Characteristics of Conflicts of Interest in Stock Recommendations

Transcription

1 Characteristics of Conflicts of Interest in Stock Recommendations Alex YiHou Huang * Department of Information Management and Finance, National Chiao Tung Unversity Chiao-Ming Cheng College of Management, Yuan Ze University ABSTRACT This paper studies characteristics of conflicts of interest in stock recommendations, and finds that analyst recommendations have an impact on recommended stocks, and that this phenomenon occurs for both traditional and revise recommendations. Moreover, the negative recommendations have a much greater impact than positive recommendations on stock returns. Findings suggest that high performance analysts are more concerned with enhancing their reputations and tend to not expose themselves to conflicts of interest when making upgrade recommendations. Furthermore, the results also suggest that concentrated ownership of individual stocks corresponds with a reduction in conflicts of interest. Keywords: Conflict of interest; recommendation; concentrated ownership; event study * Corresponding author. Address: 1001 University Road, Hsinchu City 30010, Taiwan (R.O.C.); Phone: ; ahuang1@gmail.com. This research is supported in part by the Ministry of Science and Technology of Taiwan (NSC H ).

2 1. Introduction The sell-side security analyst plays the important role as a stock market information intermediary. A review of the relevant literature shows that analyst recommendations have short-term investment value (Womack, 1996; Barber et al. 2001; Jegadeesh et al. 2004). While the impact of security analyst skill and recommendations investment value is widely understood, such analysts, along with their reports and employers are subject to conflicts of interest. Conflicts of interest exist because security analysts are dependent on investment banks or brokerage houses, which will seek to increase their revenue through accessing sensitive analysis information prior to public release. 1 For instance, Irvine et al. (2007) argue that analysts are incentivized to provide such information because their firms place a high value on their relationships with institutional clients, who will then act on this information to earn outsize profits. Shen and Chih (2009) demonstrated that analysts frequently allow their investment banking colleagues to buy shares ahead of the release of recommendations, and then sell for profit after the recommendations push up the share price. Irvine et al. (2007) and Shen and Chih (2009) both focus on trading behavior in traditional recommendations, while Loh and Stulz (2011) argue that a recommendation change event has a more significant impact on stock price than traditional recommendation events. Therefore, this study examines the trading behavior of investment banks around the release of both traditional and revision recommendations. Furthermore, this paper measures the conflicts of interest for investment banks and individual stocks based on the index provided by Shen and Chih (2009). The study seeks to extend empirical research that directly links analyst recommendations and proprietary trading. 2 1 Michaely and Womack (1999) document several sources of revenue for investment banks, such as raising bank loans, M&A consulting, investment advice and proprietary trading. 2 Irvine (2004) demonstrates that sell-side analysts have greater economic incentives to bias their recommendations than their earnings forecasts. 2

3 This paper also examines the characteristics of conflicts of interest for investment banks. Previous studies have shown that conflicts of interest are influenced by characteristics of investment banks. However, Ljungqvist et al. (2007) argues that the analyst is often under pressure to help generate brokerage revenue. Other studies, however, also show that analyst characteristics have a differential impact on stock prices (e.g., Stickel 1992; Mikhail et al. 1997; Fang and Yasuda 2009; and Loh and Stulz 2011). Therefore, conflicts of interest for investment banks are expected to be affected not only by firm characteristics but also by analyst characteristics. This study further examines whether analyst characteristics influence conflicts of interest for investment banks. Finally, we focus on conflicts of interest for individual stocks. Shen and Chih (2009) document conflicts of interest for individual stocks as affected by the characteristics of individual firms. Fan and Wong (2002) report that concentrated ownership may result in low earnings informativeness, implying information ambiguity on the part of the analyst, thus decreasing the likelihood of conflicts of interest on recommendations. The results show that both traditional and revise analyst recommendations have an impact on recommended stocks. Moreover, negative recommendations have a much greater impact than positive recommendations on stock returns. Investment banks are found to have conflicts of interest in upgrade events, suggesting that high-performance analysts are more concerned with enhancing their reputations and tend not to engage in conflicts of interest when making upgrade recommendation. The results also suggest that conflicts of interest can be reduced by concentrated ownership of individual stocks. The rest of this paper is structured as follows. Section 2 presents a literature review and proposes hypotheses. Section 3 reviews the data, discusses data constructs, defines the measures used and outlines the methodology. Section 4 presents empirical results. Conclusions are presented in Section 5. 3

4 2. Literature review and hypotheses Many recent studies have highlighted potential conflicts of interest among sell-side security analysts and investment banks. On the one hand, sell-side analyst reports tend to be overly optimistic or pessimistic for stocks with specific characteristics. For instance, Dugar and Nathan (1995) showed that analysts tend to be overly optimistic for companies that are customers of the analyst s firm. 3 Ljungqvist et al. (2007) argues that investors are not the analyst s only constituency, but also consider the interests of the firm being assessed because the analyst is often under pressure to assist in the production of brokerage revenue. On the other hand, investment banks are concerned with maintaining their client relationships. Mehran and Stulz (2007) argue that a conflict of interest could arise if a financial institution were to obtain a direct advantage, such as additional underwriting fees. Bolton et al. (2007) showed that a specialist financial institution can differentiate itself and acquire market share by providing credible information. Recent studies have found that conflicts of interest occur in investment banks not only because of the need to service clients but also due to proprietary trading activity. Juergens and Lindsey (2009) and Shen and Chih (2009) showed direct links between analyst recommendations and proprietary trading behavior. Juergens and Lindsey (2009) show that market makers employed by the analysts' firms engage in increased trading activity prior to recommendation revisions. Shen and Chih (2009) similarly found that the analyst s firm frequently buys the same recommended stocks through its broker-dealer before the release of the recommendation and sells it after. However, Loh and Stulz (2011) argue that recommendation change events have a more significant impact than traditional recommendation events. Dugar and Nathan (1995) argue 3 Eastwood and Nutt (1999) also finds that analysts both underreact to negative information and overreact to positive information. 4

5 that analysts often produce overly optimistic stock recommendations, and will only issue sell or downgrade recommendations as a result of specifically credible information. That is, negative recommendations are usually based on more substantial information content than positive recommendations. Therefore, this study expects that negative recommendations should have a stronger stock price impact than positive recommendations. Hypothesis 1: Negative recommendations have a stronger stock price impact than positive recommendations. The second research question is based on characteristics of conflicts of interest for investment banks. The literature shows that conflicts of interest for investment banks may be dependent on common characteristics. For instance, Shen and Chih (2009) show that the conflicts of interest of investment banks exhibit common characteristics such as the size and age of the investment bank, and the frequency with which recommendations are issued. However, this explanation focuses on the characteristics of the investment bank. Ljungqvist et al. (2007) argue that analysts are often under pressure to push brokerage revenue. Also, prior work shows that analyst characteristics lead to differential impact of stock recommendation changes. For example, Stickel (1992), Fang and Yasuda (2009) and Loh and Stulz (2011) report that recommendation changes made by star analysts have more impact and profitable. Mikhail et al. (1997) show that analysts improve the accuracy of their earnings forecasts as they accumulate experience. Thus, it is assumed here that conflicts of interest for investment banks are affected by analyst characteristics. Hypothesis 2: The conflicts of interest for investment banks are affected by analyst performance or experience. The last research question deals with characteristics of conflicts of interest for individual stocks. Previous studies show that the magnitude of conflicts of interest for individual stocks is affected by firm characteristics (Shen and Chih, 2009), such as trading volume, firm size, number of peer firms in the same industry, systematic risk and the number of shares held by 5

6 insiders. However, Fan and Wong (2002) argue that the concentrated ownership is associated with low earnings informativeness as ownership concentration prevents leakage of proprietary information. The low earnings informativeness implies that analysts obtain ambiguous information from the firm, thus reducing the likelihood of conflicts of interest in the recommendations. Gomes (2000) reports that a high ownership concentration can also serve as a credible indication that the controlling owner is interested in establishing a reputation for protecting minority shareholders. Gul et al. (2010) argues ownership concentration and foreign shareholding affected information quality. It is thus expected that high ownership concentrations decrease the likelihood of conflicts of interest in analyst recommendations, and thus for individual stocks. Hypothesis 3: Concentrated ownership will reduce the likelihood of conflicts of interests for individual stocks. 3. Data and methodology 3.1 Sample construction Analyst recommendation data was obtained from the Institutional Brokers Estimate System (I/B/E/S) Recommendations Detail International files from April 1995 to October Only recommendations related to the Taiwan market were collected by sorting for Country ID, Exchange ID, Currency, and Home Market Code. Next, trading information and other financial variables for companies making recommendations were obtained from the Taiwan Economic Journal (TEJ) database. Our sample included investment bank recommendations, but the two databases (I/B/E/S and TEJ) use different identifiers for investment banks. A Google search provided the respective identifiers, allowing for the investment bank data from the two databases to be integrated. The final sample contained a total of 631 recommended stocks and 11 investment banks trading 6

7 those stocks. Each of the investment banks can evaluate a given stock one or more times, so the total number of analyst recommendations in the sample was 10, Variable definitions and measures To examine the conflicts of interest, this study makes use of two measures proposed by Shen and Chih (2009). The first identifies the conflicts of interest for investment banks by computing the investment bank s net buy (sell) position for the recommended stocks eight weeks prior and eight weeks following the release of the recommendation. Net traded shares of the upgrade (buy) recommendation are defined as the buy shares minus the sell shares, divided by the mean daily trading volumes of the upgrade (buy) recommendation across the sample period to eliminate the scale problem. In contrast, the net traded shares of the downgrade (sell) recommendation are calculated as the sell shares minus the buy shares, divided by the mean daily trading volumes of the downgrade (sell) recommendation across the sample period to eliminate the scale problem. The following equation is applied: NTS i, τ = Mi Mi j j Buy _ Sharesi, τ Sell _ Sharesi, τ j= 1 j= 1 M i where i represents the i th investment bank issuing the recommendation, τ is the τ th week before or after the release of the report, j denotes the j th recommendation report (1) issued by the i th investment bank, and M i represents the total number of recommendations published by the i th recommending investment bank. The investment banks conflicts of interest are calculated using the net traded shares of all recommended stock recommended by the i th making issued investment bank during the period t to t weeks around the release of the analyst s recommendation. 4 See Table A.1 in the Appendix for more detail. 7

8 1 t CI _ IB = NTS NTS it, i, τ i, τ τ= t τ= 1 (2) Equation (2) calculates the sum of the net traded shares of the i th making issued. investment bank from eight weeks preceding the release of the analyst report and the sum of the net traded shares in the eight weeks following the release. The second part denotes the conflicts of interest for individual stocks, which are computed based on the net traded shares of the k th recommended stock for all investment banks. Similar to Eq. (1), Eq. (3) is therefore as follows: NTS k, τ = Nk Nk l l Buy _ Sharesi, τ Sell _ Sharesi, τ l= 1 l= 1 N k where k represents the k th recommended stock, τ is the τ th week before or after the (3) release of the report, l denotes the l th recommendation report, and N k represents the total number of recommendations issued regarding the k th stock by all recommending investment banks. The conflicts of interest for individual stocks are difference between the net traded shares of k th stock by all investment banks that recommend the k th stock. Equation (4) is therefore as follows: 1 t CI _ IS = NTS NTS kt, k, τ k, τ τ= t τ= 1 (4) Equation (4) calculates the sum of the total number of traded shares of the k th recommended stock eight weeks prior to the release of the analyst report and the sum of the net traded shares eight weeks following the release. This study also examines the profit made by investment banks when trading the recommended stocks. Abnormal returns and cumulative abnormal returns are both calculated based on the standard event study method, which uses the market model of ordinary least 8

9 squares (OLS) regressions for 250 trading days. The abnormal returns and cumulative abnormal returns can then be calculated over 40 trading days around the buy, sell, upgrade and downgrade recommendations. 4. Empirical results 4.1 Recommendations for net traded shares Table 1 presents net traded shares of the recommended stocks by buy and sell around recommendations. Price share statistics are provided for the 11 investment banks and 10,709 recommendations, include mean, cumulative mean, standard deviation, maximum, and minimum across all stocks for each week for a period spanning eight weeks before and after the release of the recommendations. The cumulative mean is calculated beginning in the 8 th week preceding the release of the recommendations. Panel A of Table 1 shows the mean net traded shares by the recommending investment bank around the zero before the release of the buy recommendations, and remains the status until one week after recommendation. The cumulative mean covers the eight weeks prior to the release of the report, and indicates there is no significant evidence of conflicts of interest for investment banks in buy recommendations. In Panel B of Table 1, the mean net traded shares by the recommending investment bank takes place seven weeks before the sell recommendations, and peaks two weeks before the release of the report. The cumulative mean covers the eight weeks preceding the release of the report. However, the mean net traded shares become positive immediately following the publishing of the report, and this remains true until the 5 th week following the release of the report. This result shows that the recommending investment banks become net sellers of the recommended stocks before the publication of their sell recommendations. Moreover, investment banks making a sell recommendation will buy those stocks back later. Normally, when an investment bank issues a sell recommendation, it should mean the bank considers the 9

10 share price to be unjustified in terms of risk or that it is aware of news that will adversely affect the share price. The question is why would an investment bank issue sell recommendations to investors but buyback later? Such behavior may indicate potential conflicts of interest. Insert Table 1 about here. Panel A of Table 2 shows the mean net traded shares by the recommending investment bank before and after the upgrade recommendation, and result is similar to that of a buy recommendation. The mean net traded shares indicate the eight weeks preceding the release of the upgrade recommendation. The results for the cumulative mean are consistent with those for the mean, indicating upgrade recommendations present no significant evidence to suggest the existence of conflicts of interest for investment banks. In Panel B of Table 2, the mean net traded shares by the recommending investment bank from seven weeks preceding the downgrade recommendation, and peaks one week prior to the release of the report. The cumulative mean covers the seven weeks prior to the release of the downgrade recommendation. Interestingly, the mean net traded shares become positive immediately following the publication of the downgrade recommendation, and remains positive for four weeks. This result suggests that the recommending investment bank was a net seller of the recommended stock prior to the publication of the downgrade recommendation. Furthermore, the investment banks which issued the downgrade recommendations later begin to buy these underperforming stocks, again suggesting a conflict of interest on the part of the investment banks. Insert Table 2 about here. 4.2 Benefits of recommendations This section explores whether analyst recommendations produce benefits. Table 3 reports the results of ARs and CARs of the recommended stocks over 40 trading days around the buy 10

11 recommendations. The ARs are positive at 2 trading days before the release of the buy recommendation until at 3 trading days after the recommendation, and become negative after 5 trading days following the recommendation. The highest and lowest returns respectively occur 1 trading day (0.19%) and 12 trading days (-0.21%) after the buy recommendation. Most CARs are significant beginning 1 trading day and ending 7 trading days after the recommendation, and CAR is -0.70% 20 trading days after. Graphs A and B of Fig. 1 plot the ARs and CARs around the buy recommendation. The ARs of Graph A are significant and positive from trading day -4 to +1. From trading day +5, the ARs are negative to the end of the event window. The CARs exhibit a stable trend from days -20 to -4, but change significantly from trading day -3 to +20. The results show that the buy recommendation had an impact on the recommended stock. 5 Insert Table 3 and Figure 1 about here. Table 4 presents the results of the ARs and CARs for the recommended stocks over 40 trading days around the sell recommendations. The ARs are negative starting at trading day -19, and the lowest AR is -0.74% at trading day +2. In Fig. 2, Graphs A and B respectively plot the ARs and CARs. The ARs of Graph A are significant and negative from trading days -2 to +3. In particular, the ARs are still negative and significant at trading day +18 even though recommendation is over three weeks old. The CARs are stable reduction on event windows, and CAR is -5.19% and significant at trading day +20. The results show that negative recommendations have a greater impact than positive recommendations on the recommended stock, which is consistent with hypothesis 1 and the findings of Dugar and Nathan (1995). Insert Table 4 and Figure 2 about here. Table 5 lists the ARs and CARs of the recommended stocks over 40 trading days around 5 The Wilcoxon sign and Wilcoxon sign rank tests are used because Kothari and Warner (1997) found that abnormal returns around firm-specific events are severely misspecified. The rejection frequencies using parametric tests sometimes exceed 30% when the significance level of the test is 5%. Nonparametric tests are likely to reduce misspecification. 11

12 the upgrade recommendations. Most ARs are significant from trading day -2 to +3, and become negative and significant at trading day +10. The highest and lowest returns respectively occur on trading days +1 (0.28%) and +16 (-0.26%). In addition, significant CAR clusters are found from trading days -2 to +17, and CAR is 0.57% at trading day +20. Graphs A and B of Fig. 3 plot the ARs and CARs around the upgrade events. The CARs exhibit a growth trend from trading days -14 to +6, but the CARs significance falls from trading day +7 to +20. These results and trends are similar to those for buy recommendations. Insert Table 5 and Figure 3 about here. Table 6 presents the ARs and CARs of the recommended stocks over a period of 40 trading days around the downgrade recommendations. Most ARs are significant from trading day -5 to +3, and all ARs are negative in the downgrade event window. The highest and lowest returns respectively occur at trading days -3 (-0.09%) and +1 (-0.53%). In particular, significant CAR clusters are found from trading days -19 to +20, and CAR is -5.26% at trading day +20. Graphs A and B of Fig. 4 respectively plot the ARs and CARs from downgrade events. The CARs exhibit a descending trend in the downgrade event windows. These results and trends are similar to those of the sell recommendations. The experimental results are consistent with hypothesis 1 that negative recommendations should have a stronger impact on stock prices than positive recommendations. Insert Table 6 and Figure 4 about here. 4.3 Determinants of conflicts of interest for investment banks Six variables are considered in determining conflicts of interest: the investment bank s size (BrokSize) and age (Duration), the frequency with which the investment bank issues recommendations (Times), the analyst s prior-year performance (Prior_Perf), the average analyst s experience (Experience) and the investment bank s monthly sales figure (Sale). BrokSize measured as the number of analysts employed by analyst s brokerage firm in 12

13 the year prior to issuing the revised recommendation. Duration is computed as the number of years the investment bank has been in business. Shen and Chih (2009) note that BrokSize and Duration should be negative coefficients because larger or older investment banks are more concerned about their reputations. Times is computed as the frequency with which the investment bank issues recommendations. The coefficients should be positive, because high frequency issuing of recommendations provides an increased opportunity for conflicts of interest to develop in their trading strategy. Following by Mikhail et al. (2004), Prior_Perf is the average quintile ranking of the analyst s prior performance (ranging from 0, worst, to 4, best), based on the prior-year return to the analyst s equal-weight portfolio of recommendation revisions over the event window of t= -2 to +60, where t=0 is the recommendation date. Experience is average analyst s experience measured in quarters (Lol and Stulz, 2011). The impacts of Prior_Perf and Experience are uncertain. An increase in the analyst s prior performance or experience should increase the impact on the recommended stock (Stickel, 1992; Kim and Verrecchia, 1994; Mikhail et al. 1997), and could provide increased incentive to include conflicts of interest in their trading strategy. On the other hand, increased prior performance or experience may leave the analyst increasingly concerned about maintaining his reputation, thus leaving the analyst reluctant to engage in conflicts of interest. The final variable is Sale, which is computed as the investment bank s monthly sales, and is expected to be positively correlated to the extent of conflicts of interest in the firm s trading strategy. Table 7 reports the cross-sectional regressions of the conflict of interest indices for investment banks for positive (i.e., buy and upgrade) recommendations. In Panel A of Table 7, the dependent variables are CI_IB i,1 to CI_IB i,4, which denotes a period of 1 week before to 4 weeks after the analyst recommendation. To capture more information and characteristics of the conflicts of interest indices, CI_IB i,8 is also shown. Duration is found to be significantly negative for CI_IB i,1 in buy recommendations, suggesting that longer IB periods are more 13

14 concerned with reputation and tend not to engage in conflicts of interest when making recommendations. In Panel B of Table 7, when CI_IB i,1 is the dependent variable, the coefficients for Duration and Prior_Perf are significantly negative (at the 10% and 5% level, respectively) in upgrade events. When CI_IB i,2 is the dependent variable, the coefficients for Size, Duration and Prior_Perf are significantly negative (at the 5%, 10% and 10% level, respectively). When CI_IB i,3 is the dependent variable, the coefficients for Size and Duration are significantly negative (at the 5% and 10% level, respectively). If we observe CI_IB i,4 or CI_IB i,8 for longer windows, the coefficients for Size or Duration are significantly negative (at the 5% and 10% level, respectively). The result of conflicts of interest indices in upgrade events, suggesting that larger and older investment banks are more concerned with reputation and tend not to engage in conflicts of interest when making upgrade recommendations. Insert Table 7 about here. Table 8 presents the coefficient estimates for the conflicts of interest indices for investment banks for negative (i.e., sell and downgrade) recommendations. Panel A gives result for the sell recommendations. The coefficient of most interest is that for Prior_Perf, which is positive and statistically significant 4 weeks before and after the sell recommendation (CI_IB i,4 ). The positive coefficient implies that senior analysts tend to have conflicts of interest on sell recommendations. Panel B gives the result for the downgrade recommendations. The coefficient for Duration is significantly positive in CI_IB i,4 and CI_IB i,8 (both at the 10% level), indicating that older investment banks tend to have conflicts of interest for downgrade recommendations. The coefficient for Times is significantly positive in CI_IB i,1, CI_IB i,2, CI_IB i,4 and CI_IB i,8 (at the 10%, 5%, 10% and 10% level, respectively), implying that banks which frequently issue downgrade recommendations tend to have conflicts of interest. Also, for CI_IB i,8, the t-value of Sale is significantly positive at the 5% level, indicating that investment banks with higher sales volume tend to have conflicts of 14

15 interest for downgrade recommendations. The experimental results are consistent with hypothesis 2 which states that conflicts of interest for investment banks are affected by analyst characteristics. Insert Table 8 about here. 4.4 Determinants of conflicts of interest for individual stocks Eight variables are adopted to examine the determinants of conflicts of interests for individual stocks: the average daily trading volume (Volume), the market value of the stock (Size), the frequency with which a particular stock is recommended by all investment banks (FREQ), the number of peer firms in the same industry (Number), the systematic risk of a recommended stock (Beta), the market-to-book ratio (MB), the percentage of shares held by directors or senior officers (Insider), and the percentage of shares held by foreign investors (FIH). Volume is measured as the average daily trading volume of the recommended security. Bhushan (1989a) showed that a high trading volume in a particular stock may stem from liquidity traders causing high volatility. Shen and Chih (2009) argued that an investment bank may take advantage of such conditions to earn profits. Hence, Volume is expected to be positively related to conflicts of interest. Size should be negative in sign, Lang and Lundholm (1996) argued that larger firms draw increased attention from investors, and Shen and Chih (2009) noted that, mindful of their reputations, investment banks are more vigilant when making recommendations for high-profile firms. Similar to Times, FREQ should be positive, while Number should be positive. Bhushan (1989b) and Shen and Chih (2009) reported that the cost to analysts for collecting information increases with the number of business lines a firm maintains, and high information collecting costs might reduce competition in information collecting, thus resulting in more informative recommendations and increasing the likelihood of conflicts of 15

16 interest. Beta should be negative, with a higher Beta value implying stock returns are more sensitive to market information, and relatively insensitive to firm-specific information. Thus, Shen and Chih (2009) noted that a higher Beta value makes it more difficult for investment banks to exploit firm-specific information by trading recommended stocks. MB should be positive, with Frankel et al. (2006) noting that a high MB value is often associated with younger firms with high growth potential. Such firms usually lack a financial experience, providing increased likelihood that their operating strategies will be vulnerable to conflicts of interest. The last two variables are Insider and FIH, which are proxy variables for concentrated ownership, and are expected to be negatively related to conflicts of interest. Table 9 presents the cross-sectional regressions of the conflict of interest indices for individual stocks for positive recommendations. Panel A of Table 9 lists the buy recommendations. When CI_IS k,3 and CI_IS k,4 serve as the dependent variables, the coefficient for Insider is significantly negative (at the 1% and 5% level, respectively). On the other hand, FIH is significantly negative for CI_IS k,2 and CI_IS k,8 (at the 10% and 5% level, respectively). These results suggest that when a relatively higher ratio of equity is held by insiders or foreign investors, individual stocks tend to be less vulnerable to lower conflicts of interest on buy recommendations. The experimental results are consistent with hypothesis 3 which states that concentrated ownership of individual stocks reduces the incidence of conflicts of interest. Panel B of Table 9 lists the upgrade recommendations. The coefficients for Size are significantly negative at CI_IS k,3 and CI_IS k,8 (at the 10% and 5% level, respectively), indicating that larger firms tend to have fewer conflicts of interest on upgrade recommendations. The coefficient for FREQ is significantly positive at CI_IB i,8 (at the 10% level), implying that higher frequency of upgrade recommendations tends to correlate with higher incidence of conflicts of interest. This result indicates that investment banks have more difficulty exploiting firm-specific information if the recommended stock has higher 16

17 information transparency. Insert Table 9 about here. Table 10 presents the cross-sectional regressions of the conflict of interest indices for individual stocks for negative recommendations, with results for sell and downgrade recommendations respectively displayed in Panels A and B. In Panel A, the conflicts of interest for sell recommendations are almost insignificant in all coefficients. In Panel B, Number is significant and positive in CI_IB i,1 (at the 5% level), implying that a greater number of peer firms in same industry tends to correlate with increased conflicts of interest. Beta is significant and negative from CI_IS k,1 to CI_IS k,8, (at the 1%, 5%, 5%, 5% and 10% level, respectively), suggesting that higher systematic risk makes it more difficult for investment banks build a strategy to profit from such recommendations. Judging from the above, sell recommendations appear not to be influenced by the explanatory variables. However, this view is quite unsatisfactory, and the above results are not absolute proof that stock characteristics have no impact on conflicts of interest. One possible explanation for these results is that the majority of conflicts of interest are not affected by the explanatory variables. Insert Table 10 about here. 5. Conclusions This paper studies the characteristics of conflicts of interest in stock recommendations through asking three questions: Are stock prices more heavily impacted by negative recommendations than by positive recommendations? Are conflicts of interest for investment banks affected by analyst characteristics? Will concentrated ownership reduce conflicts of interest for individual stocks? Empirical results show that individual stocks are influenced by analyst recommendations, and this holds true for both traditional (buy and sell) recommendations and revised (upgrade 17

18 and downgrade) recommendations. Moreover, negative recommendations were found to have a much greater impact on recommended stocks than positive recommendations. This paper also uses the model proposed by Shen and Chih (2009) to examine the characteristics of conflicts of interest in stock recommendations for investment banks and for individual stocks. Variables found to impact conflicts of interests for investment banks include the age and size of the investment bank, the analyst s prior-year performance, and investment bank sales volume. More specifically, the result for conflict of interest indices in upgrade events suggest that larger and older investment banks, along with high-performing analysts, are more concerned with maintaining their reputations and tend not to engage in conflicts of interest when making upgrade recommendations. On the other hand, for individual stocks, incidence of conflict of interest is affected by firm size, frequency of recommendations, the number of peer firms in the same industry, Beta, insider holdings, and foreign investor holdings. These results suggest that increased insider or foreign holdings tend to reduce the incidence of conflicts of interest on buy recommendations. 18

19 References Barber, B., Lehavy, R., McNichols, M., Trueman, B., Can investors profit from the prophets: security analyst recommendations and stock returns. Journal of Finance 55, Bhushan, R., (1989a). Collection of Information about Publicly Traded Firms: Theory and Evidence. Journal of Accounting and Economics 11, Bhushan, R., (1989b). Firm Characteristics and Analyst Following. Journal of Accounting and Economics 11, Bolton, P., Freixas, X., Shapiro, J., Conflicts of Interest, Information Provision, and Competition in the Financial Services Industry. Journal of Financial Economics 85, Dugar, A., Nathan, S., The effects of investment banking relationships on financial analysts earnings forecasts and investment recommendations. Contemporary Accounting Research 12, Eastwood, J.C., Nutt S.R., Inefficiency in Analysts Earnings Forecasts: Systematic Misreaction or Systematic Optimism. Journal of Finance 54, Fan, J.P.H., Wong, T.J., Corporate ownership structure and the informativeness of accounting earnings in East Asia. Journal of Accounting and Economics 33, Fang, L., Yasuda, A., The effectiveness of reputation as a disciplinary mechanism in sell-side research. Review of Financial Studies 22, Frankel, R., Kothari S.P., Weber. J., Determinants of the Informativeness of Analyst Research. Journal of Accounting and Economics 41, Gomes, A., Going public without governance: managerial reputation effects. Journal of Finance 55, Gul F.A., Kim, J.B., Qiu, A., Ownership concentration, foreign shareholding, audit quality, and stock price synchronicity: Evidence from China. Journal of Financial Economics 95, Irvine, P., Analysts' forecasts and brokerage-firm trading. The Accounting Review 79, Irvine, P., Lipson, M., Puckett, A., Tipping. Review of Financial Studies 20, Jegadeesh, N., Kim J., Krische S., Lee C., Analyzing the analysts: When do recommendations add value? Journal of Finance 59, Juergens, J., Lindsey, L., Getting out early: an analysis of market making activity at the recommending analyst s firm. Journal of Finance 64, Kim, O., Verrecchia, R.E., Market liquidity and volume around earnings announcements. Journal of Accounting and Economics 17, Kothari, S.P., Jerold B. Warner, Measuring Long-horizon Security Performance,Journal of Financial Economics 43,

20 Lang, M.H., Lundholm R.J Corporate Disclosure Policy and Analyst Behavior. The Accounting Review 71, Ljungqvist, A., Marston, F., Starks, L.T., Wei, K., Yan, H., Conflicts of interest in sell-side research and the moderating role of institutional investors. Journal of Financial Economics 85, Ljungqvist, A., Marston, F., Wilhelm. W., Competing for securities underwriting mandates: banking relationships and analyst recommendations. Journal of Finance 61, Loh, R.K. Stulz, R.M., When are analyst recommendation changes influential? Review of Financial Studies 24, Mehran, H., Stulz, R.M., The economics of conflicts of interest in financial institutions. Journal of Financial Economics 85, Michaely, R., and Womack, K., Conflict of Interest and the Credibility of Underwriter Analyst Recommendations. Review of Financial Studies 12, Mikhail, M., Walther, B., Willis, R., Do Security Analysts Improve Their Performance with Experience? Journal of Accounting Research 35, Mikhail, M., Walther, B., Willis, R., Do security analysts exhibit persistent differences in stock picking ability? Journal of Financial Economics 74, Shen, C.H., Chih, H.L., Conflicts of interest in the stock recommendations of investment banks and their determinants. Journal of Financial and Quantitative Analysis 44, Stickel, S.E., Reputation and Performance Among Security Analysts. Journal of Finance 47, Womack, K., Do brokerage analysts' recommendations have investment value? Journal of Finance 51,

21 Table 1. Net Traded Shares of Recommended Stocks around the Release of Buy and Sell Recommendations Net traded shares (NTS) of buy recommendations (Panel A) and sell recommendations (Panel B) from investment banks. The cumulative mean at the τth week is calculated by summing the means over the τth week and all preceding weeks. Following Shen and Chih (2009), the NTS value is divided by the average daily trading volume across the sample period (April, 1995 to October, 2013) to eliminate the scale effect Panel A. Buy Recommendations (1995/4-2013/10) Mean 0.000% 0.000% % % 0.000% 0.000% % 0.000% 0.003% 0.000% 0.000% 0.000% % 0.000% 0.000% 0.000% Cum. Mean 0.000% 0.000% % % % % % % 0.000% 0.000% 0.000% 0.000% % % % % Std. dev % 0.019% 0.024% 0.019% 0.026% 0.071% 0.025% 0.017% 0.019% 0.044% 0.030% 0.026% 0.029% 0.017% 0.020% 0.013% Max 2.532% 0.622% 0.864% 0.516% 0.994% 2.830% 0.606% 0.524% 0.896% 0.473% 0.942% 1.365% 0.257% 0.547% 0.383% 0.240% Min % % % % % % % % % % % % % % % % No. of obs Panel B. Sell Recommendations (1995/4-2013/10) Mean % % % % 0.000% % % % 0.008% 0.009% 0.004% 0.001% % 0.001% 0.003% % Cum. Mean % % % % % % % % % % % % % % % % Std. dev % 0.166% 0.155% 0.131% 0.118% 0.107% 0.452% 0.129% 0.321% 0.219% 0.214% 0.196% 0.113% 0.096% 0.099% 0.111% Max 2.632% 0.692% 0.500% 1.306% 1.306% 0.713% 0.511% 0.469% 3.242% 2.963% 2.729% 3.242% 0.228% 1.107% 1.117% 0.756% Min % % % % % % % % % % % % % % % % No. of obs

22 Table 2. Net Traded Shares of Recommended Stocks around the Release of Upgrade and Downgrade Recommendations Net traded shares (NTS) of upgrade recommendations (Panel A) and downgrade recommendations (Panel B) from investment banks. The cumulative mean at the τth week is calculated by summing the means over the τth week and all preceding weeks. Following Shen and Chih (2009), the NTS values are divided by the average daily trading volumes across the sample period (April, 1995 to October, 2013) to eliminate the scale effect Panel A. Upgrade Recommendations (1995/4-2013/10) Mean % % % % 0.000% 0.001% % % 0.005% 0.001% 0.003% 0.000% % % 0.000% 0.000% Cum. Mean % % % % % % % % 0.001% 0.002% 0.005% 0.005% 0.004% 0.003% 0.003% 0.003% Std. dev % 0.036% 0.119% 0.051% 0.037% 0.119% 0.048% 0.031% 0.036% 0.042% 0.059% 0.032% 0.045% 0.038% 0.037% 0.032% Max 0.649% 0.684% 5.726% 0.783% 0.806% 3.113% 1.081% 0.647% 0.953% 1.130% 1.565% 1.201% 0.836% 0.499% 0.641% 0.433% Min % % % % % % % % % % % % % % % % No. of obs Panel B. Downgrade Recommendations (1995/4-2013/10) Mean 0.000% % % 0.000% 0.000% 0.001% % % 0.004% 0.001% 0.001% % % % 0.000% 0.000% Cum. Mean 0.000% % % % % % % % % % % % % % % % Std. dev % 0.066% 0.046% 0.040% 0.039% 0.085% 0.026% 0.025% 0.032% 0.042% 0.045% 0.020% 0.026% 0.032% 0.033% 0.020% Max 2.830% 0.622% 0.148% 1.365% 0.733% 2.830% 0.246% 0.531% 0.451% 1.118% 1.247% 0.434% 0.193% 0.397% 0.510% 0.345% Min % % % % % % % % % % % % % % % % No. of obs

23 Table 3. Abnormal Returns and Cumulative Abnormal Returns of Recommended Firms over 40 Days around Buy Recommendations Abnormal return response to buy recommendations by investment banks for 40 days around the recommendation date. Cross-sectional t-tests, non-parametric Wilcoxon sign tests (Sign) and Wilcoxon sign rank tests (Sign rank) are applied to determine whether AR or CAR are significantly smaller than zero. Significance at the 1, 5, and 10% levels is denoted by ***, **, and *, respectively. AR CAR t-test Sign Sign rank (%) (%) t-test Sign Sign rank % *** 0.03% *** %*** * 0.13%** * % *** *** 0.08% % *** *** 0.02% ** %*** *** *** -0.08% ** %** *** *** -0.18%* *** ** % ** -0.16% %** *** -0.06% ** % *** * -0.02% % *** * 0.04% % *** ** 0.04% % *** ** 0.08% % *** * 0.11% * % *** *** 0.08% * % *** *** 0.03% * % *** 0.06% ** % *** ** 0.05% * % *** *** 0.04% ** %*** *** 0.18% ** %** *** 0.27% ** *** %*** ** 0.45%* *** *** %*** *** 0.62%*** *** *** %** *** 0.72%*** *** *** % *** ** 0.78%*** *** *** % *** ** 0.77%*** *** *** %** *** *** 0.66%** *** *** % *** *** 0.62%** *** *** %** *** *** 0.53%* *** *** %** *** *** 0.44% *** *** %* *** *** 0.37% *** *** % *** *** 0.33% *** *** %*** *** *** 0.12% *** *** %*** *** *** 0.00% *** *** %*** *** *** -0.14% ** *** %*** *** *** -0.27% ** ** %*** *** *** -0.39% * % *** *** -0.44% * * % *** *** -0.50% ** * %* *** *** -0.56% ** %*** *** *** -0.70%* * 23

24 Table 4. Abnormal Returns and Cumulative Abnormal Returns of Recommended Firms over 40 Days around Sell Recommendations Abnormal return response to buy recommendations by investment banks for 40 days around the recommendation date. Cross-sectional t-tests, non-parametric Wilcoxon sign tests (Sign) and Wilcoxon sign rank tests (Sign rank) are applied to determine whether AR or CAR are significantly smaller than zero. Significance at the 1, 5, and 10% levels is denoted by ***, **, and *, respectively. AR CAR t-test Sign Sign rank (%) (%) t-test Sign Sign rank % 0.13% %* *** *** -0.04% %** *** ** -0.23% *** ** % *** ** -0.37%* *** *** % -0.30% *** *** % -0.19% *** ** % -0.14% ** * % -0.18% *** ** % -0.18% *** ** % ** -0.24% ** ** % * -0.33% *** *** %** ** -0.53% *** *** % -0.43% *** *** %*** *** *** -0.80%* *** *** % ** ** -0.86%* *** *** % ** * -0.93%** *** *** % * -1.09%** *** *** % -1.18%** *** *** %** *** *** -1.40%*** *** *** %* * ** -1.61%*** *** *** %*** *** *** -2.05%*** *** *** %*** *** *** -2.79%*** *** *** %*** *** *** -3.25%*** *** *** % -3.42%*** *** *** % -3.40%*** *** *** %*** *** *** -3.70%*** *** *** %** *** *** -3.92%*** *** *** % -3.95%*** *** *** % -3.84%*** *** *** %*** ** *** -4.18%*** *** *** % *** ** -4.24%*** *** *** % -4.16%*** *** *** % -4.16%*** *** *** % -4.30%*** *** *** %* ** ** -4.48%*** *** *** %*** *** *** -4.84%*** *** *** % ** * -4.97%*** *** *** %* * ** -5.16%*** *** *** % -5.10%*** *** *** % -5.19%*** *** *** 24

25 Table 5. Abnormal Returns and Cumulative Abnormal Returns of Recommended Firms over 40 Days around the Upgrade Recommendations Abnormal return response to upgrade recommendations by investment banks for 40 days around the recommendation date. Cross-sectional t-tests, non-parametric Wilcoxon sign tests (Sign) and Wilcoxon sign rank tests (Sign rank) are applied to determine whether AR or CAR are significantly smaller than zero. Significance at the 1, 5, and 10% levels is denoted by ***, **, and *, respectively. AR CAR t-test Sign Sign rank (%) (%) t-test Sign Sign rank % 0.05% % * 0.14%* % ** 0.11% % *** *** 0.05% % *** *** 0.01% % *** *** -0.06% % ** -0.04% %*** 0.10% % *** 0.13% % *** ** 0.09% % * 0.11% % *** ** 0.09% %** ** 0.21% % ** 0.27% % *** 0.30% * %* 0.40%* ** % 0.44%* * % ** 0.51%** ** %** 0.64%** * *** %** 0.76%*** ** *** %*** *** 1.05%*** *** *** %*** ** 1.27%*** *** *** %*** 1.46%*** *** *** % *** * 1.48%*** *** *** % ** 1.52%*** *** *** % * 1.53%*** *** *** % *** * 1.52%*** *** *** % *** ** 1.47%*** *** *** % *** ** 1.42%*** *** *** %** *** *** 1.30%*** *** *** % *** * 1.26%*** *** *** %* *** *** 1.16%*** *** *** %* *** *** 1.07%*** ** *** % *** *** 1.01%*** *** *** % ** * 0.97%** *** *** %*** *** *** 0.71%* *** % *** * 0.70%* ** *** % 0.71%* *** *** % *** *** 0.62% *** *** % *** * 0.57% *** ** 25

26 Table 6. Abnormal Returns and Cumulative Abnormal Returns of Recommended Firms over 40 Days around Downgrade Recommendations Abnormal return response to downgrade recommendations by investment banks for 40 days around the recommendation date. Cross-sectional t-tests, non-parametric Wilcoxon sign tests (Sign) and Wilcoxon sign rank tests (Sign rank) are applied to determine whether AR or CAR are significantly smaller than zero. Significance at the 1, 5, and 10% levels is denoted by ***, **, and *, respectively. AR CAR t-test Sign Sign rank (%) (%) t-test Sign Sign rank % ** -0.01% ** %*** *** *** -0.17%** *** *** %*** *** *** -0.31%*** *** *** % *** *** -0.39%*** *** *** % *** *** -0.44%*** *** *** % *** ** -0.46%*** *** *** % *** ** -0.44%*** *** *** %* *** *** -0.53%*** *** *** % *** ** -0.58%*** *** *** % *** *** -0.66%*** *** *** % *** *** -0.74%*** *** *** %*** *** *** -0.91%*** *** *** % *** *** -0.99%*** *** *** %** *** *** -1.11%*** *** *** % *** *** -1.18%*** *** *** %** *** *** -1.31%*** *** *** %*** *** *** -1.47%*** *** *** %* *** *** -1.56%*** *** *** %*** *** *** -1.70%*** *** *** %*** *** *** -1.95%*** *** *** %*** *** *** -2.48%*** *** *** %*** *** *** -2.95%*** *** *** %*** *** *** -3.20%*** *** *** % *** *** -3.29%*** *** *** % *** *** -3.35%*** *** *** %*** *** *** -3.62%*** *** *** %*** *** *** -3.82%*** *** *** %** *** *** -3.95%*** *** *** % *** ** -4.00%*** *** *** %** *** *** -4.12%*** *** *** %*** *** *** -4.27%*** *** *** %* *** *** -4.36%*** *** *** % ** -4.42%*** *** *** % *** *** -4.49%*** *** *** %* *** *** -4.59%*** *** *** %*** *** *** -4.79%*** *** *** %** *** *** -4.89%*** *** *** %** *** *** -4.99%*** *** *** %** *** *** -5.11%*** *** *** %*** *** *** -5.26%*** *** *** 26