«DO DIVIDEND CLIENTELES EXPLAIN PRICE REACTIONS TO DIVIDEND CHANGES? Arman Kosedag, Berry College Jinhu Qian, Le Moyne College ABSTRACT Previous ...»
The International Journal of Business and Finance Research ♦ Volume 3 ♦ Number 1 ♦ 2009
DO DIVIDEND CLIENTELES EXPLAIN PRICE
REACTIONS TO DIVIDEND CHANGES?
Arman Kosedag, Berry College
Jinhu Qian, Le Moyne College
Previous studies find that stock price reactions to dividend announcements are positively related to dividend yield, consistent with the dividend-clientele hypothesis. In this paper, we argue that this yield-related clientele effect can be attributed to estimation biases in using preannouncement dividends as a proxy for market's anticipated dividends. Based on our samples constructed to mitigate the dividend estimation biases, we find that dividend yield has no additional power beyond the standardized dividend change in explaining the announcement-period excess returns. Our results are consistent with the information/signaling hypothesis, but inconsistent with the dividend-clientele hypothesis. In addition, we find that firm size remains negatively related to the price reactions to dividend changes.
JEL: G14, G35
We argue that this hypothesis has no solid theoretic base and is an empirical issue. If there is sufficient adjustment of investors in response to dividend yield changes, with some leaving a clientele replaced by others entering it, dividend yield should play little role in explaining the stock price reaction to dividend changes. As shown in this paper, their founding of a positive relationship between stock price reactions to dividend changes and dividend yield may be attributed to improper estimation of anticipated dividends in previous studies that use preannouncement dividends as a proxy for market's anticipated dividends. If markets partially anticipate a dividend increase, for example, the abnormal stock price reaction will be relatively small and reflect only the unexpected portion in the dividend change. Using preannouncement dividends as the proxy for the market's expected dividends may cause two kinds of estimation biases.
First, it underestimates anticipated dividend yield by markets and therefore associates low dividend yields with small excess returns. Second, it exaggerates the unexpected dividend increases (or dividend surprises) and thus underestimates the information effect of dividend changes, making the above spurious dividend yield effect more conspicuous. The dividend expectation is less a problem in the case of dividend decreases. Since firms seldom cut dividends, dividend decreases should contain a greater unexpected component.
In this paper, we construct two subsamples of dividend increase announcements which largely mitigate the dividend estimation problem. The first subsample is obtained by excluding those announcements with no or negative stock price reactions. We believe these dividend increases are largely anticipated by the market. If the realized dividend increase is less than what the market expects, the dividend increase announcement in fact represents a negative dividend surprise and causes stock price to decrease. Our second subsample consists of only the announcements with unusually large dividend increases, say, at least 50 percent. We conjecture that
these large dividend changes contain greater unexpected components than small dividend increases. This idea is similar to the view of Asquith and Mullins (1983) that unusual dividend policy changes such as dividend initiations are more likely to be unexpected.
Our sample closely resembles those in the previous studies. There are 7715 dividend increase and 849 dividend decrease announcements over the period 1970-2001. We find that announcement period excess returns are positively related to the magnitude of standardized dividend changes and to dividend yield for the dividend increase sample. But the excess returns are unrelated to dividend yield in the dividend decrease sample. This evidence supports our dividend expectation argument and is inconsistent with the dividendclientele hypothesis.
Our main evidence against the dividend-clientele hypothesis comes from the two subsamples. Within each subsample, when there is no control for dividend change, the higher the yield, the greater the announcement period excess return. However, no such pattern is observed after controlling for dividend change. The results from cross-sectional regressions provide further support for our argument. Although the coefficient of dividend yield is highly significant and positive in univariate regressions, it becomes insignificant and positive in the multivariate regressions with the inclusion of standardized dividend change. In addition, consistent with prior studies, firm size is found to be negatively related to excess returns. Stock price, however, is no longer associated with excess returns after controlling for the dividend change.
The rest of the paper is organized as follows. In the following section, we discuss the relevant literature. Next, we describe the data and methodology and provide summary statistics of the samples. The empirical results are reported and discussed in the following section. In addition, findings of previous studies are also replicated for comparison. The paper closes with some concluding comments.
In their seminal work, Miller and Modigliani (1961) demonstrate that, absent imperfections, a firm's dividend policy does not affect its value. Since then, challenges to this dividend irrelevance proposition have focused on imperfections. In particular, subsequent research has extensively explored the effects of tax-induced clienteles on capital asset prices and the stock price reactions to dividend announcements.
Before the implementation of the 1986 Tax Reform Act, the dividend income was taxed at a higher rate than capital gains, and this suggests a negative price impact of dividends. This idea is supported by the CAPMbased studies including Litzenberger and Ramaswamy (1979, 1980), Rosenberg and Marathe (1979), and Blume (1980). Using capital asset pricing models incorporating taxes, these studies find that, if risk is held constant, before-tax returns are an increasing function of dividend yield.
Black and Scholes (1974) and Miller and Scholes (1978) demonstrate that dividend irrelevance may hold even if there is differential taxation of dividends and capital gains. Miller and Scholes (1978) argue that the dividend receipts can be made tax exempt by laundering them with personal borrowing. Black and Scholes (1974) extend the concept of investor clienteles proposed by Miller and Modigliani (1961), i.e., low (high) yielding stocks being held by investors in high (low) marginal tax brackets if tax rates vary across investors.
They emphasize the ability of firms to adjust dividends to appeal to tax-induced investor clienteles and argue that this supply effect may account for their finding of no significant relationship between dividend yields and stock returns.
An alternative viewpoint, set forth by Litzenberger and Ramaswamy (1980) suggests that firms may make incomplete supply adjustments and individuals' portfolios may be limited by shortsale and margin restrictions.
In equilibrium, therefore, the relative prices of dividends and capital gains will reflect the tax situation of the 48 The International Journal of Business and Finance Research ♦ Volume 3 ♦ Number 1 ♦ 2009 marginal investor in the stock. Miller and Scholes (1982) criticize that the dividend yield effect found in Litzenberger and Ramaswamy (1979, 1980) may be attributed to the information biases. In response, Litzenberger and Ramaswamy (1982) use an "information free" sample and still find the yield coefficient to be positive and significant in their after-tax CAPM model.
As debate goes on, Kalay and Michaely (1983) argue that while the after-tax CAPM predicts cross-sectional return variation as a function of dividend yield, the Litzenberger and Ramaswamy test is inadvertently designed to discover whether the ex-dividend period offers unusually large risk adjusted returns. Separating the time series from the cross-sectional return variation, Kalay and Michaely cannot detect any return variation across stocks with different yields. Chen, Grundy, and Stambaugh (1990) show that the positive association between yields and returns can be explained by a time-varying risk premium correlated with yield. When they allow the risk measures to vary, the yield coefficient was found positive but insignificant.
While the evidence about the tax-induced dividend yield effect is far from conclusive, there seems to be an overall agreement that the market perceives that dividend changes convey new information about the value of firm. In their original article, Miller and Modigliani (1961) suggest that dividends may provide a vehicle for communicating management's superior information concerning their assessment of the firm's prospect. This view of "information content of dividends" is supported by the empirical evidence in numerous studies examining the price reactions to dividend changes. For example, Pettit (1972) shows that announcements of dividend increases are followed by a significant price increase and announcements of dividend decreases are followed by a significant price drop. Aharony and Swary (1980) find that these relationships hold even after controlling for contemporaneous earnings announcements. Focusing on extreme changes in dividend policy, Asquith and Mullins (1983), Healy and Palepu (1988), and Michaely, Thaler, and Womack (1995)) show that the market reacts quite severely to dividend initiations or omissions announcements.
Recently, Nissim and Ziv (2001) find that dividend changes provide information about the level of profitability in subsequent years, incremental to market and accounting data. They also find that dividend changes are positively related to earnings changes in each of the two years after the dividend change. Koch and Sun (2004) present results suggesting that changes in dividends cause investors to revise their expectations about the persistence of past earnings changes. Docking and Koch (2005) document that dividend change announcements elicit a greater change in stock price when the nature of the news (good or bad) goes against the grain of the recent market direction during volatile times Extending the previous studies, as noted earlier, Bajaj and Vijh (1990) suggest that the existence of dividend clienteles may partially explain price reactions to dividend change announcements. They argue that if marginal investors in different stocks value dividends differently, anticipated dividend yield should be associated with the price reactions to dividend change announcements. For an investor with a relatively high aversion to dividends, for example, the positive information in a dividend increase is accompanied by the negative effect of higher-than-anticipated yield. In contrast, the two effects act in the same direction for an investor with a preference for dividends. If investors with preference for dividends are marginal investors in high-yield stocks, the price reaction to dividend change should be larger, the higher the anticipated yield of the stock. This dividend-clientele hypothesis is supported by the evidence in Bajaj and Vijh (1990) and Denis, Denis, and Sarin (1994). Both studies use preannouncement dividend yield as a proxy for anticipated yield and find that the magnitude of stock price reaction to a dividend change announcement is positively related to dividend yield. In addition, Bajaj and Vijh (1990) find that stock price changes are negatively related to firm size and stock price.
Several approaches have been proposed in the literature to capture the unexpected component of dividend changes, including Lintner (1956) model, the Box-Jenkins model, and the Value Line dividend forecasts.
However, according to a study by Bar-Yosef and Sarig (1992), the measures of dividend surprises based on
these methods are not significantly correlated with the market reactions to dividend change announcements in a sample of large firms traded in the NYSE.
DATA AND METHODOLOGYIn our sample, information on dividend declarations is obtained from the CRSP NYSE/AMEX Monthly Master Files and daily rates of returns are from the CRSP Daily Master File. The time period covers from July 1970 to December 2001, excluding the latter half of 1987 to avoid the period affected by the market crash in
the October of that year. In addition, we use the following criteria to select firms:
(1) Absolute changes in consecutive regular quarterly dividends per share are greater than 10%. We require that no other type of distribution is made over the period between the two quarterly dividend announcements.
Thus, firms that pay stock dividends or special dividends and firms that split their shares during the quarter in question are not included in the sample.
(2) The dividend initiations and omissions are excluded from the sample. This is due to the difficulty involved in calculating the anticipated yield and much larger price responses to these events than those to regular dividend changes (See Asquith and Mullins (1983), Healy and Palepu (1988), and Michael, Thaler, and Womack (1995)).
(3) We choose only dividend announcements for which the announcement date precedes the ex-dividend day by at least eight trading days. The eight-day window is chosen because the ex-day effect is observed up to five days before the ex-day, as documented by Eades, Hess, and Kim (1984)).