Do analysts forecasts affect investors trading? Evidence from China s accounts data

Transcription

1 Do analysts forecasts affect investors trading? Evidence from China s accounts data Xiong Xiong, Ruwei Zhao, Xu Feng 1 China Center for Social Computing and Analytics College of Management and Economics Tianjin University, Tianjin, China Guangbin Xu Shanghai Stock Exchange, Shanghai, China Abstract We use China s accounts data to test whether analysts forecasts (including recommendations and earnings per share (EPS, hence then)) influence the trading behavior of individual and institutional investors. We find that small individual investors prefer to follow recommendations and that large individual investors and institutional investors do not like to follow recommendations. Additionally, only large individual investors incline to respond to analysts EPS forecasts. We also apply the cutoff approach to check the robustness of the results and find that the cutoff approach is a better fit for catching the trading behavior of individual investors rather than that of institutional investors. Keywords Analysts forecasts; Trading behavior; Cutoff approach; Accounts data; Chinese stock market 1. Introduction In this paper, we have studied the types of investors to be influenced by analysts forecasts. Many studies have found that analysts forecasts, e.g., recommendations or earnings per share (EPS), hold strong optimistic bias (Francis and Philbrick (1993), Easterwood and Nutt (1999), Libby et al (2008)). Due to the existence of information asymmetry, not all 1 Corresponding author: fengxu@tju.edu.cn

2 investors can identify the bias. Therefore, these investors make sub-optimal investment decisions with misleading information published by the analysts. The relationship between analysts forecasts and investors trading behavior has been checked in multiple studies (Womack (1996), Mikhail, Walther, and Will (2007), Malmendier and Shanthikumar (2007), Hirshleif et al (2008), Drake, Rees and Swanson (2011)). The investors are usually separated into two groups: individual and institutional. The cutoff approach, which uses a specific dollar value or trade size as a threshold to categorize the intraday trading activities as individual-initiated or institutional-initiated trades, has been widely used in earlier studies. Unlike these studies, we obtain the trading activities of individual and institutional investors from accounts data. Our study is supported by the Research Center of Shanghai Stock Exchange. The Research Center records trading activities of each account registered in the Shanghai Stock Exchange. The accounts data enable us to conduct our research from the following three perspectives, distinguishing our paper from the earlier studies. First, the accounts data provide a more accurate classification of the different types of investors. In the prior studies, the investors were simply divided into two groups. The accuracy of the cutoff approach adopted by Lee and Radhakrishna (2000) is still being debated (Malmendier and Shanthikumar (2007)). Shanghai Stock Exchange divides the accounts into individual and institutional based on the registered identification of a natural person or a legal entity. Specifically, the Exchange divides the individual investors into subgroups, Retail Investor, Small Investor, Median Investor, Large Investor, and Extremely Large Investor, based on the three month trading volume of an account. The institutional investors are divided into 7 subgroups: Mutual Fund, Propriety Trading, QFII, Insurance, Pension Fund, Asset Management Company, and General Institutions. The account data enables us to conduct research with more specific subgroups and have a deeper understandings of analysts influence on the different types of investors. Second, our study is conducted from the emerging markets perspective. Most of the prior studies have focused on the US market. The studies on China and other emerging markets are less frequent. The Chinese market, with 80 percent of investors as small investors, is much more immature than the developed markets. Thus, analysts may hold more bias with interest conflicts, and the information asymmetry may be more severe (Michaely and Womack (1999),

3 Ljungqvist et al (2007), Lai and Teo (2008), Moshirian, Ng, and Wu (2009), Kadan et al (2009)). From this point of view, the influence of analysts recommendations needs to be checked due to the difference between developed and emerging markets. With the accounts data, we can test whether the conclusions of prior studies still work in the Chinese market. Furthermore, this study provides suggestions for the emerging market regulators on how to better protect investors interests. Third, we can double check the analysts influence with the cutoff approach, using the intraday data. Without the cutoff check, even if the results from China accounts data are different from those of the developed markets, we cannot conclude that the results are special to the emerging markets, which may be caused by the classification error of the cutoff approach. Thus, we apply the cutoff approach with the intraday transaction data to check the difference. We find that the cutoff approach holds better accuracy in classifying individual investors but not institutional investors. The classification error explains the inconsistent results in the Chinese stock market and developed markets. With accounts data, our paper examines the analysts recommendations and EPS forecasts effect on both individual and institutional investors. We investigate the effect using investors abnormal total, buying and selling trading volumes. We find evidence that the analysts recommendations have an influence on the trading of small individual investors with significant positive coefficients. However, large individual and institutional investors prefer to unfollow the analysts recommendations, with insignificant betas at almost all levels. In addition, only large individual investors like to follow analysts EPS forecasts, with positive betas at both total and buying levels. Small individual investors and institutional investors show no evidence of following the analysts EPS forecasts, with no significant betas. All these findings are consistent with the hypothesis of naïve individual investors in prior studies but are inconsistent with the conclusions on institutional investors. The rest of the paper is organized as follows. Section 2 describes the literature review. Section 3 exhibits the data and methodology used in this study. Section 4 presents the results of the effects of analysts recommendations and EPS forecasts on individual and institutional investors behavior. Section presents the robustness check with the cutoff approach. Section 6 concludes the paper.

4 2. Literature Review Our paper is related to three strands of prior research. The first strand is the influence of analysts recommendations on the different types of investors. The second strand is the influence of analysts EPS forecasts on the different types of investors. The third strand is the cutoff approach and the classification of investors trading activities. 2.1 Analysts recommendations and investors trading Some studies have found that analysts recommendations have an effect on both individual and institutional investors and that the effect on individual investors is stronger than on institutional investors. Mikhail, Walther, and Will (2007) connected analysts forecasts with investors total trading volume and found that both small and large investors respond to analysts recommendations, with small investors trading more than institutional investors. Malmendier and Shanthikumar (2007) found that small investors by total trading volume prefer to unfollow the independent analysts recommendations, indicating that small investors are naive. Similar findings are also noted by Walker and Hatfield (1996), Chen and Cheng (2006); He, Mian, and Sangkaraguruswamy (200); Irvine, Lipson, and Puckett (2007), etc. However, some studies have also argued that institutions are seldom affected by analysts. Malmendier and Shanthikumar (2014) suggested that the recommendations have the most impact on individuals, whereas EPS forecasts have impact on institutions. Our study examines analysts influence on individuals and institutions. We use accounts data not only at the total trading level but also at the buying (selling) trading level. Our hypothesis is as follows. H1: Both individual and institutional investors are affected by analysts recommendations. However, the individual (institutional) investors are more (less) affected. 2.2 Analysts EPS forecasts and investors trading Studies of the influence of analysts EPS forecasts also suggest that small investors are naive investors. They find that individual investors cannot fully understand the fundamental asset changes introduced by the EPS and do not respond to the EPS forecasts published by analysts. In contrast, with the information advantage and sophistication,

5 institutional investors do respond to EPS forecasts. Bhattacharya (2001) finds that small traders abnormal trading is positively correlated with seasonal random walk forecast error, while the large traders trading is negatively correlated with the analysts forecast error. Hirshleif et al (2008) find that individual investors like to buy stocks regardless of whether earnings surprises are positive or negative. Battalio and Mendenhall (200) and Shanthikumar (2012) reached a similar conclusion that only large investors respond to analyst forecasts on earnings. Similar to H1, our study also examines the total, buying and selling levels. Our hypothesis is as follows. H2: Individual investors are naive investors and are not sensitive to analysts EPS forecasts. Institutional investors, who are more sophisticated, are affected by analysts EPS forecasts. 2.3 The classification of different types of investors The key issue in studying the trading behavior of different types of investors is how to identify whether a trade is initiated by an individual investor or an institutional investor. Most prior studies follow the cutoff approach proposed by Lee and Radhakrishna (2000). The cutoff approach is based on intraday transaction data. If the buy (sell) trading amount is equal to or larger than the cutoff value (maybe a dollar value or a trade size), then the buy (sell) will receive a coefficient of positive one (negative one). Otherwise, the transactions with smaller amounts obtain a coefficient of zero. Although Lee and Radhakrishna (2000) have already considered the firm size effect by the different types of investors, Malmendier and Shanthikumar (2007) still argue about the accuracy of this approach. They suggest that the reliability of Lee and Radhakrishna (2000) s study still needs to be tested. A similar conclusion is also drawn by Campbell, Ramadorai, and Schwartz (2009). Our study applies the cutoff approach to check both analysts recommendations and EPS forecasts influence on different types of investor trading, and compares the results with accounts data. Some studies also use the accounts data of institutional investors from specific investment companies (e.g., Irvine, Lipson, and Puckett (2007) and Busse, Green, Jegadeesh (2012)). Compared with these studies, the accounts data in our paper come from all of the institutional investors in the Shanghai Stock Exchange, which avoids the selection bias associated with studying only one

6 investment company. 3. Data and Methodology Our data come mainly from three databases: the RESET database, Sina Finance Tick History, and the Shanghai Stock Exchange. Analysts recommendations, analysts EPS forecasts, the number of analysts EPS forecasts, Shanghai stock market total trading volume, stocks total volume, stock market capitalization, stocks returns and stocks turnovers are daily data from RESET Financial Services. The intraday volume and strike price are from Sina Finance Tick History. The daily Buy and Sell volumes of different types of investors are from the Shanghai Stock Exchange, whereas the total assets of the brokers are from the 2013 annual report of the Securities Association of China. Our data consist of 100 stocks listed in the Shanghai Stock Exchange from July 1st, 2013 to July 1st, 2014 that have at least one analyst recommendation from agencies without trading halts. Additionally, to avoid event day bias, we eliminate the recommendations and EPS forecasts published days before and after the day that the earnings were announced by the listed companies. Accounts data from the Shanghai Stock Exchange consists of twelve different types of investors. These types are divided into two groups: institutional and individual. The Retail Investor, Small Investor, Median Investor, Large Investor and Extremely Large Investor are in the individual group, whereas the Mutual Fund, Propriety Trading, QFII, Insurance, Pension Fund, Asset Management, and General Institutions are in the institutional group. Table 1 shows the criteria for these types of investors provided by the Shanghai Stock Exchange. To simplify, we set the Retail Investor, Small Investor, and Median Investor as Small Investors and Large Investor and Extremely Large Investor as Large Investors. Large individual investors have more contacts with brokerage firms and financial advisories. Thus, it is easier for them to access inside information. For the institutional investors, we also set two subgroups. The first group, Research Investors, includes Mutual Fund, QFII and Propriety Trading. The three types of institutions not only trade but also have independent research departments responsible for publishing forecasts for other investors. The second group is called Non-research Investors and includes Insurance, Pension Fund, Asset Management, and General Institutions. These types of institutions trade only in the market and do not publish

7 forecasts. Their information may come from public information channels, such as the recommendations and EPS forecasts published by the Research Investors, or their private information channels. Additionally, the total, buying and selling abnormal trading volumes are dependent variables in models (1), (2) and (3). The abnormal trading volumes are calculated as follows. Abnormal Trading Volume m,i,t = Trading volume of firm Average trading volume i at investor m of of firm i in investor m [ analysts recommendations of analysts ] recommendations at time t with time window (t =, +) On the right side, the independent variables are analysts recommendations (Recomm), analysts recommendations with brokerages size (Recomm*InsSize), analysts recommendations with star brokerage (Recomm*InsStar) and some control variables FirmSize, MarketVolume, CAR t k, CVOL t k, and CTV t k. Thus, the regression equation is as follows. TAVolume m,i,t = α i + β re Recomm i,t + β ri Recomm InsSize i,t + β ris Recomm InsStar i,t + β fs FirmSize i,t + β mv MarketVolume t + β car CAR t k + β cvol CVOL t k + β ctv CTV t k + ε m,i,t (1) BAVolume m,i,t = α i + β re Recomm i,t + β ri Recomm InsSize i,t + β ris Recomm InsStar i,t + β fs FirmSize i,t + β mv MarketVolume t + β car CAR t k + β cvol CVOL t k + β ctv CTV t k + ε m,i,t (2) SAVolume m,i,t = α i + β re Recomm i,t + β ri Recomm InsSize i,t + β ris Recomm InsStar i,t + β fs FirmSize i,t + β mv MarketVolume t + β car CAR t k + β cvol CVOL t k + β ctv CTV t k + ε m,i,t (3) where TAVolume m,i,t, BAVolume m,i,t, SAVolume m,i,t, are the total, buying and selling log abnormal trading volumes of stock i at time t of investor m, and Recomm i,t is the average recommendation of stock i at time t, which is calculated as the arithmetic mean of recommendations at time t. The recommendations hold five levels, - Strong Sell, Sell, Hold, Buy, Strong Buy, denoted by 1, 2, 3, 4,. Recomm InsSize i,t is the average

8 recommendation multiplied by the log Size of the Brokerages, which are the log average total assets of the broker i at the end of Recomm InsStar i,t is the average recommendation multiplied by the dummy variable InsStar of stock i at time t. If the institution has been awarded the Crystal Ball Award, which is the most important award in the analysts industry of China, InsStar will be 1. Otherwise, InStar will be zero. We test the hypothesis H1 to see if β re, β ri, and β ris equal 0. A significant β re indicates sufficient evidence to conclude that there is a correlation between investors trading behavior and analysts recommendations. FirmSize i,t is the log total market capitalization of firm i at time t. MarketVolume t is the log total trading volume of the Shanghai stock exchange at time t. CAR t k is the sum of ex-ante days cumulative abnormal returns of stock i at time t. The abnormal returns are calculated as unexplained returns with the three factor model of Fama and French (1993). CVOL t k is the log sum of the ex-ante days cumulative volume of stock i at time t, and is the sum of the ex-ante days cumulative CTV t k turnover of stock i at time t. In addition, we also check the effect of the analysts EPS forecast on investors trading at total, buying and selling levels. In models (4), () and (6), the total, buying and selling abnormal trading volumes are dependent variables. On the right side, the independent variables are analysts EPS forecasts (EPS), analysts EPS forecasts with the numbers of forecasts published (EPS*Number), and some control variables: FirmSize, MarketVolume, CAR t k, CVOL t k, and CTV t k. We test the hypothesis H2 to see if β eps and β en equal 0. A significant β eps indicates sufficient evidence to conclude that there is a correlation between investors trading behavior and analysts EPS forecasts. Thus, the regression equation is as follows. TAVolume m,i,t = α i + β eps EPS i,t + β en EPS Number i,t + β fs FirmSize i,t + β mv MarketVolume t + β car CAR t k + β ctv CTV t k + ε m,i,t (4) + β cvol CVOL t k

9 BAVolume m,i,t = α i + β eps EPS i,t + β en EPS Number i,t + β fs FirmSize i,t + β mv MarketVolume t + β car CAR t k + β ctv CTV t k + ε m,i,t () + β cvol CVOL t k SAVolume m,i,t = α i + β eps EPS i,t + β en EPS Number i,t + β fs FirmSize i,t + β mv MarketVolume t + β car CAR t k + β ctv CTV t k + ε m,i,t (6) + β cvol CVOL t k where TAVolume m,i,t, BAVolume m,i,t, SAVolume m,i,t, are the total, buying and selling log abnormal trading volume of stock i at time t of investor m, and EPS i,t is the average EPS forecasts of stock i at time t, which is calculated as the arithmetic mean of recommendations at time t. EPS Number i,t is the average EPS forecasts multiplied by the number of forecasts published by different institutions. MarketVolume t, CAR t k, CVOL t k and hold the same meaning as in models (1) to (3). CTV t k, Furthermore, we use the cutoff approach to check robustness of models (1) to (6). The cutoff volumes are calculated with the intraday transaction data from Sina Finance Tick History that has the trade direction and the trading volume of each transaction. If the buy (sell) trading amount is equal to or larger than the cutoff value, the buy (sell) volume will receive a coefficient of positive one (negative one). Otherwise, the transactions with smaller amount receive a coefficient of zero. The regression models analysts recommendations and EPS forecasts on the cutoff approach data are similar to the models (1) to (6). We use Cutoff_TAV m,i,t, Cutoff_BAV m,i,t, Cutoff_SAV m,i,t for the total, buying and selling log abnormal trading volume, respectively, of stock i at time t of investor m with the cutoff approach to replace the dependent variables with account data. The independent variables are the same as in model (1) to model (6). We also check the fitness degree of the cutoff data to the real data at total, buying and selling levels. The regressions are presented as follows. Real i,t = α i + β cf Cutoff i,t + ε i,t (7) where Real i,t is the total, buying and selling volume from accounts data of stock i at time

10 t. Cutoff i,t is the total, buying and selling volume from the cutoff approach of stock i at time t. α i is the intercept for total, buying and selling volume. β cf is the estimated coefficient of Cutoff i,t, and ε i,t is the error term. We test the hypothesis that β cf = 0. A positive and significant β cf indicates sufficient evidence to conclude that there is a positive relation between cutoff data and real data. 4. Results 4.1 Descriptive Statistics Investors buy and sell Table 2 presents the description of the investors trading in China s stock market. The table shows that the small individual investors have the largest means in both buying and selling, almost 7 times those of other investors, which may be caused by the large number of small investors in the China s stock market. Together with the larger trading volume, small investors have the largest standard deviation, which indicates the highest level of impatience among investors. Large individual investors have much less trading volume than small individual investors. The standard deviation of large individual investors is much smaller than that of small individual investors, which suggests a lower level of impatience. This may result from a smaller number of large individual investors. The means of the individual investors total trading volume are roughly five times larger than those of the institutional investors in both buying and selling. The total trading volumes of the individual and institutional investors can be found in Figure 1, where it is easier to find that great gap between individual and institutional investors. This indicates that in China s market, the individual investors have more influence on stock price than institutional investors Analysts recommendations and EPS forecasts Table 3 demonstrates descriptive statistics of analysts recommendations and EPS forecasts. The average recommendation reported by the analysts is 4.36, close to the highest number.00 of the recommendation. Additionally, the median of the recommendations is 4.00, which indicates that more than half of the recommendations hold Buy recommendation. This indicates that China s analysts prefer to provide positive recommendations of stocks, which is consistent with the optimistic bias of analysts conclusion in (Francis and Philbrick

11 (1993), Easterwood and Nutt (1999), and Libby et al (2008)). Additionally, for the institution size, most institutions are large institutions, with a mean of 6.66, a median of 6.78, and a maximum of Thus, most of the recommendations and EPS forecasts come from large institutions. The median of the star institution is 1, which indicates that more than half of the recommendations come from star institutions. The EPS forecasts vary a lot in the results, with the highest standard deviation of all of the presented statistics. The mean, median and the maximum of the EPS forecasts are 1.03, 0.48 and 1.69, respectively. 4.2 Fundamental Results Investors total trading with recommendations Table 4 presents the regression results of investors total trading with analysts recommendations of model (1). The table demonstrates the betas, t-statistics and R-squares of different investors. Only small individuals and research institutions move together with the analysts recommendations, with positive coefficients and the betas significant at the 9% level. The results show a partial difference with H1 because of the insignificant coefficients of large individual investors (t-value=0.88) and non-research institutions (t-value=1.37). The results on small individual investors are consistent with the naïve individual investor hypothesis that the small individual investors are not fully aware of the optimistic bias of analysts, which is also concluded by (Malmendier and Shanthikumar (2007), Mikhail, Walther, and Will (2007)). Moreover, the coefficient of Recom on small individuals is larger than on research institutions (0.38>0.30), which is also consistent with the naïve individual investor hypothesis. However, the results regarding large individual and non-research institutional investors suggest that these investors have private information channels and are not influenced by analysts. Most importantly, the trading volume of total institutional investors is not affected by analysts recommendations (t-value=1.43), which is inconsistent with the idea that institutions also follow recommendations but supports the results of (Malmendier and Shanthikumar (2014)). The Recom*InsSize is positively significant for small and total individual investors (0.14 and 0.01, respectively), which implies that large brokers have more influence on small individual investors. The Recom*InsStar is not significant for any investors, indicating that neither individuals nor institutional investors take into account whether the analyst is a star

12 analyst. Furthermore, the R-squares of individual investors are slightly higher than those of institutional investors, with 0.37 of individual investors and 0.22 of institutional investors Investors buying and selling with recommendations Tables and 6 present the regression results of both individual and institutional investors buying and selling trading with models (2) and (3). We notice that small and total individual investors buying and selling trading are influenced by analysts recommendations. Specifically, the coefficient of recommendations on the buying of small individual investors is significantly positive (0.38), whereas the coefficient on the selling is significantly negative (-0.38). These results indicate that small investors bought stocks when recommendations increased and sold stocks when recommendations decreased. For research institutional investors, recommendations of analysts affect only their buying behavior (t-value=1.73) but have little influence on their selling behavior (t-value=-0.8). The results of the buying and selling behavior of large individual and non-research institutional investors yield the same conclusions as for the total trading volume. Recommendations cannot significantly affect the trading behavior of these two types of investors. Moreover, the total institutional investors are still not affected by analysts (t-value=0.97 on buying and on selling). The results in Tables and 6 have further verified the conclusion that small individual investors in China s stock market are not aware of analysts bias. In contrast, the impact of recommendations on large individual and institutional investors buying or selling cannot be observed in China s stock market, which is inconsistent with the conclusions on the US market. It is necessary to perform further tests to clarify whether this difference is because of the accounts data we used Investors total trading with EPS forecasts Table 7 presents the regression results of investors total trading with analysts EPS forecasts of model (4). Different from the analysts recommendation results, small and total individual investors trading was insignificantly affected by EPS forecasts (t-value=0.21 and 0.29, respectively). The results confirmed the first part of H2 that small individual investors have less understanding of fundamental asset changes and are not sensitive to EPS forecasts, which was also observed by (Bhattacharya (2001), Hirshleif et al (2008), and Shanthikumar (2012)). However, we find that both research and non-research institutions

13 trading have no relationship with the EPS forecasts (t-value=1.03 and 0.68, respectively), which rejects the second part of H2 that the institutions will be affected by the analysts EPS forecasts. The most interesting results come from the large individual investors as their trading parallels the analysts EPS forecasts, with a positive coefficient (4.89) and significant beta at the 9% level (t-value=3.60). This implies that the EPS forecasts may be an important information channel for large individual investors, which has not been revealed by prior studies. The EPS*Number showed less predictability on all types of investors trading behavior except that of large individual investors. The EPS*Number on large individual investors is positively significant at the 90% level (t-value=1.69), which implies that the large individual investors trading increases as more brokers report high EPS forecasts. Furthermore, the R-squares of regression of individual investors are higher than those of institutional investors, with 0.34 of individual investors and 0.22 of institutional investors Investors buying and selling with EPS forecasts Tables 8 and 9 present the regression results of both individual and institutional investors buying and selling trading with models () and (6). In tables 8 and 9, EPS forecasts show influence on the large individual investors buying and selling behavior. The coefficient of EPS is positively significant (t-value=2.20) on the large individual investors buying volume and negatively significant (t-value=-3.14) on their selling volume. Moreover, the EPS*Number is negatively significant with respect to the large individual investors selling behavior (t-value=-1.98) but is insignificant with respect to their buying behavior (t-value=0.1), which implies that large individual investors sold stocks when more analysts decreased their EPS forecasts. The results of small and total individual investors confirmed the individuals part of H2 that small and total individuals buying and selling are not related to EPS forecasts. The research, non-research, and total institutional investors are also not affected by the EPS forecasts of analysts. This conclusion is similar to the results drawn from total trading, which reject the institutions part of H2. Robustness Check.1 Cutoff and real data

14 The results showed that the small individual investors trading in China s stock market is consistent with the naïve individual investor hypothesis, which is also observed in prior studies. However, the results of institutional investors trading in China s stock market are inconsistent with those in the US market. It is difficult to say why the institutions in China exhibit different trading behavior than those in the US because we use account data in this study, and the prior studies use data from the cutoff approach. The errors of the cutoff approach may be a factor causing the inconsistencies between our research and prior studies. Therefore, we employ the cutoff approach on the trade-by-trade data in China s stock market to provide a robustness test of our results. Before we test models (1) to (6) with cutoff data, we test whether the cutoff data has fitness for the accounts data. We use the grid search method to seek the best cutoff value with model (7). We find that the R-square on individual investors trading increases when the cutoff value increases, whereas the R-square on institutional investors trading is nonlinear. These results indicate that the cutoff approach holds better fitness for the individual investors. Finally, we find that 00,000 CNY has the highest R-square on both individual and institutional investors trading and that 100,000 CNY has the second highest R-square on only institutional investors trading. Table 10 compares the regression results of total, buying and selling volume of the cutoff approach to those of real data with the two cutoff values. It shows that 100,000 cutoff value individual investors are not significant with respect to total, buying and selling trading, with R-squares of 0.49, 0.43 and 0.40, respectively. This may be caused by the smaller threshold value. For the 00,000 value individual investors, we find that both R-squares and t-statistics increase dramatically, from 0.49 to 0.92 and 1. to 9.17, at the total level. For the institutional investors, the R-squares are kept at a low level on the regression of total (0.3), buying (0.40) and selling (0.37) with 00,000 cutoff values. The results regarding institutional investors imply that the errors caused by the cutoff approach on institutional investors trading are significant and cannot be neglected..2 Investors trading and recommendations with cutoff data We test models (1) to (3) with cutoff data to find whether the results of recommendations are the same as those with accounts data. Table 11 presents the regression results of investors trading behavior (including total trading, buying and selling) with analysts recommendations

15 of models (1) to (3). For simplicity, we report only the results on the variables of Recom, Recom*InsSize and Recom*InsStar. With the cutoff value = 100,000 CNY, the results are different than those in Tables 4 to 6. The Recom is significant with respect to the individuals total and sell trading (t-value=1.69 and -1.74) but is insignificant with respect to the individuals buy trading (t-value=1.43). Moreover, the Recom*InSize is insignificant with respect to total, buy and sell trading, which is statistically significant at the 9% level in Tables 4 to 6. For the institutional investors, the Recom is significant with respect to total, buy and sell trading, whereas the Recom*InSize is significant with respect to total and buy trading. These results are also inconsistent with the institutions results of accounts data. With the cutoff value = 00,000 CNY, the cutoff approach s fitness on individuals trading increased. The Recom and Recom*InSize are significant on the individuals total, buy and sell trading at the 90% level, which shows the same results as the accounts data. However, there is no significant change in the fitness on institutions trading. The Recom is still significant with respect to total trading volume (t-value=1.9), and the Recom*InSize is significant with respect to buy trading (t-value=2.40)..3 Investors trading and EPS forecasts with cutoff data We also test models (4) to (6) with cutoff data to find whether the results of EPS forecasts are the same as those with accounts data. Table 12 presents the regression results of investors total, buying and selling trading with EPS forecasts of models (4) to (6). We report only the results on the variables of EPS and EPS*Number. For individuals trading, the EPS and EPS*Number are insignificant for total, buy and sell trading with the cutoff = 100,000 and 00,000, which is confirmed with the results of accounts data and the naïve individual investors hypothesis. For the institutions trading, the EPS is significant at all types of trading with two cutoff values, which shows similar results with prior studies in the US market (Bhattacharya (2001), Hirshleif et al (2008), Shanthikumar (2012), etc.) and different results from those of the accounts data in Tables 7 to Conclusion Our paper conducts a study of the effect of the analysts forecasts on the different types of investors. Using accounts data from the Shanghai Stock Exchange, we find that small

16 individual investors are inclined to accept analysts recommendations in both buying and selling. Large individual investors and institutional investors do not show much preference in following analysts recommendations. In addition, we also find that large individual investors like to respond to analysts EPS forecasts. Neither small individual or institutional investors prefer to follow analysts EPS forecasts. Our results are consistent with the naïve individual investor hypothesis that small individual investors are not fully aware of the optimistic bias of analysts, which is also observed in prior studies. However, the result regarding institutional investors suggests that the analysts forecasts have no influence on the institutions trading, which is inconsistent with the results of the US market. To find out whether the different results are caused by using accounts data or by the special characteristics of the Chinese stock market, we apply the cutoff approach used in prior studies to check the robustness of our results. We find that the results of the cutoff approach are different from those of the accounts data, especially in the institutions trading. We suggest that the cutoff approach is more suitable for studying trading behavior of individual investors rather than that of institutional investors. Reference [1] Francis, J., Philbrick, D., Analysts' decisions as products of a multi-task environment. Journal of Accounting Research, 31, [2] Easterwood, J. C., Nutt, S. R., Inefficiency in analysts' earnings forecasts: systematic misreaction or systematic optimism? Journal of Finance, 4, [3] Libby, R., Hunton, J.E., Tan, H-T., Seybert, N., Retracted: Relationship incentives and the optimistic/pessimistic pattern in analysts' forecasts. Journal of Accounting Research, 46, [4] Womack, K. L., Do brokerage analysts recommendations have investment value? Journal of Finance. 1, [] Mikhail, M. B., Walther, B. R., Willis, R. H., When Security Analysts Talk, Who Listens? The Accounting Review, 82, [6] Malmendier, U., Shanthikumar, D., Are small investors naive about incentives? Journal of Financial Economics, 8,

17 [7] Hirshleifer, D. A., Myers, J. N., Myers, L. A., Teoh, S. H., Do individual investors cause post-earnings announcement drift? Direct evidence from personal trades. The Accounting Review, 83, [8] Drake, M. S., Rees, L., Swanson, E. P., Should investors follow the prophets or the bears? Evidence on the use of public information by analysts and short sellers. The Accounting Review, 86, [9] Lee, C. M. C., and Radhakrishna, B., Inferring investor behavior: Evidence from TORQ data. Journal of Financial Markets, 3, [10] Michaely, R., Womack, K. L., Conflict of interest and the credibility of underwriter analyst recommendations. Review of Financial Studies. 12, [11] Ljungqvist, A., Marston, F., Starks, L. T., Wei, K. D., Yan, Hong, Conflicts of interest in sell-side research and the moderating role of institutional investors. Journal of Financial Economics, 8, [12] Lai, S., Teo, M., Home-biased analysts in emerging markets. Journal of Financial and Quantitative Analysis, 43, [13] Moshirian, F., Ng, D., Wu, E., The value of stock analysts' recommendations: Evidence from emerging markets. International Review of Financial Analysis, 18, [14] Kadan, O., Madureira, L., Wang, R., Zach, T., Conflicts of interest and stock recommendations: The effects of the global settlement and related regulations. Review of Financial Studies, 22, [1] Walker, M. M., Hatfield, G. B., Professional stock analysts recommendations: Implications for individual investors. Financial Services Review,, [16] Chen, Q., Chen, X., Institutional holdings and analysts' stock recommendations. Journal of Accounting, Auditing & Finance, 21, [17] He, W., Mian, G. M., Sangkaraguruswamy, S., 200. Who follows the prophets? Analysts stock recommendations and the trading response of large and small traders. Available at SSRN [18] Irvine, P., Lipson, M., Puckett, A., Tipping. Review of Financial Studies, 20, [19] Bhattacharya, N., Investors' trade size and trading responses around earnings

18 announcements: An empirical investigation. The Accounting Review, 76, [20] Battalio, R. H., Mendenhall, R. R., 200. Earnings expectations, investor trade size, and anomalous returns around earnings announcements. Journal of Financial Economics, 77, [21] Shanthikumar, D. M., Consecutive earnings surprises: Small and large trader reactions. The Accounting Review, 87, [22] Campbell, J. Y., Ramadorai, T., Schwartz, A., Caught on tape: Institutional trading, stock returns, and earnings announcements. Journal of Financial Economics, 92, [23] Busse, J. A., Green, T. C., Jegadeesh, N., Buy-side trades and sell-side recommendations: interactions and information content. Journal of Financial Markets, 1, [24] Fama, E. F., French, K. R., Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3-6. [2] Malmendier, U., Shanthikumar D., Do Security Analysts Speak in Two Tongues? Review of Financial Studies, 27,

19 Table 1. Criteria for classifications of the investors Type of Investors Trading Amount (CNY) Criteria Account Information Panel A: Individual Investors Retail Investor 0-100,000 Small Investor 100, ,000 Median Investor 1,000,000-,000,000 Large Investor,000,000-10,000,000 Extremely Large Investor >10,000,000 Panel B: Institutional Investors Mutual Fund QFII Proprietary Trading Insurance Pension Fund Asset Management Company Assigned Account by Shanghai Stock Exchange Assigned Account by Shanghai Stock Exchange Assigned Account by Shanghai Stock Exchange Assigned Account by Shanghai Stock Exchange Assigned Account by Shanghai Stock Exchange Assigned Account by Shanghai Stock Exchange General Institutions Assigned Account by Shanghai Stock Exchange Note: This table demonstrates the criteria for classifications of the investors by the Shanghai Stock Exchange. The criteria are different between individual and institutional investors. For the individual investors, the criteria are based on their three-month trading amounts. For the institutional investors, the criteria are created based on the specific assigned accounts by the Shanghai Stock Exchange due to the Registration Regulation.

20 Table 2. Descriptive statistics of the investors buying and selling Buy Sell Type of Investors Mean (million shares) Median (million shares) Maximum (million shares) Minimum (million shares) Std. Dev. Mean (million shares) Median (million shares) Maximum (million shares) Minimum (million shares) Std. Dev. Panel A: Individual Investors Small Large Total Panel B: Institutional Investors Non-Research Research Total

21 Table 3. Descriptive statistics of the analysts forecasts. Variables Mean Median Maximum Minimum Std. Dev. Recommendation Institutional Size Institutional Star EPS forecasts Number of forecasts

22 Trading Volume(1 billion shares) 6 Trading Volume of Individual & Institutional Investors 4 3 Individual Institution /7/1 2013/8/1 2013/9/1 2013/10/1 2013/11/1 2013/12/1 2014/1/1 2014/2/1 2014/3/1 2014/4/1 2014//1 2014/6/1 2014/7/1 Date Figure 1. Individual & Institutional Investors Daily Total Trading Volume

23 Table 4. The regression results of investors total abnormal trading volume with analysts recommendations Type of Investors Total Abnormal Trading Volume Recom Recom*InsSize Recom*InsStar FirmSize MarketVolume CAR t CVOL t CTV t R 2 Panel A: Individual Investor Small Investors 0.38* 0.14* * 0.93* * (2.38) (2.31) (0.98) (2.4) (9.68) (-0.76) (-2.42) (0.9) Large Investors * 0.93* * (0.88) (1.06) (0.90) (2.0) (4.40) (-1.12) (-2.04) (1.21) Total 0.14* 0.01* * 0.92* * (2.24) (2.24) (0.86) (2.1) (9.00) (-0.88) (-2.41) (0.66) Panel B: Institutional Investor Non-Research Investors * 0.4* (1.37) (1.47) (1.48) (2.29) (3.37) (1.27) (-0.27) (0.17) Research Investors 0.30* 0.12* * (1.97) (2.16) (0.4) (3.2) (1.62) (-0.84) (0.77) (-1.3) Total * 0.42* (1.43) (1.20) (1.26) (2.8) (3.24) (-0.10) (-0.13) (-0.32) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

24 Table. The regression results of investors buy volume with analysts recommendations Type of Investors Buy Abnormal Trading Volume Recom Recom*InsSize Recom*InsStar FirmSize MarketVolume CAR t CVOL t CTV t R 2 Panel A: Individual Investor Small Investors 0.38* 0.14* * 0.91* * (2.38) (2.27) (0.81) (2.27) (9.69) (-1.0) (-2.04) (0.40) Large Investors * * (0.23) (0.28) (1.43) (1.66) (.0) (-1.30) (-2.74) (0.88) Total 0.14* 0.01* * 0.92* * (2.20) (2.12) (0.66) (2.33) (9.12) (-1.4) (-2.24) (0.2) Panel B: Institutional Investor Non-Research Investors * 0.2* (1.13) (1.39) (1.31) (2.41) (2.6) (1.12) (0.01) (-1.23) Research Investors * (1.73) (1.94) (0.80) (2.66) (1.34) (0.7) (-0.23) (0.63) Total * 0.38* 0.96* (0.97) (1.40) (1.27) (2.7) (2.26) (1.96) (-0.26) (0.26) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

25 Table 6. The regression results of investors sell volume with analysts recommendations Type of Investors Sell Abnormal Trading Volume Recom Recom*InsSize Recom*InsStar FirmSize MarketVolume CAR t CVOL t CTV t R 2 Panel A: Individual Investor Small Investors -0.38* -0.14* * 0.9* * (-2.33) (-2.29) (-1.09) (2.9) (9.38) (0.09) (-2.6) (0.70) Large Investors * 0.70* (-1.17) (-1.43) (-0.21) (2.32) (3.30) (-0.86) (-0.68) (0.90) Total -0.14* -0.01* * 0.92* * (-2.24) (-2.30) (-1.01) (2.66) (8.69) (-0.09) (-2.2) (0.72) Panel B: Institutional Investor Non-Research Investors * (-1.64) (1.) (-1.01) (1.81) (3.27) (0.97) (-0.40) (1.18) Research Investors * (-0.8) (-0.94) (0.6) (2.9) (1.38) (-1.47) (1.06) (-1.48) Total * 0.48* (-1.44) (-1.63) (-0.67) (2.33) (3.43) (-0.90) (0.04) (-0.6) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

26 Table 7. The regression results of investors total volume with analysts EPS forecasts Type of Investors Total Abnormal Trading Volume EPS EPS*Number FirmSize MarketVolume CAR t CVOL t CTV t R 2 Panel A: Individual Investor Small Investors * * (0.21) (0.81) (1.77) (4.74) (1.2) (-2.07) (0.22) Large Investors 4.89* * 0.* * (3.60) (1.69) (2.12) (2.71) (-0.14) (-2.36) (-0.7) Total * * (0.29) (0.41) (1.78) (4.9) (1.3) (-2.10) (0.17) Panel B: Institutional Investor Non-Research Investors * (1.03) (1.08) (0.31) (4.12) (1.63) (-1.87) (-0.80) Research Investors * * 0.2 (0.68) (0.92) (2.18) (0.06) (0.34) (0.30) (-2.16) Total * (0.97) (1.12) (1.23) (2.99) (1.14) (-0.80) (-1.6) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

27 Table 8. The regression results of investors buy volume with analysts EPS forecasts Type of Investors Buy Abnormal Trading Volume EPS EPS*Number FirmSize MarketVolume CAR t CVOL t CTV t R 2 Panel A: Individual Investor Small Investors * * (0.32) (0.82) (1.67) (4.3) (1.08) (-2.00) (0.34) Large Investors 3.0* * (2.20) (0.1) (1.79) (4.23) (-0.78) (-1.70) (-1.71) Total * * (0.12) (0.63) (1.70) (4.41) (0.88) (-2.02) (0.01) Panel B: Institutional Investor Non-Research Investors * (0.97) (0.71) (0.4) (3.92) (0.81) (-0.80) (-1.30) Research Investors * (1.4) (0.22) (2.20) (0.17) (1.17) (-1.4) (-0.01) Total * (0.83) (0.16) (1.6) (3.0) (1.12) (-1.13) (-0.68) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

28 Table 9. The regression results of investors sell volume with analysts EPS forecasts Type of Investors Sell Abnormal Trading Volume EPS EPS*Number FirmSize MarketVolume CAR t CVOL t CTV t R2 Panel A: Individual Investor Small Investors * * (0.09) (-0.67) (1.83) (4.86) (1.87) (-2.07) (0.10) Large Investors -6.39* -1.36* * * (-3.14) (-1.98) (1.83) (4.) (1.7) (-2.12) (0.39) Total * * (-0.43) (-0.13) (2.01) (0.3) (0.44) (-2.29) (1.31) Panel B: Institutional Investor Non-Research Investors * * (-1.7) (1.63) (0.78) (2.11) (1.94) (-2.69) (1.83) Research Investors * 0.18 (-1.10) (-1.66) (1.39) (0.10) (-0.71) (0.4) (-2.41) Total * 0.18 (-1.12) (-1.61) (0.83) (1.29) (0.73) (-0.3) (-1.96) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

29 Table 10. The regression results of total volume, buying and selling with cutoff approach to real data Value(CNY) Type of Investors Total Volume R 2 Buy R 2 Sell R Individual Investors (1.0) (1.73) (1.94) 100, * 0.31* 0.33* Institutional Investors (12.29) (11.83) (11.79) 3.47* 3.17* 3.32* Individual Investors (9.17) (23.08) (2.08) 00, * 0.4* 0.48* Institutional Investors (13.82) (13.03) (12.94) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 9% level.

30 Table 11. The regression results of investors trading with cutoff approach of analysts recommendations Value(CNY) 100,000 Type of Investors Individual Institutional Total Abnormal Trading Volume Buy Abnormal Trading Volume Sell Abnormal Trading Volume Recom Recom*InsSize Recom*InsStar Recom Recom*InsSize Recom*InsStar Recom Recom*InsSize Recom*InsStar 0.26* * (1.69) (1.3) (1.2) (1.43) (1.21) (1.18) (-1.74) (-1.3) (-1.03) 0.36* 0.38* * 0.09* * (1.80) (1.8) (0.19) (1.87) (1.83) (1.62) (-1.71) (-1.2) (1.01) 0.24* 0.09* * 0.47* * -0.08* Individual (1.94) (1.88) (1.1) (2.21) (2.23) (0.36) (-1.90) (-1.8) (-1.34) 00, * * * Institutional (1.9) (1.1) (0.96) (0.67) (2.40) (0.2) (-1.04) (-1.40) (-2.07) Note: t-statistics is in parentheses, *indicates that coefficients are statistically significant at 90% level.