Taxes and Stock Returns

Transcription

1 Taxes and Stock Returns Rakesh Bali and Armen Hovakimian Extensive research exists in financial economics relating taxes and stock returns. Personal taxation of dividends at a rate higher than capital gains and corporate taxation at a lower effective rate has spawned an ocean of work. We present a synthesis of various theoretical, empirical, and regulatory issues and guide the reader through a maze of conflicting results. Students are introduced to arguments beyond what they generally see in corporate finance texts. INTRODUCTION Standard asset pricing models (e.g. CAPM and APT) introduced in the finance curricula are derived in a perfect markets framework. The relation between expected return and risk posited by these models says nothing about how expected returns should vary with taxes for a given level of risk. Through most of the twentieth century the tax rate on dividends for the highest marginal tax bracket individuals has been substantially higher than the capital gains tax rate. This fact has generated enormous interest among researchers wishing to understand the theoretical and empirical relation between pre- and after-tax returns, risk, and the differential taxation between dividends vis-à-vis capital gains. 1 The purpose of this paper is to present a synthesis of the various theoretical, empirical, and regulatory issues that have been analyzed in the literature. The after-tax version of CAPM generally predicts a linear and increasing relation between pre-tax excess returns and a stock s dividend yield and its beta. Tests of these models seek to understand the long-horizon relation between dividend yields and stock returns. The central question is whether stocks with high anticipated dividend yields earn higher pre-tax returns per unit of risk than their low yield counterparts. Another approach has been to analyze the shorthorizon relation such as the price adjustment on Baruch College, CUNY, NY We are thankful to workshop participants at Baruch College for comments and the Zicklin School of Business for summer research support. dividend days for high and low yield stocks. A plethora of research exists in this area attempting to decipher how the ex day price drop is influenced by taxes. In addition, numerous studies exist estimating the impact of changes in regulatory and tax policies on ex day price behavior. The paper is organized as follows. Section 2 discusses the theoretical models, section 3 the empirical results and section 4 concludes. THEORY It is well known that in frictionless markets, a firm s market value is independent of its dividend policy [Miller and Modigliani, 1961]. Under these conditions, the price of a stock given by the dividend discount model of stock valuation is: E D P t 0 = ( ) t t= 1 ( 1 + rt ) (1) where P 0 is the current price, r t is the expected return for t periods, and E(D t ) is the expected dividend per share at the end of period t. The basic dividend policy issue is the decision to pay dividends at one point in time versus another while keeping the financing and investment policy fixed. Since the real world is replete with frictions like taxes, transaction costs, and asymmetric information among others, the effect they may have on the value of a firm with a chosen dividend policy remains an open issue. Below we 14 Journal of Financial Education

2 summarize the directions taken by researchers in attempting to incorporate these frictions in to pricing. THE AFTER-TAX CAPM The return from holding a stock can be decomposed into price appreciation and dividends. If the holding period exceeds the minimum period required for the appreciation in price to be classified as capital gains (currently this period is one year), then an investor in a marginal tax bracket higher than the capital gains tax rate faces higher taxes on $1 received as dividends than on $1 earned as capital gains. Hence his after-tax valuation of dividends is lower than that of capital gains. Based on this intuition many versions of the after-tax CAPM have been proposed [e.g. Brennan, 1970; Long, 1977; Litzenberger and Ramaswamy, 1982] arguing that a stock s pre-tax excess return should be linearly and positively related to its dividend yield and systematic risk (beta). This long-horizon relation between return, risk, and taxes is estimated using the following tax-clientele equation: E( R ~ r ) = a + b β + c ( d r ). i f 1 1 i 1 i f (2) ~ where R i is the return on asset i, $ i is its systematic risk, d i is its dividend yield, r f is the risk-free rate and E(.) denotes expected value. In the next section we discuss the estimates of tests of equation (2). Factors which may affect an investor s preference for dividends versus capital gains include the proportion of realized capital gains exclusion, capital gains tax rate, minimum holding period required for capital gains classification, maximum deductible capital loss, his investment horizon, allowed dividend exclusion, his marginal income tax rate, restrictions on investment interest deduction, uncertainty about future changes in tax laws and the firm s dividend policy, among a host of others. Although each investor cares about maximizing his after-tax return per unit of risk, the deeper issue is if the marginal investor who determines the stock price reflects dividend aversion and if so, how to test for the existence of tax-premia in stock prices? At this point a little digression is in order. Although personal taxation may result in investor dislike for dividends, the question can be raised as to why the firms pay dividends at all. Bhattacharya (1979) argues that in the presence of asymmetric information between the managers and the investors, the management of a good firm needs to send a credible signal of its profitability to the marketplace. Since the bad firm s management cannot costlessly mimic this, such signaling would be credible. Hence there may be dividend payments in equilibrium in spite of their taxinduced disadvantage. Since our main focus is the relation between taxes and returns, we do not pursue this line of work. THE EX-DIVIDEND DAY EXPERIMENT The ex day is special in the sense that any stock purchase for a long-term buyer before this day results in the inclusion of the dividend in current income while a buyer on or after the ex day has no such obligation. Similarly a (long-term) seller after the ex day recognizes the dividend as current income, while if he sells before he excludes the dividend from current income. Elton and Gruber [1970] derive the after-tax cash flows to a seller on the cum and ex days as follows. Say, P p is his purchase price, P c is the price at which he can sell on the cum day and J c is the capital gains tax rate. If P c > P p, and the seller has held the stock long enough for the sale to be classified as capital gain, then his after-tax cash flow from selling cum equals P p + (P c - P p (1 - J c ). On the other hand, if the seller waits till the ex day to sell the price he receives is P x. Ceteris paribus, his after-tax valuation of the cash flows now equals P p + (P x - P p (1 - J c ) + (1 - J d ), where J d is his marginal tax rate on current income. Elton and Gruber claim that the equilibrium around the ex day is such that the marginal stockholders are indifferent between selling cum or ex. Equating the two after-tax cash flows they obtain 1 τ d Pc Px = P = D. 1 τ c (3) Using a sample of NYSE stocks with ex days from April 1, 1966 to March 31, 1967 they document that Fall

3 the average )P/D ratio is Using the relation J c = min (J d /2, 25%), they impute that J d = 36%. This estimate is intuitively plausible and they view this as evidence that long-term investors? dividend aversion due to taxes results in $1 in dividends being valued at 78. They further show that this ratio rises with the dividend yield (D/P c ) of the stock. Elton and Gruber interpret these findings as evidence of the existence of tax-induced dividend clienteles where high tax bracket investors hold low dividend yield stocks and hence the price drop is a smaller fraction of the dividend for low yield stocks (implying a higher tax-premium). This premium, according to the tax-clientele hypothesis, declines for high dividend yield stocks since they are held by low marginal tax bracket investors. The knowledge about the tax rates of the marginal investors in a firm s stock is important for determining the optimal dividend policy, cost of capital calculations, debt policy, and the pricing of taxable versus nontaxable obligations of the firm. Since 1970, Elton and Gruber s study has been the only study depicting a declining relation between the marginal tax rates and (short-run) dividend yields. From the market efficiency perspective, any systematic deviation in returns from zero needs to be explained. If persistent positive returns exist around the ex day due to prices reflecting tax-premia, then are these returns arbitragable by tax-neutral market participants or within trading costs? DIVIDENDS AND TRANSACTION COSTS The existence of a tax-premium in dividend pricing begs the question; Can tax-neutral traders arbitrage this premium? Short-term traders who serve as the intermediaries facilitating the market making process recognize all gains (losses) as ordinary income (loss) since they are incurred in the course of regular business activity. Kalay [1982] argues that these short-term traders or corporations, which may prefer dividends due to the dividends received deduction, have incentives to arbitrage any implicit tax-premium in security pricing beyond the transaction costs. He derives the following no-arbitrage condition: D P < α P (4) where " is the proportional transaction cost, P is the mean of the cum and ex prices, and D and )P are as defined earlier. Intuitively, the equation implies that short-term traders will buy cum and sell ex, if the difference between the dividend and the expected price decline exceeds the round-trip transaction cost. Also, they will short the stock on the cum day and cover on the ex day if the reverse is true. Kalay agrees that the ex day price drop may be less than the dividend due to taxes but is unable to explain why the imputed tax rates vary with the dividend yield. Miller and Scholes [1982] extend Kalay s argument and caution the researchers against interpreting any estimated relation between short-run dividend yields and returns as evidence for tax-clienteles effects. Scholes and Wolfson [1992] discuss the issue further and conclude, The effect on stock prices of differences in taxation of dividends and capital gains remains an open question (p.368). More recently Boyd and Jagannathan [1994] claim that short-term traders are more likely to be at the margin for high yield stocks and hence the price drop will equal the dividend amount for such stocks. For low yield stocks they posit that long-term investors are at the margin. Such arguments however cannot explain the observation in Eades et al. [1984] that ex day returns for stock dividends and splits are positive as well which could not be due to taxes. Discrete Prices and Dividends Prices on organized exchanges such as the New York Stock Exchange are limited to discrete tick multiples and the tick size has for the most part been $c or 12½. Such arrangements are necessary to provide the market maker with a minimum profit per trade for a round-trip transaction. These profits exist in equilibrium because even if the market maker just breaks even on average, the business is costly. Some measures of such costs are the unobservable costs of adverse selection (positive probability of trading against an insider) and inventory control (suboptimal portfolio position due to market making obligations) as 16 Journal of Financial Education

4 well as the observable price of a seat on the exchange 2 and order processing. The dividend amount per share D is a choice variable for the management of the firm and it need not equal an exact multiple of the tick size. Faced with sellers wanting to sell on cum day, while buyers wait till the ex day, the specialist provides liquidity by inventorying the dividends overnight. Bali and Hite [1998] argue that in such a setting the dividend is priced at D, the tick multiple just below D. For example, if a dividend of 20 is priced at 12.5, then the )P/D ratio equals Ex day abnormal returns result because the buyers and short-term traders due to taxes and transaction costs respectively do not want to pay the full price D for the dividend. Sellers, who face taxes no smaller than ( D D), willingly accept and equilibrium results. In their sample, the mean quarterly dividend is about 25 and the median is 20. So the tick size is a significant fraction of the average dividend. The pricing of dividend at D results in P D D = D (5) As D increases, the price drop is proportionally closer to the full amount. The positive correlation between dividend and dividend yield may appear like tax-clienteles. Bali and Hite also show that )P/D ratio declines between tick multiples tracing a sawtooth pattern in the data. The realized return of.18% over their sample period is even smaller than.40% that would occur if )P equalled. D The above insight is important because the implication is that both taxes and transaction costs play important roles in determining ex day dividend pricing. The difference between dividend and price drop of 6.79 is less than one tick and hence unarbitragable. The evidence does not support the tax-clientele hypothesis but is consistent with the rents that liquidity providers earn from investors with dividend aversion. Hence the claim that investors sort themselves into clienteles seems too strong. A framework with shortterm traders at the margin earning compensation for handling order flow is consistent with positive ex day returns documented for stock dividends and splits as well. Numerous researchers have examined the ex days of a wide cross-section of securities and intertemporally across different tax regimes given the importance of tax policy. Since most studies have not treated the discreteness effects explicitly their results need to be interpreted with caution. EMPIRICAL EVIDENCE Equilibrium Models Black and Scholes [1974] test for the tax-clientele hypothesis by constructing portfolios which are ex-ante mean-variance efficient but with varying dividend yields. They do not find any evidence in support of dividend clienteles and conclude that low (high) tax bracket investors do not gain anything by holding high (low) dividend yield stocks. Litzenberger and Ramaswamy [1982] claim that investors sort themselves across the dividend yield spectrum with high (low) tax bracket investors holding low (high) yield stocks. They find that the pre-tax excess returns for low dividend yield stocks have a higher slope coefficient on dividend yields than for high yield stocks. Miller and Scholes [1982] argue that these results may be related to the dividend announcement effects that convey information about the firm s cash flows and not necessarily evidence of dividend clienteles. Keim [1985] documents that generally low (high) market capitalization stocks are in the high (low) dividend yield groups. He also shows that the returns for high dividend yield stocks are substantially higher in January. For the remaining months of the year there is no significant difference in returns for the low or high yield stocks. This evidence is anomalous because the after-tax versions of CAPM do not posit any seasonality in tax-premia. Bhardwaj and Brooks [1992] claim that Keim s results may be explained by a low-price effect. If there is a selling pressure on these stocks towards the end of the year then there is a higher probability of observing a bid price in December and an ask price in early January. The observed return on these stocks will be higher than high-priced stocks because the percentage spreads and bid-ask bias are larger for these stocks. They also Fall

5 show that over the period the after transactions cost returns are lower for low-priced stocks than for high-priced stocks. Naranjo et al. [1998] document a positive relation between returns and dividend yields that is mostly driven by smaller than average market capitalization stocks. This relation persists even after accounting for the various hypothesized factors like the market risk premium, book-to-market effect, and the size effect. They used the taxable treasury yield curve and the non-taxable municipal yield curve to estimate the tax rate of the marginal investor but find no relation between the dividend yield premia and the tax rate. Fama and French [1998] estimate the cross-sectional relation between firm value, dividend policy and taxes. They do not detect any tax effect and attribute the positive correlation between firm value and dividends to the informational content of dividends. Ex Day Dividend Pricing and Regulatory Changes If personal taxes affect ex day price adjustment then changes in taxes should lead to measurable impact on the ex day relative price drop. In a significant departure from earlier studies, Lakonishok and Vermaelen [1983] estimate the effect of the Canadian tax reform act (1971) on the )P/D ratio. They find no effect of the increase in the after-tax value of the dividend caused by the act and conclude that short-term traders must be influencing ex day prices. Subsequent research has devoted little effort to model how short-term traders price the dividend in the presence of an abundance of sellers (buyers) on the cum (ex) day, but numerous studies have analyzed the effects of the elimination of fixed commissions in May 1975 and the various tax reform acts on ex day returns. Although changes in regulatory and tax policies are politico-economic decisions, financial economists are interested in the effect, if any, they may have on asset prices. The framework generally assumes the presence of various types of traders (short-term traders, long-term buyers and sellers, and corporations) in the marketplace and analyzes how these changes affect their incentives to trade. For example Karpoff and Walkling [1988] take the tax-clienteles as given and test for the impact of the elimination of fixed commissions beginning May 1, 1975 on ex day returns. They claim that this act reduced the trading costs of short-term traders who reacted by trading more aggressively and reduced the ex day tax-premium for high-yield stocks. Similar long-horizon regressions are estimated in Boyd and Jagannathan [1994] and Eades et al. [1994] concluding that the tax-premium declines. One glaring shortcoming of these papers is that they omit the fact that the closing cum and ex day prices which they use are provided by the market makers. If there are systematic patterns in the probabilities of observing a bid or an ask price on cum and ex days respectively, then the estimates may reflect reduced spreads rather than a decline in ex day tax-premium. Conrad and Conroy [1994] use a changes in order flow based argument to explain the ex day effects for stock splits. Intercorporate dividends received deduction is designed to prevent double taxation of the dividends paid out by the firm whose shares are held by the second firm since the first firm pays dividends from its after-tax income. The minimum number of days the firm must hold the stock to be eligible for this deduction was raised from 15 days to 45 days by the 1984 tax act. Clearly corporations can not be the marginal investors in these stocks because if they buy on the cum day and sell on the ex day, they will lose this deduction and pay ordinary income taxes on such dividends. This has not prevented the researchers from hypothesizing that the ex day tax premium should decline for high yield stocks following the act because it reduced the corporations incentives to indulge in dividend capture. Grammatikos [1989] finds evidence consistent with this while Eades et al. [1994] claim the opposite. The tax reform act of 1986 perhaps introduced more changes than any other act since the introduction of the federal income tax in It reduced the intercorporate dividends received deduction from 85% to 80% in 1987 and to 70% in This act also reduced the maximum personal income tax from 50% to 38.5% in 1987 and to 33.5% in Though hypotheses can be constructed either way to test for the impact of the law, Michaely [1991] concludes (p.846) the ex-day price drop is influenced primarily by short-term traders and corporate traders 18 Journal of Financial Education

6 who favor dividend income over capital gain income. Using a different approach Koski [1996] uses the closing bid and ask prices on cum and ex days in 1983 and 1988 to test for the effects of the 1984 and 1986 tax acts on various potential marginal investors. She concludes that short-term traders cannot profit if they incur the spread as the cost. Taxes fascinate us but unfortunately little can be inferred about them from ex day price behavior. In another study Skinner [1993] analyzes ten personal tax law changes over and concludes (p. 61), The results are consistent with some unknown, non-tax-induced clientele effect(s). CONCLUSIONS The relation between taxes, a firm s dividend policy, and its value remains one of intense debate and research. Systematic effects documented around the ex day, dividend announcements, and the turn-of-theyear in January seem to be related to price discreteness and spread effects. Knowledge of these may help investors reduce their transaction costs and time their buy/sell decisions, however such trading does not influence pricing at the margin. The persistent effects may be the artifacts of the data but do not present arbitrage opportunities. Long-horizon positive correlation between firm value and dividend policy may be due to signaling, but why dividends are the optimal signaling mechanism has not been proven yet. ENDNOTES 1 The highest marginal tax rate for the individual investor is 39.6 percent while the rate for capital gains is 20 percent currently. The corporate tax rate is 36 percent. Space limitations force us to focus on stocks only in this paper. We explore the relation between taxes and bonds in [Bali and Hovakimian, 1999]. 2 In August 1999, a seat on the NYSE sold for $2.65 million, up from $2.5 million in July REFERENCES Bali, Rakesh and Armen Hovakimian. Taxes and Bond Returns, working paper (1999), Baruch College, CUNY. Bali, Rakesh and Gailen Hite. Ex Dividend Day Stock Price Behavior: Discreteness or Tax-induced Clienteles?, Journal of Financial Economics, 47 (February 1998), Bhardwaj, Ravinder and Leroy Brooks. The January Anomaly: Effects of Low Share Price, Transaction Costs and Bid-Ask Bias, Journal of Finance, 47 (June 1992), Bhattacharya, Sudipto. Imperfect Information, Dividend Policy, and the Bird in the Hand Fallacy, Bell Journal of Economics, 10 (1979), Black, Fischer and Myron Scholes. The Effect of Dividend Yield and Dividend Policy on Common Stock Prices and Returns, Journal of Financial Economics, 1 (May 1974), Boyd, John and Ravi Jagannathan. Ex-dividend Price Behavior of Common Stocks, Review of Financial Studies, 7(winter 1994), Brennan, Michael. Taxes, Market Valuation and Corporate Financial Policy, National Tax Journal, 23(1970), Conrad, Jennifer and Robert Conroy. Market Microstructure and the Ex-date Return, Journal of Finance, 49 (September 1994), Eades, Kenneth, Patrick Hess, and E. Han Kim. On Interpreting Security Returns During the Ex-dividend Period, Journal of Financial Economics, 13 (March 1984), Eades, Kenneth, Patrick Hess, and E. Han Kim. Time-series Variation in Dividend Pricing, Journal of Finance, 49 (December 1994), Elton, Edwin, and Martin Gruber. Marginal Stockholder Tax Rates and the Clientele Effect, Review of Economics and Statistics, 52 (February 1970), Fama, Eugene and Kenneth French. Taxes, Financing Decisions, and Firm Value, Journal of Finance, 53 (June 1998), Grammatikos, Theoharry. Dividend Stripping, Risk Exposure, and the Effect of the 1984 Tax Reform Act on the Ex-dividend Day Behavior, Journal of Business, 62 (1989), Kalay, Avner. The Ex-dividend Day Behavior of Stock Prices: A re-examination of the clientele Fall

7 effect, Journal of Finance, 37 (September 1982), Karpoff, Jonathan, and Ralph Walkling. Short-term Trading Around Ex-dividend Days: Additional Evidence, Journal of Financial Economics, 21 (September 1988), Keim, Donald. Dividend Yields and Stock Returns: Implications of Abnormal January Returns, Journal of Financial Economics, 14 (September 1985), Koski, Jennifer L. A Microstructure Analysis of Ex-dividend Stock Price Behavior Before and After the 1984 and 1986 Tax Reform Acts, Journal of Business, 69 (1996), Lakonishok, Josef, and Theo Vermaelen. Tax Reform and Ex-dividend Behavior, Journal of Finance, 38 (September 1983), Litzenberger, Robert and Krishna Ramaswamy. The Effects of Dividends on Common Stock Prices: Tax Effects or Information Effects?, Journal of Finance 37(2, 1982), Long, John B. Efficient Portfolio Choice with Differential Taxation of Dividends and Capital Gains, Journal of Financial Economics, 5 (August 1977), Michaely, Roni. Ex-dividend Day Stock Price Behavior: The Case of the 1986 Tax Reform Act, Journal of Finance, 46 (July 1991), Miller, Merton and Franco Modigliani. Dividend Miller, Merton, and Myron Scholes. Dividends and Taxes: Some Empirical Evidence, Journal of Political Economy, 90 (December 1982), Naranjo, Andy, M. Nimalendran, and Mike Ryngaert. Stock Returns, Dividend Yields, and Taxes, Journal of Finance, 53 (December 1998), Scholes, Myron and Mark A. Wolfson. Taxes and Business Strategy: A Planning Approach, Prentice Hall (1992), New Jersey. Skinner, David. Twenty-Five Years of Tax Law Changes and Investor Response, Journal of Financial Research, 16 (Spring, 1993), Policy, Growth and the Valuation of Shares, Journal of Business, 34 (1961), Journal of Financial Education