decisions of individual investor i at branch b (i.e., the community effect) and inline image measures the other branch effect. The regression model for investor sells (not shown) is an analogy of Model (3), with buys replaced by sells.
 Investor BuysInvestor Sells
inline imaget-Statinline imaget-Statinline imaget-Statinline imaget-Stat
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Branch 10.772**(2.22)0.131(0.16)2.843***(8.65)0.500(0.71)
Branch 5−1.505***(−6.87)−2.192(−1.48)2.046***(8.15)0.243(0.13)
Branch 71.757***(5.79)−1.405**(−1.96)1.829***(6.52)−1.167*(−1.81)
Branch 8−0.786***(−5.50)−1.068(−0.97)0.327**(2.08)0.156(0.16)
Branch 121.488***(6.77)0.683(0.69)0.303(1.61)0.153(0.30)
Branch 131.720***(8.35)0.126(0.15)1.901***(10.2)−1.250**(−2.35)
Branch 14−0.169(−0.90)0.667(0.94)2.885***(17.3)1.068(1.41)
Branch 17−4.077***(−12.7)−0.268(−0.18)1.818***(5.39)−2.210(−1.54)
Branch 18−0.012(−0.08)−2.754***(−2.72)0.996***(8.08)−0.474(−0.53)
Branch 21−0.127**(−2.43)0.047(0.27)−0.580***(−11.9)0.090(0.44)
Branch 22−0.120(−1.27)0.495(0.80)−0.332***(−3.82)0.329(0.62)
Branch 23−0.015(−0.07)5.269**(2.05)0.013(0.06)−0.322(−0.09)
Branch 26−0.056(−0.54)0.177(0.37)0.906***(7.84)−0.056(−0.13)
Branch 27−1.000***(−8.60)−1.106(−0.92)−2.282***(−11.7)0.855(0.69)
Branch 29−0.449***(−3.59)0.264(0.35)−1.570***(−10.5)−0.087(−0.14)
EW diff.−0.172(−0.46)−0.062(−0.13)0.740(1.87)−0.145(−0.66)
AW diff.0.133(0.29)−0.215(−0.74)1.149**(2.25)−0.201(−0.75)

The table reveals one distinct result. The trading decisions of individuals from near branches have virtually no incremental influence on the trading decision of an individual investor from a close-by branch. In other words, individual investors from other branches, whether near or far, exhibit no differential peer effects in the trading decisions of an individual investor from another branch. While nine estimates of the inline image coefficient associated with investor buys are positive, only one is statistically significant at the 5% level. Similarly, while eight of the inline image coefficients associated with investor sells are positive, none of them is statistically significant. These findings provide compelling evidence of no community peer effects; diffusion of word-of- mouth information is only effective at the branch level.

Thus, closer geographic proximity matters for word-of-mouth to work in Chinese societies. Increased face-to-face interaction helps foster word-of-mouth communication between individual investors of close physical proximity as it is within a brokerage office, and such personal interactions become less frequent when distance between individual investors increases. In contrast, physical proximity does not appear to matter in a low-context culture, such as the United States (Hong et al., 2005; Ivkovic and Weisbenner, 2006), where the environment contains much of the information individuals need to participate and the structure of interpersonal relationships are generally loose with wide networks.10

D. Nonlocal Stocks

Thus far, we focus primarily on the base sample consisting of only SHSE stocks. As discussed earlier, our purpose for selecting these stocks is to distinguish word-of-mouth effects from local stock preference effects (Coval and Moskowitz, 1999; Huberman, 2001). However, if Chinese societies work on guanxi, their word-of-mouth influences through such social relationships should have no bearing on whether the stocks are local or nonlocal. Therefore, we test the hypothesis that there should be no differential word-of-mouth effects in local and nonlocal stock trades of individual investors. To conduct our test, we select 30 different SHSE stocks headquartered in Beijing that have the largest market turnover relative to their counterparts headquartered in the same city.

We estimate the following specification for investor buys:

  • image(4)

where Dλ is a stock dummy that takes the value one if the corporate headquarters is located in Beijing and zero otherwise. We also estimate a similar specification for investor sells. If individuals are influenced by the decisions of their peers through word of mouth, they should be indifferent in their trading of local (Shanghai) versus nonlocal (Beijing) stocks. That is, the parameter inline image in investor buys (or inline image in investor sells) should be insignificantly different from zero. Estimates of Specification (4) for investor buys and sells are presented in Table IV .

Table IV.  Word-of-Mouth Effects in Local versus Nonlocal Stock Trades The table summarizes the estimates of inline image and inline image where Y denotes either investor buying or selling, and the Buy specification is as follows: inline imagewhere Dλ is a stock dummy that takes the value of one if the corporate headquarters is located in Beijing and zero otherwise. We estimate a similar specification for investor selling. inline image and inline image capture own branch effects associated with trades of Shanghai- (local) and Beijing-headquartered (nonlocal) stocks and of only Beijing-headquartered stocks, respectively. All t-statistics in parentheses are computed using Newey-West (1987) heteroskedasticity- and autocorrelation-consistent standard errors.
 Investor BuysInvestor Sells
inline imaget-Statinline imaget-Statinline imaget-Statinline imaget-Stat
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Branch 17.492***(15.8)−0.234(−0.29)7.827***(18.3)2.837***(4.37)
Branch 25.790***(10.9)1.305*(1.73)4.876***(11.4)1.364**(1.99)
Branch 35.157***(12.3)−1.725***(−2.68)2.280***(8.28)1.270**(2.27)
Branch 47.284***(12.9)0.783(1.12)4.712***(11.5)2.690***(4.09)
Branch 53.733***(6.29)−1.895***(−3.07)2.738***(3.42)−1.458*(−1.78)
Branch 68.380***(15.9)−2.682***(−2.98)4.607***(11.9)−0.439(−0.64)
Branch 7−1.749***(−4.41)2.001***(3.44)−3.454***(−9.29)−0.364(−0.64)
Branch 82.893***(5.05)5.480***(3.95)1.890***(3.76)−0.316(−0.56)
Branch 96.398***(13.0)0.340(0.44)4.083***(8.50)1.572*(1.69)
Branch 104.654***(9.81)−0.444(−0.50)1.942***(5.93)0.710(1.05)
Branch 114.569***(11.9)−1.130(−1.31)2.169***(13.6)3.298***(4.57)
Branch 12−0.497(−1.15)2.914***(5.62)−1.506***(−5.06)0.782*(1.92)
Branch 136.247***(14.5)−1.112*(−1.68)3.906***(12.1)−0.672(−1.00)
Branch 142.610***(6.28)−1.528***(−2.92)2.986***(7.00)−0.746(−1.53)
Branch 154.494***(7.48)1.668(1.36)2.698***(8.05)3.227***(2.92)
Branch 161.964***(4.16)−1.501**(−2.41)2.066***(4.64)−1.099*(−1.81)
Branch 170.356(0.50)1.842**(2.54)0.646(0.86)1.686**(2.00)
Branch 181.575***(2.96)1.349(1.21)3.259***(6.83)−2.168***(−4.51)
Branch 190.806(1.60)2.590*(1.87)0.068(0.18)1.611***(3.07)
Branch 201.975***(3.70)−0.807(−0.94)0.498(1.06)1.754***(2.88)
Branch 210.750***(7.37)4.806***(6.44)0.318***(2.90)0.997**(2.00)
Branch 220.273(0.82)−1.929(−1.56)−0.503*(−1.69)−1.144(−1.01)
Branch 238.128***(6.55)−3.352***(−2.61)3.400**(2.02)−1.017(−0.59)
Branch 24−0.006(−0.01)1.393(1.44)0.125(0.26)2.596***(4.34)
Branch 252.981***(2.96)2.620**(2.34)0.822(0.91)2.981***(3.09)
Branch 260.900***(3.74)3.121***(2.89)2.651***(10.4)2.113***(3.54)
Branch 270.851(1.52)−1.128*(−1.77)−0.325(−0.52)1.896***(2.71)
Branch 281.755***(2.73)1.419(1.27)3.818***(7.70)−0.735(−0.72)
Branch 290.709*(1.89)4.189***(4.03)0.464(1.33)0.663(1.03)
Branch 300.298(1.02)1.367(1.63)0.429(1.37)0.546(0.84)
EW Ave.3.049***(6.29)0.657(1.60)1.808***(5.73)0.815***(2.87)
AW Ave.3.419***(4.54)0.412(1.17)1.919***(3.98)0.965***(3.07)
No. of obs.31,5582   28,4559   

A number of notable observations emerge from the table. Consistent with the results in Table II, the word-of-mouth effect is strong and significant in both investor buys and sells. It is important to emphasize that these results are not driven by the trades in Shanghai-headquartered stocks. The reason is that we also replicated the baseline specifications using trades of 30 Beijing-headquartered stocks by Shanghai investors, and we found the nonreported results to be robust. The word-of-mouth effects in investor buying of Beijing-headquartered stocks, while positive, are not significantly different from those of Shanghai-headquartered stocks. Both the equal and weighted averages of own branch coefficients associated with investor buys are statistically insignificant at conventional levels. Word-of-mouth communication seems to have similar influences on the buying decisions of an individual investor from the same branch, regardless of whether it is a local or nonlocal stock. This finding supports our hypothesis that word-of-mouth effects in investor trading decisions are independent of the type of stock.

We find statistically significant word-of-mouth effects in investor sells of Beijing-headquartered stocks by individual investors who reside in Shanghai. The equal- and accounts-weighted averages of own branch coefficients associated with only Beijing-headquartered stocks are 0.82 (t-statistic = 2.9) and 0.97 (t-statistic = 3.1) for investor sells, as opposed to 0.66 (t-statistic = 1.6) and 0.41 (t-statistic = 1.2) for investor buys. The results might imply that negative word-of-mouth information tends to have a greater influence on individual investors selling nonlocal stocks or that investors are less willing to hold on to nonlocal stocks when they receive negative information.

IV. Word-of-Mouth and Alternative Interpretations

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References

Thus far, we have established strong evidence that individual investors who place their trades through the same branch do influence each other's trades via word of mouth. However, our results can also be consistent with alternative interpretations such as within-branch-induced bias or investor preference for stocks with greater investor recognition and familiarity. In this section, we perform tests that help distinguish several alternative stories from word-of-mouth effects.

A. Across Cities

Does word-of-mouth communication only influence the trading decisions of investors who live in Shanghai, a large financial city? While we believe this is unlikely, we verify our results by replicating our above study using similar information on individual investors who live in a different city. We select Beijing, a smaller and a nonfinancial city that is about 1,093 kilometers from Shanghai. In contrast to Shanghai, there are fewer brokerage branch offices located in Beijing. Using the same criteria as we did in selecting branches in Shanghai, we select 20 Beijing brokerage branches from 20 different brokerage firms. As in the preceding section, we also select 30 different SHSE stocks that are all headquartered in Beijing and that have the largest market turnover relative to their counterparts headquartered in the same city. The unreported summary statistics indicate that the number of brokerage accounts at the 20 Beijing branches ranges from 1,419 to 7,067; their average number is 2,776 as compared to 4,036 for Shanghai. The total number of buy and sell trades executed by all individual investors at the 20 branches are 104,028 and 87,711, respectively. The total market value of buys is RMB 2.601 billion and of sells is RMB 2.827 billion. While the sample of different Beijing brokerage branches is smaller than our baseline Shanghai sample, the number of brokerage accounts in the former is large enough for our analysis. To conserve space, we do not report the detailed regression results, but instead we provide a brief discussion below.

In general, the results are broadly consistent with those of Table II. We find strong word-of-mouth effects that are also more pronounced in investor buying than investor selling. Almost all of the estimates of own branch buy coefficients are positive, and 12 of the 20 coefficients are statistically significant at the 5% level and 2 at the 10% level. The difference between the own and other branch coefficients remains mainly positive and statistically significant. The accounts- and equal-weighted averages of the differences between own and other branch coefficients for investor buys are both 1.25 with t-statistics of 4.7 and 6.4, respectively. In addition, we find that 12 of the own branch sell coefficients are positive, and 7 of them are statistically significant at conventional levels. The accounts- and equal-weighted average differentials for investor sells are 1.02 (t-statistic = 4.2) and 1.14 (t-statistic = 3.1). The results suggest that the trading decision of an individual from a Beijing branch is more significantly affected by the trading decision of other individuals from the same branch than by the trading decisions of individuals from other branches located in the same city. Overall, the evidence of word-of-mouth effects in investor trading decisions is independent of the city where the individual investors are located.

B. Within-Branch Bias

There are a couple of possible factors that might induce within branch bias and, hence, generate the trading patterns of individual investors that we have found in the preceding section. Even though they offer no stock recommendations, brokerage firms in Mainland China frequently release firm-specific or industry-wide information as well as their in-house research reports to their brokerage account holders. Individuals might react similarly to such information or reports and trade the same stock. This herd behavior should not be mistakenly interpreted as an outcome of word-of-mouth communication. Second, there are social factors (correlated social effects) that might contribute to the observed correlated decisions of individual investors at the same branch. For instance, in Mainland China, employees of a large-listed, particularly state-owned firm generally receive housing benefits from the company. The subsidized housing is typically located in a neighborhood near their workplace. Naturally, these employees would open their brokerage accounts at a brokerage branch in their neighborhood. When they exhibit preferences for stocks whose companies they work for, such noncommunicative behavioral patterns are also consistent with those of word-of-mouth effects.

To control for any within branch bias, we estimate Specifications (1) and (2) by stock.11 This approach permits us to examine how individuals trade a particular stock relative to their peers at the same branch and how they trade relative to their peers from the remaining 29 branches. If trading behavior is induced by social factors or brokerage-specific research information, then our earlier finding of correlated trading should only be evident in certain stocks. Table V summarizes the equal- and accounts-weighted averages of the differences between own and other branch coefficients by stock. The last row of the table demonstrates the aggregate weighted average differential.

Table V.  Word-of-Mouth Effects and within-Branch Bias The table reports own branch effects and other branch effects in investor trading decisions by stock. The investor buys and sells results are reported separately. We summarize the regression results by averaging the differences between own branch and other branch coefficients across 30 branches. EW Diff. shows the equal-weighted average differences between own and other branch coefficients across 30 branches. AW Diff. shows the accounts-weighted differentials, with the number of accounts in each branch as the weight. All t-statistics are reported in parentheses. The last row reports the aggregate weighted average differential.
 Investor BuysInvestor Sells
EW Difft-StatAW Diff.t-StatEW Difft-StatAW Difft-Stat
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Stock 15.648***(5.22)5.517***(4.76)2.781***(2.78)3.158***(3.22)
Stock 27.165***(6.54)6.319***(6.78)1.585**(1.99)1.805*(1.87)
Stock 34.832***(6.75)4.786***(5.73)1.770**(2.41)2.065***(2.64)
Stock 43.091***(3.03)3.109**(2.37)−0.716(−0.45)0.629(0.46)
Stock 55.853***(5.77)7.211***(4.87)2.162***(3.12)2.043***(2.93)
Stock 64.055***(5.58)4.634***(4.24)2.446***(3.34)2.681***(2.75)
Stock 73.112***(4.36)3.756***(3.75)5.592***(4.54)5.404***(4.31)
Stock 83.268***(3.35)3.263***(2.82)1.466**(2.52)1.847***(3.36)
Stock 91.670(1.59)2.580**(2.01)2.468***(2.84)3.026***(3.08)
Stock 105.466***(5.67)5.507***(5.00)3.175**(2.55)2.714***(3.07)
Stock 112.317**(2.12)2.651**(2.35)2.220***(2.69)2.331**(2.24)
Stock 123.641***(4.13)3.584***(3.78)0.915(0.68)2.248**(2.25)
Stock 133.312***(3.21)4.539***(3.80)3.277***(4.64)3.440***(4.18)
Stock 144.951***(5.02)5.076***(4.17)3.450***(3.52)3.733***(3.11)
Stock 154.519***(5.61)4.451***(4.99)4.312(1.56)2.488***(2.68)
Stock 165.187***(6.09)5.529***(5.33)2.608***(2.99)3.347***(3.77)
Stock 173.737***(3.15)4.225***(3.30)−0.204(−0.27)−0.301(−0.38)
Stock 183.165***(2.65)2.777**(2.53)1.698*(1.91)2.431**(2.20)
Stock 196.401***(6.60)6.672***(4.86)6.252***(5.12)6.428***(4.63)
Stock 206.114***(7.25)6.656***(6.05)0.612(0.67)0.275(0.33)
Stock 214.342***(4.50)4.695***(3.54)0.779(0.94)0.460(0.63)
Stock 222.320**(2.29)2.522*(1.76)2.168***(2.79)1.979**(2.36)
Stock 235.621***(4.91)6.324***(4.15)1.108(1.39)1.713**(2.10)
Stock 242.116***(3.62)2.684***(3.16)4.048***(9.38)3.896***(7.69)
Stock 253.835***(6.47)4.318***(4.67)2.838***(4.02)3.020***(3.29)
Stock 267.178***(6.42)7.188***(6.02)4.412***(5.91)4.590***(4.49)
Stock 274.912***(4.78)5.258***(3.95)4.804***(4.57)5.759***(4.25)
Stock 282.856***(2.77)3.294***(3.00)0.539(0.68)0.526(0.71)
Stock 293.633***(4.20)4.818***(3.91)1.840***(2.68)1.735**(2.06)
Stock 305.964***(4.67)5.575***(4.52)3.588***(3.93)4.253***(3.59)
Aggregate Average4.343***(15.8)4.797***(17.8)2.466***(8.24)2.657***(9.08)

Overall, the results of Table V are in line with the baseline regression results. Based on both equal- and account-weighting measures for investor buys, the own branch coefficient, on average, exceeds the other branch coefficient with aggregate weighted average differentials between 4.34 (t-statistic = 15.8) and 4.80 (t-statistic = 17.8). Similarly, we find that most of the weighted average differentials for investor sells are positive, and their aggregate weighted average differentials are 2.57 (t-statistic = 8.2) and 2.66 (t-statistic = 9.1), respectively. Therefore, these results indicate that our earlier evidence of word-of-mouth effects is not attributed to the within-branch bias induced by specific stock preferences of individual investors, or by their correlated trading due to branch-specific releases of information or research reports

C. Familiarity and Visibility

There are behavioral patterns of individual investors that can produce buying and selling patterns similar to those induced by word of mouth. It is possible that individuals from the same branch merely invest in the same stocks because those are the stocks that they are more familiar with, or that have greater visibility in the markets. For example, the Merton's (1987) investor-recognition hypothesis states that investors do not have equal information and they invest only in those stocks that they know about. Several international and US studies find strong support of the Merton hypothesis (Kang and Stulz, 1997; Amihud, Mendelson, and Uno, 1999; Bailey, Chung, and Kang, 1999; Foerster and Karolyi, 1999; Huberman, 2001).

To address the above issue, we test whether word-of-mouth effects are driven by familiarity or by stocks with greater investor recognition or visibility. Drawn from the existing literature, we suggest three different stock characteristics as proxies for visibility/recognition and familiarity. They are a stock's index membership, firm size, and a stock's listing period on the exchange. Out of the 30 sample stocks, nine are the members of the popular SHSE30 index, while the rest are not. To separate the 30 stocks based on firm size and the listing period, we first sort all stocks traded on the SHSE based on their size and listing period, respectively. We then assign stocks to large and small groups or long and short listing period groups by using the median stock of each sort as a breakpoint. As a result, we have 26 large stocks and four small stocks, and 21 long listing stocks and nine short listing stocks.

Test results of the three stock characteristics using baseline Specifications (1) and (2) are summarized in Table VI. Panels A to C of the table present equal- and accounts-weighted average differences between own and other branch effects associated with the respective characteristic. The overall result of the table demonstrates no evidence that visibility or investor recognition contributes to the word-of-mouth effect in investor trading decisions.

Table VI.  Word-of-Mouth Effects and Visibility/Investor Recognition The table presents results of own branch effects versus other branch effects with investor recognition and visibility controls. We use three proxies for visibility/investor recognition: 1) stock membership of the SHSE30 index, 2) firm size, and 3) the listing period of a stock. For 1), we divide stocks into two groups according to whether the stock is a member of SHSE30. For 2) and 3), we rank all stocks on the SHSE into two groups according to the market capitalization of the stock (firm size) and the listing period of the stock, respectively. The median stock determines the breakpoint for large versus small stocks or short versus long listing periods. Then, we assign each of 30 stocks based on the breakpoints. We use baseline Models (1) and (2) to estimate the own versus other branch coefficients for each characteristic-based sample. We also test the differences in the weighted average coefficients between two subgroups associated with each visibility/recognition proxy variable. The “EW” displays the equal-weighted average difference between own and other branch coefficients across the two subgroups, together with its associated t-statistic in parentheses. The “AW” differential displays the accounts-weighted differential, with the number of accounts in each branch as the weight.
 Investor BuysInvestor Sells
EWt-StatAWt-StatEWt-StatAWt-Stat
 
Panel A. SHSE30 Membership
SHSE303.602***(7.56)3.988***(4.91)2.193***(5.61)2.248***(4.42)
Non-SHSE303.531***(6.50)3.859***(4.61)1.388***(4.29)1.535***(3.02)
Difference0.071(0.10)0.129(0.11)0.805(1.59)0.712(0.97)
Panel B. Firm Size
Large3.529***(6.78)3.902***(4.65)1.577***(4.92)1.680***(3.42)
Small3.995***(7.06)4.175***(5.50)2.087***(3.79)2.700***(3.45)
Difference−0.466(−0.61)−0.273(−0.24)−0.510(−0.80)−1.020(−1.10)
Panel C. Stock Listing Period
  1. ***Significant at the 0.01 level.

Long3.655***(6.83)4.045***(4.66)1.702***(5.20)1.803***(3.58)
Short3.866***(7.75)4.164***(5.47)1.436***(4.22)1.684***(3.25)
Difference−0.211(−0.29)−0.119(−0.10)0.266(0.56)0.118(0.16)

D. Calendar Effects

Here we test whether our results are driven by a particular month of the year by reestimating the baseline regressions by month. For each month, we estimate the own and other branch coefficients and their differences associated with investor buys and investor sells. Then, we compute the Fama and Macbeth (1973) cross-sectional averages of the accounts- and equal-weighted differentials. For investor buys, all the unreported own branch coefficients are greater than their other branch counterparts, and nine of the weighted average differentials are statistically significant at the 5% level. The Fama and Macbeth (1973) cross-sectional equal- and accounts-weighted differential averages are 3.59 (t-statistic = 6.3) and 3.67 (t-statistic = 6.6), respectively. For investor sells, all the own branch coefficients are also greater than their other branch counterparts, but only six of the differentials are statistically significant at conventional levels. The Fama and Macbeth (1973) cross-sectional equal- and accounts-weighted differential averages are 2.54 (t-statistic = 3.2) and 2.63 (t-statistic = 3.6), respectively. The results show no evidence of any calendar month effects.

Overall, our results suggest that word-of-mouth communication plays an important role in the correlated trading of investors. Notwithstanding this, it is important to note that our evidence does not rule out the possibility that such a phenomenon can also be consistent with both existing rational and behavioral explanations. It is plausible that investors who maintain brokerage accounts at the same branch have the same information source, thus contributing to their coordinated trades. Such “herding” behavior reflects investor rational decisions as predicted in Daniel, Hirshleifer, and Subrahmanyam (1998). Alternatively, the observed correlated trading by investors may perhaps reflect the behavior of noisy traders. Recent studies by Barber, Odean, and Zhu (2009a, 2009b) and Jackson (2003) indicate that individual investors are noisy traders in equity markets and that they trade on noise as if it were information. However, distinguishing these various competing explanations for the correlated trading by individual investors is beyond the current scope of the study. We will leave this for future research.

V. Conclusion and Discussion

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References

This study employs a new and interesting data set to examine direct peer effects, via word of mouth, in the trading decisions of groups of individual investors who placed their trades largely in person at their brokerage branch office located in the same city, Shanghai. The key advantage of our data is that they allow us not only to directly measure peer effects but also to construct various tests that distinctly discriminate between word-of-mouth effects and alternative interpretations. More importantly, as our study is based on an emerging market with different social and cultural structures than those of developed markets, it provides a comparison of the findings of peer effects with those of US markets.

Results indicate strong word-of-mouth effects in the trading decisions of individual investors who are in the same room at the time they place their trades. The influence of word-of-mouth information is more pronounced in investor buys than investor sells of locally headquartered stocks. The evidence is robust to several alternative interpretations associated with familiarity or stock visibility, within branch bias, local bias, location, firm size, and calendar effects. In contrast to existing US studies, we find evidence of peer effects only at the branch level, but not at the city level. The distinction of these findings suggests that in Chinese societies that rely heavily on constant face-to-face social interactions (guanxi), physical proximity is essential for the working of word-of-mouth communication.

Finally, our results offer several asset pricing implications. With a preponderance of inexperienced individual investors in Mainland Chinese markets, individual investors’ excessive reliance on word-of-mouth (or unverified) information might increase the divergence of asset valuations. It is also likely that stronger word-of-mouth propensities in investor buys situations amplify the magnitude of the speculative component of asset prices. This implication is in accord with the recent evidence provided by Mei, Scheinkman, and Xiong (2005). Their study demonstrates that a significant fraction of the price difference between the dual-class shares may be due to the fact that trading in A-Share markets is more likely to be driven by speculation than by liquidity. Possibly peer effects, via word of mouth, contribute in part to this investor speculative trading behavior.     ▪

Footnotes
  • 1

    Glaeser and Scheinkman (2000) provide a detailed survey of the literature.

  • 2

    The ethnographical study of the SHSE by Hertz (1998, chap. 6) details her observation of the extent of social interactions among Chinese individual investors in Shanghai.

  • 3

    The data are similar to the New York Stock Exchange's consolidated equity audit trail data.

  • 4

    Glaeser et al. (1996) and Glaeser (2004) employ this relation to show, respectively, aggregate outcomes (e.g., crimes) and individual decisions (e.g., residence in cities).

  • 5

    A-Shares are common stocks accessible only to local Chinese investors, whereas B-Shares were initially accessible only to foreign investors until June 2001, a period during which the Chinese markets fully liberalized the B-share market to all domestic investors. Chinese firms may choose to list their stock on overseas exchanges, such as the H- Shares listings on the Hong Kong Stock Exchange (Jia, Sun, and Tong, 2005).

  • 6
  • 7

    RMB is the Chinese currency, reminbi (or Chinese yuan). Since 1995, RMB was officially fixed at RMB 8.28 to one US dollar. But on July 21, 2005, China abandoned the 10-year-long US dollar peg to allow its currency to fluctuate against the dollar.

  • 8

    We also employed an alternative specification: inline imagewhere inline imageis inline image the equal weighted average buy-ratio across all branches other than branch b. A similar specification is conducted for sells. The results of these alternative models yielded qualitatively similar to those reported in the paper.

  • 9

    At a given branch, we first calculate the monthly buy intensity and sell intensity across our entire sample period. The buy (sell) intensity is measured as the total buy (sell) value of the 30 sample SHSE stocks divided by the total buy (sell) value of all SHSE stocks placed through a branch. We then measure the pairwise correlation of buys and sells between any two branches. There are 435 pairs for 30 branches.

  • 10

    See Hall (1963) for his theory of proxemics that argues that human perceptions of space are molded and patterned by culture. For example, people will maintain varying degrees of personal distance depending on the social setting and their cultural backgrounds.

  • 11

    We have also incorporated stock dummy variables in Specifications (1) and (2) as alternative specifications to control for specific stock effects. The results are almost identical to those reported in Table V, thereby implying that neither the stock preference of individual investors nor group psychology plays a significant role in the base results.

References

  1. Top of page
  2. Abstract
  3. I. Related Literature and Hypotheses Development
  4. II. Data Description and Sample Selection
  5. III. Word-of-Mouth Effects in Investor Trading Decisions
  6. IV. Word-of-Mouth and Alternative Interpretations
  7. V. Conclusion and Discussion
  8. References
  • Amihud, Y., M. Haim, and J. Uno, 1999, “Number of Shareholders and Stock Prices: Evidence from Japan,” Journal of Finance 54, 1169-1184.
  • Bailey, W., Y.P. Chung, and J. Kang, 1999, “Foreign Ownership Restrictions and Equity Price Premiums: What Drives the Demand for Cross-Border Investments Journal of Financial and Quantitative Analysis 34, 489-512.
  • Banerjee, A.V., 1992, “A Simple Model of Herd Behavior,” Quarterly Journal of Economics 107, 797-817.
  • Banerjee, A.V. and D. Fudenberg, 2004, “Word-of-Mouth Learning,” Games and Economic Behavior 46, 1-22.
  • Barber, B., T. Odean, and N. Zhu, 2009a, “Systematic Noise,” Journal of Financial Markets 12, 547-569.
  • Barber, B., T. Odean, and N. Zhu, 2009b, “Do Retail Trades Move Markets Review of Financial Studies 22, 151-186.
  • Bikhchandani, S., D. Hirshleifer, and I. Welch, 1992, “A Theory of Fads, Fashion, Custom, and Cultural Change as Information Cascade,” Journal of Political Economy 100, 992-1026.
  • Bris, A., W. Goetzmann, and N. Zhu, 2007, “Efficiency and the Bear, Short Sales and Markets around the World,” Journal of Finance 62, 1029-1079.
  • Brown, J.J. and P.H. Reingen, 1987, “Social Ties and Word-of-Mouth Referral Behavior,” Journal of Consumer Research 14, 350-362.
  • Brown, J.R., Z. Ivkovic, P.A. Smith, and S. Weisbenner, 2005, “Neighbors Matter: The Geography of Stock Market Participation,” University of Illinois Working Paper.
  • Cao, H.H. and D. Hirshleifer, 2002, “Taking the Road Less Traveled: Does Conversation Eradicate Pernicious Cascades?” Ohio State University Working Paper.
  • Chen, B., X. Li, and Y. Du, 2002, “A Survey of Individual Investors in Chinese Stock Markets,” ShenZhen Stock Exchange Research Report.
  • Chiao, C., 1982, “Guanxi: A Preliminary Conceptualization” in K.S.Yang and C.I.Wen, Eds., The Sinicization of Social and Behavioral Science Research in China, Taipei , Academia Sinica (in Chinese), 345-360.
  • Coval, J. and T. Moskowitz, 1999, “Home Bias at Home: Local Equity Preference in Domestic Portfolios,” Journal of Finance 54, 2045-2073.
  • Curran, P.J., E. Stice, and L. Chassin, 1997, “The Relation between Adolescent Alcohol Use and Peer Alcohol Use: A Longitudinal Random Coefficients Model,” Journal of Consulting and Clinical Psychology 65, 130-140.
  • Daniel, K., D. Hirshleifel, and A. Subrahmanyam, 1998, “Investor Psychology and Security Market Under- and Overreactions,” Journal of Finance 53, 1839-1886.
  • Dishion, T.J., J. McCord, and F. Poulin, 1999, “When Interventions Harm: Peer Groups and Problem Behavior,” American Psychologist 54, 755-764.
  • Ellison, G. and D. Fudenberg, 1993, “Rules of Thumb for Social Learning,” Journal of Political Economy 101, 93-126.
  • Ellison, G. and D. Fudenberg, 1995, “Word-of-Mouth Communication and Social Learning,” Quarterly Journal of Economics 110, 93-125.
  • Fama, E.F. and J.D. MacBeth, 1973, “Risk, Return, and Equilibrium: Empirical Tests,” Journal of Political Economy 71, 753-755.
  • Farth, J., A. Tsui, K. Xin, and B. Cheng, 1998, “The Influence of Relational Demography and Guanxi: The Chinese Case,” Organization Science 9, 471-488.
  • Feng, L. and M. Seasholes, 2004, “Correlated Trading and Location,” Journal of Finance 59, 2117-2144.
  • Foerster, S.R. and G.A. Karolyi, 1999, “The Effects of Market Segmentation and Investor Recognition on Asset Prices: Evidence from Foreign Stocks Listing in the United States,” Journal of Finance 54, 981-1013.
  • Glaeser, E.L., 2004, “Cities and Social Interactions,” Harvard University Working Paper.
  • Glaeser, E.L., B. Sacerdote, and J.A. Scheinkman, 1996, “Crimes and Social Interactions,” Quarterly Journal of Economics 111, 507-548.
  • Glaeser, E.L. and J.A. Scheinkman, 2000, “Non-Market Interactions,” NBER Working Paper No. 8053.
  • Hall, E.T., 1963, “A System for the Notation of Proxemic Behavior,” American Anthropologist 65, 1003-1026.
  • Hertz, E., 1998, The Trading Crowd: An Ethnography of the Shanghai Stock Market, Cambridge , UK , Cambridge University Press.
  • Hong, H., J. Kubik, and J. Stein, 2004, “Stock Interaction and Stock-Participation,” Journal of Finance 59, 137-163.
  • Hong, H., J. Kubik, and J. Stein, 2005, “Thy Neighbor's Portfolio: Word-of-Mouth Effects in the Holdings and Trades of Money Managers,” Journal of Finance 60, 2801-2824.
  • Huberman, G., 2001, “Familiarity Breeds Investment,” Review of Financial Studies 14, 659-680.
  • Ivkovic, Z. and S. Weisbenner, 2005, “Local Does as Local Is: Information Content of the Geography of Individual Investors’ Common Stock Investments,” Journal of Finance 60, 267-306.
  • Ivkovic, Z. and S. Weisbenner, 2006, “Information Diffusion Effects in Individuals’ Common Stock Purchases: Covet thy Neighbors’ Investment Choices,” Review of Financial Studies 20, 1327-1357.
  • Jackson, A., 2003, “The Aggregate Behavior of Individual Investors,” London Business School Working Paper.
  • Jencks, C. and S.M. Mayer, 1990, “The Social Consequences of Growing Up in a Poor Neighborhood” in L.E.LynnJr. and M.G.H.McGeary, Eds., Inner-City Poverty in the United States, Washington , DC , National Academy Press.
  • Jia, J., Q. Sun, and H.S. Tong, 2005, “Privatization through an Overseas Listing: Evidence from China's H-Shares Firms,” Financial Management 34, 5-26.
  • Kang, J. and R.M. Stulz, 1997, “Why Is There a Home Bias? An Analysis of Foreign Portfolio Equity Ownership in Japan,” Journal of Financial Economics 46, 3-28.
  • Katz, L.F., J.F. Kling, and J.B. Liebman, 2001, “Moving to Opportunity in Boston: Early Results of a Randomized Mobility Experiment,” Quarterly Journal of Economics 116, 607-654.
  • Lam, D. and X. Lin, 2003, “Guanxi and Word-of-Mouth,” Proceedings of the Australia and New Zealand Marketing Association Conference 2003, Adelaide, Australia.
  • Leung, T.K.P., Y.H. Wong, and J.L.M, Tam, 1995, “Adaptation and the Relationship Building Process in the People's Republic of China (PRC),” Journal of International Consumer Marketing 8, 7-19.
  • Mei, J., J.A. Scheinkman, and W. Xiong, 2005, “Speculative Trading and Stock Prices: An Analysis of Chinese A-B Share Premia,” New York University and Princeton University Working Paper.
  • Merton, R.C., 1987, “A Simple Model of Capital Market Equilibrium with Incomplete Information,” Journal of Finance 42, 483-510.
  • Money, R.B., M.C. Gilly, and J.L. Graham, 1998, “Explorations of National Culture and Word-of-Mouth Referral Behavior in the Purchase of Industrial Services in the United States and Japan,” Journal of Marketing 62, 76-87.
  • Morris, M.W. and K. Peng, 1994, “Culture and Cause: American and Chinese Attributions for Social and Physical Events,” Journal of Personality and Social Psychology 67, 949-971.
  • Newey, W.K. and K.D. West, 1987, “A Simple Positive Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica 5, 703-705.
  • O’Keefe, H. and W.M. O’Keefe, 1997, “Chinese and Western Behavioral Differences: Understanding the Gaps,” International Journal of Social Economics 24, 190-196.
  • Osland, G.E., 1989, “Doing Business in China: A Framework for Cross-Cultural Understanding,” Marketing Intelligence & Planning 8, 4-14.
  • Shiller, R.J., 1995, “Conversation, Information and Herd Behavior,” American Economic Review 85, 181-185.
  • SHSE Fact Book, 2001, Available at: http://www.sse.com.cn/en_us/cs/about/factbook/factbook_us2001.pdf.
  • Valliant, P.M., 1995, “Personality, Peer Influence and Use of Alcohol and Drugs by First-Year University Students,” Psychological Reports 77, 401-402.
  • Zhu, N., 2002, “The Local Bias of Individual Investors,” Yale ICF Working Paper No. 02-30.