Stock price synchronicity and the role of analyst: Do analysts generate firm-specific vs. market-wide information?

Stock price synchronicity and the role of analyst: Do analysts generate firm-specific vs. market-wide information? Yongsik Kim * Abstract This paper provides empirical evidence that analysts generate firm-specific information. While empirical findings in previous literatures show the positive relation between the stock price synchronicity and analyst coverage, we find that the positive relation might come from the interaction effect between the cyclicality of firm performance and analyst activity. Motivated by the model in Jin and Myers (2006) that outside investors have limited information and consequently inside managers have an incentive to capture a part of the firm s operating cash flow, we find the role of analyst in uncovering the captured cash flow captured by inside managers. After controlling for this interaction effect between analyst activity and cyclicality of firm performance, stock return synchronicity decreases as analyst coverage increases. Furthermore, when the analysts forecast dispersion is high, both effects that stock return synchronicity decreases with analyst coverage and increases with interaction between analyst and cyclicality are reduced. JEL classification: G14. Keywords: Stock price synchronicity; Analyst coverage; Information efficiency * Securities Derivatives R&D Center, Korea Exchange, 76 Yeouinaru-ro, Yeongdeungpo-gu, Seoul, Korea, 07329, Tel.: +82 2-3774-9725; Email: yongkim@krx.co.kr

1. Introduction Many previous literatures have studied the relation between stock price synchronicity and analyst coverage. Some studies argue that the analysts generate market-wide information and show that the stock price synchronicity is positively related with the analyst coverage (Piotroski and Roulstone, 2004; Chan and Hameed, 2006). The other studies also argue that the analysts produce firm-specific information using theoretical models (Admatia and Pfleiderer, 1986; Diamond and Verrecchia, 1981; Grossman and Stiglitz, 1980) and provide evidence that the stock price synchronicity is negatively related with analyst coverage (Brown, 1978; O Brien, 1988; O Brien and Bhushan, 1990). Based on these conflict empirical findings, we investigate the relation between the stock price synchronicity and the analyst coverage and provide evidence that analyst generate firm-specific information. We also find that the positive relation between the stock price synchronicity and analyst coverage in previous studies might come from the interaction effect between the cyclicality of firm performance and analyst activity. Our paper is motivated by the model in Jin and Myers (2006) that outside investors have limited information and consequently inside managers have an incentive to capture a part of the firm s operating cash flow. The amount of captured cash flow depends on the hidden information. More specifically, inside mangers would capture more (less) cash flows when the hidden firm-specific information is positive (negative). Based on this model, we find the role of analyst in uncovering the firm-specific cash flow captured by inside managers. For example, if analysts uncover the captured cash flow of procyclical firms, new earning information will be aligned with the market movement and consequently the synchronicity will increase. After controlling for this interaction effect between analyst activity and cyclicality of firm performance, we find that stock return synchronicity decreases as analyst coverage increases. Furthermore, we find that both effects that stock return synchronicity decreases with analyst coverage and increases with interaction between analyst and cyclicality are reduced when the analysts forecast dispersion is high.

2. Empirical methodology and results 2.1 Data and variables Our sample includes all ordinary common shares (codes 10 and 11) on the NYSE and AMEX. We use data for stock prices, trading number, shares outstanding and market index from the Center for Research in Securities Prices (CRSP) database and accounting data from COMPUSTAT. We also obtain the number of analysts and forecasts data for each company in the Institutional Brokers Estimate (I/B/E/S). Our sample covers the period from 1986 to 2015. To measure the stock price synchronicity (Synch), we follow Morck et al. (2000). We first regress individual firms weekly stock returns on market index return at the end of every year, and then Synch is calculated as, Synch i,t = log ( R2 1 R2), (1) where R 2 is the coefficient of determination for firm i in year t. Since Synch increases as R 2 increases, a firm with high Synch indicates that it is highly correlated with the market. For the explanatory variables, we include analyst coverage, firm size, trading volume, and firm age following Chan and Hameed (2006) and Dasgupta et al. (2010). Chan and Hameed (2006) show that stock price synchronicity is positively related with firm size, trading volume. Dasgupta et al. (2010) also find that firm age is positively associated with stock price synchronicity because the market learns more about firm s time-invariant intrinsic quality. Since previous studies have mentioned that Synch is endogenous with Analyst, we also consider simultaneous equation model. Chan and Hameed (2006) report that stock return volatility is significantly explain the variation of the analyst coverage as well as firm size, trading volume, and stock price synchronicity. The analyst coverage (Analyst) is defined as the natural log of one plus number of analysts for individual firm. We count the number of analysts as the number of earnings forecasts during a given calendar year. Note that we set zero for the number of analysts, if the firms do not have earnings

forecasts. As we previously discussed, we mainly investigate the effect of Analyst on Synch after controlling for the interaction influence between the analyst coverage and the cyclicality of firm performance. Given that stock prices immediately reflect information about the firm performance in the market, we use market beta (Beta) as a proxy for the cyclicality of firm performance. Since we would like to measure relative degree of cyclicality effect among firms rather than just try to figure out how much the individual firm responds to the market, we adopt post-ranking beta instead of preranking beta. To obtain Beta, we follow the constructing procedure suggested by Fama and French (1992). First, we estimated pre-ranking betas on 60 monthly returns (minimum 24 monthly returns) in June of year y and we double sort the individual stocks into the deciles of pre-ranking beta and size using NYSE break points. We then calculate the post-ranking monthly returns of the 10-by-10 portfolios for the next 12 months, from July of year y to June of year y+1. We then estimate postranking betas for these portfolios using the full sample period with the CRSP value-weighted portfolio market index. The β effect on R 2 suggested by Dasgupta et. al (2010) is another reason that we include Beta as well as interaction term in the estimation model. For firm size (Size), we use the firm s market capitalization and trading volume (Volume) is the number of shares traded relative to the number of shares outstanding in natural log form, respectively. While Chan and Hameed (2006) include the log of trading number in millions of shares as a proxy for Volume, we use share turnover ratio due to the high correlation with firm size. 1 Firm age (Age) is the number of months since initial public offering (IPO) in natural log form. 2.2 Empirical results In this section, we mainly investigate the effect of analyst coverage and interaction influence between analyst coverage and cyclicality of firm performance. Table 1 provides the descriptive 1 Chan and Hameed (2006) also point out the high correlation problem between firm size and trading number. They show that the main results are robust even if firm size is excluded.

statistics of the independent variables. Panel A reports the summary statistics, including means and standard deviations, as well as the 5 th, 25 th, 50 th, 75 th, and 95 th percentile values. The mean and median value of the number of analyst is 5.96 and 3.0, respectively and the largest number is 50. Since we include the firms with zero analysts, the 5 th and 25 th percentile value is zero. Panel B reports the number and age of our sample firms by year. The average age of firms has generally increased over the past 30 years from 204.2 to 288.8 months. As of 2015, 29.6% and 13.8% of sample firms that have lasted less than 10 years and more than 50 years, respectively. We also report the correlation matrix of explanatory variables in Panel C. The estimated correlation coefficient between Volume and Size is 0.452 that is relatively low. As we previously discussed, we use share turnover ratio as a proxy for Volume due to the high correlation. The correlation coefficient between trading number and firm size is 0.794. For Analyst and Size, the correlation coefficient is relatively high, 0.635, which is consistent with Piotroski and Roulstone (2004) and Chan and Hameed (2006). They mention that the demand for analyst service increases as firm size increases because private information about a larger firm is more valuable and there are likely to be more shareholders in larger firms. 2.2.1 Effect of firms cyclicality and analyst activity In this section, we first replicate Chan and Hameed (2006) with US stock data, and then we reinvestigate the effect of analyst coverage on stock price synchronicity after controlling for the interaction influence between the cyclicality of firm performance and analyst coverage. We set the first GMM estimation model for firm i in year t as follows: Synch i,t = α + β 1 Analyst i,t + β 2 Size i,t + β 3 Volume i,t + λ ExcDum i,t + 29 79 l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (2) where Synch i,t is the stock price synchronicity for firm i in year t estimated based on equation (1), Analyst i,t is the analyst coverage. Size and Volume is firm size and turnover ratio in log format and ExcDum i,t, YearDum i,t, and SICDum i,t are dummy variables controlling for exchange, year, and

industry effects, respectively. All standard errors are clustered at an industry level. We report the GMM estimation results of equation (2) in Model 1 of Table 2. Consistent with Piotroski and Roulstone (2004) and Chan and Hameed (2006), the coefficient on Analyst is positive and significant at the one percent level. This result supports the hypothesis of the previous literatures that analyst generates market-wide information. The coefficient on Size is also positive and significant at the one percent level that is similar with the empirical finding of Piotroski and Roulstone (2004). While Chan and Hameed (2006) also predict the positive sign on firm size because the stock market indices are value weighted, their empirical results using international data show insignificant negative coefficient on firm size. The coefficient on Volume is significantly positive supporting the hypothesis that greater trading activity increases the speed of price adjustments. We re-implement the GMM estimation by additionally including Age and Beta following Dasgupta et al. (2010) as follows: Synch i,t = α + β 1 Analyst i,t + β 2 Size i,t + β 3 Volume i,t + β 4 Age i,t + β 5 Beta i,t + 29 79 λ ExcDum i,t + l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (3) Age i,t and Beta i,t are firm age and post-ranking beta for firm i in year t, respectively. We report the estimation results in Model 2. The coefficients on Age and Beta are positive and also significant that is consistent with the findings of Dasgupta et al. (2010). The positive relation between Synch and Analyst is still significant at one percent level event after controlling for the effect of Age and Beta. The results reported in Model 1 and Model 2 are replication of Chan and Hameed (2006) using US stock data with additional control variables. Based on these findings, we re-examine the relation between stock price synchronicity and analyst coverage after controlling for the interaction influence between the cyclicality of firm performance and analyst coverage. We set the estimation model as follows: Synch i,t = α + β 1 Analyst i,t + β 2 Size i,t + β 3 Volume i,t + β 4 Age i,t + β 5 Beta i,t + β 6 29 Analyst i,t Beta i,t + λ ExcDum i,t + l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (4) 79

where Analyst i,t Beta i,t is interaction term between Analyst and Beta for firm i in year t. The estimation results are reported in Model 3 and Model 4. The coefficient on the interaction term is positive and significant at the one percent level that supports our hypothesis that analysts uncover the firm s real cash flow that inside manager may want capture (Jin and Myers, 2006). If analysts uncover the captured cash flow of procyclical firms, the information of uncovered cash flow is likely to be aligned with the market. Therefore the more analysts follow firms that have high cyclicality, the higher stock price synchronicity will be estimated, while analysts gather firm-specific information and disseminate it to public. The sign of estimated coefficient on Analyst changes from positive to negative after controlling the effect of interaction between cyclicality of firm performance and analyst coverage. This result also supports the possibility that the positive relation between the stock price synchronicity and analyst coverage come from the interaction effect. 2.2.2 Effect of earnings forecasts dispersion Higher dispersion of earnings forecasts indicates less agreement about future earnings among analysts. This implies that investors would have uncertainty about firms cash flow uncovered by analysts, therefore the effect of interaction between the cyclicality of firm performance and analyst activity as well as analyst coverage itself on stock price synchronicity would be reduced. 2 The earnings forecasts dispersion is calculated as the standard deviation of analysts forecasts normalized by the mean value of forecasts, and divided by the square root of the number of analysts following Jin and Myers (2006). We then set dispersion dummy variable (DispersDum) equal to one if the calculated dispersion is higher than the average in the same year. Firms with zero analyst coverage or dispersion are excluded. We modify the GMM estimation model (4) by including an interaction term between analyst coverage and dispersion dummy variable (Analyst DisperDum) and a triple interaction term involving analyst coverage, cyclicality of firm performance, and dispersion dummy 2 Chan and Hameed (2006) show that the dispersion of earning forecasts make the effect of analyst coverage unclear, that is, it lessens the stock price synchronicity under their hypothesis.

(Analyst Beta DisperDum) as follows: Synch i,t = α + β 1 Analyst i,t + β 2 Analyst i,t DisperDum i,t + β 3 Size i,t + β 4 Volume i,t + β 5 Age i,t + β 6 Beta i,t + β 7 Analyst i,t Beta i,t + β 8 Analyst i,t Beta i,t DisperDum i,t + 29 79 λ ExcDum i,t + l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (5). We show the estimation results of equation (5) in Table 3. Model 1 includes Analyst DisperDum term only and Model 2 includes both of interaction terms, Analyst DisperDum and Analyst Beta DisperDum. The estimated coefficient on Analyst DisperDum in Model 1 is positive and significant at the one percent level and the significance of the estimated coefficient on Analyst and Analyst Beta is still significant and the signs remain unchanged. In Model 2, we find the coefficient on Analyst Beta DisperDum is negative and significant at the one percent level, while the coefficients on other variables are still significant. These results support our hypothesis that investors will have uncertainty about uncovered cash flow when there is less agreement about earnings forecasts among analysts and consequently the effects of analyst activity and interaction influence between firm s cyclicality and analyst activity would be reduced. 2.3 Robustness check In this section, we verify the robustness of the estimation results reported in previous section by adopting alternative market index and implementing the regression simultaneously. 2.3.1 Alternative market index We first re-calculate the stock price synchronicity based on equation (1) using S&P 500 index instead of CRSP value-weighted index and then implement similar GMM estimation. Table 4 shows the re-estimation results of equation (4) and (5) with the re-estimated stock price synchronicity

(Synch_SnP). All the estimated results are consistent with Table 2 and 3 that the coefficients on Analyst are all significantly significant and those on Analyst Beta are positive and significant. We also find that the dispersion of earnings forecasts make the effect of analyst unclear. The results in Table 4 confirm that the analyst coverages generate firm-specific information and the interaction effect between firms procyclicality and analyst activity is aligned to the market movement. 2.3.2 Simultaneous equation model Since previous studies have mentioned that the stock price synchronicity is endogenous with the analyst coverage, the estimated coefficients reported in Table 2 4 likely to be biased. We therefore set the simultaneous equation model using Synch and Analyst as endogenous variables. The equation for Analyst is defined as follows following Chan and Hameed (2006): Analyst i,t = α + β 1 Synch i,t + β 2 Size i,t + β 3 Volume i,t + β 4 Volatility i,t + λ 29 79 ExcDum i,t + l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (6) where Volatility i,t is the return volatility for firm i in year t. We then estimated the coefficients using three-stage least squares estimation between equation (4) and equation (6), equation (5) and equation (6), respectively. Table 5 shows the estimation results of simultaneous equation model. All the results are robust that the signs of the coefficients under three-stage least squares are same to those under individual GMM estimation. Chan et al. (2013) show that stock price synchronicity affects stock market liquidity. As the analyst coverage and stock market liquidity affect each other simultaneously, we additionally implement the simultaneous regression estimation by modifying the equation (6) and add the equation for the stock market illiquidity as follows: Analyst i,t = α + β 1 Synch i,t + β 2 Size i,t + β 3 Volume i,t + β 4 Volatility i,t + β 5 29 Amihud i,t + λ ExcDum i,t + l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (7) 79

Amihud i,t = α + β 1 Analyst i,t + β 2 Synch i,t + β 3 Size i,t + β 4 Volume i,t + λ 29 79 ExcDum i,t + l=1 δ l YearDum i,t + m=1 φ m SICDum i,t + ε i,t (8) where Amihud i,t is Amihud (2002) illiquidity measure for firm i in year t. We report the results in Table 6. In consistent with the previous results, we confirm that all the signs and significances of the estimated coefficients are robust. 3. Conclusion In this paper, we investigate whether analyst generate firm-specific information or marketwide information. Based on the model suggested by Jin and Myers (2006) that outside investors have limited information and consequently inside managers have an incentive to capture a part of the firm s operating cash flow, we analyze the role of analyst in uncovering the firm-specific cash flow captured by inside managers. In consequence, we find evidence that analyst generate firm-specific information rather than market-wide information. The stock price synchronicity is negatively related with the analyst coverage when we control for the interaction effect between the cyclicality of firm performance and analyst activity. Furthermore, we find that both effects that stock return synchronicity decreases with analyst coverage and increases with interaction between analyst and cyclicality are reduced when the analysts forecast dispersion is high. These results are robust when we calculate stock price synchronicity with alternative market index or regress the estimation model simultaneously to avoid potential endogeneity problem.

References Admati, A., Pfleiderer, P., 1986. A monopolistic market for information. Journal of Economic Theory 39, 400 438 Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series effects. Journal of Financial Markets 5, 31 56 Brown, S.L., 1978. Earnings changes, stock prices, and market efficiency. Journal of Finance 33, 17 28 Chan, K., Hameed, A., 2006. Stock price synchronicity and analyst coverage in emerging markets. Journal of Financial Economics 80, 115 147 Chan, K., Hameed, A., Kang, W., 2013. Stock price synchronicity and liquidity. Journal of Financial Markets 16, 414 438 Dasgupta, S., Gan, J., Gao, N., 2010. Transparency, price informativeness, and stock return synchronicity: Theory and Evidence. Journal of Financial and Quantitative Analysis 45, 1189 1220 Diamond, D., Verrecchia, R., 1981. Information aggregation in a noisy rational expectations economy. Journal of Financial Economics 9, 221 235 Grossman, S., Stiglitz, J., 1980. On the impossibility of informationally efficient markets. American Economic Review 70, 393 408 Jin, L., Myers, S.C., 2006. R 2 around the world: New theory and new tests. Journal of Financial Economics 79, 257 292 Liu, M.H., 2011. Analysts incentives to produce industry-level versus firm-specific information. Journal of Financial and Quantitative Analysis 46(3), 757 784 Morck, R., Yeung, B., Yu, W., 2000. The information content of stock markets: why do emerging

markets have synchronous stock price movements? Journal of Financial Economics 34, 307 344 O Brien, P., 1988. Analysts forecasts as earnings expectations. Journal of Accounting and Exconomics 10, 53 83 O Brien, P., 1990. Forecast accuracy of individual analysts in nine industries. Journal of Accounting Research 28, 286 304 Piotroski, J., Roulstone, D., 2004. The influence of analysts, institutional investors, and insiders on the incorporation of market, industry, and firm-specific information into stock prices. Accounting Review 79, 1119 1151

Table 1. Correlation Matrix Synch Analyst Size Volume Age Volatility Beta Synch 1 Analyst 0.3495 1 Size 0.4709 0.6353 1 Volume 0.3495 0.432 0.4522 1 Age 0.1467 0.1596 0.2273 0.0376 1 Volatility -0.1046-0.0874-0.1396 0.0363-0.054 1 Beta 0.1303 0.0314-0.083 0.2465-0.1659 0.0613 1

Table 2. Effect of firms cyclicality and analyst activity Model 1 Model 2 Model 3 Model 4 Synch Synch Synch Synch Analyst 0.0305*** 0.0318*** -0.348*** -0.0956** (2.677) (2.915) (-8.564) (-2.303) Size 0.346*** 0.358*** 0.362*** 0.359*** (24.51) (28.57) (25.21) (28.14) Volume 0.170*** 0.0927*** 0.119*** 0.0938*** (10.14) (6.231) (7.121) (6.299) Age 0.0905*** 0.0913*** (6.760) (6.840) Beta 0.805*** 0.642*** (15.33) (12.91) Analyst Beta 0.352*** 0.119*** (9.996) (3.189) Constant -7.089*** -8.772*** -7.454*** -8.591*** (-36.81) (-49.51) (-39.48) (-42.61) Observations 56,161 53,510 53,510 53,510 R-squared 0.386 0.399 0.395 0.399 *** p<0.01, ** p<0.05, * p<0.1

Table 3. Effect of earnings forecasts dispersion Model 1 Model 2 Synch Synch Analyst -0.0837** -0.123*** (-1.967) (-2.834) Analyst DispersDum 0.0253*** 0.176*** (2.823) (3.519) Size 0.352*** 0.354*** (26.88) (27.23) Volume 0.0774*** 0.0767*** (5.031) (4.985) Age 0.0806*** 0.0872*** (5.768) (6.776) Beta 0.663*** 0.714*** (12.20) (13.83) Analyst Beta 0.106*** 0.142*** (2.797) (3.646) Analyst Beta DispersDum -0.137*** (-3.324) Constant -0.995*** -8.323*** (-28.37) (-29.69) Observations 48,508 48,508 R-squared 0.340 0.408 *** p<0.01, ** p<0.05, * p<0.1

Table 4. Robustness Check (1) - Alternative market index Model 1 Model 2 Model 3 Model 4 Synch_SnP Synch_SnP Synch_SnP Synch_SnP Analyst -0.348*** -0.0889** -0.0793* -0.125*** (-9.485) (-2.305) (-1.939) (-2.971) Size 0.370*** 0.366*** 0.360*** 0.361*** (27.24) (31.08) (29.50) (29.88) Analyst DispersDum 0.0188** 0.176*** (2.046) (3.702) Volume 0.111*** 0.0859*** 0.0694*** 0.0693*** (6.515) (5.597) (4.421) (4.407) Age 0.0967*** 0.0830*** 0.0834*** (6.472) (5.150) (5.045) Beta 0.658*** 0.666*** 0.694*** (13.60) (12.52) (12.47) Analyst Beta 0.350*** 0.112*** 0.103*** 0.147*** (10.84) (3.243) (2.833) (3.831) Analyst Beta DispersDum -0.140*** (-3.644) Constant -7.659*** -8.835*** -0.151*** 1.311*** (-44.63) (-44.04) (-12.82) (34.47) Observations 53,510 53,510 48,508 48,508 R-squared 0.390 0.394 0.319 0.286 *** p<0.01, ** p<0.05, * p<0.1

Table 5. Robustness Check (2) Three-stage least squares with two simultaneous equation model Model 1 Model 2 Model 3 Synch Synch Synch Analyst -61.27*** -57.42** -46.36*** (-2.835) (-2.545) (-2.733) Analyst DispersDum 0.972** 19.89*** (2.532) (2.698) Size 3.302*** 3.251*** 2.431*** (3.758) (3.297) (3.613) Volume 2.366*** 2.239*** 1.563*** (3.293) (2.899) (3.110) Age 0.198** 0.228*** 0.359*** (2.452) (2.638) (3.368) Beta -64.93*** -63.80** -41.46*** (-2.680) (-2.409) (-2.584) Analyst Beta 48.47*** 45.35** 37.67*** (2.749) (2.473) (2.676) Analyst Beta DispersDum -17.18*** (-2.694) Constant 43.99** 46.45* 23.80* (2.043) (1.877) (1.729) Observations 53,510 48,508 48,508 *** p<0.01, ** p<0.05, * p<0.1

Table 6. Robustness Check (3) Three-stage least squares with three simultaneous equation model Model 1 Model 2 Model 3 Synch Synch Synch Analyst -68.91** -71.05** -57.54** (-2.317) (-2.063) (-2.516) Analyst DispersDum 1.027* 23.54** (1.949) (2.520) Size 2.983*** 3.278** 2.536*** (2.622) (2.334) (2.932) Volume 1.917** 2.031** 1.473*** (2.401) (2.161) (2.634) Age 0.111 0.192* 0.354*** (1.115) (1.770) (2.696) Beta -79.68** -85.91** -56.64** (-2.298) (-2.038) (-2.477) Analyst Beta 56.57** 58.02** 47.91** (2.319) (2.059) (2.514) Analyst Beta X DispersDum -20.37** (-2.521) Constant 65.36** 75.24* 42.40** (2.041) (1.828) (2.057) Observations 50,760 45,961 45,961 *** p<0.01, ** p<0.05, * p<0.1