The Effect of Taxes on Stock Prices: Case for Australia s Full Imputation Scheme

Transcription

1 Preliminary. Comments welcome. The Effect of Taxes on Stock Prices: Case for Australia s Full Imputation Scheme Jaywon Lee * 2 nd Year Ph.D. Student Graduate School of Business Columbia University April 29, 2003 * 311 Uris Hall, 3022 Broadway, New York, NY 10027; jl2018@columbia.edu. This paper originates from a course project in Professor Deen Kemsley s Tax Factors and Business Decisions class in the Columbia Business School s MBA program. I wish to thank Deen Kemsley, Bjorn Jorgensen, Doron Nissim and Kenton Yee for valuable comments. All errors are my own.

2 Abstract This article investigates the controversy concerning the effect of dividend yields on common stock returns. This issue is quite timely since the Bush administration currently proposing the elimination of corporate dividends from income taxation. Australia s Imputation Scheme of 1986 offers an ideal controlled experiment setting for assessing such effects. The dividend tax imputation of Australia created an atmosphere where dividend is tax-preferred to net income transferred to retained earnings on the balance sheet. Dividend per share and the payout ratio have increased on average post-imputation. In addition, the cumulative abnormal returns (CAR) show that the tax changes affected high-dividend and low-dividend stocks differently pre- and post- imputation. Some implications regarding the stock market anticipation of tax law changes and government policies are given. 1

3 1. Introduction The effect of taxes on stock prices has long been debated. Specifically, there has been substantial controversy over the effect of dividend yields on stock returns. Some empirical research has found evidence of positive association between common stock returns and dividend yields while others have found contrary evidence leading to other factors. Corporations pay taxes on the dividends they issue. At the taxpayer level, dividends are taxed again. In many cases, dividend tax rates can be many percentages higher than those applied to capital gains. Depending on the amount of this imbalance, the capital market may result in a bias on the part of investor clienteles toward shares that provide capital growth versus shares that provide high dividend yield. This paper investigates the effect of differential taxation of dividends on stock returns. In addition to providing additional evidence regarding the effect of taxes on stock prices, this paper attempts to shed some light on the current policy debate regarding the Bush administration s tax cut proposal. Washington advocates the abolition of the double taxation of stock dividends. The chairman of the White House Council of Economic Advisers, R. Glenn Hubbard, argues that eliminating dividend taxes could lead to a 20% surge in equity prices (Blouin, 2003). Many other scholars and practitioners remain skeptical about such measures. Therefore, examining regimes where similar tax changes have occurred should be a meaningful exercise. Particularly, evaluating regimes with a similar institutional background as the U.S. will prove to be much more fruitful. For instance, La Porta et al. (2000) document common traits among common law countries in regards to dividend policies as compared to civil law counterparts. Poterba and Summers (1984) use British 2

4 data to examine the effects of dividend taxes on investors relative valuation of dividends and capital gains. Their study was possible since there have been two radical changes in British tax policy. Amoako-Adu (1983) provides additional empirical evidence on the effect of the differential taxation of dividends and capital gains on stock prices by analyzing the Canadian Tax Reform. This paper analyzes the dividend tax policy debate using Australia s new company tax system known as imputation announced in December of 1986 and put into effect in July of Surprisingly, there has not been much empirical research documented regarding dividend yield and stock returns using the imputation scheme in Australia even though such a setting creates a fruitful quasi-experimental research opportunity. This paper examines the effect of tax change on stock prices by investigating the case for Australia s full imputation system. Studying the relationship between dividends and stock price movements during different tax regimes offers an ideal controlled experiment for assessing the effects of taxes on investors valuation of dividends (Poterba and Summers 1984). This article consists of seven sections. Section 1 is the introduction and points of departure. Section 2 provides the reviews of past literature regarding dividend policy and tax effects on stock returns. Section 3 examines Australia s imputation scheme in more detail. Data and methodologies are outlined in section 4, and the results are presented in section 5. In section 6, regression analysis is carried out for some important variables that may influence the common stock returns together with other robustness checks. Finally in section 7, the main conclusions are given and some outlines for further research are presented. 3

5 2. Past Literature on Dividends In the dividend displace property, Modigliani and Miller (1961) argue that the dividend policy should be irrelevant to investors. Namely, the dollar dividend should be worth a dollar to shareholders. In the traditional view, Brennan (1970) posits that capital gains are preferred over dividends in terms of tax since it imposes unnecessary dividend tax on shareholders who would rather pay taxes on capital gains. Miller and Scholes (1978), in their tax irrelevance hypothesis, maintain that personal taxes are immaterial for the marginal investor. Thus, dividend taxes do not influence the cost of capital. In the tax capitalization hypothesis (King, 1977), dividend taxes are applied to all earnings of the firm including the reinvested portion of earnings firms will eventually pay out reinvested earnings as taxable dividends. This implies that the personal tax on distributions is unavoidable, and the present value of the tax is independent of the timing of dividend payouts. In sum, the above theories create a dividend puzzle. In the same line, there has been considerable controversy concerning the effect of dividend yields and stock returns. The debate centers on whether or not the positive association between common stock returns and dividend yields reported in a number of empirical studies can be attributed to information effects (Litzenberger and Ramaswamy, 1982). Harris, Hubbard and Kemsley (2001) examine the hypothesis that dividend taxes are capitalized into share prices by focusing on investors implicit valuations of retained earnings versus paid-in equity. Harris, Hubbard and Kemsley provide evidence that differences in dividend tax rates across US tax regimes are associated with predictable 4

6 differences in the magnitude of the implied tax discount for retained earnings. Although the US has changed its top rate on dividends ranging from 28 to 70 percent, the US stock market has never experienced full imputation of dividend taxes. Therefore, scholars have turned their attention to the international arena. Examining foreign countries tax reforms and their effects may provide valuable policy implications for the US capital market. La Porta et al. (2000) provide a comprehensive international comparison of agency problems and dividend policies, and compare the two agency models of dividends. Using British data, PS find clear evidence that taxes affect the equilibrium relationship between dividend yields and market returns. They suggest that taxes affect the security market equilibrium and deepen the puzzle of why firms pay dividends. Amoako-Adu shows that the Canadian Tax Reform and its amendments affected highyield and low-yield stocks differently. Harris, Hubbard and Kemsley state that Australia provides a generous tax imputation credit system that can lead to a negative tax rate on dividends for the marginal investor. They test the tax capitalization hypothesis using US data and international data, including data from Australia. Benge (1997) presents a formal model of an optimizing firm to examine the effects of Australia s full imputation system. He derives an expression for the cost of capital for corporate investment. Westfarmers Limited of Australia provides a nice corporate example of the change in the dividend policy due to the imputation scheme (Moreton, 1993). Australia s imputation credits caused the cost of using external equity to be lower than using internal equity (retained earnings) for this firm. Consequently, the paid-in-capital portion of 5

7 shareholders equity continually increased while the retained earnings portion continually declined. This paper examines Australia s full imputation scheme. This study provides additional empirical evidence on the effect of the differential taxation of dividends and capital gains on stock prices. However, this paper differs from previous literature in that it shows a different reaction of the stock market with relation to a tax change announcement. Depending on how much information is disseminated into the public before the announcement is actually made, firms anticipation of the tax law change differs. In relation to the amount of information available, the stock price is effected by such information. This effect may be visible while examining the trend in abnormal returns pre- and post- announcement. Some government policy implications may be provided from the analysis. 3. Australia s Imputation Scheme and Its Anticipated Effect on Stock Prices On December 19, 1986, the Hon. P.J. Keating, M.P. and Treasurer, announced an end to the policy of double taxation of dividends paid to Australian shareholders of domestic companies (Moreton, 1993). For Australian shareholders, this tax reform eliminated the corporate tax on net income distributed as dividends. This implies that the new tax system would treat the tax paid by companies as personal income tax withholding by corporations on behalf of their shareholders. Before the imputation system, dividends were effectively taxed twice (which is the current US system) while net income transferred to retained earnings on the balance sheet was taxed only once. However, under the new imputation scheme, net income 6

8 distributed as dividends would carry a tax credit equal to the amount of tax paid by the company on that income. Moreton (1993) documents that, under the old system, the Australian Treasury taxed a company s net income at a rate of 46% while shareholders receiving dividends were subsequently taxed on their dividend income at personal income tax rates up to 60%. Under the new system, the tax rate was raised to 49% of net income while the maximum personal tax rate was lowered to 49%. As a result of these tax rate changes, it can be expected that changes will occur in the demand for various stocks depending on their level of dividends. Investors with lower marginal tax rates may be expected to increase their demand for high-dividend stocks because of the lower effective tax on dividends, while for those with higher marginal tax rates, the effect of the capital gains tax must be weighed against the increased dividend tax (Amoako-Adu, 1983). This new company tax system, known as imputation, became the law on July 1, However, the capital markets do react to the contents of the anticipated legal changes (Amoako-Adu, 1983). In this setting, the proposed tax changes were first announced on December 19, Thus, this paper considers December 1986 as the announcement date, which will serve as the event month that separates the pre- and postimputation eras. Amoako-Adu states that as relative prices adjusted to reflect the net effect of the tax changes around the time of the announcement, prices of high-dividend stocks should have increased as a result of the reduction of the effective tax on dividend income, while 7

9 the net effect on the low-yield stocks might have been trivial because the new capital gains tax might have offset the increased tax on dividend income. 4. Data and Methodology The original Australian sample for the empirical test consists of Australian listed firms reported in the 2003 Thompson Financial s Datastream file, which this study covers for the time period. This study defines the pre-imputation era to be in the years 1981 through Since the imputation scheme was announced in December of 1986, and since Australian firms annual fiscal year end is June 30, this study defines the post-imputation era to be after 1986, namely from 1987 to To examine stock price reactions, 87 Australian Stock Exchange (ASX) stocks which had continuous monthly data from June 1985 to June 1988 on the Datastream file were used. Additional data on ASX stock indexes and their weights were also taken from Datastream. The total sample consists of 1,697 observations. Table 1 presents the descriptive statistics of the sample in terms of dividend per share (DPS), dividend yield, the dividend payout ratio, the stock price, and the debt to total assets ratio. A review of Table 1 shows that DPS gradually increases after 1986 showing a clear distinction between the DPS amount pre- and post- imputation regimes. The mean of the DPS variable increases from a A$0.08 to A$0.10 range to a A$0.13 to A$0.19 range, pre- and post-imputation, respectively. In Figure 1, it can be easily seen that the tax rate change affected the DPS figure. There is a clear difference between the pre- and post- imputation era in terms of the level of the DPS. 8

10 The payout ratio, defined as the dividend over earnings, also increased similar to DPS implying that the increase in DPS is not due merely to an increase in earnings, but rather due to an increase in the actual amount of dividends paid after The dividend yield and the debt to assets ratio do not show a clear trend. However, it can be inferred that firms gradually increased their dividends after 1986 while their stock prices fluctuated. The stable trend in the debt to assets ratio suggests that changes in taxes do not affect firms capital structure. Unfortunately, the retained earnings and the total stockholder s equity variables could not be verified in the data files. Therefore, further analysis of the changes in corporate capital structure due to tax law modifications could not be carried out. The dividend yield expresses the dividend per share as a percentage of the share price in the fiscal year ending on June 30, The dividend yield in June 1986 is calculated for all the stocks in the sample. Similar to Amoako-Adu s study, the top 25 high-yield stocks were used to form the high-yield portfolio while the bottom 25 lowyield stocks were used to form the low-yield portfolio. A control portfolio is formed for a robustness check to control for different factors such as the clientele effect. The control portfolio consists of 25 firms in the middle range in terms of dividend yield rankings based on June of Nonetheless, this approach to measuring abnormal returns makes no attempt to control for systematic risk. This introduces a bias if systematic risk differs between the high- and low- yield portfolios, since the conclusions are based on comparisons of abnormal returns of the two portfolios (Bernard and Thomas, 1989). This study tests for such possibility as a sensitivity analysis. As a summary measure, the three-year event 9

11 window average beta for the high-yield portfolio is much lower than the low-yield portfolio s beta of To evaluate the long-run performance of two portfolios, two measures are used following the methodology implemented in Ritter (1991): (1) cumulative abnormal returns (CAR) calculated monthly, and (2) 3-year buy and hold returns (BHR) for both high and low portfolios and the control portfolio. In a three year window from June 1985 to June 1988, the initial return period is defined to be month 17 (July 1985). Returns for 35 months are calculated for both CAR and BHR ending on month +18 (June 1988). Again, December 1986 is month 0 representing the announcement event. Monthly benchmark-adjusted returns are calculated as the raw return (rit) on a stock minus the monthly benchmark return (rmt) for the corresponding month. The benchmarks used are (1) the ASX All Industrials Price Index, (2) the ASX Share Price Index, and (3) the ASX All Ordinaries Price Index, which is Australia s premier market indicator. The results for monthly benchmark-adjusted returns only show a little difference in the trends. Thus, the result for only the ASX All Ordinaries Price Index adjusted return is presented in this study. Following Ritter (1991), the abnormal return is calculated as the benchmarkadjusted return for stock i in event month t is defined as ar it = r it r mt. The average benchmark-adjusted return on a portfolio of n stocks for event month t is the weighted average of the benchmark-adjusted returns: AR t wi = TW ar it. 10

12 where wi representing the individual firm weight in the portfolio and TW representing the total weight of the portfolio. The data on individual firm weights is extracted from Datastream using the market capitalization variable. Market capitalization is defined as the share price multiplied by the number of ordinary shares in issue. The cumulative abnormal returns from month 17 to month +18 is the summation of the average benchmark-adjusted returns: CAR , + 18 = AR t. 17 Since the three portfolios have continuous and non-missing data for all 35 monthly returns, monthly rebalancing is not required. Therefore, as in the CAR equation above, summing abnormal returns over time does not lead to a bias in the returns cumulated over long periods as indicated in Blume and Stambaugh (1983). They report that biases may arise in computed returns in portfolios requiring rebalancing since new firms can lead to an upward bias in the returns cumulated over long periods. Similarly, as an alternative to the use of CAR, this study computes the three-year buy and hold returns (BHR), defined as R i = + 18 ( 1+ rit 17 ). 5. Results Abnormal returns for the three year window consist of 35 months, since the first observation (June 1985) is lost in calculating the second month return. Table 2 reports the benchmark-adjusted returns (ARt), the cumulative abnormal returns (CAR-18,t), and the three-year buy and hold returns (BHRt) for the 35 month time frame. Results are reported 11

13 for both high- and low- yield portfolios. Again, December 1986 serves as the event month 0 (announcement month) while June 1985 (June 1988) is month 18 (+18). In Figure 2, the plots of the CAR show that the tax changes affected high- and low- yield portfolios differently. In the pre-imputation regime (month 18 to month 0), the high-yield portfolio shows a declining trend in CAR while the low-yield portfolio shows an increasing trend. The high-dividend stocks decrease to 10% in CAR while the low-dividend stocks rise to 22% in CAR. The portfolio of high-dividend stocks started reacting to the tax change around the announcement month (December 1986), which implies that the announcement was a surprise to the stock market. As shown in Figure 2, the CAR for the high-yield portfolio rose from 10% in month 0 to 11% in month 18 increasing 21% in CAR. The portfolio of low-dividend stocks shows somewhat of an opposite reaction. After climbing to a CAR of 22% pre-imputation, the CAR increases to 31% in the peak of month 4. Then, the CAR starts declining, ending up at 10% in month 18 a decline of 21% in CAR. With perfect hindsight, going long in low-dividend stocks pre-imputation and going long in high-dividend stocks (while going short in the low-dividend stocks) post-imputation would have yielded a CAR of 43%. Examining the control portfolio, as expected, there seems to be no particular pattern of abnormal returns pre- and post- imputation. The trend seems random, and the fluctuations are clustered above and below zero CAR. This implies that the CAR results for the high- and low- yield portfolios are mainly due to taxes and not due to other factors such as the clientele effect (Elton and Gruber, 1970; Litzenberger and Ramaswamy, 1982). 12

14 Some government policy implications can be derived from this study. Examining Figure 2, the reaction from the market seems to be triggered around the event month (December 1986). Although the law was in effect beginning in month 7 (July 1987), the stock market reacted when the announcement was made in month 0 (December 1986). However, it seems that there was relatively low leakage of information regarding the tax change to the stock market before the official announcement was actually made. Comparing the results of this study to Amoako-Adu s, there is a clear distinction between the two event studies. Amoako-Adu covers the Canadian tax reform and its effect on stock prices. In Amoako-Adu s article (1983), the capital market shows a clear trend in anticipation of both the announcement and the effective date of the reform when the tax change became law. The high-yield firms in Amoako-Adu s study, show the same ascending pattern in CAR before and after the announcement date. In this study, however, the trend reverses at the announcement month implying that the announcement was a surprise. As for Canada, the information regarding the potential tax reform must have been leaked or debated far before the actual announcement or the effective date. Therefore, some misleading information regarding changes in tax laws may cause a disruption in the capital market, even if it does not become law. In addition, leakage of information regarding a tax law change may cause a frenzy even before it actually becomes law or before it is announced. 6. Regression Analysis and Some Robustness Checks Fama and French (1993) identify common risk factors in the returns on stocks and bonds. They outline three stock-market factors: an overall market factor, factors related to 13

15 firm size and book-to-market equity. However, in this study, the book-to-market variable could not be verified in Datastream files. Fortunately, market values and betas were documented. This section attempts to explain the variability of adjusted returns and the dividend yield together with some common risk factors that may explain stock returns. Therefore the following regression is designed. The market value (MV) variable which represents firm size will be implemented in the equation together with the dividend yield and firm beta to explain benchmark adjusted returns in the regression below: ar it = α + β DY + β MV + β Beta + ε it where the benchmark-adjusted return for stock i in event month t is defined as ar it = r r. DY represents the dividend yield which is the dividend per share as a it mt percentage of the share price. MV represents the market value derived by the share price multiplied by the number of ordinary shares in issue. Beta is individual firm beta given in Datastream files. Table 4 illustrates above regression results for both pre- and postimputation periods. The most noticeable result in Table 4 is that the coefficients of the dividend yield variable rise in all six models after the tax law change, implying that dividend yields gain incremental power in explaining the variability of benchmark-adjusted returns after dividend tax imputations are applied. Surprisingly, there are no clear pattern for the other two variables. Next, following Bernard and Thomas (1989), this study investigates whether shifts in betas may explain announcement drift. Ball, Kothari and Watts (1993) suggest that betas increase for firms with high unexpected earnings and decrease for firms with 14

16 low unexpected earnings. Table 5 outlines the shifts in betas over the three year event time frame. To examine the shift in betas over time, six half year windows are allocated to each portfolio in estimating betas. The first window starts from June 1985 (month 18) and ends in December 1985 (month 13). To calculate betas, the following regression based on the Sharpe-Lintner-Mossin CAPM is implemented: r r = a + b( r r ) + e it ft mt ft it where rit is the raw return on a stock while rmt is the monthly benchmark return for the corresponding month. The benchmark used is the ASX All Ordinaries Price Index, which is Australia s premier market indicator. rft represents the risk free return which is Australia s benchmark 10 year bond rate. A review of Table 5 reveals that betas are quite volatile in the Australian stock market. Many of the figures are below 1, implying that the market index is also volatile. Surprisingly, the betas shift in the opposite of the expected direction further deepening the dividend puzzle. For high-yield firms, the betas actually decline after the tax law change even with the high-yield portfolio s increase in abnormal returns which is counter-intuitive if risk was to explain the rise in stock returns. On the contrary, the betas for low-yield firms increase post-imputation. However, due to the volatility of these measures, interpreting betas and returns with these results should be cautioned. Nevertheless, the betas in this analysis fails to explain the magnitude of the rise and decline of CAR pre- and postimputation for both low- and high- yield portfolios. 7. Conclusions and Further Research 15

17 Australia s imputation scheme may have created an environment where dividends are tax-preferred to net income transferred to retained earnings on the balance sheet. Dividend per share has increased on average after the imputation credits. In addition, the CAR plots show that the tax changes affected high- and low- yield portfolios differently. In the pre-imputation regime (month 18 to month 0), the high-yield portfolio shows a declining trend in CAR while the low-yield portfolio shows an increasing trend. This trend reverses around month 0 and continues in the post-imputation regime (month 1 to month +18). This is somewhat surprising since the Australian stock market did not show much lag or an anticipated reaction to the tax reform announcement. Comparison of the other balance sheet items such as retained earnings and paidin-capital would provide an important evidence of the tax effects of the imputation scheme. Data availability plays a crucial role here since data on ASX listed firms prior to 1986 is not easily verifiable. Although many of the international articles reviewed in this paper unearthed numerous facts and implications regarding dividends, generalizing it to other countries should be cautioned due to the institutional differences among various countries. Rajan and Zingales (1995) state that it is necessary that we gain a deeper understanding of the effects of institutional differences in the international environment. Thus, it will be noteworthy to devise an empirical study with a normative approach to dividend policies in an international setting. In a different turn, it would be meaningful in answering other crucial research questions regarding international capital structure: (1) how do firms manage their capital structure in an international setting? and (2) what can be deduced from international data 16

18 regarding capital structure? It will be worthwhile to examine the benefits of reinvesting, internal versus external cost of equities and dividend policies. 17

19 References: Amoako-Adu, B. (1983), The Canadian Tax Reform and Its Effect on Stock Prices: A Note. The Journal of Finance, 38, 5, Ball, R., Kothari SP, and Watts RL (1993), Economic-Determinants of the Relation between Earnings Changes and Stock Returns, Accounting Review, 68 (3): Benge, M. (1997), Taxes, Corporate Financial Policy and Investment Decisions in Australia. The Economic Record, 73 (220), Bernard, L., and Thomas, J. (1989), Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium?. Journal of Accounting Research, 27, Blouin, J. (2003), Shareholder Taxes and Stock Prices. Working Paper, University of North Carolina Chapel Hill. Blume, M., and R. Stambaugh. (1983), Biases in Computed Returns: An Application to the Size Effect. Journal of Financial Economics (June 1983): Brennan, M. (1970), Taxes, Market Valuation and Corporate Financial Policy. National Tax Journal, Elton, E. and Gruber, M. (1970), Marginal Stockholder Tax Rates and the Clientele Effect, Review of Economics and Statistics, February, Fama, E. and French, K. (1993), Common Risk Factors in the Returns on Stocks and Bonds, Journal of Financial Economics, 33, Harris, T., Hubbard, G., and Kemsley, D. (2001), The Share Price Effects of Dividend Taxes and Tax Imputation Credits. Journal of Public Ecocomics, 79, King, M. A. (1977), Public Policy and the Corporation. Chapman and Hall, London. La Porta, R., Lopez-De-Silanes, F., Shleifer, A., and Vishny, R. (2000), Agebcy Problems and Dividend Policies around the World, The Journal of Finance, Vol LV, 1, Litzenberger, R. and Ramaswamy, K. (1982), The Effect of Dividends on Common Stock Prices Tax Effects or Information Effects?, The Journal of Finance, 37, 2, Miller MH and Scholes MS. (1978), Dividends and Taxes, Journal of Financial Economics, 6 (4):

20 Modigliani, F. and M.H. Miller (1961), Dividend Policy, Growth and the Valuation of Sales, Journal of Business, 34: Moreton, P. (1993), Westfarmers Limited: The Dividend Decision. Case No Harvard Business School, Boston, October 15. Poterba, J. and Summers, L. (1984), New Evidence That Taxes Affect the Valuation of Dividends, The Journal of Finance, 39, 5, Rajan, R.G. and Zingales, L. (1995), What Do We Know About Capital Structure? Some Evidence From International Data, Journal of Finance, 50, Ritter, J. (1991), The Long-Run Performance of Initial Public Offerings, The Journal of Finance, VOL. XLVI, 1,

21 Note on Source: Data for the tables and figures that follow are all taken from the 2003 Thompson Financial s Datastream file, which this study covers, for the time period. Table 1 Descriptive Statistics for the ASX Listed Firms The dividend per share (DPS) is the current annualised dividend rate. The DPS reflects the 'indicated annual dividend'. The dividend yield expresses the dividend per share as a percentage of the share price. The dividend payout ratio is the amount of dividend paid over earnings. The currency for DPS and the average stock price is the Australian dollar (A$). Year N Dividend Per Share (A$) Dividend Yield (%) Dividend Payout Ratio Average Stock Price (A$) Debt to Total Assets % % % % % % % % % % % Total 1697 Figure 1 Average Dividend Per Share for ASX Listed Firms from DPS

22 Table 2 Abnormal Returns for ASX Firms from June 1985 to June 1988 Average adjusted returns (AR), cumulative average returns (CAR), and 3-year buy and hold returns (BHR) in percent for 36 month event period. December 1986 serves as the event month 0 while June 1985 (June 1988) is month 18 (+18). Month High-Yield Portfolio Low-Yield Portfolio t ARt (%) CAR-18,t (%) 3-Year BHR (%) ARt (%) CAR-18,t (%) 3-Year BHR (%)

23 Figure 2 Cumulative Abnormal Returns from June 1985 to June 1988 Two CAR series are plotted for the 36 month event period. December 1986 serves as the event month 0 while June 1985 (June 1988) is month 18 (+18). The solid line represents the high-yield portfolio while the dotted line depicts the low-yield portfolio. Cumulative Abnormal Returns (CAR) High Yield CAR Low Yield CAR Month

24 Table 3 Abnormal Returns for the Control Portfolio from June 1985 to June 1988 Average adjusted returns (AR), cumulative average returns (CAR), and 3-year buy and hold returns (BHR) in percent for 36 month event period. December 1986 serves as the event month 0 while June 1985 (June 1988) is month 18 (+18). The control portfolio was constructed with firms in the middle range in terms of the dividend yield rankings. Month ARt (%) CAR-18,t (%) 3-Year BHR (%)

25 Figure 3 Cumulative Abnormal Returns for the Control Portfolio from June 1985 to June 1988 The CAR series are plotted for the 36 month event period. December 1986 serves as the event month 0 while June 1985 (June 1988) is month 18 (+18). The control portfolio was constructed with firms in the middle range in terms of the dividend yield rankings. Control Portfolio CAR Month

26 Table 4 Pre-Imputation OLS Versus Post-Imputation OLS Results ar it = α + β DY + β MV + β Beta + ε it Adjusted return for stock i in event month t is defined as arit = rit rmt. DY represents the dividend yield which is the dividend per share as a percentage of the share price. MV represents the market value derived by the share price multiplied by the number of ordinary shares in issue. Please refer to the appendix for the Beta calculation method used by Datastream. t-statistics are in parentheses. Pre-Imputation ( ) Post-Imputation ( ) Model α β 1 β β 2 3 α β 1 β β (1.952) (4.244) E E-05 (1.613) (0.721) (3.511) (0.922) E E (-1.528) (-1.134) (5.474) (-0.798) (-1.632) (8.807) (5.057) (-2.783) (10.242) (-4.092) E E-05 (4.958) (-2.765) (-0.396) (10.291) (-3.904) (-1.363) E E (1.543) (-2.185) (-1.074) (2.481) (4.632) (-3.401) (-1.464) (0.744) 25

27 Table 5 Beta Estimates by Dividend Yield Category r r = a + b( r r ) + e it ft mt ft it Regression is based on the Sharpe-Lintner-Mossin CAPM following Bernard and Thomas (1989). See Ball, Kothari, and Watts (1993) for details. rit is the raw return on a stock while rmt is the monthly benchmark return for the corresponding month. The benchmarks used is the ASX All Ordinaries Price Index which is the Australia s premier market indicator. rft represents the risk free return which is Australia s benchmark 10 year bond rate. The shift in beta can be easily seen for three portfolios. Low (High) represents the lowdividend (high-dividend) stock firms. Results for the control portfolio is also given. Standard errors (SD) are in parentheses. Pre-Imputation (Month 18,0) Post-Imputation (Month +1,+18) Portfolio (-18,-13) (-12,-7) (-6,0) (+1,+6) (+7,+12) (+13,+18) High (0.124) (0.120) (0.114) (0.294) (0.040) (0.147) Low (0.398) (0.229) (0.381) (0.664) (0.069) (0.244) Control (0.150) (0.154) (0.162) (0.356) (0.053) (0.162) 26

28 Appendix Variable Definitions from Datastream Dividend Per Share (DPS): This is the current annualised dividend rate. It is intended to represent the anticipated payment over the following 12 months and for that reason may be calculated on a rolling 12-month basis, or as the "indicated" annual amount. Special or once-off dividends are generally excluded. Dividends per share are displayed gross, inclusive of local tax credits where applicable where dividends per share are displayed net. The dividend per share reflects the 'indicated annual dividend', which is derived by multiplying the latest payment by the frequency of payment. A latest quarterly payment, for example, is multiplied by 4 Earnings Per Share (EPS): This is the latest annualised rate that may reflect the last financial year or be derived from an aggregation of interim period earnings. Dividend Yield (DY): The dividend yield expresses the dividend per share as a percentage of the share price. The underlying dividend is calculated according to the same principles as data type DPS (Dividend per share, current rate) in that it is based on an anticipated annual dividend and excludes special or once-off dividends. The dividend yield is based on gross dividends (including tax credits) where available, for which dividends per share are displayed on a net basis. The dividend yield excludes the tax credit applicable to domestic investors only. Market Capitalization (MV): Market value on Datastream is the share price multiplied by the number of ordinary shares in issue. The amount in issue is updated whenever new tranches of stock are issued or after a capital change. For companies with more than one class of equity capital, the market value is expressed according to the individual issue. Market value is displayed in millions of units of local currency. Price (P): The current price on Datastream s equity programs is the latest price available from the appropriate market in primary units of currency. It is the previous day s closing price from the default exchange except where more recent or real-time prices are available, as listed in the Data sources & updating procedures section of this help system. The current prices taken at the close of market are stored each day. These stored prices are adjusted for subsequent capital actions, and this adjusted figure then becomes the default price offered on all Research programs. Prices are generally based on last trade or an official price fixing. For stocks which are listed on more than one exchange within a country, default prices are taken from the primary exchange of that country (note that this is not necessarily the home exchange of the stock). Australian Stock Exchange (ASX) All Ordinaries index: The ASX All Ordinaries index is Australia's premier market indicator. BETA: The beta factor of a stock relates movements in its price to movements in the market as a whole. Over a period it expresses the relative movement of the price against the market, showing the likely relative change for a given market movement and whether the stock is prone to under- or over-react. In order to display beta calculations, at least 2½ years of data are required. There are many ways of calculating beta factors. The method adopted by Datastream is described here. The beta factor is derived by performing a least squares regression between adjusted prices of the stock and the corresponding Datastream market index. The historic beta so derived is then adjusted using Bayesian techniques to predict the probable behavior of the stock price on the basis that any extreme behavior in the past is likely to average out in the future. This adjusted value, or "forecast" beta, is represented by the BETA datatype. The Datastream beta factor is calculated using stock prices and market indices as the only variables. 27