- Email: [email protected]
On the differential market reaction to dividend initiations
The Quarterly Review of Economics and Finance 40 (2000) 263–277
On the differential market reaction to dividend initiations Zhenhu Jina** a
Illinois Wesleyan University, P.O. Box 2900 Bloomington, IL 67102-2900, USA
JEL Classifications: G12, G14, G32, G35 Keywords: Dividend initiation; Payout policy; Event study; Security pricing
Abstract The finance literature documents substantial positive stock price reaction to dividend initiations. Most dividend initiation studies focus on the average positive reaction; however, 40 percent of the firms that initiate dividends experience negative abnormal returns at announcement. This paper focuses on the apparent heterogeneity in the stock price reaction to dividend initiation. I find that the observed negative market reaction reflects the market’s economic assessment of the impact of the event on these firms, and that it is not caused by anticipation or confounding events. The result is also supported by the fact that the market reaction to dividend initiation for these firms is negatively related to initial dividend yield. Both the positive and negative observed reactions are consistent with conventional arguments regarding the information content of dividends, and their role in mitigating agency problems. © 2000 Bureau of Economic and Business Research, University of Illinois. All rights reserved.
1. Introduction Although studies1 report that 30 to 40 percent of the firms that initiate dividends experience negative abnormal returns at announcement, most research focuses on the positive average reaction rather than the apparent heterogeneity in the stock price reaction to dividend initiation. This paper addresses whether there are two fundamentally different types of dividend initiating firms, with dividend initiation being assessed by the market as a
* Tel.: ⫹1-309-556-3570. E-mail address: [email protected]. 1062-9769/00/$ – see front matter © 2000 Bureau of Economic and Business Research, University of Illinois. All rights reserved. PII: S 1 0 6 2 - 9 7 6 9 ( 9 9 ) 0 0 0 5 2 - 6
264
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
positive event for one type, and a negative event for the other. The key issue is to determine whether the observed negative reaction for a subset of the firms actually reflects an assessment by the market that initiation is a value-decreasing event due to firm specific factors. A Chow test is first conducted to determine whether the firms in the two subgroups are significantly different, and the result of the test rejects the hypothesis, at one percent significance level, that there is no difference between the two groups. Cross-sectional regressions using variables that proxy for the various determinants of the dividend initiation announcement reaction are then performed to test the relationship between market reaction and these firm specific factors. The regressions are run on the whole sample as well as two subgroups: Group P, firms with positive market reaction, and Group N, firms with negative market reaction. The whole sample regression results indicate that market reaction to dividend initiation announcements is significantly related to many of the firm specific variables in the model. The results of the two subsample regressions show that the Cumulative Abnormal Returns (CAR) are significantly negatively related to the initial dividend for Group N, but are significantly positively related for Group P. Direct evidence about the firms that have negative announcement returns is then collected from the Wall Street Journal Index, Moody’s Industrial Manual, and the annual reports of the firms in question to see whether the regression results can be supported. The information gathered from these sources corroborates, in general, the regression results. All these results clearly show that market reaction to dividend initiation announcements is affected by firm specific variables and that there might indeed be firms for which dividend initiation is viewed by the market as a negative event due to firm-specific factors. The paper proceeds as follows. In section one, I present the theory and hypothesis along with a short review of the literature. Section two presents the data and the empirical methodology. Section three presents the specific models and results. Finally, section four summarizes the results and conclusions.
1. Theoretical background Miller and Modigliani (1961) demonstrate that firm value is independent of dividend policy in a perfect capital market setting. However, they speculate that when information asymmetry exists, dividend announcements may be a means of transmitting information from well informed insiders to outsiders. Bhattacharya (1979), Asquith and Mullins (1986), John and Williams (1985) and Miller and Rock (1985) develop models consistent with this “signaling hypothesis.” In contrast, Easterbrook (1984) and Jensen (1986) use agency theory to argue that dividend payments can reduce agency costs. These two hypotheses have been examined empirically,2 and results of these studies are summarized briefly in the following paragraphs. There is general agreement in the empirical and theoretical literature in finance that, on average, dividend initiations have a positive impact on firm value. Many researchers argue that unanticipated initiation announcements serve as an optimistic signal of the firm’s future performance. In addition, the initiation of a dividend program puts in place a mechanism for
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
265
periodic future information releases, that serves to reduce information asymmetry. Asquith and Mullins (1983) point out conceptually that the abnormal return associated with dividend initiation can be decomposed into two components, a yield effect and an initiation effect. The initiation effect reflects the value resulting from the sequence of future dividend payments and captures the benefits from initiating a dividend program per se. The yield effect consists of the value change resulting from the magnitude of the initial dividend and captures the information content of the dividend payment. Agency theorists Jensen (1986) and Easterbrook (1984) identify a second factor that can potentially explain the positive wealth effects of dividend initiation. Jensen argues that dividend payments reduce the agency cost arising from managers’ incentives to use free cash flow to invest in negative NPV projects. Not only does a dividend payment reduce the cash available to managers for potential investment in negative NPV projects today, but to the extent that managers are reluctant to reduce dividends, it also offers the prospect of reduced availability of cash for such investment in future years. Easterbrook argues that a dividend payment today increases the probability that the firm will need to seek external financing from the capital market in the future. Thus, there is a higher probability of managers being exposed to the monitoring associated with external financing. This higher probability of future monitoring reduces the extent to which managers will deviate from stockholder wealth maximization, thus reducing the costs of the agency conflict between managers and stockholders. As in the information content argument, there is both a yield effect, and an initiation effect associated with a dividend initiation announcement. The yield effect consists of the reduction in agency costs resulting from the initial dividend; the initiation effect consists of the present value of the reductions in agency costs expected from the implicit sequence of future dividends. Both information content and agency cost arguments identify potential benefits of dividend initiation. However, the actual economic impact of initiation will consist of these benefits netted against the costs associated with payment of the initial dividend, as well as the implicit sequence of future dividends. These costs include the tax penalty on dividends (relative to capital gains), the increase in expected financing costs (due to the higher probability of external financing) and any costs associated with a reduction in financial slack (Black, 1976; Bhattacharya, 1979; Myers and Majluf, 1984). Thus, for a particular firm, if the market judges the benefits of initiation to be smaller than the costs, we would expect to observe negative abnormal returns at announcement.3 Furthermore, market reaction to a particular firm’s dividend initiation announcement can also be affected by market anticipation. Abnormal returns will reflect only the unanticipated component of the announcement. If the initiation announcement is fully anticipated, and the initial dividend is less than expected, the abnormal return at announcement could be negative, although initiation itself is a value-increasing event. Finally, although formal dividend signaling models suggest that unexpected dividend increases are always good news, the literature has long recognized the possibility that, for some firms, a dividend increase could signal a lack of profitable investment opportunities. By this argument, an unexpected dividend increase would convey an unanticipated deteri-
266
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
oration in investment opportunities, causing the market to revise the value of the firm downward. Take note that conventional signaling models, as well as the Jensen and Easterbrook agency arguments, predict that the announcement effect should be positively related to the initial dividend. Under these arguments, even when the total announcement effect is negative, there is still a benefit of dividend payment that is positively related to the dividend amount. Only if the dividend itself is bad news, because it conveys a deterioration in investment opportunities, do we expect to see the announcement returns negatively related to the initial dividend.
2. Hypotheses, data and methodology 2.1. Hypotheses The question addressed by this paper is whether dividend initiation is assessed by the market to be a value-increasing event for some firms and a value-decreasing event for others. The fact that 30 to 40 percent of dividend initiating-firms experience negative abnormal returns does not, in itself, establish that such a dichotomy exists. In an event study, some sample firms could exhibit negative abnormal returns ex post even if the average impact of the event is positive for all firms. The abnormal return at announcement can be negative when the expected economic impact of dividend initiation is positive for several reasons, such as market anticipation of the event, errors in estimating abnormal return, and confounding events.4 Thus, the key issue addressed in this paper is whether the negative abnormal returns observed at announcement for a subset of dividend-initiating firms reflect the market’s negative assessment of the event due to firm specific economic factors. Specifically, the following two hypotheses are examined: H1: Firms in Group P and Group N are from the same population and there are no fundamental differences between the two groups. H2: The negative market reaction to dividend initiation announcements is caused by anticipation, estimation error or confounding events and is not related to firm-specific, economic factors.
Chow test and cross-section multivariate regression models are used to test the hypotheses. If a new dividend program is viewed as value-increasing for one group of firms and value-decreasing for another for economic reasons, the firms in the two groups should be significantly different in some important aspects. At the same time, if the CAR for Group N are negative, on average, due to anticipation or estimation errors or other confounding events, there should be little relation between the observed CARs and the variables proxying for the determinants of the dividend initiation announcement reaction. As such, the adjusted R-squared of the regression for Group N should be lower than that for Group P and that the announcement CAR for Group N should not be significantly related to the initial dividend.
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
267
2.2. The data The initial data set consists of firms on the Center for Research in Security Prices (CRSP) NYSE/AMEX monthly file that paid a cash dividend for the first time in their history, or resumed paying cash dividends after 10 years without dividends, between 1973 to 1993.5 These events are verified using the Wall Street Journal Index. The firms in the sample also pass the following screens. 1. Common stock daily returns starting from 300 days before the initiation announcement are available from the CRSP daily returns file. 2. There are no confounding announcements such as earnings reports within five days of the initiation announcement. This is done so as not to compound market reaction to the announcement of dividend initiation with earnings announcement. While doing so might introduce some bias, it is mitigated by the fact that earnings volatility and earnings changes are two of the variables used in the regression model. 3. There are no major corporate events within the 300 days before the initiation announcement that would significantly change the nature or risk level of the firm, such as a merger or restructuring. 4. The firm is included in the Compustat Annual or Research files. 5. The firm’s proxy statement in the year of initiation, and a year before and after, are available. The screening process yields a sample of 157 firms. 2.3. Estimation of firm announcement reaction Abnormal returns at announcement are estimated by employing an expanded market model, including an industry index. The coefficients are estimated using 300 days’ return data prior to two days before the announcement. In particular, the abnormal return of Firm K at time t (ARkt) is defined as: ARkt ⫽ Rkt ⫺ (␣k ⫹ kRmt ⫹ 2kRind,t),
(1)
where Rkt is the observed return of Firm K on day t; Rmt is the return on the CRSP equally weighted index; and Rind,t is the return on an equally weighted portfolio of firms in the same industry as Firm K on day t. Firms are considered to be in the same industry if the first three digits of the four-digit Standard Industry Classification Code (SIC) are the same.6 Following Asquith and Mullins and other researchers, the market reaction to Firm K’s initiation announcement is defined as the 2-day CARk. It is the sum of Firm K’s abnormal return from day t-1 to day t. Day t is the day that the initial dividend announcement appeared in the Wall Street Journal. Thus, the 2-day CAR is computed as the sum of Firm K’s abnormal return on days t-1 and t. The basic assumption here is that the market immediately reassesses the company following the announcement of the new dividend program and reacts to it. (As a precaution, 3, 5 and 10 day CARs are also computed to see whether market reaction centers on the two days, and the answer is affirmative. The results, therefore, are based on the 2-day CARs.) Of the 157 firms in the sample, 102 firms (64.9%) have positive
268
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
cumulative abnormal returns and 55 firms (35.2%) have negative abnormal returns. The average 2-day CAR for all 157 firms is 2.98%, with a median of 1.43% and a standard deviation of 4.66%. The average CAR for the 102 firms that have positive returns is 6.16%, with a median of 4.26% and standard deviation of 4.13%. The average CAR for the 55 firms that have negative returns is ⫺2.88%. The median and standard deviations are ⫺1.94% and 1.48% respectively. 2.4. Proxies for the determinants of initiation abnormal returns Cross-sectional regressions are performed on the full group and the two subgroups to examine the relationship between the variables that proxy for the determinants of initiation abnormal return and the observed announcement reaction and the explanatory power of these variables in the two subgroups. The proxies used are as follows: 1. Firm Size (SIZE): The firm size is measured as the natural log of the market value of equity as of two days before the announcement. Zeghal (1983), Eddy and Seifert (1988) and Mitra and Owers (1990) argue that firm size is a good proxy for the degree of publicly available information about a firm; the larger the firm, the greater the availability of information. Thus, the value of instituting a mechanism for periodic information releases may be greater for small firms than for large firms. The sign of the coefficient is expected to be negative. 2. Earnings Volatility7 (EARNVOL): Earnings volatility is estimated as the standard deviation of earnings per share over the 16 quarters immediately preceding the initiation announcement. If part of the benefit of a regular dividend program is to provide an information-releasing mechanism, then the information provided by quarterly dividend announcements will be more valuable for firms with less predictable earnings. On the other hand, for firms with stable earnings, the value of the additional information from the dividend program may not be as significant. The sign is expected to be positive. 3. Institutional Holdings8 (INST): The fraction of outstanding shares held by institutions is used as a proxy for the intensity of monitoring that the firm is subjected to by institutions. It is hypothesized that the agency costs arising from the manager-stockholder conflict are smaller when institutions monitor the firm more closely. Thus, the benefit of agency cost reduction may be smaller for firms with large institutional holdings. An alternative interpretation of this variable is that heavy institutional holdings are associated with greater information availability about the firm, reducing the benefit from periodic future information releases. Both interpretations predict a negative coefficient. 4. Board Ownership9 (BOARD): The percentage of shares owned by members of the board of directors is used here as a proxy of the board’s monitoring activities. Higher board ownership should result in more intensive board monitoring, less serious agency problems, less benefit from the reduction of agency costs, and thus a greater likelihood of initiation being value-decreasing.
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
269
5. Tobin’s Q ratio (Q10): This ratio is widely used, as in this study, as a proxy for the firm’s future growth opportunities. The higher the ratio before dividend initiation, the greater is the present value of growth opportunities assessed by the market before initiation. It is reasonable to assume that the more investment opportunities a firm has, the more cash it needs to invest in the future. If the dividend program is interpreted by the market as a signal of deterioration in investment opportunities, the negative effect may be larger for firms with higher Tobin’s Q ratio. The sign is expected to be negative. 6. Dividend Yield11 (DIV): Dividend yield is computed as the initial dividend amount divided by the share price two days before the initiation announcement. To the extent that the initiation announcement is unanticipated, the signaling and agency cost arguments imply that the abnormal return at announcement is positively related to the dividend yield. On the other hand, the deterioration in the investment opportunities argument implies a negative relation. 7. Pre-Announcement CAR (PRECAR): The CAR for each firm, from day ⫺20 through day ⫺2, is used to control for market anticipation of the initiation announcement. If observed negative abnormal returns result from market anticipation, PRECAR should be positive, and negatively related to the 2-day announcement CAR. 8. Earnings Change12 (EARNCH): This is a dummy variable, taking on a value of 1 if the quarterly earnings announcement just prior to dividend initiation represents an increase over the quarterly earnings four quarters ago, and 0 otherwise. Since an increase in earnings can lead to anticipation of dividend initiation, this variable is also used as a proxy for market anticipation; however, this variable may also proxy for corroboration effects between earnings and dividend announcements. By this interpretation, the abnormal return at announcement is larger for firms with a prior earnings increase. 2.5. Chow test To test whether the variables in the two subgroups are statistically different, a Chow test is performed. The procedure is defined as: Fc ⫽
共ESS t ⫺ EES p ⫺ ESS n兲/K 共ESS p ⫹ ESS n兲/共N ⫺ 2K兲
(2)
where ESSt ⫽ the sum of squared residuals obtained from the regression using the total sample; ESSp ⫽ the sum of squared residuals obtained from the regression using only the firms with positive announcement returns; ESSn ⫽ the sum of squared residuals obtained from the regression using only the firms with negative announcement returns; K ⫽ the number of regression coefficients; and N ⫽ the number of observations in the total sample. If, as expected, there is indeed a subset of firms for which dividend initiation announcements is a value-decreasing event due to their firm specific factors, then the variables proxying for these factors should be different from those of the firms for which dividend initiation is a value-increasing event. The purpose of this test procedure is to see whether the null hypothesis (that the variables in the two groups are not different) can be rejected.
270
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
Table 1 The means and standard deviations (in parentheses) of the variables of interest for the full sample, and for subgroups based on whether the CAR at announcement is positive (Group P), or negative CAR (Group n) Variable
Full sample (N ⫽ 157)
Group P Car ⬎ 0 (n ⫽ 102)
Group N Car ⬍ 0 (n ⫽ 55)
SIZE
25.90 (1.46) 0.21 (0.18) 11.52 (16.08) 19.68 (19.14) 0.95 (0.68) 0.83 (0.59) 3.07 11.12) 0.69 (0.43) 2.98 (6.66)
15.92 (1.29) 0.25 (0.18) 7.08 (8.83) 19.67 (19.18) 0.97 (0.78) 0.90 (0.57) ⫺0.02 (12.97) 0.79 (0.29) 6.16 (6.13)
64.69* (1.32) 0.14* (0.15) 19.76* (18.50) 19.72 (19.06) 0.94 (0.45) 0.70** (0.62) 9.08* (14.25) 0.51* (0.49) ⫺2.88* (2.48)
EARNVOL INST BOARD Q DIV PRECAR EARNCH Car
* P ⫽ .01, Group P vs. Group N. ** P ⫽ .05, Group P vs. Group N.
3. Empirical results 3.1. Preliminary analysis Table 1 reports the mean and the standard deviation of the variables used in the crosssectional regressions, for the entire sample, and separately for firms that experience a positive CAR at announcement (Group P), and those that experience a negative CAR (Group N).13 A comparison across the two groups gives a preliminary indication regarding the extent to which firms with negative CARs differ from those with positive CARs, in terms of ex ante characteristics. If negative CARs stem from market anticipation, estimation errors and confounding events, we expect the two subsamples to differ only with respect to variables that proxy for market anticipation. While the CARs at announcement differ significantly across the two groups, this is only to be expected since the groups are formed on the basis of announcement CARs. The subsample values of pre-announcement CARs (PRECAR) suggest that negative announcement CARs may result solely from market anticipation. Firms with positive announcement CARs exhibit mean value of PRECAR that are not materially different from zero, and insignificant. In contrast, firms with negative announcement CARs exhibit large positive mean value of pre-announcement CAR. In fact, the sum of pre-announcement and announcement CARs is almost identical for the two groups, and is positive in both cases. This
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
271
indicates that announcement CARs are negative for Group N only because of “overanticipation,” and that the total economic impact of initiation is the same for both groups. The proportion of firms with positive earnings changes prior to the announcement; however, is greater for Group P. This shows that initiation is more likely to be anticipated for Group P than for Group N. The two groups are also significantly different with respect to several other variables, such as firm size, earnings volatility, institutional holdings and earnings change. The directions of the differences are consistent with theoretical predictions and tend to explain the difference in announcement CARs. For instance, announcement CARs tend to be negative for larger firms (with greater availability of public information) and those with less volatile earnings. In both cases, the value of future information releases is likely to be low. Negative announcement CAR firms are also characterized by large institutional holdings. Thus, these are firms where agency costs are likely to be low to begin with because of closer institutional monitoring. Overall, negative announcement CARs tend to occur when the benefits of dividend initiation are likely to be small; the economic impact of the dividend initiation is affected by the perceived value of the dividend program as a future information release mechanism; and the extent to which the manager/stockholder agency problem is mitigated. The above results are suggestive of factors that determine the magnitude of CARs at the announcement of a dividend initiation program, and explain the incidence of negative CARs. However, it is necessary to examine the influence of each factor, controlling for other factors. Cross-sectional regressions are conducted for this purpose. Another motivation is to compare the explanatory powers of the two regressions for the two subsamples. If they are roughly the same, then there are reasons to believe that the negative announcement CARs are driven by the underlying economic variables to the same extent as positive announcement CARs, rather than being caused by estimation errors or confounding events. 3.2. Regression analysis Cross-sectional regressions are used to examine the relationship between the dependent variable, the 2-day CARs and the independent variables listed in Table 1. The results of the full sample and the two subsample regressions are reported in Table 2. For the full sample, the coefficient of firm size is significant and negative, consistent with the hypothesis that the larger the firm, the more information is available about the firm and the less valuable the dividend program is as an information releasing mechanism. As expected, the coefficient of earnings volatility is positive and significant at 1%. The higher the earnings volatility, the less information earnings announcements convey to the market in terms of the firm’s future prospects and the more valuable it is to have a dividend program to provide another information releasing mechanism. The coefficient of institutional holdings is negative and significant, consistent with both the agency’s cost reduction and information effect hypotheses. Dividend yield has a positive and significant coefficient, consistent with the notion that if a new dividend program is good news, the greater the amount of the initial dividend, the better. The coefficient of PRECAR is negative and significant, as expected. The observed market reaction inversely correlates with the degree of anticipation. The coefficient
272
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
Table 2 Results of multi-variate, cross-sectional regressions
Variables Intercept SIZE EARNVOL INST BOARD Q DIV PRECAR EARNCH R2 F-Statistics
Full
Group P
Group N
157 0.15* (0.029; 2.09) ⫺0.01*b (0.032; 2.23) 0.07** (0.054; 2.91) ⫺0.08* (0.029; 2.10) 0.01 (0.003; 0.68) 0.01 (0.002; 0.55) 1.76* (0.030; 2.23) ⫺0.06* (0.026; 2.01) 0.02* (0.026; 1.98) 0.34 11.15**
102 0.19* (0.044; 2.06) ⫺0.01* (0.040; 1.98) 0.02 (0.004; 0.63) 0.02 (0.000; 0.08) 0.03 (0.000; 0.13) 0.03 (0.002; 0.43) 4.40** (0.149; 4.04) ⫺0.03 (0.007; 0.80) ⫺0.01 (0.003; 0.49) 0.23 4.82**
55 0.05 (0.037; 1.33) ⫺0.00 (0.039; 1.37) 0.018 (0.034; 1.27) ⫺0.01** (0.125; 2.57) 0.02 (0.010; 0.68) 0.08 (0.024; 1.07) ⫺2.33** (0.325; 4.71) ⫺0.02 (0.051; 1.57) 0.01 (0.067; 1.82) 0.31 3.98**
Partial R-squares and T-statistics are reported in the parentheses. the subgroups are formed based on CAR. Group P consists of 102 firms with positive CARs and Group N consists of 55 firms with negative CARs.
of earnings change is positive and significant, supporting the notion of corroboration effects between earnings and dividend announcements. The coefficients of percentage of shares owned by the Board of Directors and Tobin’s Q ratio are insignificant. Overall, the regression explains 34% of the cross-sectional variation in announcement CARs. For the two subsample regressions, the coefficient of the dividend yield is significant and positive for Group P as predicted by dividend signaling and agency cost arguments but it is significant and negative for Group N. The coefficients of firm size in Group P and institutional holdings in Group N are significant and have the expected signs. The regressions explain 23% of the cross-sectional variation in announcement CARs for Group P and 31% for Group N. Two aspects of these results provide important evidence to reject the H2 and to support the contention that the observed negative announcement CARs for Group N indeed reflect the market’s assessment of the economic impact of initiation. First, the different signs of the yield coefficients show that the initial dividend appears to convey very different information to the market for the two groups. For Group P, the positive sign indicates that the dividend payment, in and of itself, is viewed as value-increasing, and the value increase is positively related to the initial dividend amount. However, for Group N, the dividend, in and of itself, is viewed as value-decreasing, and a larger initial dividend results in a greater decrease in value. Additionally, it has been well documented that the signaling effect, the dividend yield effect, and the agency cost reduction are the main reasons why many firms experience positive market reaction when the announcements of initial dividends are made. The fact that the value of R-squared for Group N is greater than that of Group P in the two subsample regressions indicates that the variables proxying for the determinants of the economic impact of initiation yield higher explanatory power for Group N than for Group P. This shows that the dependent variable is more closely related to the independent variables in Group N than that in Group P and that the negative reactions are due to economic reasons and not
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
273
Table 3 Results of the Chow test 8.9338** ⫽
共0.43679 ⫺ 0.018598 ⫺ 0.271248兲/8 共0.018598 ⫹ 0.271248兲 / 共157 ⫺ 16兲
estimation errors. This further supports the notion that the negative announcement returns for these firms are due to economic reasons and not estimation errors. 3.3. The Chow test The Chow test is performed to test H1. To say that the dividend initiation is a valueincreasing event for some firms and a value-decreasing event for others, it is necessary to establish that the two groups of firms are fundamentally different. The result of the Chow test is reported in Table 3. It rejects the null hypothesis that there is no difference between the variables in Group P and Group N at the 1% significance level. This clearly shows that there are indeed two groups of firms with different characteristics that result in different market reaction to dividend initiation announcements. 3.4. Analysis of direct evidence Direct evidence from the 55 firms in Group N is collected and analyzed to see whether the regression results can be corroborated. The sources of the direct evidence are news stories from the Wall Street Journal Index, the Moody’s Industrial Manual and the annual reports of the firms in question. A summary of the evidence collected is reported in Table 4. The events reported are: acquisitions (firms that acquired other firms within two years after Table 4 Summary of the direct evidence gathered form the 55 firms that experience negative market reaction after dividend initiation
Acquisition Target Seasoned issue Loss Dividend suspension Debt
Acquisition
Target
Seasoned issue
Loss
Divided suspension
Debt
20
0 11
1 0 6
1 1 0 13
1 0 0 1 3
2 2 1 4 0 12
The numbers in the diagonal cells are the total number of firms in which the particular events occur. The numbers in other cross cells indicate the number of firms in which two events occur around the same time when initiation announcements are made. For instance, there are altogether 20 firms that were engaged in acquisitions. None of the 20 firms was a take-over target, one issued new shares, one reported loss, one subsequently suspended dividends and two issued new debt.
274
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
dividend initiation); targets (firms which were takeover targets); seasoned issues (firms that issued additional shares); losses (firms which reported losses within a year after initiation was announced); dividend suspension (firms that suspended dividend payments less than two years after initiation); debt (firms that issued significant new debt shortly after). The numbers reported are the total number of firms in which a particular event took place. In cases where two events occurred within one firm, the numbers are reported in the cross-cells. There is no case where three or more events took place within one firm. Within two years of starting a new dividend program, 20 out of the 55 firms (36%) acquired other firms, either in part or as a whole. This may provide indirect evidence that a possible reason the firms started dividend programs was due to a lack of internal growth opportunities. If the market viewed these firms as growth companies and interpreted the dividend initiation as a signal that the firms no longer had the kind of growth prospects they used to have, the response would be negative. Eleven firms (20%) in the group were takeover targets around the time of dividend initiation announcements. It may not serve the best interest of the shareholders if management initiated a new dividend program for defensive purposes; thus, the market reaction could also be negative. As shown in Table 4, there is no overlap between the firms that were takeover targets and those that acquired other firms. This means that 56% of the firms that experienced negative market reaction either acquired other firms or became a takeover target shortly after dividend initiation. There are six firms in the group that issued new shares and 12 firms that issued new debt, with one firm doing both. As discussed previously, one benefit of a dividend program is the signaling effect in that it provides a mechanism with which to send credible signals to the market that the management is confident about future cash flow. The fact that these firms were also raising money around the same time mitigates the positive signaling effect and may result in negative market reaction. Among the 13 firms in this group that reported a loss, four firms also issued new debt, one firm acquired another firm, and one firm was a takeover target. Starting a new dividend program while incurring losses certainly does not send a credible signal to the investors. The market may also be suspicious of the management’s motivation in face of the loss. Finally, three firms had to omit paying dividends within less than two years of initiation due to financial difficulties. In retrospect, the decision to start a regular dividend program may not have been prudent for these firms, and the market’s negative reaction to initiation may have anticipated this imprudence.
4. Conclusion and summary This paper examines the apparent heterogeneity in the stock price reaction to dividend initiation announcements. It offers evidence that dividend-initiating firms fall into two distinct categories, with initiation being a value-increasing event for one and a valuedecreasing event for the other due to different firm specific characteristics. The results of this study strongly suggest that market reaction to dividend initiation announcements be based on the net effects of the costs and benefits of a new dividend program, which differ from firm to firm. It shows that a seemingly positive corporate event, such as starting a new dividend
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
275
program, could very well be perceived by the market as having a negative impact on the value of the company. The findings of this paper may have implications for future studies on other corporate events, such as stock repurchase, stock split, seasoned new issue, dividend omission, layoffs and so on. It will also be interesting to study whether heterogeneous market reactions to these events exist, and how market reactions are affected by various firm specific factors. The answers to these questions will certainly deepen our understanding of the market implications of these important corporate events.
Acknowledgment I want to express my gratitude to the two anonymous referees for their constructive comments and suggestions. I am also grateful to my colleagues Mona Gardner and Robert Leekley for their valuable help during the revision process.
Notes 1. For example, Asquith and Mullins (1983), Benesh et al. (1984), Healy and Palepu (1988), Venkatesh (1989), Mitra and Owers (1990), Rimbey and Officer (1992). 2. Pettit (1972), Aharony and Swary (1980), Asquith and Mullins (1983), Brickley (1983), Zeghal (1983), Dielman and Oppenheimer (1984), Kalay and Lowenstein (1985), Eades et al. (1986), Eddy and Seifert (1988), Lang and Litzenberger (1989), Mitra and Owers (1990). 3. One possible explanation for why such a firm initiates dividends is that the managers had judged the benefits to exceed the costs. Note that such a disagreement between the managers and the market can happen even when both parties have unbiased beliefs. 4. Although initiation announcements are screened for confounding events in the Wall Street Journal Index, some of those events may not be detected in this screening. Further, intraindustry information transfers, whereby an announcement by one firm affects returns for other firms in the same industry, may also be considered confounding events. 5. The firm also had to be listed for at least two years before paying dividends. These are the same criteria used by Asquith and Mullins (1983) in their study. 6. The number of firms in the industry index ranges from nine to 55, with an average of 20. 7. Source: Compustat 8. Source: Standard & Poor’s Security Owner’s Guide. 9. Source: proxy statements. 10. Source: Compustat. The calculation of Tobin’s Q is based on an algorithm contained in McConnell and Servaes (1990). The market value of the firm is the sum of the market value of equity and the book value of debt. The replacement value of plant and equipment is calculated by approximating the plant’s age and then adjusting its book
276
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
value for inflation. The replacement value of inventory is assumed to equal book value, unless the firm used last-in-first-out (LIFO) accounting. If LIFO is used, the book value is adjusted for inflation. The replacement value of the firm is computed as the book value of assets minus the book value of inventories minus the book value of net plant plus replacement value of inventory plus replacement value of plant and equipment. 11. Source: The initial dividend is obtained from the CRSP monthly file and confirmed through The Wall Street Journal Index. The share price is collected from the NYSE/ AMEX Daily Stock Record. 12. Source: Compustat 13. For the entire sample, the correlation coefficients across the independent variables range from a minimum of ⫺0.19 and a maximum of 0.17, indicating that multicollinearity is not a problem.
References Aharony, J., & Swary, I. (1980). Quarterly dividend and earnings announcements and stockholders’ returns: an empirical analysis. Journal of Finance 35, 1–12. Asquith, P., & Mullins, D. (1983). The impact of initiating dividend payments on shareholders’ wealth. Journal of Business 46, 77–96. Asquith, P., & Mullins, D. (1986). Signaling with dividends, stock repurchases and equity issues. Financial Management (Autumn) , 27– 44. Bhattacharya, S. (1979). Imperfect information, dividend policy and ‘the bird in the hand fallacy’. Bell Journal of Economics 10(Spring), 259 –270. Benesh, G., Keown, A., & Pinkerton, J. (1984). An examination of market reaction to substantial shifts in dividend policy. Journal of Financial Research 7, 131–142. Black, F. (1976). The dividend puzzle. Journal of Portfolio Management 2(2), 72–77. Brickley, J. (1983). Shareholder wealth, information signaling and the specially designated dividend. Journal of Financial Economics 12, 187–209. Dielman, T. E., & Oppenheimer, H. R. (1984). An examination of investor behavior during periods of large dividend changes. Journal of Financial and Quantitative Analysis 19(2), 197–216. Eades, K., Hess, P., & Kim, E. (1986). Market rationality and dividend announcements. Journal of Financial Economics 14, 581– 604. Easterbrook, F. (1984). Two agency-cost explanations of dividends. American Economic Review 74(September), 650 – 659. Eddy, A., & Seifert, B. (1988). Firm size and dividend announcements. The Journal of Financial Research 4(Winter), 295–302. Healy, P., & Palepu, K. (1988). Earnings information conveyed by dividend initiation and omissions. Journal of Financial Economics 21, 149 –175. Jensen, M. (1986). Agency costs of free cash flow, corporate finance and takeovers. American Economic Review 76(May), 323–329. John, K., & Williams, J. (1985). Dividends, dilution and taxes: a signaling equilibrium. Journal of Finance 40(September), 1053–1070. Kalay, A., & Loewenstein, L. (1985). Predictable events and excess returns: the case of dividend announcements. Journal of Financial Economics 14, 423– 449. Lang, L., & Litzenberger, R. (1989). Dividend announcements: cash flow signaling vs. free cash flow hypothesis? Journal of Financial Economics 24, 181–191.
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
277
McConnell, J. J., & Servaes, H. (1990). Additional evidence on equity ownership and corporate value. Journal of Financial Economics 27(2), 595– 612. Miller, M., & Modigliani, F. (1961). Dividend policy, growth, and the valuation of shares. Journal of Business 34, 411– 433. Miller, M., & Rock, K. (1985). Dividend policy under asymmetric information. Journal of Finance 40(September), 1031–1052. Mitra, D., & Owers, J. (1990). The information content of dividend initiations in asymmetrical information environments. Working paper presented at the 1990 FMA Meetings. Myers, S. C., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have. NBER Working Paper: 1396 (July). Pettit, R. R. (1972). Dividend announcements, security performance, and capital market efficiency. Journal of Finance 5(December), 993–1007. Rimbey, J., & Officer, D. (1992). The market response to subsequent dividend actions of dividend-initiating and -omitting firms. Quarterly Journal of Business and Economics 31(Winter), 3–20. Venkatesh, P.C. (1989). The impact of dividend initiation on the information content of earnings announcements and return volatility. Journal of Business 62, 175–197. Zeghal, D. (1983). Firm size and the informational content of financial statements. Journal of Financial Economics 12(September), 299 –310.
On the differential market reaction to dividend initiations Zhenhu Jina** a
Illinois Wesleyan University, P.O. Box 2900 Bloomington, IL 67102-2900, USA
JEL Classifications: G12, G14, G32, G35 Keywords: Dividend initiation; Payout policy; Event study; Security pricing
Abstract The finance literature documents substantial positive stock price reaction to dividend initiations. Most dividend initiation studies focus on the average positive reaction; however, 40 percent of the firms that initiate dividends experience negative abnormal returns at announcement. This paper focuses on the apparent heterogeneity in the stock price reaction to dividend initiation. I find that the observed negative market reaction reflects the market’s economic assessment of the impact of the event on these firms, and that it is not caused by anticipation or confounding events. The result is also supported by the fact that the market reaction to dividend initiation for these firms is negatively related to initial dividend yield. Both the positive and negative observed reactions are consistent with conventional arguments regarding the information content of dividends, and their role in mitigating agency problems. © 2000 Bureau of Economic and Business Research, University of Illinois. All rights reserved.
1. Introduction Although studies1 report that 30 to 40 percent of the firms that initiate dividends experience negative abnormal returns at announcement, most research focuses on the positive average reaction rather than the apparent heterogeneity in the stock price reaction to dividend initiation. This paper addresses whether there are two fundamentally different types of dividend initiating firms, with dividend initiation being assessed by the market as a
* Tel.: ⫹1-309-556-3570. E-mail address: [email protected]. 1062-9769/00/$ – see front matter © 2000 Bureau of Economic and Business Research, University of Illinois. All rights reserved. PII: S 1 0 6 2 - 9 7 6 9 ( 9 9 ) 0 0 0 5 2 - 6
264
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
positive event for one type, and a negative event for the other. The key issue is to determine whether the observed negative reaction for a subset of the firms actually reflects an assessment by the market that initiation is a value-decreasing event due to firm specific factors. A Chow test is first conducted to determine whether the firms in the two subgroups are significantly different, and the result of the test rejects the hypothesis, at one percent significance level, that there is no difference between the two groups. Cross-sectional regressions using variables that proxy for the various determinants of the dividend initiation announcement reaction are then performed to test the relationship between market reaction and these firm specific factors. The regressions are run on the whole sample as well as two subgroups: Group P, firms with positive market reaction, and Group N, firms with negative market reaction. The whole sample regression results indicate that market reaction to dividend initiation announcements is significantly related to many of the firm specific variables in the model. The results of the two subsample regressions show that the Cumulative Abnormal Returns (CAR) are significantly negatively related to the initial dividend for Group N, but are significantly positively related for Group P. Direct evidence about the firms that have negative announcement returns is then collected from the Wall Street Journal Index, Moody’s Industrial Manual, and the annual reports of the firms in question to see whether the regression results can be supported. The information gathered from these sources corroborates, in general, the regression results. All these results clearly show that market reaction to dividend initiation announcements is affected by firm specific variables and that there might indeed be firms for which dividend initiation is viewed by the market as a negative event due to firm-specific factors. The paper proceeds as follows. In section one, I present the theory and hypothesis along with a short review of the literature. Section two presents the data and the empirical methodology. Section three presents the specific models and results. Finally, section four summarizes the results and conclusions.
1. Theoretical background Miller and Modigliani (1961) demonstrate that firm value is independent of dividend policy in a perfect capital market setting. However, they speculate that when information asymmetry exists, dividend announcements may be a means of transmitting information from well informed insiders to outsiders. Bhattacharya (1979), Asquith and Mullins (1986), John and Williams (1985) and Miller and Rock (1985) develop models consistent with this “signaling hypothesis.” In contrast, Easterbrook (1984) and Jensen (1986) use agency theory to argue that dividend payments can reduce agency costs. These two hypotheses have been examined empirically,2 and results of these studies are summarized briefly in the following paragraphs. There is general agreement in the empirical and theoretical literature in finance that, on average, dividend initiations have a positive impact on firm value. Many researchers argue that unanticipated initiation announcements serve as an optimistic signal of the firm’s future performance. In addition, the initiation of a dividend program puts in place a mechanism for
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
265
periodic future information releases, that serves to reduce information asymmetry. Asquith and Mullins (1983) point out conceptually that the abnormal return associated with dividend initiation can be decomposed into two components, a yield effect and an initiation effect. The initiation effect reflects the value resulting from the sequence of future dividend payments and captures the benefits from initiating a dividend program per se. The yield effect consists of the value change resulting from the magnitude of the initial dividend and captures the information content of the dividend payment. Agency theorists Jensen (1986) and Easterbrook (1984) identify a second factor that can potentially explain the positive wealth effects of dividend initiation. Jensen argues that dividend payments reduce the agency cost arising from managers’ incentives to use free cash flow to invest in negative NPV projects. Not only does a dividend payment reduce the cash available to managers for potential investment in negative NPV projects today, but to the extent that managers are reluctant to reduce dividends, it also offers the prospect of reduced availability of cash for such investment in future years. Easterbrook argues that a dividend payment today increases the probability that the firm will need to seek external financing from the capital market in the future. Thus, there is a higher probability of managers being exposed to the monitoring associated with external financing. This higher probability of future monitoring reduces the extent to which managers will deviate from stockholder wealth maximization, thus reducing the costs of the agency conflict between managers and stockholders. As in the information content argument, there is both a yield effect, and an initiation effect associated with a dividend initiation announcement. The yield effect consists of the reduction in agency costs resulting from the initial dividend; the initiation effect consists of the present value of the reductions in agency costs expected from the implicit sequence of future dividends. Both information content and agency cost arguments identify potential benefits of dividend initiation. However, the actual economic impact of initiation will consist of these benefits netted against the costs associated with payment of the initial dividend, as well as the implicit sequence of future dividends. These costs include the tax penalty on dividends (relative to capital gains), the increase in expected financing costs (due to the higher probability of external financing) and any costs associated with a reduction in financial slack (Black, 1976; Bhattacharya, 1979; Myers and Majluf, 1984). Thus, for a particular firm, if the market judges the benefits of initiation to be smaller than the costs, we would expect to observe negative abnormal returns at announcement.3 Furthermore, market reaction to a particular firm’s dividend initiation announcement can also be affected by market anticipation. Abnormal returns will reflect only the unanticipated component of the announcement. If the initiation announcement is fully anticipated, and the initial dividend is less than expected, the abnormal return at announcement could be negative, although initiation itself is a value-increasing event. Finally, although formal dividend signaling models suggest that unexpected dividend increases are always good news, the literature has long recognized the possibility that, for some firms, a dividend increase could signal a lack of profitable investment opportunities. By this argument, an unexpected dividend increase would convey an unanticipated deteri-
266
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
oration in investment opportunities, causing the market to revise the value of the firm downward. Take note that conventional signaling models, as well as the Jensen and Easterbrook agency arguments, predict that the announcement effect should be positively related to the initial dividend. Under these arguments, even when the total announcement effect is negative, there is still a benefit of dividend payment that is positively related to the dividend amount. Only if the dividend itself is bad news, because it conveys a deterioration in investment opportunities, do we expect to see the announcement returns negatively related to the initial dividend.
2. Hypotheses, data and methodology 2.1. Hypotheses The question addressed by this paper is whether dividend initiation is assessed by the market to be a value-increasing event for some firms and a value-decreasing event for others. The fact that 30 to 40 percent of dividend initiating-firms experience negative abnormal returns does not, in itself, establish that such a dichotomy exists. In an event study, some sample firms could exhibit negative abnormal returns ex post even if the average impact of the event is positive for all firms. The abnormal return at announcement can be negative when the expected economic impact of dividend initiation is positive for several reasons, such as market anticipation of the event, errors in estimating abnormal return, and confounding events.4 Thus, the key issue addressed in this paper is whether the negative abnormal returns observed at announcement for a subset of dividend-initiating firms reflect the market’s negative assessment of the event due to firm specific economic factors. Specifically, the following two hypotheses are examined: H1: Firms in Group P and Group N are from the same population and there are no fundamental differences between the two groups. H2: The negative market reaction to dividend initiation announcements is caused by anticipation, estimation error or confounding events and is not related to firm-specific, economic factors.
Chow test and cross-section multivariate regression models are used to test the hypotheses. If a new dividend program is viewed as value-increasing for one group of firms and value-decreasing for another for economic reasons, the firms in the two groups should be significantly different in some important aspects. At the same time, if the CAR for Group N are negative, on average, due to anticipation or estimation errors or other confounding events, there should be little relation between the observed CARs and the variables proxying for the determinants of the dividend initiation announcement reaction. As such, the adjusted R-squared of the regression for Group N should be lower than that for Group P and that the announcement CAR for Group N should not be significantly related to the initial dividend.
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
267
2.2. The data The initial data set consists of firms on the Center for Research in Security Prices (CRSP) NYSE/AMEX monthly file that paid a cash dividend for the first time in their history, or resumed paying cash dividends after 10 years without dividends, between 1973 to 1993.5 These events are verified using the Wall Street Journal Index. The firms in the sample also pass the following screens. 1. Common stock daily returns starting from 300 days before the initiation announcement are available from the CRSP daily returns file. 2. There are no confounding announcements such as earnings reports within five days of the initiation announcement. This is done so as not to compound market reaction to the announcement of dividend initiation with earnings announcement. While doing so might introduce some bias, it is mitigated by the fact that earnings volatility and earnings changes are two of the variables used in the regression model. 3. There are no major corporate events within the 300 days before the initiation announcement that would significantly change the nature or risk level of the firm, such as a merger or restructuring. 4. The firm is included in the Compustat Annual or Research files. 5. The firm’s proxy statement in the year of initiation, and a year before and after, are available. The screening process yields a sample of 157 firms. 2.3. Estimation of firm announcement reaction Abnormal returns at announcement are estimated by employing an expanded market model, including an industry index. The coefficients are estimated using 300 days’ return data prior to two days before the announcement. In particular, the abnormal return of Firm K at time t (ARkt) is defined as: ARkt ⫽ Rkt ⫺ (␣k ⫹ kRmt ⫹ 2kRind,t),
(1)
where Rkt is the observed return of Firm K on day t; Rmt is the return on the CRSP equally weighted index; and Rind,t is the return on an equally weighted portfolio of firms in the same industry as Firm K on day t. Firms are considered to be in the same industry if the first three digits of the four-digit Standard Industry Classification Code (SIC) are the same.6 Following Asquith and Mullins and other researchers, the market reaction to Firm K’s initiation announcement is defined as the 2-day CARk. It is the sum of Firm K’s abnormal return from day t-1 to day t. Day t is the day that the initial dividend announcement appeared in the Wall Street Journal. Thus, the 2-day CAR is computed as the sum of Firm K’s abnormal return on days t-1 and t. The basic assumption here is that the market immediately reassesses the company following the announcement of the new dividend program and reacts to it. (As a precaution, 3, 5 and 10 day CARs are also computed to see whether market reaction centers on the two days, and the answer is affirmative. The results, therefore, are based on the 2-day CARs.) Of the 157 firms in the sample, 102 firms (64.9%) have positive
268
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
cumulative abnormal returns and 55 firms (35.2%) have negative abnormal returns. The average 2-day CAR for all 157 firms is 2.98%, with a median of 1.43% and a standard deviation of 4.66%. The average CAR for the 102 firms that have positive returns is 6.16%, with a median of 4.26% and standard deviation of 4.13%. The average CAR for the 55 firms that have negative returns is ⫺2.88%. The median and standard deviations are ⫺1.94% and 1.48% respectively. 2.4. Proxies for the determinants of initiation abnormal returns Cross-sectional regressions are performed on the full group and the two subgroups to examine the relationship between the variables that proxy for the determinants of initiation abnormal return and the observed announcement reaction and the explanatory power of these variables in the two subgroups. The proxies used are as follows: 1. Firm Size (SIZE): The firm size is measured as the natural log of the market value of equity as of two days before the announcement. Zeghal (1983), Eddy and Seifert (1988) and Mitra and Owers (1990) argue that firm size is a good proxy for the degree of publicly available information about a firm; the larger the firm, the greater the availability of information. Thus, the value of instituting a mechanism for periodic information releases may be greater for small firms than for large firms. The sign of the coefficient is expected to be negative. 2. Earnings Volatility7 (EARNVOL): Earnings volatility is estimated as the standard deviation of earnings per share over the 16 quarters immediately preceding the initiation announcement. If part of the benefit of a regular dividend program is to provide an information-releasing mechanism, then the information provided by quarterly dividend announcements will be more valuable for firms with less predictable earnings. On the other hand, for firms with stable earnings, the value of the additional information from the dividend program may not be as significant. The sign is expected to be positive. 3. Institutional Holdings8 (INST): The fraction of outstanding shares held by institutions is used as a proxy for the intensity of monitoring that the firm is subjected to by institutions. It is hypothesized that the agency costs arising from the manager-stockholder conflict are smaller when institutions monitor the firm more closely. Thus, the benefit of agency cost reduction may be smaller for firms with large institutional holdings. An alternative interpretation of this variable is that heavy institutional holdings are associated with greater information availability about the firm, reducing the benefit from periodic future information releases. Both interpretations predict a negative coefficient. 4. Board Ownership9 (BOARD): The percentage of shares owned by members of the board of directors is used here as a proxy of the board’s monitoring activities. Higher board ownership should result in more intensive board monitoring, less serious agency problems, less benefit from the reduction of agency costs, and thus a greater likelihood of initiation being value-decreasing.
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
269
5. Tobin’s Q ratio (Q10): This ratio is widely used, as in this study, as a proxy for the firm’s future growth opportunities. The higher the ratio before dividend initiation, the greater is the present value of growth opportunities assessed by the market before initiation. It is reasonable to assume that the more investment opportunities a firm has, the more cash it needs to invest in the future. If the dividend program is interpreted by the market as a signal of deterioration in investment opportunities, the negative effect may be larger for firms with higher Tobin’s Q ratio. The sign is expected to be negative. 6. Dividend Yield11 (DIV): Dividend yield is computed as the initial dividend amount divided by the share price two days before the initiation announcement. To the extent that the initiation announcement is unanticipated, the signaling and agency cost arguments imply that the abnormal return at announcement is positively related to the dividend yield. On the other hand, the deterioration in the investment opportunities argument implies a negative relation. 7. Pre-Announcement CAR (PRECAR): The CAR for each firm, from day ⫺20 through day ⫺2, is used to control for market anticipation of the initiation announcement. If observed negative abnormal returns result from market anticipation, PRECAR should be positive, and negatively related to the 2-day announcement CAR. 8. Earnings Change12 (EARNCH): This is a dummy variable, taking on a value of 1 if the quarterly earnings announcement just prior to dividend initiation represents an increase over the quarterly earnings four quarters ago, and 0 otherwise. Since an increase in earnings can lead to anticipation of dividend initiation, this variable is also used as a proxy for market anticipation; however, this variable may also proxy for corroboration effects between earnings and dividend announcements. By this interpretation, the abnormal return at announcement is larger for firms with a prior earnings increase. 2.5. Chow test To test whether the variables in the two subgroups are statistically different, a Chow test is performed. The procedure is defined as: Fc ⫽
共ESS t ⫺ EES p ⫺ ESS n兲/K 共ESS p ⫹ ESS n兲/共N ⫺ 2K兲
(2)
where ESSt ⫽ the sum of squared residuals obtained from the regression using the total sample; ESSp ⫽ the sum of squared residuals obtained from the regression using only the firms with positive announcement returns; ESSn ⫽ the sum of squared residuals obtained from the regression using only the firms with negative announcement returns; K ⫽ the number of regression coefficients; and N ⫽ the number of observations in the total sample. If, as expected, there is indeed a subset of firms for which dividend initiation announcements is a value-decreasing event due to their firm specific factors, then the variables proxying for these factors should be different from those of the firms for which dividend initiation is a value-increasing event. The purpose of this test procedure is to see whether the null hypothesis (that the variables in the two groups are not different) can be rejected.
270
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
Table 1 The means and standard deviations (in parentheses) of the variables of interest for the full sample, and for subgroups based on whether the CAR at announcement is positive (Group P), or negative CAR (Group n) Variable
Full sample (N ⫽ 157)
Group P Car ⬎ 0 (n ⫽ 102)
Group N Car ⬍ 0 (n ⫽ 55)
SIZE
25.90 (1.46) 0.21 (0.18) 11.52 (16.08) 19.68 (19.14) 0.95 (0.68) 0.83 (0.59) 3.07 11.12) 0.69 (0.43) 2.98 (6.66)
15.92 (1.29) 0.25 (0.18) 7.08 (8.83) 19.67 (19.18) 0.97 (0.78) 0.90 (0.57) ⫺0.02 (12.97) 0.79 (0.29) 6.16 (6.13)
64.69* (1.32) 0.14* (0.15) 19.76* (18.50) 19.72 (19.06) 0.94 (0.45) 0.70** (0.62) 9.08* (14.25) 0.51* (0.49) ⫺2.88* (2.48)
EARNVOL INST BOARD Q DIV PRECAR EARNCH Car
* P ⫽ .01, Group P vs. Group N. ** P ⫽ .05, Group P vs. Group N.
3. Empirical results 3.1. Preliminary analysis Table 1 reports the mean and the standard deviation of the variables used in the crosssectional regressions, for the entire sample, and separately for firms that experience a positive CAR at announcement (Group P), and those that experience a negative CAR (Group N).13 A comparison across the two groups gives a preliminary indication regarding the extent to which firms with negative CARs differ from those with positive CARs, in terms of ex ante characteristics. If negative CARs stem from market anticipation, estimation errors and confounding events, we expect the two subsamples to differ only with respect to variables that proxy for market anticipation. While the CARs at announcement differ significantly across the two groups, this is only to be expected since the groups are formed on the basis of announcement CARs. The subsample values of pre-announcement CARs (PRECAR) suggest that negative announcement CARs may result solely from market anticipation. Firms with positive announcement CARs exhibit mean value of PRECAR that are not materially different from zero, and insignificant. In contrast, firms with negative announcement CARs exhibit large positive mean value of pre-announcement CAR. In fact, the sum of pre-announcement and announcement CARs is almost identical for the two groups, and is positive in both cases. This
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
271
indicates that announcement CARs are negative for Group N only because of “overanticipation,” and that the total economic impact of initiation is the same for both groups. The proportion of firms with positive earnings changes prior to the announcement; however, is greater for Group P. This shows that initiation is more likely to be anticipated for Group P than for Group N. The two groups are also significantly different with respect to several other variables, such as firm size, earnings volatility, institutional holdings and earnings change. The directions of the differences are consistent with theoretical predictions and tend to explain the difference in announcement CARs. For instance, announcement CARs tend to be negative for larger firms (with greater availability of public information) and those with less volatile earnings. In both cases, the value of future information releases is likely to be low. Negative announcement CAR firms are also characterized by large institutional holdings. Thus, these are firms where agency costs are likely to be low to begin with because of closer institutional monitoring. Overall, negative announcement CARs tend to occur when the benefits of dividend initiation are likely to be small; the economic impact of the dividend initiation is affected by the perceived value of the dividend program as a future information release mechanism; and the extent to which the manager/stockholder agency problem is mitigated. The above results are suggestive of factors that determine the magnitude of CARs at the announcement of a dividend initiation program, and explain the incidence of negative CARs. However, it is necessary to examine the influence of each factor, controlling for other factors. Cross-sectional regressions are conducted for this purpose. Another motivation is to compare the explanatory powers of the two regressions for the two subsamples. If they are roughly the same, then there are reasons to believe that the negative announcement CARs are driven by the underlying economic variables to the same extent as positive announcement CARs, rather than being caused by estimation errors or confounding events. 3.2. Regression analysis Cross-sectional regressions are used to examine the relationship between the dependent variable, the 2-day CARs and the independent variables listed in Table 1. The results of the full sample and the two subsample regressions are reported in Table 2. For the full sample, the coefficient of firm size is significant and negative, consistent with the hypothesis that the larger the firm, the more information is available about the firm and the less valuable the dividend program is as an information releasing mechanism. As expected, the coefficient of earnings volatility is positive and significant at 1%. The higher the earnings volatility, the less information earnings announcements convey to the market in terms of the firm’s future prospects and the more valuable it is to have a dividend program to provide another information releasing mechanism. The coefficient of institutional holdings is negative and significant, consistent with both the agency’s cost reduction and information effect hypotheses. Dividend yield has a positive and significant coefficient, consistent with the notion that if a new dividend program is good news, the greater the amount of the initial dividend, the better. The coefficient of PRECAR is negative and significant, as expected. The observed market reaction inversely correlates with the degree of anticipation. The coefficient
272
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
Table 2 Results of multi-variate, cross-sectional regressions
Variables Intercept SIZE EARNVOL INST BOARD Q DIV PRECAR EARNCH R2 F-Statistics
Full
Group P
Group N
157 0.15* (0.029; 2.09) ⫺0.01*b (0.032; 2.23) 0.07** (0.054; 2.91) ⫺0.08* (0.029; 2.10) 0.01 (0.003; 0.68) 0.01 (0.002; 0.55) 1.76* (0.030; 2.23) ⫺0.06* (0.026; 2.01) 0.02* (0.026; 1.98) 0.34 11.15**
102 0.19* (0.044; 2.06) ⫺0.01* (0.040; 1.98) 0.02 (0.004; 0.63) 0.02 (0.000; 0.08) 0.03 (0.000; 0.13) 0.03 (0.002; 0.43) 4.40** (0.149; 4.04) ⫺0.03 (0.007; 0.80) ⫺0.01 (0.003; 0.49) 0.23 4.82**
55 0.05 (0.037; 1.33) ⫺0.00 (0.039; 1.37) 0.018 (0.034; 1.27) ⫺0.01** (0.125; 2.57) 0.02 (0.010; 0.68) 0.08 (0.024; 1.07) ⫺2.33** (0.325; 4.71) ⫺0.02 (0.051; 1.57) 0.01 (0.067; 1.82) 0.31 3.98**
Partial R-squares and T-statistics are reported in the parentheses. the subgroups are formed based on CAR. Group P consists of 102 firms with positive CARs and Group N consists of 55 firms with negative CARs.
of earnings change is positive and significant, supporting the notion of corroboration effects between earnings and dividend announcements. The coefficients of percentage of shares owned by the Board of Directors and Tobin’s Q ratio are insignificant. Overall, the regression explains 34% of the cross-sectional variation in announcement CARs. For the two subsample regressions, the coefficient of the dividend yield is significant and positive for Group P as predicted by dividend signaling and agency cost arguments but it is significant and negative for Group N. The coefficients of firm size in Group P and institutional holdings in Group N are significant and have the expected signs. The regressions explain 23% of the cross-sectional variation in announcement CARs for Group P and 31% for Group N. Two aspects of these results provide important evidence to reject the H2 and to support the contention that the observed negative announcement CARs for Group N indeed reflect the market’s assessment of the economic impact of initiation. First, the different signs of the yield coefficients show that the initial dividend appears to convey very different information to the market for the two groups. For Group P, the positive sign indicates that the dividend payment, in and of itself, is viewed as value-increasing, and the value increase is positively related to the initial dividend amount. However, for Group N, the dividend, in and of itself, is viewed as value-decreasing, and a larger initial dividend results in a greater decrease in value. Additionally, it has been well documented that the signaling effect, the dividend yield effect, and the agency cost reduction are the main reasons why many firms experience positive market reaction when the announcements of initial dividends are made. The fact that the value of R-squared for Group N is greater than that of Group P in the two subsample regressions indicates that the variables proxying for the determinants of the economic impact of initiation yield higher explanatory power for Group N than for Group P. This shows that the dependent variable is more closely related to the independent variables in Group N than that in Group P and that the negative reactions are due to economic reasons and not
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
273
Table 3 Results of the Chow test 8.9338** ⫽
共0.43679 ⫺ 0.018598 ⫺ 0.271248兲/8 共0.018598 ⫹ 0.271248兲 / 共157 ⫺ 16兲
estimation errors. This further supports the notion that the negative announcement returns for these firms are due to economic reasons and not estimation errors. 3.3. The Chow test The Chow test is performed to test H1. To say that the dividend initiation is a valueincreasing event for some firms and a value-decreasing event for others, it is necessary to establish that the two groups of firms are fundamentally different. The result of the Chow test is reported in Table 3. It rejects the null hypothesis that there is no difference between the variables in Group P and Group N at the 1% significance level. This clearly shows that there are indeed two groups of firms with different characteristics that result in different market reaction to dividend initiation announcements. 3.4. Analysis of direct evidence Direct evidence from the 55 firms in Group N is collected and analyzed to see whether the regression results can be corroborated. The sources of the direct evidence are news stories from the Wall Street Journal Index, the Moody’s Industrial Manual and the annual reports of the firms in question. A summary of the evidence collected is reported in Table 4. The events reported are: acquisitions (firms that acquired other firms within two years after Table 4 Summary of the direct evidence gathered form the 55 firms that experience negative market reaction after dividend initiation
Acquisition Target Seasoned issue Loss Dividend suspension Debt
Acquisition
Target
Seasoned issue
Loss
Divided suspension
Debt
20
0 11
1 0 6
1 1 0 13
1 0 0 1 3
2 2 1 4 0 12
The numbers in the diagonal cells are the total number of firms in which the particular events occur. The numbers in other cross cells indicate the number of firms in which two events occur around the same time when initiation announcements are made. For instance, there are altogether 20 firms that were engaged in acquisitions. None of the 20 firms was a take-over target, one issued new shares, one reported loss, one subsequently suspended dividends and two issued new debt.
274
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
dividend initiation); targets (firms which were takeover targets); seasoned issues (firms that issued additional shares); losses (firms which reported losses within a year after initiation was announced); dividend suspension (firms that suspended dividend payments less than two years after initiation); debt (firms that issued significant new debt shortly after). The numbers reported are the total number of firms in which a particular event took place. In cases where two events occurred within one firm, the numbers are reported in the cross-cells. There is no case where three or more events took place within one firm. Within two years of starting a new dividend program, 20 out of the 55 firms (36%) acquired other firms, either in part or as a whole. This may provide indirect evidence that a possible reason the firms started dividend programs was due to a lack of internal growth opportunities. If the market viewed these firms as growth companies and interpreted the dividend initiation as a signal that the firms no longer had the kind of growth prospects they used to have, the response would be negative. Eleven firms (20%) in the group were takeover targets around the time of dividend initiation announcements. It may not serve the best interest of the shareholders if management initiated a new dividend program for defensive purposes; thus, the market reaction could also be negative. As shown in Table 4, there is no overlap between the firms that were takeover targets and those that acquired other firms. This means that 56% of the firms that experienced negative market reaction either acquired other firms or became a takeover target shortly after dividend initiation. There are six firms in the group that issued new shares and 12 firms that issued new debt, with one firm doing both. As discussed previously, one benefit of a dividend program is the signaling effect in that it provides a mechanism with which to send credible signals to the market that the management is confident about future cash flow. The fact that these firms were also raising money around the same time mitigates the positive signaling effect and may result in negative market reaction. Among the 13 firms in this group that reported a loss, four firms also issued new debt, one firm acquired another firm, and one firm was a takeover target. Starting a new dividend program while incurring losses certainly does not send a credible signal to the investors. The market may also be suspicious of the management’s motivation in face of the loss. Finally, three firms had to omit paying dividends within less than two years of initiation due to financial difficulties. In retrospect, the decision to start a regular dividend program may not have been prudent for these firms, and the market’s negative reaction to initiation may have anticipated this imprudence.
4. Conclusion and summary This paper examines the apparent heterogeneity in the stock price reaction to dividend initiation announcements. It offers evidence that dividend-initiating firms fall into two distinct categories, with initiation being a value-increasing event for one and a valuedecreasing event for the other due to different firm specific characteristics. The results of this study strongly suggest that market reaction to dividend initiation announcements be based on the net effects of the costs and benefits of a new dividend program, which differ from firm to firm. It shows that a seemingly positive corporate event, such as starting a new dividend
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
275
program, could very well be perceived by the market as having a negative impact on the value of the company. The findings of this paper may have implications for future studies on other corporate events, such as stock repurchase, stock split, seasoned new issue, dividend omission, layoffs and so on. It will also be interesting to study whether heterogeneous market reactions to these events exist, and how market reactions are affected by various firm specific factors. The answers to these questions will certainly deepen our understanding of the market implications of these important corporate events.
Acknowledgment I want to express my gratitude to the two anonymous referees for their constructive comments and suggestions. I am also grateful to my colleagues Mona Gardner and Robert Leekley for their valuable help during the revision process.
Notes 1. For example, Asquith and Mullins (1983), Benesh et al. (1984), Healy and Palepu (1988), Venkatesh (1989), Mitra and Owers (1990), Rimbey and Officer (1992). 2. Pettit (1972), Aharony and Swary (1980), Asquith and Mullins (1983), Brickley (1983), Zeghal (1983), Dielman and Oppenheimer (1984), Kalay and Lowenstein (1985), Eades et al. (1986), Eddy and Seifert (1988), Lang and Litzenberger (1989), Mitra and Owers (1990). 3. One possible explanation for why such a firm initiates dividends is that the managers had judged the benefits to exceed the costs. Note that such a disagreement between the managers and the market can happen even when both parties have unbiased beliefs. 4. Although initiation announcements are screened for confounding events in the Wall Street Journal Index, some of those events may not be detected in this screening. Further, intraindustry information transfers, whereby an announcement by one firm affects returns for other firms in the same industry, may also be considered confounding events. 5. The firm also had to be listed for at least two years before paying dividends. These are the same criteria used by Asquith and Mullins (1983) in their study. 6. The number of firms in the industry index ranges from nine to 55, with an average of 20. 7. Source: Compustat 8. Source: Standard & Poor’s Security Owner’s Guide. 9. Source: proxy statements. 10. Source: Compustat. The calculation of Tobin’s Q is based on an algorithm contained in McConnell and Servaes (1990). The market value of the firm is the sum of the market value of equity and the book value of debt. The replacement value of plant and equipment is calculated by approximating the plant’s age and then adjusting its book
276
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
value for inflation. The replacement value of inventory is assumed to equal book value, unless the firm used last-in-first-out (LIFO) accounting. If LIFO is used, the book value is adjusted for inflation. The replacement value of the firm is computed as the book value of assets minus the book value of inventories minus the book value of net plant plus replacement value of inventory plus replacement value of plant and equipment. 11. Source: The initial dividend is obtained from the CRSP monthly file and confirmed through The Wall Street Journal Index. The share price is collected from the NYSE/ AMEX Daily Stock Record. 12. Source: Compustat 13. For the entire sample, the correlation coefficients across the independent variables range from a minimum of ⫺0.19 and a maximum of 0.17, indicating that multicollinearity is not a problem.
References Aharony, J., & Swary, I. (1980). Quarterly dividend and earnings announcements and stockholders’ returns: an empirical analysis. Journal of Finance 35, 1–12. Asquith, P., & Mullins, D. (1983). The impact of initiating dividend payments on shareholders’ wealth. Journal of Business 46, 77–96. Asquith, P., & Mullins, D. (1986). Signaling with dividends, stock repurchases and equity issues. Financial Management (Autumn) , 27– 44. Bhattacharya, S. (1979). Imperfect information, dividend policy and ‘the bird in the hand fallacy’. Bell Journal of Economics 10(Spring), 259 –270. Benesh, G., Keown, A., & Pinkerton, J. (1984). An examination of market reaction to substantial shifts in dividend policy. Journal of Financial Research 7, 131–142. Black, F. (1976). The dividend puzzle. Journal of Portfolio Management 2(2), 72–77. Brickley, J. (1983). Shareholder wealth, information signaling and the specially designated dividend. Journal of Financial Economics 12, 187–209. Dielman, T. E., & Oppenheimer, H. R. (1984). An examination of investor behavior during periods of large dividend changes. Journal of Financial and Quantitative Analysis 19(2), 197–216. Eades, K., Hess, P., & Kim, E. (1986). Market rationality and dividend announcements. Journal of Financial Economics 14, 581– 604. Easterbrook, F. (1984). Two agency-cost explanations of dividends. American Economic Review 74(September), 650 – 659. Eddy, A., & Seifert, B. (1988). Firm size and dividend announcements. The Journal of Financial Research 4(Winter), 295–302. Healy, P., & Palepu, K. (1988). Earnings information conveyed by dividend initiation and omissions. Journal of Financial Economics 21, 149 –175. Jensen, M. (1986). Agency costs of free cash flow, corporate finance and takeovers. American Economic Review 76(May), 323–329. John, K., & Williams, J. (1985). Dividends, dilution and taxes: a signaling equilibrium. Journal of Finance 40(September), 1053–1070. Kalay, A., & Loewenstein, L. (1985). Predictable events and excess returns: the case of dividend announcements. Journal of Financial Economics 14, 423– 449. Lang, L., & Litzenberger, R. (1989). Dividend announcements: cash flow signaling vs. free cash flow hypothesis? Journal of Financial Economics 24, 181–191.
Z. Jin / The Quarterly Review of Economics and Finance 40 (2000) 263–277
277
McConnell, J. J., & Servaes, H. (1990). Additional evidence on equity ownership and corporate value. Journal of Financial Economics 27(2), 595– 612. Miller, M., & Modigliani, F. (1961). Dividend policy, growth, and the valuation of shares. Journal of Business 34, 411– 433. Miller, M., & Rock, K. (1985). Dividend policy under asymmetric information. Journal of Finance 40(September), 1031–1052. Mitra, D., & Owers, J. (1990). The information content of dividend initiations in asymmetrical information environments. Working paper presented at the 1990 FMA Meetings. Myers, S. C., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have. NBER Working Paper: 1396 (July). Pettit, R. R. (1972). Dividend announcements, security performance, and capital market efficiency. Journal of Finance 5(December), 993–1007. Rimbey, J., & Officer, D. (1992). The market response to subsequent dividend actions of dividend-initiating and -omitting firms. Quarterly Journal of Business and Economics 31(Winter), 3–20. Venkatesh, P.C. (1989). The impact of dividend initiation on the information content of earnings announcements and return volatility. Journal of Business 62, 175–197. Zeghal, D. (1983). Firm size and the informational content of financial statements. Journal of Financial Economics 12(September), 299 –310.