- Email: [email protected]
Demographics, dividend clienteles and the dividend premium
The Quarterly Review of Economics and Finance 51 (2011) 368–375
Contents lists available at SciVerse ScienceDirect
The Quarterly Review of Economics and Finance journal homepage: www.elsevier.com/locate/qref
Demographics, dividend clienteles and the dividend premium King Fuei Lee ∗ Schroder Investment Management, 65 Chulia Street, #46-00 OCBC Centre, 049513, Singapore
a r t i c l e
i n f o
Article history: Received 26 December 2010 Received in revised form 7 August 2011 Accepted 26 August 2011 Available online 3 September 2011 JEL classification: G000 G350 C320 Keywords: Dividend policy Demographics Dividend premium Dividend clienteles
a b s t r a c t The catering theory of dividends proposed that corporate dividend policy is driven by prevailing investor demand for dividend payers, and that managers cater to investors by paying dividends when the dividend premium is high. While earlier research found that the dividend premium is not driven by traditional clienteles derived from market imperfections such as taxes, transaction costs, or institutional investment constraints, we find empirical evidence that demographic clienteles are an important source of the timevarying demand for dividend payers. In particular, we find that, as consistent with the behavioural lifecycle theory and the marginal opinion theory of stock price, the dividend premium is positively driven by demographic clientele variation represented by changes in the proportion of the older population. Our results are robust when controlled for the factors of investor sentiment, signalling, agency costs, tax clienteles, time trend, business cycle fluctuations and varying sample periods. © 2011 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.
1. Introduction Modigliani and Miller (1961) proposed in their theory of dividend policy irrelevance that in a world of perfect information, full capital mobility, no taxes and no agency costs, the dividend policy of a company should have no impact on its value. However these assumptions rarely hold true in the real world. The catering theory of dividend policy (Baker & Wurgler, 2004a) relaxes the assumption of market efficiency and proposed that managers cater to investors by paying dividends when investors put a stock price premium on payers, and by not paying dividends when investors prefer nonpayers. According to the catering theory, corporate dividend policy is therefore driven by prevailing investor demand for dividend payers which is represented by the dividend premium, defined as the log difference in the average market-to-book ratio of dividend payers to nonpayers. Empirical evidence has generally been supportive. Baker and Wurgler (2004b) studied US companies on the COMPUSTAT between the period of 1963–2000, and found that the “disappearing dividends” phenomena observed by Fama and French (2001) can be largely explained by the catering theory. Li and Lie (2006) found similar catering effects among US firms when they exam-
∗ Tel.: +65 6535 3411; fax: +65 6535 3486. E-mail address: [email protected]
ined changes in corporate payout ratios to the market dividend premium, while Ferris, Sen, and Yui (2006) extended the analysis to the UK where they concluded that a shift in catering incentives most likely drove the declining propensity to pay dividends over the sub-period of 1998–2002. Despite strong empirical evidence of the importance of dividend premium as a determinant of dividend policy, there has been sparse research on the factors driving the dividend premium. Liu and Shan (2007) attempted to fill this gap by examining the relation of dividend premium to proxies of agency costs and signalling motivations, namely the differences in cash holdings and in future profitability between dividend payers and nonpayers respectively. Yet, their positive findings are viewed as inconclusive since their analysis is conducted without consideration for time trends which substantially affects their conclusions. In fact, the notion that the dividend premium reflects time-varying contracting problems is rejected by Baker and Wurgler (2004a) as being inconsistent with the observation of improving corporate governance and declining propensity of dividend payout in the 1960s. Baker and Wurgler (2004a) also attempted to identify whether traditional dividend clienteles are the source of the time-varying demand for dividend payers. They were however unable to match up the dividend premium to any plausible proxies for clienteles. For example, when they included a tax control variable in their multivariate regression, they found that the added tax control variable did not appear to impact the dividend premium much, and thus
1062-9769/$ – see front matter © 2011 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.qref.2011.08.004
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
369
rejected tax clientele as a driver of the dividend premium. They also rejected transaction costs clienteles as an explanation, while ruling out institutional investor clienteles because of the difficulty in reconciling the rise in institutional ownership since the 1980s with the time-varying pattern in the 1960s. This paper investigates the effects of changes in demographic clienteles on the dividend premium. We hypothesise that in a world where stock prices are determined by marginal opinion (Smith, 1967; Williams, 1938) and dividend preferences of retail investors are influenced by behavioural life-cycle considerations (Thaler & Shefrin, 1988), demographic variations may induce changes in demographic clienteles that then drives the dividend premium. In particular, we find strong empirical evidence that the dividend premium is positively related to changes in the proportion of the older-to-younger population. Our results are robust when controlled for the factors of investor sentiment, signalling, agency costs, tax clienteles, business cycle fluctuations and time trend. This paper contributes to current literature by adding to the understanding of the drivers of the dividend premium through an examination of demographic clientele changes as a source of the time-varying demand for dividend payers. To our knowledge, there has not been any work done in this aspect. The rest of this paper is structured as follows: Section 2 discusses the behavioural life-cycle hypothesis and the marginal opinion theory of stock price, and introduces our hypothesis. Section 3 describes the data sample and the methodology pursued. The empirical findings are reported in Section 4, while robustness tests are conducted in Section 5. Section 6 concludes the paper.
In traditional asset pricing theory, the distinction between the average investor and the marginal investor is usually not important as the marginal investor in every asset market is also assumed to be the same broadly diversified representative investor i.e. the average investor. Shleifer (2000) however highlighted that when investors’ trading schedules are not flat, then the trading of participating investors can affect asset prices. In other words, it is the marginal investor, rather than the average or representative investor, who affects asset prices. While the emphasis on the distinction between the marginal investor and the average investor as a driver of asset prices is only more commonly adopted in recent research (Allen & Gale, 1994; Bae, Yamada, & Ito, 2008; Gabaix, Krishnamurthy, & Vigneron, 2007; He & Krishnamurthy, 2008a, 2008b), the concept of marginal opinion as the determinant of stock prices was in fact first raised by Williams (1938) and subsequently extended by Smith (1967). According to the marginal opinion theory, in a market comprising of a number of interested parties who each possess an opinion as to the worth of the stock, the price of the stock is not set by the majority, regardless of how overwhelming it is, but by the last owner. This means that marginal opinion will determine the market price. If the stock prices of dividend-paying and non-paying stocks are determined by the marginal investor rather than the average investor, then it means that rather than being related to the absolute demographic structure which reflects the average investor opinion, dividend premium should be related to changes in the demographic structure which proxies for the marginal investor opinion i.e. demographic clientele variation.
2. Demographic clientele variations and the dividend premium
2.3. Hypothesis of dividend premium and changes in demographic clienteles
2.1. Behavioural life-cycle theory and the dividend preferences of older investors
We hypothesise that in a world where stock prices are determined by the marginal investor and where components of wealth are mentally treated as being non-fungible, the preference for dividend-paying stocks by older investors means that the dividend premium should be positively related to changes in the proportion of the older population, a proxy for the marginal investor opinion. In essence, the larger the increase in the proportion of the older population is, the greater the marginal investor preference for dividend-paying stocks and hence the higher the dividend premium. Our hypothesis is therefore that the time-varying demand for dividend payers is driven by variations in the demographic clienteles.
According to the behavioural life-cycle theory (Thaler & Shefrin, 1988), households treat components of their wealth as nonfungible. In particular, wealth is assumed to be broken into three mental accounts, namely current income, current assets and future income, with the temptation to spend being greatest for current income and least for future income. The behavioural life-cycle theory therefore hypothesises that in the later stage of a household’s life cycle when they reach retirement and begin to dis-save, the investor perception of the non-fungibility between dividends and capital gains should lead to a preference for dividend-paying stocks by older investors for consumption purposes. Empirical evidence has generally been supportive. Graham and Kumar (2006) studied the stock holdings and trading behaviour of 77,995 households over the period of 1991–1996 and found that, compared to younger investors, older investors allocate a greater proportion of their equity portfolios to dividend paying stocks. This suggests that senior investors have a greater preference for dividends. 2.2. Marginal opinion theory of stock price Given the dividend preference of older investors, it is reasonable to conjecture that when the general population has a greater proportion of older people, then the greater degree of buying of dividend-paying stocks by these senior investors should lead to a high dividend premium. Such a conjecture however assumes that stock prices are determined by the opinion of the average investor. If share prices are instead determined by the marginal investor, then the implications can be quite different.
3. Data sample and methodology This section briefly discusses the data sources and the variables’ definitions. The dividend premium PD−ND is the difference between the logs of the value-weighted market-to-book ratios for dividend payers to nonpayers, and the data is downloaded from the website of Jeffrey Wurgler.1 Following Graham and Kumar (2006), we use the older-to-younger ratio Old/Young, defined as the proportion of population aged above 65 to those aged under 45, as the variable representing the demographic structure. Demographic variation is therefore expressed as the annual change in the older-to-younger ratio dOld/Young dOld/Youngt = Old/Youngt − Old/Youngt−1
1
Available at http://pages.stern.nyu.edu/∼jwurgler/.
(1)
2 3 4
Available at http://www.census.gov. Available at http://www.ctj.org. Available at http://www.bea.gov.
3.59 1.68 6.30 −0.80 47 3.36 2.02 7.20 −1.90 47 27072.00 30727.77 103823.00 1.00 47 1984.00 760.00 13.71 678.81 2007.00 2209.00 1961.00 1.00 47 47 0.60 0.24 1.00 0.27 47 −61.69 75.60 88.67 −188.97 43 50.67 36.75 239.55 15.80 43 8.64 7.23 23.53 −10.91 47 85.27 8.76 101.40 70.06 47 0.41 1.63 2.76 −3.77 47
Fig. 1 shows the time series plots of the value-weighted dividend premium and the demographic variation variable. It is observed that while both variables are not perfectly synchronous, they are visibly positively related to each other. Indeed it can be seen from the correlation matrix in Table 2 that the contemporaneous correlation between dividend premium and the demographic variation measure of annual change in older-to-younger ratio is 0.457 at 5% significance level. Table 1 shows the descriptive statistics of the dividend premium, the demographic variation variables as well as the control variables, while Table 2 shows their unit root test statistics and the correlation matrix. The unit root test employed here is the
Table 1 Descriptive statistics, 1961–2007.
4. Empirical findings
−4.79 18.26 31.23 −44.43 47
where ˛i is the regression coefficient of explanatory variable i and εt is the random disturbance term.
Mean Standard deviation Maximum Minimum Number of observations
(2)
Annual change in older-to-younger ratio
+ ˛4 Cash/At−1 + ˛5 TAXt + ˛6 ECONt + ˛7 t + εt
Ann chg in prime consumers to prime savers ratio
PtD–ND = ˛0 + ˛1 dOld/Youngt + ˛2 CEFDt + ˛3 E/At+1
0.08 0.62 1.08 −1.10 47
Older to younger ratio
Closedend fund discount
Profitability Cash premium premium
Capital gains tax to dividend income tax ratio
Year
Time trend squared
Time trend cubed
Real GDP growth
Real PCE growth
Real PDI growth
We also employ an alternative measure of the demographic variation variable dConsumers/Savers defined as the annual change in the prime consumers-to-prime savers ratio in our robustness test, where prime consumers are persons aged above 65 while prime savers are persons aged from 45 to 65. The US population data used for the calculations of the two measures of demographic variations is downloaded from the US Census Bureau2 website. In our other robustness checks, we include measures of investor sentiment, signalling, agency costs, tax clienteles, time trends and business cycle fluctuations as control variables. Our chosen measure of investor sentiment is the closed-end fund discount CEFD which is the value-weighted discount on closed-end funds. Data for the variable is obtained from the website of Jeffrey Wurgler. Following Liu and Shan (2007), we also include measures of signalling and agency costs as control variables. The measure of signalling used is the profitability premium E/A which is defined as the difference between the natural logs of the value-weighted future returns-onassets ratios for dividend payers to nonpayers, while the measure of agency costs is the cash premium Cash/A which is the difference between the natural logs of the value-weighted cash-to-asset ratios for dividend payers to nonpayers. Data for both variables are obtained from Liu and Shan (2007). The measure of tax clienteles TAX is represented in our study as the ratio of capital gains tax versus dividend income tax, and the data is obtained from the website of Citizens for Tax Justice.3 As for measures of business cycle fluctuations ECON, we separately use real gross domestic product (GDP) growth, real personal consumption expenditure (PCE) growth and real private domestic investment (PDI) growth as control variables. The data for the business cycle indicators are obtained from the US Department of Commerce Bureau of Economic Analysis.4 To allow for a more flexible trend in the diagnosis of the effects of time t, we also separately use the linear time trend (year), time trend squared (t2 ) and time trend cubed (t3 ) as control variables. The time period employed in this study is from 1961 to 2007 which represents the period for which the data for dividend premium is available. Following the methodology of Liu and Shan (2007), we employ multivariate OLS regression to estimate the relation. The regression is expressed as
4.69 8.95 29.50 −17.70 47
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Value-weighted dividend premium
370
Table 2 Unit root statistics and correlation matrix, 1961–2007. Unit root
* ** ***
Ann chg in prime consumers to prime savers ratio
KPSS: trend stationarity
0.423*
0.189**
1.000
0.107
0.394***
0.173
0.124
0.457***
0.124
0.117
0.141
0.122*
0.343
0.146*
−0.081
−0.427***
−0.030
0.439* 0.505**
0.105 0.182**
0.535*** −0.181
0.748*** −0.277*
0.500*** −0.066
– 0.459* 0.443* 0.307 0.259 0.182
– 0.151** 0.150** 0.125* 0.132* 0.130*
−0.546*** −0.464*** −0.398*** 0.171 0.125 0.082
−0.664*** −0.714*** −0.688**** −0.061 −0.087 −0.105
−0.303** −0.143 −0.010 0.066 0.114 −0.001
0.293
Valueweighted dividend premium
Annual change in older-toyounger ratio
KPSS: level stationarity
−0.039 0.348**
0.394*** 1.000 0.150
0.719*** 0.261*
0.457*** 0.150
Older to younger ratio
−0.039 0.719***
Closedend fund discount
Profitability Cash premium premium
Capital gains tax to dividend income tax ratio
Year
Time trend squared
Time trend cubed
Real PDI growth
−0.081
0.535***
−0.181
−0.546***
−0.464***
−0.398***
0.171
0.125
0.082
0.261*
−0.427***
0.748***
−0.277*
−0.664***
−0.714***
−0.688***
−0.061
−0.087
−0.105
0.065
−0.030
0.500***
−0.066
−0.303**
−0.143
−0.010
0.066
0.114
−0.001
−0.213
−0.232
−0.084
0.130
0.014
0.101
−0.101
−0.101
1.000
0.431***
−0.315**
0.271*
0.111
−0.081
−0.229
−0.269*
0.431******
1.000
−0.207
0.248
0.163
−0.054
−0.109
−0.123
−0.207
1.000
−0.450***
−0.105
−0.315***
Real PCE growth
0.348**
1.000
0.065
Real GDP growth
0.407***
0.469***
0.495***
−0.175
−0.024
−0.190
0.271* 0.111
0.248 0.163
−0.450*** −0.105
1.000 −0.509***
−0.509*** 1.000
−0.945*** 0.667***
−0.905*** 0.634***
−0.839*** 0.604***
0.197 −0.203
0.176 −0.289*
0.063 −0.078
−0.081 −0.229 −0.269* −0.213 −0.232 −0.084
−0.054 −0.109 −0.123 0.130 0.014 0.101
0.407*** 0.469*** 0.495*** −0.175 −0.024 −0.190
−0.945*** −0.905*** −0.839*** 0.197 0.176 0.063
0.667*** 0.634*** 0.604*** −0.203 −0.289* −0.078
1.000 0.970*** 0.919*** −0.212 −0.192 −0.063
0.970*** 1.000 0.986*** −0.174 −0.151 −0.056
0.919*** 0.986*** 1.000 −0.155 −0.126 −0.057
−0.212 −0.174 −0.155 1.000 0.882*** 0.875***
−0.192 −0.151 −0.126 0.882*** 1.000 0.701***
−0.063 −0.056 −0.057 0.875*** 0.701*** 1.000
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Value-weighted dividend premium Older-to-younger ratio (chg) Prime consumersto-prime savers ratio (chg) Older to younger ratio Closed-end fund discount Profitability premium Cash premium Capital gains tax to dividend income tax ratio Year Time trend squared Time trend cubed Real GDP growth Real PCE growth Real PDI growth
Correlation matrix
Significance level at 10%. Significance level at 5%. Significance level at 1%.
371
372
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Fig. 1. Time series plots of dividend premium and annual change in older-toyounger ratio, 1961–2007.
Fig. 2. Time series plots of dividend premium and annual change in prime consumers-to-prime savers ratio, 1961–2007.
5. Robustness checks and future work Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test (Kwiatkowski, Phillips, Schmidt, & Shin, 1992) which uses the null hypotheses of linear stationarity and trend stationarity respectively. It can be seen that for the demographic variation variables and the control variables, the unit root tests generally support the null hypotheses of linear stationarity and trend stationarity. For the dividend premium variable, the KPSS tests appear to reject the notion of stationarity, a finding that is similar to that of Baker and Wurgler (2004a). As they highlighted, there are theoretical reasons to expect the dividend premium to be stationary and not grow without bound. The practical message is therefore to also examine the robustness of the regression results to the inclusion of time trends. The theoretical backing for and our econometric findings of stationarity properties possessed by the variables of dividend premium, demographic variation and the control variables supports the employment of ordinary least squares regression techniques to test our hypothesis. Column 1 of Table 3 shows the OLS regressions of the value-weighted dividend premium against the annual change in older-to-younger ratio over the period of 1961–2007. It can be seen from our regression results that the annual change in the older-to-younger ratio is a positive and statistically significant determinant of the dividend premium. This means that the dividend premium is high when the proportion of older population to younger population increases, while the premium is low when the proportion of older population to younger population falls. Column 2 of Table 3 shows the multivariate regression with the inclusion of a linear time trend. It can be seen that the annual change in older-to-younger ratio continues to be positively related to the dividend premium at 1% significance level. The regression coefficient of the time trend is also highly significant at 1% level. In order to allow for a more flexible diagnosis of the time trend effects, we also perform multivariate regressions using as control variables the time trend squared and the time trend cubed. The results are shown in Columns 3 and 4 of Table 3 respectively. The regression coefficients of the annual change in older-to-younger ratio remains positively related to the dividend premium at 1% significance level in both regressions. The regression coefficients of the time trend variables also continue to be highly significant at 1% level, although the nature of the relation appears to have reversed from negative to positive when the time trend cubed variable is used as the control variable. Our results therefore support our hypothesis that the timevarying demand for dividend payers is positively related to variations in the demographic structure and our findings are robust to the inclusion of time trend.
5.1. Alternative definition of demographic variation variable While our measure of demographic structure is adopted from Graham and Kumar (2006), there are alternative definitions of the demographic structure used in other research. In their analysis of the effects of demographic structure on asset prices in Asia, Eskesen, Lueth, and Syed (2008) have defined the demographic structure as the ratio of prime consumers (aged 65+) to prime savers (aged 40–65). This definition of demographic structure is also used by Krüger (2004). We have therefore adopted the definition of Eskesen et al. (2008) as an alternative definition of the demographic structure and calculated the equivalent demographic variation variable as the annual change in the prime consumers-to-prime savers ratio. Fig. 2 shows the time series plots of the dividend premium to the annual change in the prime consumers-to-prime savers ratio. It can be seen that the two variables appear to be strongly positively related. In fact Table 2 shows that the correlation between them is 0.394 with 5% significance level. The result of the multivariate regression is shown in Column 5 of Table 3. It can be seen that the alternative definition of demographic variation remains an important determinant of the dividend premium at 1% significance level. Our earlier finding is therefore robust to the alternative definition of the demographic variation measure.
5.2. Investor sentiment In trying to identify the drivers of the dividend premium, Baker and Wurgler (2004a) found initial support for a sentiment-based explanation. In particular, they compared the dividend premium to the closed-end fund discount, a measure of investor sentiment that is also favoured by Zweig (1973) and Lee, Shleifer, and Thaler (1991), and concluded that their results provide affirmative support for a sentiment interpretation. We therefore include the closed-end fund discount as a control variable in our multivariate regression to investigate the possibility that the demographic variation measure is only serving as a proxy for investor sentiment. Column 6 of Table 3 shows the results. It can be seen that demographic variation remains an important determinant of dividend premium at 1% significance level even with the inclusion of investor sentiment as an additional explanatory variable. The closed-end fund discount variable is also positively correlated to dividend premium at 5% significance level, a finding that is consistent with the conclusion of Baker and Wurgler (2004a).
Table 3 Regressions of dividend premium against annual change in older-to-younger ratio and control variables, 1961–2007. Dependent Variable
Column (1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
Value-weighted dividend premium Explanatory variables Increase in older-to-younger ratio Increase in prime consumers to prime savers ratio
Closed-end fund discount Profitability premium Cash premium Capital gains tax to dividend income tax ratio Year Time trend squared Time trend cubed Real GDP growth Real PCE growth Real PDI growth Constant R2 Observations
3.603***
4.473***
5.083***
–
4.892***
3.208*
5.013***
5.019***
5.029***
5.125***
–
2.800
8.668**
2.616*
(3.448) –
(2.606) –
(3.305) –
(3.771) –
– 11.555***
(3.480) –
(1.905) –
(3.377) –
(3.375) –
(3.336) –
(3.424) –
– –
(0.991) –
(1.992) –
(1.684) –
– –
– –
– –
– –
(2.883) –
– –
– –
– –
– –
– –
– –
– −0.082
– –
– –
– –
– –
– –
– –
– –
– –
– 0.809**
– –
– –
– –
– –
– –
(−0.263) –
– –
– –
– 0.824**
– –
– –
– –
– –
– –
(2.553) –
– 0.066
– –
– –
– –
– –
– –
– –
– –
(2.409) 0.16**
– – – –
– – – –
– – – –
– – – –
– – – –
– – – –
(0.894) 0.110*** (2.668) –
– – – −11.599
– – – –
– – – –
– – – –
– – – –
– – – –
– – – –
(2.109) −0.004 (−0.036) 28.276**
– – – – – – – – – – – – – −6.882*** (−2.785) 0.209 47
– −0.597*** (−3.637) – – – – – – – – – – 1179.046*** (3.616) 0.392 47
– – – −0.011*** (−3.376) – – – – – – – – 1.716 (0.507) 0.372 47
– – – – – 0.00*** (−3.277) – – – – – – −0.528 (−0.178) 0.364 47
– – – – – – – – – – – – – −5.692** (−2.282) 0.156 47
– – – – – – – – – – – – – −13.770*** (−3.861) 0.311 47
– – – – – – – – – – – – – −2.173 (−0.532) 0.377 43
(−1.143) – – – – – – – – – – – – 0.068 (0.01) 0.232 47
– – – – – – – 1.282 (1.07) – – – – −11.144** (−2.378) 0.229 47
– – – – – – – – – 0.806 (0.551) – – −9.736* (−1.693) 0.214 47
– – – – – – – – – – – 0.168 (0.615) −7.669*** (−2.741) 0.216 47
– – – – – – – – – – – – – 2.175 (0.082) 0.002 47
– −1.123 (−1.602) – – – – – – – – – – 2215.184 (1.600) 0.338 25
– −1.804** (−2.154) – – – – – – – – – – 3589.081** (2.147) 0.196 22
(2.024) −1.003* (−1.753) – – – – 0.874 (0.837) – – – – 1948.577* (1.729) 0.560 43
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Older-to-younger ratio
5.125***
Multivariate regressions of dividend premium against measures of demographic variation, investor sentiment, signalling, agency costs and time trend. PtD–ND = ˛0 + ˛1 dOld/Youngt + ˛2 CEFDt + ˛3 E/At+1 + ˛4 Cash/At−1 + ˛5 TAXt + ˛6 ECONt + ˛7 t + εt
373
The dividend premium PD−ND is the difference between the logs of the value-weighted market-to-book ratios for dividend payers to nonpayers, and is downloaded from the website of Jeffrey Wurgler. The demographic variation measures are given by dOld/Young and dConsumers/Savers which represent the annual change in older-to-younger ratio and the annual change in prime consumers-to-prime savers ratio respectively. Investor sentiment is represented by the closed-end fund discount CEFD which is obtained from the website of Jeffrey Wurgler. The profitability premium E/A is the difference between the natural logs of the value-weighted returns on assets ratios for dividend payers to nonpayers, while cash premium Cash/A is the difference between the natural logs of the value-weighted cash-to-asset ratios for dividend payers to nonpayers. Both variables are obtained from Liu and Shan (2007). TAX is the ratio of marginal capital gains tax rate to marginal dividend income tax rate and the data is obtained from the website of Citizens for Tax Justice. ECON represents the measures of business cycle fluctuations which are real gross domestic product (GDP) growth, real personal consumption expenditure (PCE) growth and real private domestic investment (PDI) growth. t is the time trend variable and is separately measured as the year, time trend squared and time trend cubed. Note: t-Statistics are in parentheses. * Significance level at 10%. ** Significance level at 5%. *** Significance level at 1%.
374
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
5.3. Agency costs and signalling
5.5. Business cycle fluctuations
While the idea of the dividend premium reflecting time-varying contracting problems is rejected by Baker and Wurgler (2004a), Liu and Shan (2007) found evidence that the dividend premium is higher when the need to mitigate the agency problem is greater. In particular, they examined the relation of dividend premium to proxy measures of signalling and agency costs, namely the profitability premium and the cash premium respectively. They found that the 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. They interpret this as investors valuing dividend payers with a higher premium when dividend payers have more cash and fewer profitable future investment projects than nonpayers, and concluded that this is consistent with the agency costs theory of dividends. It is worth noting that in the correlation matrix in Table 2, the cash premium variable is highly negatively correlated to the time trend at −0.945 with 5% significance level. Given the lack of inclusion of a time trend in the analysis of Liu and Shan (2007), it is highly likely that their finding of the significance of the cash premium variable is only a reflection of the importance of the time trend, and that the cash premium variable only served as a proxy for the time trend in their regression. While the conclusions of Liu and Shan (2007) are best viewed as inconclusive, we have nevertheless included the profitability premium and cash premium as control variables for completeness. Column 7 of Table 3 shows the regression of the dividend premium against the annual change in older-to-younger ratio, the profitability premium and the cash premium. It can be seen that demographic variation remains an important determinant of dividend premium at 10% significance level while the cash premium variable is highly significant at 1% level. The profitability premium is not significant.
To control for the possible time-varying effects of business cycle fluctuations on the dividend premium, we also use three measures of business activity as control variables: annual real gross GDP growth, annual real PCE growth, and annual real PDI growth. Columns 9–11 of Table 3 then shows the results of the multivariate regressions using each of the measures respectively as control variables. It can be seen that for all three regressions, the annual change in the older-to-younger ratio is shown to be strongly positively related to the dividend premium, while the coefficients for all three measures of business cycle indicators are not significant. Our findings therefore support the existence of demographic dividend clienteles and our results are robust to the inclusion of business cycle fluctuation considerations.
5.4. Tax clienteles Because companies have a choice between paying dividends and doing share buybacks, while investors typically face differential taxation on their dividend income versus capital gains, it is therefore conceivable that tax clienteles may form in response to this, and this may impact the dividend premium. Empirical studies (Brav, Graham, Harvey, & Michaely, 2005; Eckbo & Verma, 1994; Michaely, Thaler, & Womack, 1995) have generally not found the existence of tax-based preferences for dividends over repurchases. In fact, Baker and Wurgler (2004a) included a tax variable, measured as one minus the average marginal income rate, divided by one minus the average marginal long-term capital gains tax rate, in their study and were unable to find any evidence that the added tax control variable impacted the dividend premium. They hence rejected tax clienteles as a driver of the dividend premium. Despite previous research, we have, for completeness, included an alternatively defined tax control variable, measured as the ratio of the capital gains tax to the dividend income tax for the top bracket, in our robustness checks. The results of the multivariate regression are shown in Column 8 of Table 3, and it continues to highlight that the demographic variation variable is positively related to the dividend premium at 1% significance level. Our tax control variable is however not significant, a finding that is similar to Baker and Wurgler (2004a).
5.6. Absolute demographic ratio While our hypothesis is built upon the marginal opinion theory of stock price where the clear distinction between the role of the marginal investor versus the average investor in price-setting is emphasized, we remain conscious that traditional asset pricing theory assumes that the average investor and the marginal investor is the same. If this is true, then a relation between the absolute older-to-younger ratio and the dividend premium should exist. We therefore also perform a multivariate regression using the absolute older-to-younger ratio as the explanatory variable. Column 12 of Table 3 shows the results. It can be seen that the regression coefficient is not significant, a finding that is consistent with our adoption of the marginal opinion theory of stock price in our hypothesis. 5.7. Varying time periods It has been proposed that a visual examination of the time series plots of the demographic variation variable and the dividend premium in Fig. 1 appears to suggest that much of the positive relation between the two variables is accruing to the time period of 1961–1985 while the relation seems to be weak post-1986. Although there are no a priori reasons in our hypothesis to believe that the nature of relationship should alter after 1986, we nevertheless perform separate multivariate regressions for the two time periods. Our results are shown in Columns 13 and 14 of Table 3 for the time periods of 1961–1985 and 1986–2007 respectively. It can be seen that contrary to initial suspicions, the demographic variation variable is in fact strongly positively related to the dividend premium at 5% significance level for the time period of post-1986. On the other hand, the relation between the two variables appears to be weaker for the period of 1961–1985. While our results should go some way in allaying fears that the relation is only driven by the 1961–1985 time period, we also advise readers not to interpret too deeply into the results due to the severe reduction in data points used to estimate both regressions given the time period is being split in half. Column 15 of Table 3 shows the regression of the dividend premium against the demographic variation variable and all the control variables. It should be noted that for this combined multivariate regression, we have chosen to use only one out of the three available measures in each of the categories of time trend and business cycle fluctuations as control variables. This is because, as shown in the correlation matrix in Table 2, the measures within each category are highly correlated to each other. This means that using all three measures in each of those categories may risks introducing multicollinearity into the regression. We have therefore chosen to use linear time trend and real GDP growth as the representative control variables for the categories of time trend and
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
business cycle fluctuations respectively. It can be seen that demographic variation is positively related to dividend premium at 10% significance level while the closed-end fund discount is also significant at 5% level. The cash premium variable however loses its significance when a time trend is included, thus confirming our earlier suspicion that the cash premium variable is only a proxy for the time trend in the analysis of Liu and Shan (2007). Our robustness checks have therefore shown that changes in the demographic clientele is an important determinant of the dividend premium even when controlled for investor sentiment, signalling, agency costs, tax clienteles, time trend, business cycle fluctuations and varying sample periods.5 5.8. Future work While not within the scope of the paper, we believe that following the conclusions of our paper establishing the link between the dividend premium and changes in demographic clienteles, future work can be focused on an investigation of how the effects of demographic clientele variations are translated to actual corporate dividend policy. There is also the potential to explore the effects of demographic clientele changes on the stock performance of dividend-paying companies versus non-dividend-paying companies. 6. Conclusion The catering theory of dividends proposed that corporate dividend policy is driven by prevailing investor demand for dividend payers, and that managers cater to investors by paying dividends when the dividend premium is high. While Baker and Wurgler (2004a) found that the dividend premium is not driven by traditional clienteles derived from market imperfections such as taxes, transaction costs, or institutional investment constraints, we hypothesise that changes in the demographic clientele can be an important source of the time-varying demand for dividend payers. In particular, we conduct multivariate regressions and find empirical evidence that demographic variation as represented by the annual change in the older-to-younger ratio is a significant determinant of the dividend premium. This is consistent with our hypothesis, as well as the behavioural life-cycle hypothesis and the marginal opinion theory of stock price. Our findings are robust to the inclusion of control variables of investor sentiment, signalling, agency costs, tax clienteles, business cycle fluctuations and time trend, and to alternative definitions of the demographic structure as well as to varying time periods.
5 While we have tried to include as many control variables in our robustness tests as we can, the lack of available time series data means that other plausible determinants of dividend policy and the dividend premium, such as corporate leverage, may not be adequately addressed in our study.
375
References Allen, F., & Gale, D. (1994). Limited market participation and volatility of asset prices. American Economic Review, 84, 933–955. Bae, K. H., Yamada, T., & Ito, K. (2008). Interaction of investor trades and market volatility: Evidence from the Tokyo stock exchange. Pacific-Basin Finance Journal, 16, 370–388. 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. Brav, A., Graham, J., Harvey, C., & Michaely, R. (2005). Payout policy in the 21st century. Journal of Financial Economics, 77, 483–527. Eckbo, B. E., & Verma, S. (1994). Managerial share ownership, voting power, and cash dividend policy. Journal of Corporate Finance, 1, 33–62. Eskesen, L., Lueth, E., & Syed, M. (2008). The graying of Asia: Demographics, capital flows, and financial markets. IMF Regional Economic Outlook: Asia and Pacific, 41–57. Fama, E., & French, K. (2001). Disappearing dividends: Changing firm characteristics or lower propensity to pay? Journal of Financial Economics, 60, 3–44. Ferris, S., Sen, N., & Yui, H. (2006). God save the queen and her dividends: Corporate payouts in the U.K. Journal of Business, 79, 1149–1173. Gabaix, X., Krishnamurthy, A., & Vigneron, O. (2007). Limits of arbitrage: Theory and evidence from the mortgage-backed securities market. Journal of Finance, 62, 557–595. Graham, J., & Kumar, A. (2006). Do dividend clienteles exist? Evidence on dividend preferences of retail investors. Journal of Finance, 61, 1305–1336. He, Z., & Krishnamurthy, A. (2008a). A model of capital and crises. NBER Working Paper No. 14366. He, Z., & Krishnamurthy, A. (2008b). Intermediary asset pricing. NBER Working Paper No. 14517. Kwiatkowski, D., Phillips, P. C. B., Schmidt, P., & Shin, Y. (1992). Testing the null of stationarity against the alternative of a unit root. Journal of Econometrics, 54, 159–178. Krüger, D. (2004). The effects of demographic changes on aggregate savings: Some implications from the life cycle model. In Capital markets in the long term: Demography, economic development and funded pension systems. Frankfurt am Main: Center for Financial Studies and Allianz. Lee, C. M., Shleifer, A., & Thaler, R. (1991). Investor sentiment and the closed-end fund puzzle. Journal of Finance, 46, 75–110. Li, W., & Lie, E. (2006). Dividend changes and catering incentives. Journal of Financial Economics, 80, 293–308. Liu, L., & Shan, L. (2007). What is dividend premium? In Financial Management Association annual meeting 2007 Online conference proceedings, Michaely, R., Thaler, R., & Womack, K. (1995). Price reactions to dividend initiations and omissions: Overreaction and drift? Journal of Finance, 50, 573–608. Modigliani, F., & Miller, M. (1961). Dividend policy, growth, and the valuation of shares. Journal of Business XXXIV, 235–264. Shleifer, A. (2000). Inefficient markets: An introduction to behavioral finance. Oxford: Oxford University Press. Smith, R. G. E. (1967). The ‘marginal opinion’ theory of stock price. Financial Analysts Journal, 23, 127–132. Thaler, R. H., & Shefrin, H. H. (1988). The behavioral life-cycle hypothesis. Economic Inquiry, 26, 609–643. Williams, J. (1938). The theory of investment value. Cambridge, MA: Harvard University Press. Zweig, M. E. (1973). An investor expectations stock price predictive model using closed-end fund premiums. Journal of Finance, 28, 67–87.
Contents lists available at SciVerse ScienceDirect
The Quarterly Review of Economics and Finance journal homepage: www.elsevier.com/locate/qref
Demographics, dividend clienteles and the dividend premium King Fuei Lee ∗ Schroder Investment Management, 65 Chulia Street, #46-00 OCBC Centre, 049513, Singapore
a r t i c l e
i n f o
Article history: Received 26 December 2010 Received in revised form 7 August 2011 Accepted 26 August 2011 Available online 3 September 2011 JEL classification: G000 G350 C320 Keywords: Dividend policy Demographics Dividend premium Dividend clienteles
a b s t r a c t The catering theory of dividends proposed that corporate dividend policy is driven by prevailing investor demand for dividend payers, and that managers cater to investors by paying dividends when the dividend premium is high. While earlier research found that the dividend premium is not driven by traditional clienteles derived from market imperfections such as taxes, transaction costs, or institutional investment constraints, we find empirical evidence that demographic clienteles are an important source of the timevarying demand for dividend payers. In particular, we find that, as consistent with the behavioural lifecycle theory and the marginal opinion theory of stock price, the dividend premium is positively driven by demographic clientele variation represented by changes in the proportion of the older population. Our results are robust when controlled for the factors of investor sentiment, signalling, agency costs, tax clienteles, time trend, business cycle fluctuations and varying sample periods. © 2011 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.
1. Introduction Modigliani and Miller (1961) proposed in their theory of dividend policy irrelevance that in a world of perfect information, full capital mobility, no taxes and no agency costs, the dividend policy of a company should have no impact on its value. However these assumptions rarely hold true in the real world. The catering theory of dividend policy (Baker & Wurgler, 2004a) relaxes the assumption of market efficiency and proposed that managers cater to investors by paying dividends when investors put a stock price premium on payers, and by not paying dividends when investors prefer nonpayers. According to the catering theory, corporate dividend policy is therefore driven by prevailing investor demand for dividend payers which is represented by the dividend premium, defined as the log difference in the average market-to-book ratio of dividend payers to nonpayers. Empirical evidence has generally been supportive. Baker and Wurgler (2004b) studied US companies on the COMPUSTAT between the period of 1963–2000, and found that the “disappearing dividends” phenomena observed by Fama and French (2001) can be largely explained by the catering theory. Li and Lie (2006) found similar catering effects among US firms when they exam-
∗ Tel.: +65 6535 3411; fax: +65 6535 3486. E-mail address: [email protected]
ined changes in corporate payout ratios to the market dividend premium, while Ferris, Sen, and Yui (2006) extended the analysis to the UK where they concluded that a shift in catering incentives most likely drove the declining propensity to pay dividends over the sub-period of 1998–2002. Despite strong empirical evidence of the importance of dividend premium as a determinant of dividend policy, there has been sparse research on the factors driving the dividend premium. Liu and Shan (2007) attempted to fill this gap by examining the relation of dividend premium to proxies of agency costs and signalling motivations, namely the differences in cash holdings and in future profitability between dividend payers and nonpayers respectively. Yet, their positive findings are viewed as inconclusive since their analysis is conducted without consideration for time trends which substantially affects their conclusions. In fact, the notion that the dividend premium reflects time-varying contracting problems is rejected by Baker and Wurgler (2004a) as being inconsistent with the observation of improving corporate governance and declining propensity of dividend payout in the 1960s. Baker and Wurgler (2004a) also attempted to identify whether traditional dividend clienteles are the source of the time-varying demand for dividend payers. They were however unable to match up the dividend premium to any plausible proxies for clienteles. For example, when they included a tax control variable in their multivariate regression, they found that the added tax control variable did not appear to impact the dividend premium much, and thus
1062-9769/$ – see front matter © 2011 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.qref.2011.08.004
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
369
rejected tax clientele as a driver of the dividend premium. They also rejected transaction costs clienteles as an explanation, while ruling out institutional investor clienteles because of the difficulty in reconciling the rise in institutional ownership since the 1980s with the time-varying pattern in the 1960s. This paper investigates the effects of changes in demographic clienteles on the dividend premium. We hypothesise that in a world where stock prices are determined by marginal opinion (Smith, 1967; Williams, 1938) and dividend preferences of retail investors are influenced by behavioural life-cycle considerations (Thaler & Shefrin, 1988), demographic variations may induce changes in demographic clienteles that then drives the dividend premium. In particular, we find strong empirical evidence that the dividend premium is positively related to changes in the proportion of the older-to-younger population. Our results are robust when controlled for the factors of investor sentiment, signalling, agency costs, tax clienteles, business cycle fluctuations and time trend. This paper contributes to current literature by adding to the understanding of the drivers of the dividend premium through an examination of demographic clientele changes as a source of the time-varying demand for dividend payers. To our knowledge, there has not been any work done in this aspect. The rest of this paper is structured as follows: Section 2 discusses the behavioural life-cycle hypothesis and the marginal opinion theory of stock price, and introduces our hypothesis. Section 3 describes the data sample and the methodology pursued. The empirical findings are reported in Section 4, while robustness tests are conducted in Section 5. Section 6 concludes the paper.
In traditional asset pricing theory, the distinction between the average investor and the marginal investor is usually not important as the marginal investor in every asset market is also assumed to be the same broadly diversified representative investor i.e. the average investor. Shleifer (2000) however highlighted that when investors’ trading schedules are not flat, then the trading of participating investors can affect asset prices. In other words, it is the marginal investor, rather than the average or representative investor, who affects asset prices. While the emphasis on the distinction between the marginal investor and the average investor as a driver of asset prices is only more commonly adopted in recent research (Allen & Gale, 1994; Bae, Yamada, & Ito, 2008; Gabaix, Krishnamurthy, & Vigneron, 2007; He & Krishnamurthy, 2008a, 2008b), the concept of marginal opinion as the determinant of stock prices was in fact first raised by Williams (1938) and subsequently extended by Smith (1967). According to the marginal opinion theory, in a market comprising of a number of interested parties who each possess an opinion as to the worth of the stock, the price of the stock is not set by the majority, regardless of how overwhelming it is, but by the last owner. This means that marginal opinion will determine the market price. If the stock prices of dividend-paying and non-paying stocks are determined by the marginal investor rather than the average investor, then it means that rather than being related to the absolute demographic structure which reflects the average investor opinion, dividend premium should be related to changes in the demographic structure which proxies for the marginal investor opinion i.e. demographic clientele variation.
2. Demographic clientele variations and the dividend premium
2.3. Hypothesis of dividend premium and changes in demographic clienteles
2.1. Behavioural life-cycle theory and the dividend preferences of older investors
We hypothesise that in a world where stock prices are determined by the marginal investor and where components of wealth are mentally treated as being non-fungible, the preference for dividend-paying stocks by older investors means that the dividend premium should be positively related to changes in the proportion of the older population, a proxy for the marginal investor opinion. In essence, the larger the increase in the proportion of the older population is, the greater the marginal investor preference for dividend-paying stocks and hence the higher the dividend premium. Our hypothesis is therefore that the time-varying demand for dividend payers is driven by variations in the demographic clienteles.
According to the behavioural life-cycle theory (Thaler & Shefrin, 1988), households treat components of their wealth as nonfungible. In particular, wealth is assumed to be broken into three mental accounts, namely current income, current assets and future income, with the temptation to spend being greatest for current income and least for future income. The behavioural life-cycle theory therefore hypothesises that in the later stage of a household’s life cycle when they reach retirement and begin to dis-save, the investor perception of the non-fungibility between dividends and capital gains should lead to a preference for dividend-paying stocks by older investors for consumption purposes. Empirical evidence has generally been supportive. Graham and Kumar (2006) studied the stock holdings and trading behaviour of 77,995 households over the period of 1991–1996 and found that, compared to younger investors, older investors allocate a greater proportion of their equity portfolios to dividend paying stocks. This suggests that senior investors have a greater preference for dividends. 2.2. Marginal opinion theory of stock price Given the dividend preference of older investors, it is reasonable to conjecture that when the general population has a greater proportion of older people, then the greater degree of buying of dividend-paying stocks by these senior investors should lead to a high dividend premium. Such a conjecture however assumes that stock prices are determined by the opinion of the average investor. If share prices are instead determined by the marginal investor, then the implications can be quite different.
3. Data sample and methodology This section briefly discusses the data sources and the variables’ definitions. The dividend premium PD−ND is the difference between the logs of the value-weighted market-to-book ratios for dividend payers to nonpayers, and the data is downloaded from the website of Jeffrey Wurgler.1 Following Graham and Kumar (2006), we use the older-to-younger ratio Old/Young, defined as the proportion of population aged above 65 to those aged under 45, as the variable representing the demographic structure. Demographic variation is therefore expressed as the annual change in the older-to-younger ratio dOld/Young dOld/Youngt = Old/Youngt − Old/Youngt−1
1
Available at http://pages.stern.nyu.edu/∼jwurgler/.
(1)
2 3 4
Available at http://www.census.gov. Available at http://www.ctj.org. Available at http://www.bea.gov.
3.59 1.68 6.30 −0.80 47 3.36 2.02 7.20 −1.90 47 27072.00 30727.77 103823.00 1.00 47 1984.00 760.00 13.71 678.81 2007.00 2209.00 1961.00 1.00 47 47 0.60 0.24 1.00 0.27 47 −61.69 75.60 88.67 −188.97 43 50.67 36.75 239.55 15.80 43 8.64 7.23 23.53 −10.91 47 85.27 8.76 101.40 70.06 47 0.41 1.63 2.76 −3.77 47
Fig. 1 shows the time series plots of the value-weighted dividend premium and the demographic variation variable. It is observed that while both variables are not perfectly synchronous, they are visibly positively related to each other. Indeed it can be seen from the correlation matrix in Table 2 that the contemporaneous correlation between dividend premium and the demographic variation measure of annual change in older-to-younger ratio is 0.457 at 5% significance level. Table 1 shows the descriptive statistics of the dividend premium, the demographic variation variables as well as the control variables, while Table 2 shows their unit root test statistics and the correlation matrix. The unit root test employed here is the
Table 1 Descriptive statistics, 1961–2007.
4. Empirical findings
−4.79 18.26 31.23 −44.43 47
where ˛i is the regression coefficient of explanatory variable i and εt is the random disturbance term.
Mean Standard deviation Maximum Minimum Number of observations
(2)
Annual change in older-to-younger ratio
+ ˛4 Cash/At−1 + ˛5 TAXt + ˛6 ECONt + ˛7 t + εt
Ann chg in prime consumers to prime savers ratio
PtD–ND = ˛0 + ˛1 dOld/Youngt + ˛2 CEFDt + ˛3 E/At+1
0.08 0.62 1.08 −1.10 47
Older to younger ratio
Closedend fund discount
Profitability Cash premium premium
Capital gains tax to dividend income tax ratio
Year
Time trend squared
Time trend cubed
Real GDP growth
Real PCE growth
Real PDI growth
We also employ an alternative measure of the demographic variation variable dConsumers/Savers defined as the annual change in the prime consumers-to-prime savers ratio in our robustness test, where prime consumers are persons aged above 65 while prime savers are persons aged from 45 to 65. The US population data used for the calculations of the two measures of demographic variations is downloaded from the US Census Bureau2 website. In our other robustness checks, we include measures of investor sentiment, signalling, agency costs, tax clienteles, time trends and business cycle fluctuations as control variables. Our chosen measure of investor sentiment is the closed-end fund discount CEFD which is the value-weighted discount on closed-end funds. Data for the variable is obtained from the website of Jeffrey Wurgler. Following Liu and Shan (2007), we also include measures of signalling and agency costs as control variables. The measure of signalling used is the profitability premium E/A which is defined as the difference between the natural logs of the value-weighted future returns-onassets ratios for dividend payers to nonpayers, while the measure of agency costs is the cash premium Cash/A which is the difference between the natural logs of the value-weighted cash-to-asset ratios for dividend payers to nonpayers. Data for both variables are obtained from Liu and Shan (2007). The measure of tax clienteles TAX is represented in our study as the ratio of capital gains tax versus dividend income tax, and the data is obtained from the website of Citizens for Tax Justice.3 As for measures of business cycle fluctuations ECON, we separately use real gross domestic product (GDP) growth, real personal consumption expenditure (PCE) growth and real private domestic investment (PDI) growth as control variables. The data for the business cycle indicators are obtained from the US Department of Commerce Bureau of Economic Analysis.4 To allow for a more flexible trend in the diagnosis of the effects of time t, we also separately use the linear time trend (year), time trend squared (t2 ) and time trend cubed (t3 ) as control variables. The time period employed in this study is from 1961 to 2007 which represents the period for which the data for dividend premium is available. Following the methodology of Liu and Shan (2007), we employ multivariate OLS regression to estimate the relation. The regression is expressed as
4.69 8.95 29.50 −17.70 47
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Value-weighted dividend premium
370
Table 2 Unit root statistics and correlation matrix, 1961–2007. Unit root
* ** ***
Ann chg in prime consumers to prime savers ratio
KPSS: trend stationarity
0.423*
0.189**
1.000
0.107
0.394***
0.173
0.124
0.457***
0.124
0.117
0.141
0.122*
0.343
0.146*
−0.081
−0.427***
−0.030
0.439* 0.505**
0.105 0.182**
0.535*** −0.181
0.748*** −0.277*
0.500*** −0.066
– 0.459* 0.443* 0.307 0.259 0.182
– 0.151** 0.150** 0.125* 0.132* 0.130*
−0.546*** −0.464*** −0.398*** 0.171 0.125 0.082
−0.664*** −0.714*** −0.688**** −0.061 −0.087 −0.105
−0.303** −0.143 −0.010 0.066 0.114 −0.001
0.293
Valueweighted dividend premium
Annual change in older-toyounger ratio
KPSS: level stationarity
−0.039 0.348**
0.394*** 1.000 0.150
0.719*** 0.261*
0.457*** 0.150
Older to younger ratio
−0.039 0.719***
Closedend fund discount
Profitability Cash premium premium
Capital gains tax to dividend income tax ratio
Year
Time trend squared
Time trend cubed
Real PDI growth
−0.081
0.535***
−0.181
−0.546***
−0.464***
−0.398***
0.171
0.125
0.082
0.261*
−0.427***
0.748***
−0.277*
−0.664***
−0.714***
−0.688***
−0.061
−0.087
−0.105
0.065
−0.030
0.500***
−0.066
−0.303**
−0.143
−0.010
0.066
0.114
−0.001
−0.213
−0.232
−0.084
0.130
0.014
0.101
−0.101
−0.101
1.000
0.431***
−0.315**
0.271*
0.111
−0.081
−0.229
−0.269*
0.431******
1.000
−0.207
0.248
0.163
−0.054
−0.109
−0.123
−0.207
1.000
−0.450***
−0.105
−0.315***
Real PCE growth
0.348**
1.000
0.065
Real GDP growth
0.407***
0.469***
0.495***
−0.175
−0.024
−0.190
0.271* 0.111
0.248 0.163
−0.450*** −0.105
1.000 −0.509***
−0.509*** 1.000
−0.945*** 0.667***
−0.905*** 0.634***
−0.839*** 0.604***
0.197 −0.203
0.176 −0.289*
0.063 −0.078
−0.081 −0.229 −0.269* −0.213 −0.232 −0.084
−0.054 −0.109 −0.123 0.130 0.014 0.101
0.407*** 0.469*** 0.495*** −0.175 −0.024 −0.190
−0.945*** −0.905*** −0.839*** 0.197 0.176 0.063
0.667*** 0.634*** 0.604*** −0.203 −0.289* −0.078
1.000 0.970*** 0.919*** −0.212 −0.192 −0.063
0.970*** 1.000 0.986*** −0.174 −0.151 −0.056
0.919*** 0.986*** 1.000 −0.155 −0.126 −0.057
−0.212 −0.174 −0.155 1.000 0.882*** 0.875***
−0.192 −0.151 −0.126 0.882*** 1.000 0.701***
−0.063 −0.056 −0.057 0.875*** 0.701*** 1.000
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Value-weighted dividend premium Older-to-younger ratio (chg) Prime consumersto-prime savers ratio (chg) Older to younger ratio Closed-end fund discount Profitability premium Cash premium Capital gains tax to dividend income tax ratio Year Time trend squared Time trend cubed Real GDP growth Real PCE growth Real PDI growth
Correlation matrix
Significance level at 10%. Significance level at 5%. Significance level at 1%.
371
372
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Fig. 1. Time series plots of dividend premium and annual change in older-toyounger ratio, 1961–2007.
Fig. 2. Time series plots of dividend premium and annual change in prime consumers-to-prime savers ratio, 1961–2007.
5. Robustness checks and future work Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test (Kwiatkowski, Phillips, Schmidt, & Shin, 1992) which uses the null hypotheses of linear stationarity and trend stationarity respectively. It can be seen that for the demographic variation variables and the control variables, the unit root tests generally support the null hypotheses of linear stationarity and trend stationarity. For the dividend premium variable, the KPSS tests appear to reject the notion of stationarity, a finding that is similar to that of Baker and Wurgler (2004a). As they highlighted, there are theoretical reasons to expect the dividend premium to be stationary and not grow without bound. The practical message is therefore to also examine the robustness of the regression results to the inclusion of time trends. The theoretical backing for and our econometric findings of stationarity properties possessed by the variables of dividend premium, demographic variation and the control variables supports the employment of ordinary least squares regression techniques to test our hypothesis. Column 1 of Table 3 shows the OLS regressions of the value-weighted dividend premium against the annual change in older-to-younger ratio over the period of 1961–2007. It can be seen from our regression results that the annual change in the older-to-younger ratio is a positive and statistically significant determinant of the dividend premium. This means that the dividend premium is high when the proportion of older population to younger population increases, while the premium is low when the proportion of older population to younger population falls. Column 2 of Table 3 shows the multivariate regression with the inclusion of a linear time trend. It can be seen that the annual change in older-to-younger ratio continues to be positively related to the dividend premium at 1% significance level. The regression coefficient of the time trend is also highly significant at 1% level. In order to allow for a more flexible diagnosis of the time trend effects, we also perform multivariate regressions using as control variables the time trend squared and the time trend cubed. The results are shown in Columns 3 and 4 of Table 3 respectively. The regression coefficients of the annual change in older-to-younger ratio remains positively related to the dividend premium at 1% significance level in both regressions. The regression coefficients of the time trend variables also continue to be highly significant at 1% level, although the nature of the relation appears to have reversed from negative to positive when the time trend cubed variable is used as the control variable. Our results therefore support our hypothesis that the timevarying demand for dividend payers is positively related to variations in the demographic structure and our findings are robust to the inclusion of time trend.
5.1. Alternative definition of demographic variation variable While our measure of demographic structure is adopted from Graham and Kumar (2006), there are alternative definitions of the demographic structure used in other research. In their analysis of the effects of demographic structure on asset prices in Asia, Eskesen, Lueth, and Syed (2008) have defined the demographic structure as the ratio of prime consumers (aged 65+) to prime savers (aged 40–65). This definition of demographic structure is also used by Krüger (2004). We have therefore adopted the definition of Eskesen et al. (2008) as an alternative definition of the demographic structure and calculated the equivalent demographic variation variable as the annual change in the prime consumers-to-prime savers ratio. Fig. 2 shows the time series plots of the dividend premium to the annual change in the prime consumers-to-prime savers ratio. It can be seen that the two variables appear to be strongly positively related. In fact Table 2 shows that the correlation between them is 0.394 with 5% significance level. The result of the multivariate regression is shown in Column 5 of Table 3. It can be seen that the alternative definition of demographic variation remains an important determinant of the dividend premium at 1% significance level. Our earlier finding is therefore robust to the alternative definition of the demographic variation measure.
5.2. Investor sentiment In trying to identify the drivers of the dividend premium, Baker and Wurgler (2004a) found initial support for a sentiment-based explanation. In particular, they compared the dividend premium to the closed-end fund discount, a measure of investor sentiment that is also favoured by Zweig (1973) and Lee, Shleifer, and Thaler (1991), and concluded that their results provide affirmative support for a sentiment interpretation. We therefore include the closed-end fund discount as a control variable in our multivariate regression to investigate the possibility that the demographic variation measure is only serving as a proxy for investor sentiment. Column 6 of Table 3 shows the results. It can be seen that demographic variation remains an important determinant of dividend premium at 1% significance level even with the inclusion of investor sentiment as an additional explanatory variable. The closed-end fund discount variable is also positively correlated to dividend premium at 5% significance level, a finding that is consistent with the conclusion of Baker and Wurgler (2004a).
Table 3 Regressions of dividend premium against annual change in older-to-younger ratio and control variables, 1961–2007. Dependent Variable
Column (1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
Value-weighted dividend premium Explanatory variables Increase in older-to-younger ratio Increase in prime consumers to prime savers ratio
Closed-end fund discount Profitability premium Cash premium Capital gains tax to dividend income tax ratio Year Time trend squared Time trend cubed Real GDP growth Real PCE growth Real PDI growth Constant R2 Observations
3.603***
4.473***
5.083***
–
4.892***
3.208*
5.013***
5.019***
5.029***
5.125***
–
2.800
8.668**
2.616*
(3.448) –
(2.606) –
(3.305) –
(3.771) –
– 11.555***
(3.480) –
(1.905) –
(3.377) –
(3.375) –
(3.336) –
(3.424) –
– –
(0.991) –
(1.992) –
(1.684) –
– –
– –
– –
– –
(2.883) –
– –
– –
– –
– –
– –
– –
– −0.082
– –
– –
– –
– –
– –
– –
– –
– –
– 0.809**
– –
– –
– –
– –
– –
(−0.263) –
– –
– –
– 0.824**
– –
– –
– –
– –
– –
(2.553) –
– 0.066
– –
– –
– –
– –
– –
– –
– –
(2.409) 0.16**
– – – –
– – – –
– – – –
– – – –
– – – –
– – – –
(0.894) 0.110*** (2.668) –
– – – −11.599
– – – –
– – – –
– – – –
– – – –
– – – –
– – – –
(2.109) −0.004 (−0.036) 28.276**
– – – – – – – – – – – – – −6.882*** (−2.785) 0.209 47
– −0.597*** (−3.637) – – – – – – – – – – 1179.046*** (3.616) 0.392 47
– – – −0.011*** (−3.376) – – – – – – – – 1.716 (0.507) 0.372 47
– – – – – 0.00*** (−3.277) – – – – – – −0.528 (−0.178) 0.364 47
– – – – – – – – – – – – – −5.692** (−2.282) 0.156 47
– – – – – – – – – – – – – −13.770*** (−3.861) 0.311 47
– – – – – – – – – – – – – −2.173 (−0.532) 0.377 43
(−1.143) – – – – – – – – – – – – 0.068 (0.01) 0.232 47
– – – – – – – 1.282 (1.07) – – – – −11.144** (−2.378) 0.229 47
– – – – – – – – – 0.806 (0.551) – – −9.736* (−1.693) 0.214 47
– – – – – – – – – – – 0.168 (0.615) −7.669*** (−2.741) 0.216 47
– – – – – – – – – – – – – 2.175 (0.082) 0.002 47
– −1.123 (−1.602) – – – – – – – – – – 2215.184 (1.600) 0.338 25
– −1.804** (−2.154) – – – – – – – – – – 3589.081** (2.147) 0.196 22
(2.024) −1.003* (−1.753) – – – – 0.874 (0.837) – – – – 1948.577* (1.729) 0.560 43
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
Older-to-younger ratio
5.125***
Multivariate regressions of dividend premium against measures of demographic variation, investor sentiment, signalling, agency costs and time trend. PtD–ND = ˛0 + ˛1 dOld/Youngt + ˛2 CEFDt + ˛3 E/At+1 + ˛4 Cash/At−1 + ˛5 TAXt + ˛6 ECONt + ˛7 t + εt
373
The dividend premium PD−ND is the difference between the logs of the value-weighted market-to-book ratios for dividend payers to nonpayers, and is downloaded from the website of Jeffrey Wurgler. The demographic variation measures are given by dOld/Young and dConsumers/Savers which represent the annual change in older-to-younger ratio and the annual change in prime consumers-to-prime savers ratio respectively. Investor sentiment is represented by the closed-end fund discount CEFD which is obtained from the website of Jeffrey Wurgler. The profitability premium E/A is the difference between the natural logs of the value-weighted returns on assets ratios for dividend payers to nonpayers, while cash premium Cash/A is the difference between the natural logs of the value-weighted cash-to-asset ratios for dividend payers to nonpayers. Both variables are obtained from Liu and Shan (2007). TAX is the ratio of marginal capital gains tax rate to marginal dividend income tax rate and the data is obtained from the website of Citizens for Tax Justice. ECON represents the measures of business cycle fluctuations which are real gross domestic product (GDP) growth, real personal consumption expenditure (PCE) growth and real private domestic investment (PDI) growth. t is the time trend variable and is separately measured as the year, time trend squared and time trend cubed. Note: t-Statistics are in parentheses. * Significance level at 10%. ** Significance level at 5%. *** Significance level at 1%.
374
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
5.3. Agency costs and signalling
5.5. Business cycle fluctuations
While the idea of the dividend premium reflecting time-varying contracting problems is rejected by Baker and Wurgler (2004a), Liu and Shan (2007) found evidence that the dividend premium is higher when the need to mitigate the agency problem is greater. In particular, they examined the relation of dividend premium to proxy measures of signalling and agency costs, namely the profitability premium and the cash premium respectively. They found that the 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. They interpret this as investors valuing dividend payers with a higher premium when dividend payers have more cash and fewer profitable future investment projects than nonpayers, and concluded that this is consistent with the agency costs theory of dividends. It is worth noting that in the correlation matrix in Table 2, the cash premium variable is highly negatively correlated to the time trend at −0.945 with 5% significance level. Given the lack of inclusion of a time trend in the analysis of Liu and Shan (2007), it is highly likely that their finding of the significance of the cash premium variable is only a reflection of the importance of the time trend, and that the cash premium variable only served as a proxy for the time trend in their regression. While the conclusions of Liu and Shan (2007) are best viewed as inconclusive, we have nevertheless included the profitability premium and cash premium as control variables for completeness. Column 7 of Table 3 shows the regression of the dividend premium against the annual change in older-to-younger ratio, the profitability premium and the cash premium. It can be seen that demographic variation remains an important determinant of dividend premium at 10% significance level while the cash premium variable is highly significant at 1% level. The profitability premium is not significant.
To control for the possible time-varying effects of business cycle fluctuations on the dividend premium, we also use three measures of business activity as control variables: annual real gross GDP growth, annual real PCE growth, and annual real PDI growth. Columns 9–11 of Table 3 then shows the results of the multivariate regressions using each of the measures respectively as control variables. It can be seen that for all three regressions, the annual change in the older-to-younger ratio is shown to be strongly positively related to the dividend premium, while the coefficients for all three measures of business cycle indicators are not significant. Our findings therefore support the existence of demographic dividend clienteles and our results are robust to the inclusion of business cycle fluctuation considerations.
5.4. Tax clienteles Because companies have a choice between paying dividends and doing share buybacks, while investors typically face differential taxation on their dividend income versus capital gains, it is therefore conceivable that tax clienteles may form in response to this, and this may impact the dividend premium. Empirical studies (Brav, Graham, Harvey, & Michaely, 2005; Eckbo & Verma, 1994; Michaely, Thaler, & Womack, 1995) have generally not found the existence of tax-based preferences for dividends over repurchases. In fact, Baker and Wurgler (2004a) included a tax variable, measured as one minus the average marginal income rate, divided by one minus the average marginal long-term capital gains tax rate, in their study and were unable to find any evidence that the added tax control variable impacted the dividend premium. They hence rejected tax clienteles as a driver of the dividend premium. Despite previous research, we have, for completeness, included an alternatively defined tax control variable, measured as the ratio of the capital gains tax to the dividend income tax for the top bracket, in our robustness checks. The results of the multivariate regression are shown in Column 8 of Table 3, and it continues to highlight that the demographic variation variable is positively related to the dividend premium at 1% significance level. Our tax control variable is however not significant, a finding that is similar to Baker and Wurgler (2004a).
5.6. Absolute demographic ratio While our hypothesis is built upon the marginal opinion theory of stock price where the clear distinction between the role of the marginal investor versus the average investor in price-setting is emphasized, we remain conscious that traditional asset pricing theory assumes that the average investor and the marginal investor is the same. If this is true, then a relation between the absolute older-to-younger ratio and the dividend premium should exist. We therefore also perform a multivariate regression using the absolute older-to-younger ratio as the explanatory variable. Column 12 of Table 3 shows the results. It can be seen that the regression coefficient is not significant, a finding that is consistent with our adoption of the marginal opinion theory of stock price in our hypothesis. 5.7. Varying time periods It has been proposed that a visual examination of the time series plots of the demographic variation variable and the dividend premium in Fig. 1 appears to suggest that much of the positive relation between the two variables is accruing to the time period of 1961–1985 while the relation seems to be weak post-1986. Although there are no a priori reasons in our hypothesis to believe that the nature of relationship should alter after 1986, we nevertheless perform separate multivariate regressions for the two time periods. Our results are shown in Columns 13 and 14 of Table 3 for the time periods of 1961–1985 and 1986–2007 respectively. It can be seen that contrary to initial suspicions, the demographic variation variable is in fact strongly positively related to the dividend premium at 5% significance level for the time period of post-1986. On the other hand, the relation between the two variables appears to be weaker for the period of 1961–1985. While our results should go some way in allaying fears that the relation is only driven by the 1961–1985 time period, we also advise readers not to interpret too deeply into the results due to the severe reduction in data points used to estimate both regressions given the time period is being split in half. Column 15 of Table 3 shows the regression of the dividend premium against the demographic variation variable and all the control variables. It should be noted that for this combined multivariate regression, we have chosen to use only one out of the three available measures in each of the categories of time trend and business cycle fluctuations as control variables. This is because, as shown in the correlation matrix in Table 2, the measures within each category are highly correlated to each other. This means that using all three measures in each of those categories may risks introducing multicollinearity into the regression. We have therefore chosen to use linear time trend and real GDP growth as the representative control variables for the categories of time trend and
K.F. Lee / The Quarterly Review of Economics and Finance 51 (2011) 368–375
business cycle fluctuations respectively. It can be seen that demographic variation is positively related to dividend premium at 10% significance level while the closed-end fund discount is also significant at 5% level. The cash premium variable however loses its significance when a time trend is included, thus confirming our earlier suspicion that the cash premium variable is only a proxy for the time trend in the analysis of Liu and Shan (2007). Our robustness checks have therefore shown that changes in the demographic clientele is an important determinant of the dividend premium even when controlled for investor sentiment, signalling, agency costs, tax clienteles, time trend, business cycle fluctuations and varying sample periods.5 5.8. Future work While not within the scope of the paper, we believe that following the conclusions of our paper establishing the link between the dividend premium and changes in demographic clienteles, future work can be focused on an investigation of how the effects of demographic clientele variations are translated to actual corporate dividend policy. There is also the potential to explore the effects of demographic clientele changes on the stock performance of dividend-paying companies versus non-dividend-paying companies. 6. Conclusion The catering theory of dividends proposed that corporate dividend policy is driven by prevailing investor demand for dividend payers, and that managers cater to investors by paying dividends when the dividend premium is high. While Baker and Wurgler (2004a) found that the dividend premium is not driven by traditional clienteles derived from market imperfections such as taxes, transaction costs, or institutional investment constraints, we hypothesise that changes in the demographic clientele can be an important source of the time-varying demand for dividend payers. In particular, we conduct multivariate regressions and find empirical evidence that demographic variation as represented by the annual change in the older-to-younger ratio is a significant determinant of the dividend premium. This is consistent with our hypothesis, as well as the behavioural life-cycle hypothesis and the marginal opinion theory of stock price. Our findings are robust to the inclusion of control variables of investor sentiment, signalling, agency costs, tax clienteles, business cycle fluctuations and time trend, and to alternative definitions of the demographic structure as well as to varying time periods.
5 While we have tried to include as many control variables in our robustness tests as we can, the lack of available time series data means that other plausible determinants of dividend policy and the dividend premium, such as corporate leverage, may not be adequately addressed in our study.
375
References Allen, F., & Gale, D. (1994). Limited market participation and volatility of asset prices. American Economic Review, 84, 933–955. Bae, K. H., Yamada, T., & Ito, K. (2008). Interaction of investor trades and market volatility: Evidence from the Tokyo stock exchange. Pacific-Basin Finance Journal, 16, 370–388. 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. Brav, A., Graham, J., Harvey, C., & Michaely, R. (2005). Payout policy in the 21st century. Journal of Financial Economics, 77, 483–527. Eckbo, B. E., & Verma, S. (1994). Managerial share ownership, voting power, and cash dividend policy. Journal of Corporate Finance, 1, 33–62. Eskesen, L., Lueth, E., & Syed, M. (2008). The graying of Asia: Demographics, capital flows, and financial markets. IMF Regional Economic Outlook: Asia and Pacific, 41–57. Fama, E., & French, K. (2001). Disappearing dividends: Changing firm characteristics or lower propensity to pay? Journal of Financial Economics, 60, 3–44. Ferris, S., Sen, N., & Yui, H. (2006). God save the queen and her dividends: Corporate payouts in the U.K. Journal of Business, 79, 1149–1173. Gabaix, X., Krishnamurthy, A., & Vigneron, O. (2007). Limits of arbitrage: Theory and evidence from the mortgage-backed securities market. Journal of Finance, 62, 557–595. Graham, J., & Kumar, A. (2006). Do dividend clienteles exist? Evidence on dividend preferences of retail investors. Journal of Finance, 61, 1305–1336. He, Z., & Krishnamurthy, A. (2008a). A model of capital and crises. NBER Working Paper No. 14366. He, Z., & Krishnamurthy, A. (2008b). Intermediary asset pricing. NBER Working Paper No. 14517. Kwiatkowski, D., Phillips, P. C. B., Schmidt, P., & Shin, Y. (1992). Testing the null of stationarity against the alternative of a unit root. Journal of Econometrics, 54, 159–178. Krüger, D. (2004). The effects of demographic changes on aggregate savings: Some implications from the life cycle model. In Capital markets in the long term: Demography, economic development and funded pension systems. Frankfurt am Main: Center for Financial Studies and Allianz. Lee, C. M., Shleifer, A., & Thaler, R. (1991). Investor sentiment and the closed-end fund puzzle. Journal of Finance, 46, 75–110. Li, W., & Lie, E. (2006). Dividend changes and catering incentives. Journal of Financial Economics, 80, 293–308. Liu, L., & Shan, L. (2007). What is dividend premium? In Financial Management Association annual meeting 2007 Online conference proceedings, Michaely, R., Thaler, R., & Womack, K. (1995). Price reactions to dividend initiations and omissions: Overreaction and drift? Journal of Finance, 50, 573–608. Modigliani, F., & Miller, M. (1961). Dividend policy, growth, and the valuation of shares. Journal of Business XXXIV, 235–264. Shleifer, A. (2000). Inefficient markets: An introduction to behavioral finance. Oxford: Oxford University Press. Smith, R. G. E. (1967). The ‘marginal opinion’ theory of stock price. Financial Analysts Journal, 23, 127–132. Thaler, R. H., & Shefrin, H. H. (1988). The behavioral life-cycle hypothesis. Economic Inquiry, 26, 609–643. Williams, J. (1938). The theory of investment value. Cambridge, MA: Harvard University Press. Zweig, M. E. (1973). An investor expectations stock price predictive model using closed-end fund premiums. Journal of Finance, 28, 67–87.