What is Dividend Premium? Laura Yue Liu Assistant Professor of Finance College of Business and Economics P.O. Box 6848 California State University, Fullerton Fullerton, CA 92834 yueliu@fullerton.edu (714) 278-8426 Liwei Shan Assistant Professor of Finance Research Institute of Economics and Management Southwestern University of Finance and Economics Chengdu, China, 610074 lshan@swufe.edu.cn Tel: (86) 28-8709-9204
What is Dividend Premium? Abstract: Defined by Baker and Wurgler (2004a), dividend premium is the difference between the average market-to-book ratio of dividend payers and non-payers. We study what dividend premium is by examining two explanations, agency explanation and signaling explanation. Using U.S. industrial firms from 1962 to 2004, we find evidence consistent with the agency theory that dividend premium is higher when the need to mitigate the agency problem is greater. In particular, dividend premium is positively related to the difference in cash holdings at the beginning of the year between dividend payers and nonpayers, and is negatively related to the difference in future profitability between dividend payers and nonpayers. In other words, investors value dividend payers with a higher premium when dividend payers have more cash and fewer profitable future investment projects than nonpayers. 1. Introduction Baker and Wurgler (2004a) propose a catering theory of dividends to explain the decision to pay dividends. They argue that firms initiate dividends to cater to investors demand for dividends when the overall dividend premium in the stock market is high. They define dividend premium as the difference between the average market-to-book ratio of dividend payers and non-payers. Consistent with their theory, Baker and Wurgler (2004a, 2004b) and Li and Lie (2006) find that dividend premium has significant explanatory power of dividend initiations, dividend changes and changes in the propensity to pay dividends. Despite the importance of dividend premium in the payout literature, what dividend is remains unknown. We propose and test two theories that could explain the dividend premium. The first theory is agency theory. The agency theory predicts that investors will value dividend payers with a higher premium when the need to mitigate the potential agency problem is greater. Another theory is signaling theory, which predicts that investors will value dividend payers with a higher premium when the need to signal future profitability is greater. 1
Following Baker and Wurgler (2004a and 2004b), we select our sample of firms from Compustat and CRSP between 1962 and 2004. We find evidence consistent with the agency theory that investors place a higher premium on dividend payers when dividend payers have greater needs to mitigate the potential agency problem. In particular, dividend premium is positively related to cash difference at the beginning of the year between dividend payers and nonpayers, and is negatively related to the difference in future profitability between dividend payers and nonpayers. Our results suggest that investors value dividend payers with a higher premium when dividend payers have more cash at the beginning of the year and fewer profitable investments relative to nonpayers, a situation where the potential cash-induced agency problem is more severe for dividend payers. Our study contributes to the literature in two ways. First, we contribute to the developing literature on dividend premium. We find that the dividend premium is related to the need to mitigate the potential agency problem. Our research also adds to our understanding of investors reaction to dividend policy, and therefore, contributes to the payout literature. 2. Hypothesis development Baker and Wurgler (2004a) develop a catering theory of dividends and argue that firms initiate dividends depending on the prevailing investor demand for dividend. Consistent with their theory, they find that the overall initiation rate of dividends in the economy is related to the lagged overall dividend premiums in the stock market. In another paper, Baker and Wurgler (2004b) find that their measure of overall dividend premiums in the stock market have significant explanatory powers of changes in the 2
propensity to pay dividends even after controlling for changes in firm characteristics. Building on Baker and Wurgler s study, Li and Lie (2006) find that dividend changes are also associated with the dividend premium. Despite their studies, dividend premium remains unknown. Up to now, there is no explanation for what the dividend premium is and why the dividend premium fluctuates overtime. There are two theories that could explain the dividend premium. The first on is agency theory. According to Jensen (1986) and Easterbrook (1984), agency theory argues that paying out cash as dividends can mitigate the potential cash-induced agency problem between managers and shareholders. Therefore, our first hypothesis says that the dividend premium is positively related to the lagged cash premium, which is the difference in cash holdings at the beginning of the year between dividend payers and nonpayers. Another explanation is the signaling theory. The payout literature has a signaling explanation for dividends, which says that firms pay dividends to signal their better future prospects in order to differentiate themselves from their peers. 1 Therefore, our second hypothesis says that the dividend premium is positively related to the future profitability premium, which is the difference in future profitability between dividend payers and nonpayers. 3. Data Following Baker and Wurgler (2004a) and Fama and French (2001), we select our sample of firms from publicly traded firms (share code 10 or 11) with available data in Compustat and CRSP from 1962 to 2004. We exclude utilities (SIC 4900-4949), financial firms (SIC 6000-6999) and firms with book value of equity below $250,000 or total 1 For signaling theory and evidence in payout literature, see Bhattacharya (1979), Miller and Rock (1985), John and Williams (1985), Nissim and Ziv (2001), Aharony and Swary (1980), Asquith and Mullins (1983), and Healy and Palepu (1988). 3
assets below $500,000. Market-to-book (M/B) is the book value of assets (data6) minus the book value of equity plus the market value of equity (data24*data25) all divided by the book value of assets. The book value of equity is stockholders equity (data216) (or first available of data60 plus data130 or data6 minus data181) minus preferred stock liquidating value (data10) (or first available of data56 or data130) plus balance sheet deferred taxes and investment tax credit (data35) if available and minus post retirement assets (data330) if available. A firm is defined as a dividend payer if it has positive dividends per share by the ex date (data26), else it is a non-payer. The overall dividend premium for each year is the difference between the natural logs of the dividend payers and nonpayers book-value-weighted average M/B ratios. This variable is downloaded from Jeffery Wurgler s website with year 2004 data filled in by us following Baker and Wurgler (2004a). We create a measure of profitability premium (E/A) of dividend payers relative to nonpayers. We follow Fama and French (2001) s definition of profit and calculate the profitability premium as the difference between the natural logs of the dividend payers and nonpayers book-value-weighted average earnings before interest (data18+data15+data50 (if available)) over total assets. Similarly, we also create a measure of cash premium (Cash/A) of dividend payers relative to nonpayers, defined as the difference between the natural logs of the dividend payers and nonpayers bookvalue weighted average cash (data1) to assets (data6) ratios. Table 1 summarizes dividend premium, lagged cash premium (Cash/A t-1 ), and future profitability premium (E/A t+1 ) from 1962 to 2004. 4
4. Empirical Tests We estimate regressions with the following specification: Dividend Premium = α 0 + α1cash/at 1 + α2e/at + 1 + α3other Variables + εt. Results are reported in Table 2. T-statistics are reported in parentheses under coefficients. T-statistics are calculated by using standard errors that are robust to heteroskedasticity and serial correlation up to four lags. We begin with testing the agency theory and the signaling theory separately by estimating univariate regressions. The dependent variable is dividend premium and the right-hand-side variable is lagged cash premium (Cash/A t-1 ), and future profitability premium (E/A t+1 ) respectively. Consistent with the agency theory, Model (1) indicates that the dividend premium is positively associated with lagged cash premium between dividend payers and nonpayers. The coefficient is positive and significant at 1% level. This suggests that firms with potentially more severe agency problem are valued higher by investors if they pay out dividends than if they do not. Results from model (2) do not support the signaling theory. The coefficient of future profitability premium is insignificant. The agency theory predicts that firms with more cash and fewer growth opportunities should pay out more dividends. Therefore, we create an interaction of lagged cash premium with growth premium. Similar to other variables, growth premium is the difference in the natural logs of book-value weighted average growth opportunities between dividend payers and nonpayers. Because the dependent variable, dividend premium, is defined by using M/B ratio, we avoid using the same ratio to measure growth opportunities. Specifically, we follow Fama and French (2001) and use growth rate in 5
assets (da/a) as a measure of growth opportunities, where da/a is defined as total assets (data6) minus lagged total assets all divided by total assets. In a few years, the bookvalue weighted average da/a ratios are negative. In these cases, we use the negative of the natural logs of the absolute values of da/a ratios instead. The agency theory predicts a negative sign of the interaction of lagged cash premium with growth premium. Model (3) in Table 2 presents the result in this regression. The coefficient of the interaction has a negative sign as predicted, but it is not significant at the conventional level. The coefficient of lagged cash premium is still positive and significant at 1% level. The signaling theory predicts that firms that have greater degree of information asymmetry and better future prospects should pay dividends as a signal. Therefore, we create an interaction of future profitability premium with information asymmetry premium. The information asymmetry premium is defined as the difference between the natural logs of dividend payers and nonpayers book-value weighted average standard deviation of daily CRSP stock returns. Following Bhagat and Frost (1986), Dierkens (1991), and Fee and Thomas (1999), we interpret greater daily stock return volatility as higher levels of asymmetric information problems. The signaling theory predicts a positive coefficient of the interaction of future profitability premium with information asymmetry premium. Model (4) summarizes the result of the regression with this interaction. In contrast to the signaling theory, both coefficients of profitability premium and the interaction are significantly negative. However, this negative relation between dividend premium and future profitability premium is consistent with the agency theory. Firms that run out of profitable investment projects have lower future profitability. Investors will value these firms with a higher premium if they pay dividends to mitigate 6
the potential agency problem. Similarly, the significantly negative interatction of future profitability premium and information asymmetry is also consistent with agency theory. Given the same degree of information asymmetry, investors will value dividend payers with a higher premium if payers have lower future profitability because these payers run of out profitable investment projects and therefore they pay out dividends to mitigate the potential agency problem. Previous studies have documented that dividend payers are relatively larger than nonpayers. Therefore, we create a size premium to control for size difference between dividend payers and non-payers. This size premium is the difference between the natural logs of the dividend payers and nonpayers book-value weighted average total assets. In Model (5) and (6), we include the size premium as a control variable. Results of Model (5) and (6) are similar to results of Model (3) and (4). The coefficient of lagged cash premium is significantly positive and the coefficient of future profitability premium is significantly negative. In Model (7), we include all right-hand-side variables in the previous six models as right-hand-side variables. The result is consistent with the agency theory. The coefficient of lagged cash premium is positive and significant at 1% level, and the coefficient of future profitability premium is significantly negative. The coefficient of interaction of future profitability premium and information asymmetry premium is also significantly negative. Overall, results in Table 2 indicate that investors value dividend payers more when payers have higher cash holdings at the beginning of the year and lower future profitability relative to nonpayers. In these situations, dividend payers can mitigate the potential agency problem by paying out cash as dividends. 7
As robustness tests, instead of directly using Baker and Wurgler s dividend premium data, we follow their method and recalculate the entire series of dividend premium using our sample firms. Our dividend premiums are very close to Baker and Wurgler s results with a correlation of 0.93. Using our dividend premiums, we generate very similar regression results as in Table 2. 5. Conclusion Current literature on catering theory of dividends [Baker and Wurgler (2004a, 2004b), Li and Lie (2006)] finds that dividend initiations and dividend changes are related to the dividend premium. However, what dividend premium is and why dividend premium varies through time remains a puzzle in the literature. We test two explanations of dividend premium in this study, the agency theory and the signaling theory. Using U.S. industrial firms from 1962 to 2004, we find evidence consistent with the agency theory that dividend premium is higher when the need to mitigate the agency problem is greater. Our results suggest that investors value dividend payers with a higher premium when dividend payers have more cash at the beginning of the year and fewer profitable investment opportunities relative to nonpayers. This study contributes to the dividend premium literature by uncovering a relation between dividend premium and the need to reduce agency costs of free cash flow. 8
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. W., 1983. The impact of initiating dividend payments on shareholders wealth. Journal of Business 56, 77-96. Baker, M., Wurgler, J., 2004a. A catering theory of dividends. Journal of Finance 59, 1125 1165. Baker, M., Wurgler, J., 2004b. Appearing and disappearing dividends: the link to catering incentives. Journal of Financial Economics 73, 271 288. Bhagat, S., Frost, P.A., 1986. Issuing costs to existing shareholders in competitive and negotiated underwritten public utility equity offerings. Journal of Financial Economics 15, 233-260. Bhattacharya, S., 1979. Imperfect information, dividend policy, and the bird in the hand fallacy. Bell Journal of Economics 10, 259 270. Dierkens, N., 1991. Information asymmetry and equity issues. Journal of Financial and Quantitative Analysis 26, 181-200. Easterbrook, F., 1984. Two agency-cost explanations of dividends. American Economic Review 74, 650-659. Fama, E.F., French, K.R., 2001. Disappearing dividends: changing firm characteristics or lower propensity to pay? Journal of Financial Economics 60, 3 44. Fee, C. E., Thomas S., 1999. Corporate diversification, asymmetric information, and firm value: Evidence from stock market trading characteristics. Working Paper. Michigan State University. Healy, P., Palepu K., 1988. Earnings information conveyed by dividend initiations 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, 323-329. John, K., Williams, J., 1985. Dividends, dilution, and taxes: A signalling equilibrium. Journal of Finance 40, 1053 1070. Li, W., Lie E., 2006. Dividend changes and catering incentives. Journal of Financial Economics 80, 293-308. 9
Miller, M.H., Rock, K., 1985. Dividend policy under asymmetric information. Journal of Finance 40, 1031 1051. Nissim, D., Ziv, A., 2001. Dividend changes and future profitability. Journal of Finance 56, 2111 2133. 10
Table 1 Descriptive statistics of variables from 1962 to 2004. This table summarizes descriptive statistics of our sample from 1962 to 2004. Dividend premium is obtained from Jeffery Wurgler s website with year 2004 data filled in by us following Baker and Wurgler (2004). It is calculated as the difference between the natural logs of dividend payers and nonpayers book-value-weighted average market-to-book ratio. A firm is a payer in year t if it has positive dividends per share (data26). E/A t+1 is the profitability premium at year t+1, calculated as the difference between the natural logs of dividend payers and nonpayers earnings before interest (data18+data15+data50 (if available)) over total assets. Cash/A t-1 is the lagged cash premium at year t-1, calculated as the difference between the natural logs of dividend payers and nonpayers book-value-weighted average cash (data1) over total assets. Year Dividend Premium E/A t+1 Cash/A t-1 1962 34.89 44.80 11.38 1963 32.92 48.74 88.67 1964 35.64 62.61 68.12 1965 22.65 28.09 43.50 1966 5.37 37.46 46.98 1967-17.23 29.64 33.63 1968-18.79 15.80 5.45 1969-3.80 55.29-24.32 1970 16.05 57.13-4.51 1971 18.16 50.42 1.07 1972 26.57 29.61-0.74 1973 25.87 33.93-1.69 1974 13.20 42.63 4.84 1975 15.61 34.75 0.32 1976 15.59 23.95 20.84 1977 4.58 30.73 8.08 1978-4.96 17.19-12.45 1979-14.28 17.26-25.28 1980-22.11 28.20-12.27 1981-24.93 46.00-33.92 1982-16.90 56.92-55.46 1983-26.20 41.47-71.47 1984-12.51 56.77-86.98 1985-11.03 81.60-80.35 1986-7.32 42.79-93.48 1987-7.78 26.16-104.81 1988-7.81 31.98-115.74 1989-8.66 33.69-110.57 1990-1.02 63.81-120.30 1991-4.58 42.77-126.12 1992-5.32 38.89-147.65 1993-11.48 46.47-125.95 1994-7.47 52.68-141.66 1995-15.07 52.25-136.09 1996-9.43 51.46-148.76 1997-4.82 43.05-156.47 1998 1.44 40.49-138.83 1999-33.17 87.99-153.22 2000-20.56 142.49-164.94 2001-1.57 239.55-188.97 2002 5.37 76.98-159.78 2003-3.51 44.24-144.93 2004-9.92 50.07-97.90 Std deviation 17.20 36.75 75.60 Mean -1.36 50.67-61.69 Median -4.96 43.05-71.47 11
Table 2 Regressions of Dividend Premium on Lagged Cash Premium and Future Profitability Premium: Year 1962-2004 Results of OLS regressions with the following specification are presented. Dividend Premium α + α Cash/A + α E/A + α Cash/A da/a t = 0 1 t 1 2 t + 1 3 t 1 t + α4e/at + 1 Std + α5size + εt The dependent variable is the dividend premium, calculated as the difference between the natural logs of dividend payers and nonpayers book-value-weighted average market-to-book ratio for each year. This variable is downloaded from Jeffery Wurgler s website with year 2004 data filled in by us following Baker and Wurgler (2004). A firm is a payer in year t if it has positive dividends per share (data26). T-statistics are in parentheses. E/A t+1 is the profitability premium at year t+1, calculated as the difference between the natural logs of dividend payers and nonpayers earnings before interest (data18+data15+data50 (if available)) over total assets. Cash/A t-1 is the lagged cash premium at year t-1, calculated as the difference between the natural logs of dividend payers and nonpayers book-value-weighted average cash (data1) over total assets. Size is the size premium, calculated as the difference between the natural logs of dividend payers and nonpayers book-value-weighted average total assets. da/a is the growth opportunity premium, calculated as the difference between the natural logs of dividend payers and nonpayers bookvalue-weighted average total assets (data6) minus lagged total assets all divided by total assets. Std is the information asymmetry premium measured as the difference between the natural logs of dividend payers and nonpayers book-value-weighted average standard deviation of CRSP daily stock returns. T-statistics are reported in parentheses under coefficients. T-statistics use standard errors that are robust to heteroskedasticity and serial correlation up to four lags. (1) (2) (3) (4) (5) (6) (7) t Cash/A t-1 0.126*** (3.20) 0.120*** (2.96) 0.184*** (3.92) 0.133*** (2.76) E/A t+1-0.049 (-1.09) -1.053*** (-3.02) -0.898** (-2.58) -0.641** (-2.31) Cash/A t-1 da/a t -0.000 (-1.14) -0.000 (-0.74) -0.000 (-1.56) E/A t+1 Std t -0.020*** (-2.89) -0.018** (-2.64) -0.013** (-2.33) Size t 0.077 (0.87) -0.127 (-1.53) -0.078 (-0.89) Constant 6.400 (1.21) 1.151 (0.20) 6.57 (1.22) -7.400 (-1.67) -30.89 (-1.20) 45.36* (1.91) 21.800 (0.409) R 2 0.305 0.011 0.314 0.246 0.272 0.360 0.446 N 43 43 43 43 43 43 43 12