Insider trading returns and dividend signals

Insider trading returns and dividend signals

International Review of Economics and Finance 20 (2011) 421–429 Contents lists available at ScienceDirect International Review of Economics and Fina...

214KB Sizes 0 Downloads 73 Views

International Review of Economics and Finance 20 (2011) 421–429

Contents lists available at ScienceDirect

International Review of Economics and Finance j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / i r e f

Insider trading returns and dividend signals Louis T.W. Cheng a,⁎, Wallace N. Davidson III b, T.Y. Leung c a b c

School of Accounting and Finance, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Finance Department, Southern Illinois University, Carbondale, IL 62903, USA Department of Accountancy, City University of Hong Kong, Kowloon Tong, Hong Kong

a r t i c l e

i n f o

Article history: Received 12 November 2010 Available online 25 November 2010 JEL classification: D82 G14 M41

Keywords: Dividends Earnings Insider trading Information asymmetry

a b s t r a c t The literature shows that insider trading activities and dividends contain information content and serve as signals to firm value. If insider return is a proxy for information asymmetry, we should expect a positive relation between dividends and insider returns. Using a sample of unambiguous (good and bad) news concerning earnings and dividend announcements from Hong Kong firms, we show that information asymmetry is stronger for bad news firms with insider sales than good news firms with insider purchases. In addition, we improve the methodology of Khang and King [Khang, K., & King, T. H. D. (2006). Does dividend policy relate to cross-sectional variation in information asymmetry? Evidence from returns to insider trades. Financial Management, 35, 71–94] and provide evidence that dividend is a credible signal for measuring information asymmetry. © 2010 Elsevier Inc. All rights reserved.

1. Theoretical background The analysis of insider trading is important on many levels. For example, it conveys insights into the strong form of market efficiency. In addition, insider trading conveys information to the market and may help alleviate information asymmetry. In this role, it signals management and/or board opinions about the future profitability of the firm. Corporate insiders possess information on their firms' latest developments, expected performance, and ability to pay dividend. Insiders likely have information long before it is released to the public. Research has shown that firms often leak forthcoming good news prior to a formal announcement but shy away from leaking bad news until absolutely necessary (Begley & Fischer, 1998; Chambers & Penman, 1984; Givoly & Palmon, 1982). According to the “information model theory”, insider trading should be profitable (Givoly & Palmon, 1985; Jaffe, 1974; Lin & Howe, 1990; Seyhun, 1986). Insiders are expected to possess private and price-sensitive information about firm value. As a result, insider purchases (sales) provide signals of future price increases (decreases), (Fishe & Robe, 2004; Jeng, Metrick, & Zeckhauser, 2003). Early insider trading studies on the US market document significant abnormal returns (Finnerty, 1976; Jaffe, 1974; Seyhun, 1986). More recent studies report a smaller market reaction (Lakonishok & Lee, 2001) which may be due to the imposition of stricter insider trading rules. Studies in other markets such as the UK (Gregory, Matatko, & Tonks, 1997; Pope, Morris, & Peel, 1990) and Canada (Lee & Bishara, 1989) also report varying degrees of profitability for insiders. However, Eckbo and Smith (1998) find zero (or negative) returns for insiders in the Oslo Stock Exchange but conclude that the performance of the insiders is conditional on the evaluation approach. Recently, Daher and Mirman (2007) and Wang, Wang, and Ren (2009) extend the theoretical model from Jain and Mirman (2000) and examine the real and financial effects of insider trading with correlated signals. They find that insider trading is related to market structure. Furthermore, Wang et al. (2009) demonstrates that the manager's profits may decrease or increase depending on the variances in the real and financial sectors. Although there are some studies showing that the market reacts less to insider ⁎ Corresponding author. Tel.: + 852 2766 7140; fax: + 852 2356 9550. E-mail addresses: afl[email protected] (L.T.W. Cheng), [email protected] (W.N. Davidson), [email protected] (T.Y. Leung). 1059-0560/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.iref.2010.11.016

422

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

trading (e.g., Eckbo & Smith, 1998; Lakonishok & Lee, 2001), Ma, Sun, and Tang (2009) find evidence that the insiders (particularly the board chairmen) are information-motivated traders and the market tends to react to insider trading gradually in the filing and publication periods rather than immediately in the trading period. In the Hong Kong stock market, Wong, Cheung, and Wu (2000) examine the abnormal returns associated with insider trading and conclude that prices increase (decrease) following insider purchases (sales) especially for small firms. Zhu, Chang, and Pinegar (2002) also report that inside buyers can earn significant abnormal returns. Cheng, Firth, Leung, and Rui (2006) show that increased insider trading impairs liquidity by increasing spread and reducing depth. Cheng and Leung (2008) find that there are significant insider purchases before the announcements of good earnings and dividend news; and significant insider sales before bad earnings and dividend news. As insiders possess private information about the firm, insider trading activities before a corporate announcement should not be a rare phenomenon (Clarke, Dunbar, & Kahle, 2001; Fidrmuc, Goergen, & Renneboog, 2006). These studies document that insiders buy before favorable announcements and sell before unfavorable announcements. It is well-documented that earnings (including earnings forecasts) and dividends provide signals about the future performance of firms (Asquith & Mullins, 1983; Bajaj & Vijh, 1995; Basyah & Hartigan, 2007; Brown, 1978; Nissim & Ziv, 2001; Rendleman, Jones, & Latane, 1982). Pre-disclosed earnings and dividend information is a price-sensitive information that insiders may use to time their trading for profits. Therefore, there are a number of studies examining the intensity of insider trading activities before the announcements of earnings (Allen & Ramanan, 1995; Park, Jang, & Loeb, 1995; Sivakumar & Vijayakumar, 2001; Sivakumar & Waymire, 1994; Udpa, 1996) and dividends (John & Lang, 1991). A recent study by Khang and King (2006) uses insider returns as a proxy for information asymmetry between insiders and outsiders to examine the relation between dividends and information asymmetry and finds a negative relation between dividends and returns to insider trades. We hypothesize that insider returns serve as an effective measure to validate the signaling effect of dividend announcements. Khang and King (2006) use insider returns as a proxy for information asymmetry between insiders and outsiders to evaluate whether dividend payout serves as a credible signal for information asymmetry. They find a negative relation between dividends and insider returns and conclude that dividends may not be a credible signal for information asymmetry. We re-examine the signaling effect of dividend using insider returns as a validation tool with a modified methodology. These methodological enhancements include 1) providing alternative measurements of dividend changes; 2) controlling for earnings management; and 3) employing simultaneous earnings/dividend announcements to eliminate the potential confounding effects of earnings on dividends. We find a positive correlation between insider returns and dividends when insiders trade firm securities. We, thus, find that insider returns are an effective measure validating the signaling effect of dividends. The rest of the paper is organized as follows: Section 2 discusses the data and methodology. Section 3 describes the results, and the conclusion is reported in Section 4. 2. Data and methodology 2.1. Earnings and dividend data We obtain stock price data and financial statement data from the PACAP database, Datastream, and the Securities Trading Record Journal issued by the Hong Kong Exchanges. We limit our analysis to industrial companies in the PACAP database. During our sample period, 1993 through 2003, we find 6152 simultaneous annual earnings and dividend announcements. In this study, we focus on the examination of the three unambiguous announcement signals with stronger information content of good and bad news rather than the three ambiguous signals which contain both good and bad news. We exclude the ambiguous announcements because the theoretical direction of the mixed effects is unclear, and the net effects on the market reaction may be weak as the effects of good and bad news cancel each other out. We have 653 simultaneous earnings and dividend announcements, and we are able to match them with 3329 insider trading events. We discuss the insider trades next. 2.2. Insider trading data We obtain our insider trading data from the Inside Trade Asia database maintained by Primark (from 1993 to April 2000) and from the website of the Hong Kong Exchanges (from May 2000 to 2003). The Listing Rules of the Hong Kong Exchanges require directors to report their securities transactions within five business days (three business days from April 2003 onwards) from the day they place the transactions. The trading information would be disclosed on the Securities (Disclosure of Interest) Daily Summary — Directors'/ Chief Executives' Notification Report issued by the Hong Kong Exchanges. Although the changes in the share-holding of the directors can be the result of exercising options, warrants, bonus warrants and rights, the issue of bonus shares, the conversion of bonds and debentures, special and scrip dividends, stock splits as well as gifts, we only include those inside transactions which increase and decrease the shareholdings of the directors through open market purchase and sale of shares in our study (Lin & Howe, 1990). Our focus is on the relation between insider trading returns and dividend signals. Therefore, we need to select the insider trades after the previous but prior to the current announcements of annual earnings and dividends. Based on this criterion, we initially identify 17,342 inside transactions. We apply two data-cleaning conditions to our sample. First, to avoid potential confounding effects on market reaction, we exclude those cases where the firms have made other corporate announcements such as seasoned equity offering, stock split, and consolidations during our −40 to + 40 examination period, and we also eliminate cases with missing data. At this point, we have 8454 insider trading records remaining in the sample.

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

423

Individual insiders within each firm have their own timing and reasons to trade, and different insiders may conduct transactions on the same day or within a short period of time. Mixed trades create conflicting and unclear signals to the market. Also, trade reversals may imply a higher likelihood of trades due to liquidity rather than information. Lee (1997) argues that a pure insider trading pattern provides a clearer signal than a mixed trading pattern. Therefore, we focus on pure insider sales and purchases and exclude the mixed transactions. In other words, insider events that contain mixed transactions within the −40 to +40 periods are excluded as well. This second restriction leaves us with 3329 insider trading events for our 653 forthcoming earnings and dividend announcements. 2.2.1. Methodology To test our hypothesis that there exists a positive correlation between insider returns and dividends, we first need to employ event study methodology to evaluate the abnormal share price reaction of insider trades. The event date, t = 0, is the date when the insiders conduct their securities transactions in the market. To estimate our parameters, we use the market model. The parameter estimation period covers 150 days from t = − 250 to t = − 101. We estimate insider trading returns over a 40-day period between t = +1 and t = + 40.1 The abnormal return (ARit) for sample firm i on day t is defined as: ARit = ½Retit –ðα + β × MRett Þ × θ

ð1Þ

Retit is the actual returns of firm i on day t. MRett is the market return on day t. θ is a variable for the direction of trade which takes the value of +1 if trade j is a “Buy” transaction and −1 if trade j is a “Sell” transaction. Seyhun (1986) and Lin and Howe (1990) suggest that insiders make abnormal profits when share price increases after insider purchase or when share price decreases after insider sales. Therefore, for the purpose of aggregation and comparison, the abnormal returns of the “Sell” transactions are multiplied by −1. Two factors are considered when determining the appropriate event windows. First, we have to include sufficient time to capture the full market reaction. In this study, we need to examine the cumulative abnormal returns (CAR) after the insider trading activity. Thus post-event windows are the focus. We first plot the CARs to determine how long the upward drift persists until the price adjustment is completed. The plots show that most adjustments are completed in + 40 days, and as a result, we select the time intervals of +1 to +20 and + 1 to + 40 for our examination. The second factor we consider in determining the post-event window is related to sample size maximization. As we mention earlier, we employ two conditions to filter our insider trading events. The longer the post-event periods, the more the insider events would be excluded due to the existence of more confounding events and mixed trades. After careful consideration, we believe that the +1 to +40 windows can achieve an optimal balance between sample maximization and data-cleaning. As it is possible for a given firm to have many days with pure insider sales and purchases, the abnormal returns calculated based on daily insider trades may be biased towards firms having more daily transactions. To avoid this potential bias, we combine abnormal returns (by taking simple average) of all pure insider sales and purchases within the same year for each firm into one firm-based CAR. We have 653 observations in our dataset, of which 394 are pure insider purchases (Buy sample) and 259 are pure insider sales (Sell sample). When we divide our sample into good news firms and bad news firms, we have 263 firm-based events (or CARs for each event window) with forthcoming unambiguous good news (Good News sample) and 390 firm-based events with unambiguous bad news (Bad News sample). We will define how we classify good news and bad news in the next section. 2.3. Measuring information asymmetry using insider returns Khang and King (2006) use insider returns as a proxy for information asymmetry between insiders and outsiders to evaluate if the dividend payout serves as a credible signal for information asymmetry. They propose that firms with stronger information asymmetry should have larger insider returns. Similarly, if dividends are a signal for information asymmetry, then firms with greater information asymmetry would need a larger dividend or a larger unexpected dividend as a signal. If insider returns are a good measure of information asymmetry, there should be a positive relation between dividends and returns to insider trades. However, they find a negative relation between dividends and returns to insider trades. Their conclusion is that dividend may not be a credible signal for information asymmetry. We believe that the negative relation between dividends and returns to insider trades found by Khang and King (2006) can be methodology specific. We discuss these methodological issues and our modifications later. 2.4. Controlling for potential confounding effect of earnings Research has shown that earnings announcements provide a signal to the market (Brown, 1978; Rendleman et al., 1982). Unexpected earnings, earnings changes, and earnings forecast have all been documented to contain useful information and serve as signals for future profitability of firms. As earnings and dividend announcements are normally made closely or even simultaneously, studying a dividend signal without controlling for the effect of earnings can be problematic. Kane, Lee, and Marcus (1984) and Conroy, Eades, and Harris (2000) show that earnings and dividend signals are collaborative (interactive). 1 On average, in Hong Kong, there are 23 business day in a month. We choose our event window intervals to be 20-day or 40-day periods are because 20-day is close to a one-month period and 40-day is close to a two-month period.

424

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

Let us suppose Firm A has poor earnings this year but for some reason decides to pay a large dividend payout. The Khang and King (2006) method would treat the announcement the same as it would Firm B, which has strong earnings and the same high dividend payout. However, the level of information asymmetry for Firm A and Firm B may be very different because the information in the dividend announcement is contaminated by the information in the earnings announcement. Therefore, to study the signaling effect of dividends, we have to control for the confounding effect of earnings. In the US, simultaneous announcements of earnings and dividends are not very common. For example, Kane et al. (1984) find only 10 of 352 dividend announcements between 1979 and 1981 occurred within ten days of an earnings announcement. In addition, Chang and Chen (1991) show that the time gap between earnings and dividends vary from 1 day to more than 20 days, and only 48.21% of firms announce earnings and dividends within 10 days of each other. As the time gap for earnings and dividend announcements vary substantially among firms in the US, it is not easy to control for the effects of earnings on dividend signals. Contemporaneous announcements of earnings and dividends are more common in Hong Kong. So, we are able to control for the effect of earnings by using contemporaneous announcements of earnings and dividends in Hong Kong firms. The combined signal of simultaneous earnings and dividend announcements allows us to better classify the signal type. In short, we hypothesize a positive relation between information asymmetry and the dividend signal. In measuring signals, we use three possible choices. First, we follow Khang and King (2006) and use just actual earnings and dividends. Second, we use percent change of earnings and dividends relative to prior year's data. Finally, we use our expectation models to measure the earnings and dividend surprises. 2.5. Earnings and dividend model construction In addition to earnings and dividend payouts (as in Khang & King, 2006), we also use abnormal change in earnings and dividends as signals. A positive (negative) change in earnings is a signal of improved (deteriorated) performance. Similarly, a positive change (negative change or zero payout) in dividend distribution is a favorable (an unfavorable) signal to the market about the firm's confidence and ability to maintain future cash outflows. We use earnings per share to measure corporate earnings. Dividend is measured in terms of dividend per share.2 We cannot use analyst forecast as a benchmark to measure the market expectation for earnings and dividend surprises for our sample firms because in Hong Kong, analyst forecast reports are usually made for large firms only. We follow Cheng, Fung, and Leung (2007) and Cheng and Leung (2008) to estimate the market expectations for earnings and dividends. First, we compute the expected earnings per share and dividend per share for the current year by multiplying last year's estimate. We use the previous year's industry adjustment to compute last year's estimate. The previous year's industry adjustment is the percentage change in the mean earnings/dividend per share for all firms in industry k for year y − 1 and year y − 2.     V E Viy = Við y−1Þ × 1 + gy

ð2Þ

Viy is a variable which represents earnings per share (eiy) or dividend per share (diy) of firm i in year y. E(Viy) represents the expected change in earnings per share or dividend per share of firm i in year y. eiy is the realized earnings per share for firm i in year y. diy is the realized dividend per share of firm i in year y. The variable, gVy , is the industry adjustment factor for earnings (the percentage change of the mean earnings per share for all firms of industry k in year y − 1) or for dividends (the percent change of the mean dividend per share for all firms of industry k in year y − 1). We denote industry as k (utilities, properties, consolidated enterprises, industrial, hotels and miscellaneous). y represents the year number from 1993 to 2003. Once we estimate our earnings and dividend expectations, we then compute our standardized abnormal earnings (Aeiy) and dividends (Adiy). The standardized abnormal earnings (dividends) is the difference between the realized earnings (eiy) (dividends, diy) and expected earnings (E(eiy)) (dividends, E(diy)) scaled by expected earnings (share price of firm i on day t or Pit)3:   eiy −E eiy   Aeiy = E eiy

Adiy =

  diy −E diy Pit

ð3Þ

ð4Þ

Using Eqs. (3) and (4), we can compute the abnormal earnings and dividends. Then we can categorize these announcements into one of the six groups, three unambiguous announcement signals (Abnormal Earnings–Dividend Increase; Abnormal Earnings–Dividend Decrease; and Abnormal Earnings Decrease–Dividend Zero) and three ambiguous announcement signals 2 Firms can distribute cash to their shareholders in the form of interim, final or special cash dividends. It is not uncommon for firms to declare extra distribution through special cash or bonus dividends besides the regular interim or final dividends. While these extra special cash payouts are usually “one-off” distribution, to avoid misclassification of the announcement type of dividend as a result of the irregular cash payment, we use only the regular final dividend to calculate the dividend per share for the year. 3 As there are many cases that the dividend per share of the previous year is zero, to avoid having a naught figure as the denominator, we scale the magnitude difference between the dividend per share of year y and year y − 1 by the share price.

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

425

(Abnormal Earnings Decrease–Dividend Increase; Abnormal Earnings Increase–Dividend Decrease; and Abnormal Earnings Increase–Dividend Zero). As we want to focus on the three unambiguous announcement signals with stronger information content of either good or bad news rather than the three ambiguous signals, we exclude the ambiguous announcements. Finally, we define the firms with forthcoming Abnormal Earnings–Dividend Increase as our Good News sample. Also, we combine firms with forthcoming Abnormal Earnings–Dividend Decrease and Abnormal Earnings Decrease-Dividend Zero into our Bad News sample.4 These Good News and Bad News samples form the dataset for our hypothesis testing. 2.5.1. Regression analysis We use regression model (5a–c) to test our hypothesis. CAR = α0 + β1 e + β2 d + β3 MJAccrual + β4 LnAsset + β5 CFOPS + β6 DE

ð5aÞ

CAR = α0 + β1 echg + β2 dchg + β3 MJAccrual + β4 LnAsset + β5 CFOPS + β6 DE

ð5bÞ

CAR = α0 + β1 Ae + β2 Ad + β3 MJAccrual + β4 LnAsset + β5 CFOPS + β6 DE

ð5cÞ

CAR is the cumulative abnormal insider returns measured over various event windows. We use firm-based CARs instead of eventbased observation for the regression analysis to avoid bias due to the over-weightings of firms that have frequent insider trading activities. 2.5.1.1. Key variables. We use three measures of earnings and dividends: the original earnings per share, the change in earnings per share and the abnormal component of earnings per share. Similar variations are also applied to dividends. e is earnings per share. d is dividend per share. echg is the percentage change in earnings per share. dchg is the percentage change in dividend per share. Ae is the abnormal change in earnings per share with an industry adjustment. Ad is the abnormal change in dividend per share with an industry adjustment. MJAccrual is the measure of abnormal accrual using modified version of Jones (1991) model. We control for accruals because while earnings and dividends can be good measures of firms' economic value, they may be affected by the extent to which managers exercise their discretion over financial reporting choices (Dechow, 1994; Holthausen, 1990). In the accounting literature, many studies have shown that accruals can be used for income- and dividend-smoothing purposes (e.g., Agarwal, Chomsisengphet, Liu, & Rhee, 2007; Burgstahler & Dichev, 1997; Kasanen, Kinnunen, & Niskanen, 1996), compensation strategies (e.g., Gaver, Gaver, & Austin, 1995), issue of seasoned equity offers (e.g., Rangan, 1998), bond issuance (Liu, Ning, & Davidson, 2008), and mergers and acquisitions strategies (e.g., Erickson & Wang, 1999). Therefore, we include MJAccrual as a control variable. The literature shows that there is a negative relation between the level of discretionary accruals and informativeness of reported earnings. Sloan (1996) and Xie (2001) show a negative relation between abnormal returns and discretionary accruals. We use the modified version of the Jones (1991) model (Dechow, Sloan, & Sweeney, 1995; DeFond & Park, 1997) to compute the cross-sectional regression parameters for each industry (utilities, properties, consolidated enterprises, industrials, hotels and miscellaneous) in each year (from 1993 to 2003). 2.5.1.2. Other control variables. We include three variables, LnAsset, CFOPS and DE, in the regression model to control for the firm specific characteristics. LnAsset is log value of total assets which is used to control for the size effect (Banz, 1981) to avoid model misspecification. CFOPS is operating cash flow per share. We use CFOPS as our measure of financial liquidity (Healy, Palepu, & Ruback, 1992). DE is debt to equity ratio, which is used to measure the risk level of financial leverage (Grullon, Michaely, & Swaminathan, 2002). 3. Results We first conduct some descriptive analyses on the insider trading activities by examining the event CARs of the Good News and Bad News subsamples. Based on the prior research that firms “leak” good news early but postpone the release of bad news until absolutely necessary (Begley & Fischer, 1998; Chambers & Penman, 1984; Givoly & Palmon, 1982), we expect more information asymmetry around bad news than good news. Thus, firms with forthcoming good news should have smaller information asymmetry than firms with forthcoming bad news. In other words, we expect a smaller CAR (in terms of magnitude) after the insider trading announcements for the Good News sample than for the Bad News sample. Table 1 contains the CARs for two intervals for both good news and bad news samples. The main focus of our investigation should be the post-event windows. The post insider trading CARs for the Good News sample of 263 events (2.15% for +1 to +20 and 3.72% for +1 to + 40) are normally larger than those for the Bad News sample of 390 events (1.40% for + 1 to +20 and 2.59% for + 1 to +40).5 Yet, there are no significant differences in the post insider trading CARs between the Good News and Bad News samples. 4 For all of our analyses (plotting of CAR paths and measuring market reaction) in this study, we focus on Good News sample and Bad News sample. We combine the subsamples of Abnormal Earnings–Dividend Decrease and Abnormal Earnings Decrease–Dividend Zero into Bad News sample. But for robustness purpose, we also perform the analyses on the two subsamples of Abnormal Earnings–Dividend Decrease and Abnormal Earnings Decrease–Dividend Zero separately. The results are all qualitatively the same. In addition, we also conduct similar analyses for each of the three subsamples of ambiguous news (Abnormal Earnings Decrease–Dividend Increase, Abnormal Earnings Increase–Dividend Decrease, Abnormal Earnings Increase–Dividend Zero). The results are available upon request. 5 We repeat the same analysis with insider trading events without aggregating into firm-based events, results are very similar and are reported in Appendix 1.

426

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

Table 1 CAR of insider trading activity for forthcoming Good News and Bad News. The Good News sample includes the events with Abnormal Earnings–Dividend Increase. Bad News sample includes events with Abnormal Earnings–Dividend Decrease and Abnormal Earnings Decrease–Dividend Zero. Good News (Buy) Sample includes the event with good news (abnormal earnings increase and abnormal dividend increase) and insider purchase. Bad News (Sell) includes the event is with bad news (abnormal earnings decrease and abnormal dividend decrease and dividend zero) and insider sale. Difference shows the mean CAR difference between Good News sample and Bad News sample or between Good News (Buy) sample and Bad News (Sell) sample. A positive (negative) value for CAR Difference indicates that there is a higher (lower) value for Good News (Good News (Buy)) Sample than that for Bad News (Bad News (Sell) Sample. [N]

Good News

[263]

Bad News

[390]

Difference

+ 1 to + 20

+ 1 to + 40

(t-values)

(t-values)

0.0215 (3.01)** 0.0140 (1.98)* 0.0075 (0.75)

0.0372 (3.64)** 0.0259 (2.30)* 0.0113 (0.74)

[N]

Good News (Buy)

[156]

Bad News (Sell)

[152]

Difference

+ 1 to + 20

+ 1 to + 40

(t-values)

(t-values)

0.0210 (2.19)* 0.0406 (3.30)** − 0.0196 (− 1.26)

0.0331 (2.54)** 0.1058 (5.22)** − 0.0727 (3.01)**+

* Significant at 5% (parametric test). ** Significant at 1% (parametric test). + Significant at 5% (non-parametric test for mean difference comparison).

However, if insiders actually take advantage of the forthcoming good news, they should buy instead of sell. Thus, the insider sales before the good news may be due to other reasons such as liquidity and are not related to the future good news. Similar logic can apply to the insider purchases before the bad news. In order to evaluate on insider trades that are relevant to the forthcoming news and examine information asymmetry through the post insider trading CARs, we repeat our analysis by focusing on CARs from

Table 2 Regression analysis for the Buy sample. Dependent variables are insider trading abnormal returns (CARs) for various time intervals. e is earnings per share. d is dividend per share. echg is the percentage change in earnings per share for year y and year y − 1. dchg is the percentage change in dividend per share for year y and year y − 1. Ae is the abnormal change in earnings per share. Ad is the abnormal change in dividend per share. LnAsset is log value of total assets. MJAccrual is the measure of abnormal accrual using modified version of Jones (1991) model. CFOPS is operating cash flow per share. DE is debt to equity. t-values are adjusted for heteroskedasticity using White's procedure (1980). Coefficient (t-value) + 1 to + 20 Intercept e d

0.1431 (2.25) 0.0223 (1.01) − 0.0194 (− 0.25)

echg

+ 1 to + 40 0.1469 (2.36)

Ae Ad

LnAsset CFOPS DE Adjusted R2 F p-value Number of observations * Significant at 0.05 level. ** Significant at 0.01 level.

0.2179 (2.28) 0.0264 (0.65) − 0.0365 (− 0.33)

0.0094 (1.70) 0.2429 (1.38)

dchg

MJAccrual

0.1346 (2.19)

− 0.0936 (− 1.73) − 0.0095 (− 2.14)* 0.0106 (0.50) − 0.0087 (− 0.80) 0.0134 1.8813 0.08 394

− 0.1048 (− 1.96)* − 0.0093 (− 2.19)* 0.0096 (0.54) − 0.0069 (− 0.63) 0.0407 3.7589 0.00 394

0.2455 (2.71)

0.2178 (2.49)

0.0173 (2.46)* 0.5344 (2.43)* 0.0092 (1.75) 0.2001 (3.18)** − 0.1046 (− 1.96)* − 0.0085 (− 2.02)* 0.0092 (0.52) − 0.0040 (− 0.36) 0.0472 4.2224 0.00 394

− 0.1481 (− 1.83) − 0.0158 (− 2.29)* 0.0493 (1.23) − 0.0102 (− 0.61) 0.0438 3.9853 0.00 394

− 0.1832 (− 2.48)* − 0.0169 (− 2.59)** 0.0400 (1.23) − 0.0059 (− 0.36) 0.0989 8.1559 0.00 394

0.0164 (2.51)* 0.5421 (4.62)** − 0.1855 (− 2.54)* − 0.0151 (− 2.36)* 0.0380 (1.16) 0.0027 (0.17) 0.1304 10.7747 0.00 394

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

427

insider purchases preceding good news (Good News (Buy) sample with 156 events) and insider sales preceding bad news (Bad News (Sell) sample with 152 events). Good News (Buy) Sample includes event with good news (abnormal earnings increase and abnormal dividend increase) and insider purchase. Bad News (Sell) Sample includes event with bad news (abnormal earnings decrease and abnormal dividend decrease and dividend zero) and insider sale. The post insider trading CARs for the Good News (Buy) sample (2.10% for +1 to +20 and 3.31% for +1 to +40) are smaller than those for the Bad News (Sell) sample (4.06% for +1 to +20 and 10.58% for + 1 to +40). We find a significant difference in CARs between the Good News (Buy) and Bad News (Sell) samples in the +1 to + 40 interval. These findings provide support to our expectation that the post-event price reactions to insider trading are stronger for bad news than for good news. This shows that the information asymmetry is stronger for Bad News (Sell) Sample than for Good News (Buy) Sample. In other words, there is a bigger information asymmetry for firms with insider sales before poor earnings–dividend announcements. Our hypothesis proposes that both insider returns and dividend announcements serve as effective signals for information asymmetry. Khang and King (2006) find a negative relation between dividends and insider returns and conclude that dividends may not be a credible signal for information asymmetry. We re-examine the signaling effect of dividend with a modified methodology. We have three methodological modifications to the method used by Khang and King (2006). First, in addition to using the dividend measure as in Khang and King, we use dividend per share change (dchg) and abnormal change in dividend per share (Ad) to reflect the information content of dividends. Second, we control for earnings management in our model. Third, we use Hong Kong data to better control for the effects of earnings due to the simultaneous nature of these dividends and earnings announcements. In Table 2, we provide the estimated regression results for the Buy sample for testing our hypothesis. For the dividend per share (d) models, none of the estimated coefficients are significant. However, for the dchg models, the estimated coefficient for the interval + 1 to +40 is positive and significant. More importantly, the estimated coefficients for the Ad models (both +1 to + 20 and + 1 to +40) are positive and significant. This result indicates that insider returns and dividends demonstrate a positive correlation, an observation which is consistent with our hypothesis. Table 3 contains the results for the Sell sample. None of the dividend measures are significant in the Sell sample. Many previous studies show that insider purchases are mainly motivated by insider profit through the purchase of undervalued stock while insider sales may be driven by liquidity reasons rather than by

Table 3 Regression analysis for the Sell sample. Dependent variables are insider trading abnormal returns (CARs) for various time intervals. e is earnings per share. d is dividend per share. echg is the percentage change in earnings per share for year y and year y − 1. dchg is the percentage change in dividend per share for year y and year y − 1. Ae is the abnormal change in earnings per share. Ad is the abnormal change in dividend per share. LnAsset is log value of total assets. MJAccrual is the measure of abnormal accrual using modified version of Jones (1991) model. CFOPS is operating cash flow per share. DE is debt to equity. t-values are adjusted for heteroskedasticity using White's procedure (1980). Coefficient (t-value) + 1 to + 20 Intercept e d

− 0.3235 (− 3.03) − 0.0076 (− 0.33) − 0.0107 (− 0.19)

+ 1 to + 40 − 0.3085 (− 3.32)

dchg Ae Ad

LnAsset CFOPS DE Adjusted R2 F p-value Number of observations ** Significant at 0.01 level.

− 0.5384 (− 3.46) 0.0236 (0.55) − 0.0520 (− 0.52)

− 0.0014 (− 0.21) − 0.2173 (− 0.75)

echg

MJAccrual

− 0.3049 (− 3.25)

0.0204 (2.68)** − 0.0965 (− 1.56) 0.0028 (0.21) 0.0030 (0.19) 0.0308 2.3414 0.03 259

0.0190 (3.02)** − 0.0956 (− 1.56) 0.0042 (0.30) 0.0038 (0.24) 0.0329 2.4330 0.03 259

− 0.5330 (− 3.74)

− 0.5193 (− 3.65)

0.0084 (0.83) 0.0652 (0.09) − 0.0007 (− 0.11) − 0.0041 (− 0.03) 0.0189 (2.99)** − 0.1000 (− 1.62) 0.0011 (0.08) 0.0034 (0.21) 0.0301 2.3066 0.03 259

0.0319 (2.96)** − 0.2843 (− 2.58)** 0.0048 (0.20) 0.0034 (0.14) 0.0643 3.9201 0.00 259

0.0319 (3.34)** − 0.2953 (− 2.71)** 0.0020 (0.09) 0.0019 (0.08) 0.0680 4.1015 0.00 259

0.0101 (1.10) 0.1820 (0.72) 0.0313 (3.28)** − 0.2939 (− 2.64)** − 0.0005 (− 0.02) 0.0006 (0.02) 0.0732 4.3557 0.00 259

428

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

information (Gregory et al., 1997; Lakonishok & Lee, 2001; Lin & Howe, 1990; Seyhun, 1986). This may be the reason why there is no significant relation between insider returns and dividend signal in the Sell sample. 4. Conclusion The Hong Kong market provides a good opportunity to examine information content of unambiguous corporate news because companies announce earnings and dividends at the same time. We classify announcements of earnings and dividend increases as good news and earnings and dividend decreases (or dividend zero) as bad news. We first examine the market reaction to insider trades that occur 40 days around the earnings/dividend announcements. We find that the market reacts more to the insider trades when there is a subsequent earnings/dividend decrease (bad news) than when there is an increase (good news). The insider trades contain more information and relieve more information asymmetry prior to the bad news than when there is subsequent good news. More importantly, we make methodological improvements (relative to Khang and King (2006)) and reexamine the relation between insider returns and dividends. These enhancements include 1) providing alternative measurements of dividend changes; 2) controlling for earnings management; and 3) employing simultaneous earnings/dividend announcements to eliminate the potential confounding effects of earnings on dividends. As a result, we find evidence consistent with the dividend signaling hypothesis. Prior research may not have been able to find this result largely because dividend and earnings announcements do not always occur simultaneously in many countries. In short, using data from firms with two separate announcements may confound the empirical results of the dividend signal. Acknowledgments We would like to thank Hongxia Wang for her assistance on this manuscript. All errors should be attributed to the authors. The authors acknowledge the financial support (Project 7002620) from Department of Accountancy, City University of Hong Kong. Appendix 1. CARs of insider trading activity for good news and bad news.(using insider trading events) The Good News sample includes the events with Abnormal Earnings-Dividend Increase. Bad News sample includes events with Abnormal Earnings–Dividend Decrease and Abnormal Earnings Decrease–Dividend Zero. Good News (Buy) Sample includes the event with good news (abnormal earnings increase and abnormal dividend increase) and insider purchase. Bad News (Sell) includes the event is with bad news (abnormal earnings decrease and abnormal dividend decrease and dividend zero) and insider sale. Difference shows the mean CAR difference between Good News sample and Bad News sample or between Good News (Buy) sample and Bad News (Sell) sample. A positive (negative) value for CAR Difference indicates that there is a higher (lower) value for Good News (Good News (Buy)) Sample than that for Bad News (Bad News (Sell) Sample. Panel A: Forthcoming Good News and forthcoming Bad News N

Good News Bad News Difference

[1383] [1946]

+ 1 to + 20

+ 1 to + 40

CAR

t-statistics

CAR

t-statistics

0.0139 −0.0036 0.0175

6.73** −1.32 5.11**++

0.0228 −0.0053 0.0281

6.53** −1.12 4.76**++

Panel B: Forthcoming Good News with Buy and forthcoming Bad News with Sell N

Good News (Buy) Bad News (Sell) Difference

[1002] [437]

+ 1 to + 20

+ 1 to + 40

CAR

t-statistics

CAR

t-statistics

0.0149 0.0396 − 0.0248

6.38** 5.93** − 3.50**++

0.0238 0.1184 − 0.0946

6.24** 9.96** − 7.58**++

*Significant at 5% (parametric test). **Significant at 1% (parametric test). + Significant at 5% (non-parametric test for mean difference comparison). ++ Significant at 1% (non-parametric test for mean difference comparison).

References Agarwal, S., Chomsisengphet, S., Liu, C., & Rhee, S. G. (2007). Earnings management behaviors under different economic environments: Evidence from Japanese banks. International Review of Economics and Finance, 16, 429−443. Allen, S., & Ramanan, R. (1995). Insider trading, earnings changes and stock prices. Management Science, 41, 653−668. Asquith, P., & Mullins, D. W. (1983). The impact of initiating dividend payments on shareholder' wealth. Journal of Business, 56, 77−96. Bajaj, M., & Vijh, A. M. (1995). Trading behavior and the unbiasedness of the market reaction to dividend announcement. Journal of Finance, 50, 255−279. Banz, R. (1981). The relationship between return and market value of common stocks. Journal of Financial Economics, 9, 3−18.

L.T.W. Cheng et al. / International Review of Economics and Finance 20 (2011) 421–429

429

Basyah, M., & Hartigan, J. C. (2007). Analyst earnings forecast revisions and the persistence of antidumping relief. International Review of Economics and Finance, 16, 383−399. Begley, J., & Fischer, P. E. (1998). Is there information in an earnings announcement delay? Review of Accounting Studies, 3, 347−363. Brown, S. L. (1978). Earnings changes, stock prices and market efficiency. Journal of Finance, 33, 17−28. Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics, 24, 99−126. Chambers, A. E., & Penman, S. H. (1984). Timeliness of reporting and the stock price reaction to earnings announcements. Journal of Accounting Research, 22, 21−47. Chang, S. J., & Chen, S. N. (1991). Information effects of earnings and dividend announcements on common stock returns: Are they interactive? Journal of Economics and Business, 43, 179−192. Cheng, L. T. W., Firth, M., Leung, T. Y., & Rui, O. M. (2006). The effects of insider trading on liquidity. Pacific-Basin Finance Journal, 14, 467−483. Cheng, L. T. W., Fung, H. G., & Leung, T. Y. (2007). Information effects of dividends: Evidence from the Hong Kong market. Review of Quantitative Finance and Accounting, 28, 23−54. Cheng, L. T. W., & Leung, T. Y. (2008). Is there information content from insider trading activities preceding earnings and dividend announcements in Hong Kong. Accounting and Finance, 48, 417−437. Clarke, J., Dunbar, C., & Kahle, K. M. (2001). Long-run performance and insider trading in completed and canceled seasoned equity offerings. Journal of Financial and Quantitative Analysis, 36, 415−430. Conroy, R. M., Eades, K. M., & Harris, R. S. (2000). A test of the relative pricing effects of dividends and earnings: Evidence from simultaneous announcements in Japan. Journal of Finance, 55, 1199−1227. Daher, W., & Mirman, L. J. (2007). Market structure and insider trading. International Review of Economics and Finance, 16, 306−331. Dechow, P. M. (1994). Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals. Journal of Accounting and Economics, 18, 3−42. Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995). Detecting earnings management. The Accounting Review, 70, 193−226. DeFond, M. L., & Park, C. W. (1997). Smoothing income in anticipation of future earnings. Journal of Accounting and Economics, 23, 115−139. Eckbo, B. E., & Smith, D. C. (1998). The conditional performance of insider trades. Journal of Finance, 53, 467−498. Erickson, M., & Wang, S. W. (1999). Earnings management by acquiring firms in stock for stock mergers. Journal of Accounting and Economics, 27, 149−176. Fidrmuc, J., Goergen, M., & Renneboog, L. (2006). Insider trading, news releases and ownership concentration. Journal of Finance, 61, 2931−2973. Finnerty, J. E. (1976). Insiders’ activity and insider information: A multivariate analysis. Journal of Financial and Quantitative Analysis, 11, 205−215. Fishe, R. P. H., & Robe, M. A. (2004). The impact of illegal insider trading in dealer and specialist markets: Evidence from a natural experiment. Journal of Financial Economics, 71, 461−488. Gaver, J., Gaver, K., & Austin, J. (1995). Additional evidence on bonus plans and income management. Journal of Accounting and Economics, 18, 3−28. Givoly, D., & Palmon, D. (1982). Timeliness of annual earnings announcements: Some empirical evidence. The Accounting Review, 57, 486−508. Givoly, D., & Palmon, D. (1985). Insider trading and the exploitation of inside information: Some empirical evidence. Journal of Business, 58, 69−87. Gregory, A., Matatko, J., & Tonks, I. (1997). Detecting information from directors' trades: Signal definition and variable size effects. Journal of Business Finance and Accounting, 24, 309−342. Grullon, G., Michaely, R., & Swaminathan, B. (2002). Are dividend changes a sign of firm maturity? Journal of Business, 75, 387−424. Healy, P. M., Palepu, K. G., & Ruback, R. S. (1992). Does corporate performance improve after mergers? Journal of Financial Economics, 31, 135−175. Holthausen, R. W. (1990). Accounting method choice: Opportunistic behavior, efficient contracting and information perspectives. Journal of Accounting and Economics, 12, 207−218. Jaffe, J. (1974). Special information and insider trading. Journal of Business, 47, 410−428. Jain, N., & Mirman, L. J. (2000). Real and financial effects of insider trading with correlated signals. Economic Theory, 16, 333−353. Jeng, L. A., Metrick, A., & Zeckhauser, A. (2003). Estimating the returns to insider trading: A performance-evaluation perspective. The Review of Economics and Statistics, 85, 453−471. John, K., & Lang, L. H. P. (1991). Insider trading around dividend announcements: Theory and evidence. Journal of Finance, 46, 1361−1389. Jones, J. (1991). Earnings management during import relief investigation. Journal of Accounting Research, 29, 193−228. Kane, A., Lee, Y. K., & Marcus, A. (1984). Earnings and dividend announcements: Is there a corroboration effect? Journal of Finance, 39, 1091−1099. Kasanen, E., Kinnunen, J., & Niskanen, J. (1996). Dividend-based earnings management: Empirical evidence from Finland. Journal of Accounting and Economics, 22, 283−312. Khang, K., & King, T. H. D. (2006). Does dividend policy relate to cross-sectional variation in information asymmetry? Evidence from returns to insider trades. Financial Management, 35, 71−94. Lakonishok, J., & Lee, I. (2001). Are insider trades informative? Review of Financial Studies, 14, 79−111. Lee, I. (1997). Do firms knowingly sell overvalued equity? Journal of Finance, 52, 1439−1466. Lee, M. H., & Bishara, H. (1989). Recent Canadian experience on the profitability of insider trades. Financial Review, 24, 235−249. Lin, J., & Howe, J. S. (1990). Insider trading in the OTC market. Journal of Finance, 45, 1273−1284. Liu, Y., Ning, Y., & Davidson, W. N. (2008). Earnings management and the pricing of new debt. Working Paper. : University of New Hampshire. Ma, Y., Sun, H., & Tang, A. P. (2009). Do insiders have insider tracks: An examination of Wall Street Journal's inside track column? International Review of Economics and Finance, 18, 520−530. Nissim, D., & Ziv, A. (2001). Dividend changes and future profitability. Journal of Finance, 56, 2111−2133. Park, S., Jang, H. J., & Loeb, M. P. (1995). Insider trading activity surrounding annual earnings announcements. Journal of Business Finance and Accounting, 22, 587−614. Pope, P. F., Morris, R. C., & Peel, D. A. (1990). Insider trading: Some evidence of market efficiency and directors' share dealings in Great Britain. Journal of Business Finance and Accounting, 17, 359−380. Rangan, S. (1998). Earnings management and the performance of seasoned equity offerings. Journal of Financial Economics, 56, 101−122. Rendleman, R. J., Jones, C. P., & Latane, H. A. (1982). Empirical anomalies based on unexpected earnings and importance of risk adjustment. Journal of Financial Economics, 10, 269−287. Seyhun, H. N. (1986). Insiders' profits, costs of trading and market efficiency. Journal of Financial Economics, 16, 189−212. Sivakumar, K., & Vijayakumar, J. (2001). Insider trading, analysts' forecast revisions, and earnings changes. Journal of Accounting, Auditing and Finance, 16, 167−187. Sivakumar, K., & Waymire, G. (1994). Insider trading following material news events: Evidence from earnings. Financial Management, 23, 23−32. Sloan, R. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71, 289−315. Udpa, S. C. (1996). Insider trading and the information content of earnings. Journal of Business Finance and Accounting, 23, 1069−1095. Wang, L. F. S., Wang, Y., & Ren, S. (2009). Stackelberg financial-leader in insider trading model. International Review of Economics and Finance, 18, 123−131. White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817−838. Wong, M. C. S., Cheung, Y. L., & Wu, L. (2000). Insider trading in the Hong Kong stock market. Asia-Pacific Financial Markets, 7, 275−288. Xie, H. (2001). The mispricing of abnormal accruals. The Accounting Review, 76, 357−373. Zhu, J., Chang, E. C., & Pinegar, J. M. (2002). Insider trading in Hong Kong: Concentrated ownership versus the legal environment. Working paper. : Brigham Young University.