The Effects of Dividend Taxes on Equity Prices: A Re-examination of the 1997 U.K. Tax Reform
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
  • | 2 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

We re-examine the extent to which personal taxes on dividends are capitalized into the equity prices of domestic firms, using data from around the time of the 1997 U.K. dividend tax reform, which removed a significant tax credit for an important group of investors: U.K. pension funds. The tax-adjusted CAPM suggests that the impact should depend on an average of dividend tax rates across all investors, and that U.K. pension funds should reduce their holdings of the previously tax-favored asset: U.K. equities. Given that U.K. pension funds are small relative to the total size of the world capital market, a small open economy-type argument implies that the main effect of the reform would be to reduce U.K. pension funds' ownership of U.K. equities, with little impact on their price. We present evidence which is consistent with these hypotheses. We discuss why previous research (Bell and Jenkinson, 2002) reached a different conclusion.

Abstract

We re-examine the extent to which personal taxes on dividends are capitalized into the equity prices of domestic firms, using data from around the time of the 1997 U.K. dividend tax reform, which removed a significant tax credit for an important group of investors: U.K. pension funds. The tax-adjusted CAPM suggests that the impact should depend on an average of dividend tax rates across all investors, and that U.K. pension funds should reduce their holdings of the previously tax-favored asset: U.K. equities. Given that U.K. pension funds are small relative to the total size of the world capital market, a small open economy-type argument implies that the main effect of the reform would be to reduce U.K. pension funds' ownership of U.K. equities, with little impact on their price. We present evidence which is consistent with these hypotheses. We discuss why previous research (Bell and Jenkinson, 2002) reached a different conclusion.

I. Introduction

Recent years have seen important reforms of dividend taxation in several OECD countries. In 2003, the USA moved away from a pure classical system, introducing preferential personal tax rates for dividend income. In 1999, Ireland moved in the opposite direction, replacing a partial imputation system with a classical system, in which dividend income is taxed at the shareholder’s full marginal income tax rate. The operation of imputation systems in European Union countries has been held to be inconsistent with the EU Treaty in a series of rulings by the European Court of Justice, generally involving discrimination against either foreign shareholders or foreign corporations. Several EU countries have modified or abandoned their imputation systems as a result. Both Germany and the U.K. now tax dividends in a similar way to the current US system, with dividends subject to personal income tax, but not taxed at the shareholder’s full marginal income tax rate.2 In Ireland, Germany and the U.K., revenue raised from these dividend tax increases has been used to finance reductions in the corporate income tax rate.

For any assessment of these reforms, an important question is whether dividend taxes have any significant effect on corporate investment decisions. The public economics literature suggests two distinct reasons why dividend taxation may have little relevance for corporate investment. The “new view” or “tax capitalization” hypothesis accepts that dividend taxes will be capitalized into share prices, but emphasizes that taxes on dividends will still have no effect on the cost of capital or investment, if the marginal source of finance used by firms is retained earnings and the dividend tax rate is constant. The classical statements of this result are in King (1974) and Auerbach (1979), and the intuition is straightforward. Suppose that dividend income is taxed at rate m, so that the shareholder gives up net income of $(1 - m) to finance an investment of $1 from retained earnings. When the return from the investment—say $(1 + r)—is paid out as a dividend, the shareholder receives $(1 - m)(1 + r), so that the post-tax rate of return (r) is the same as the pre-tax rate of return. For investment financed from retained earnings, the dividend tax acts as a cash flow tax, and is neutral with respect to the investment decision. Of course, this argument does not hold if the marginal source of finance is new equity, and the return is paid out at least partly in the form of a dividend.3 It is possible therefore that dividend taxation acts as a disincentive to investment by immature firms, which are more reliant on new equity finance.4 But since most investment is undertaken by mature corporations and since most of their investment is financed from retained earnings, dividend taxation is expected to have only a limited effect on aggregate investment.

A different argument that leads to broadly the same conclusion is the “tax irrelevance” view, which suggests that dividend taxes have little or no effect on share prices. As first stated by Miller and Scholes (1978), the argument is that share prices are determined by the trading decisions of large, tax-exempt financial institutions like pension funds, that do not pay tax on their dividend income. If the “marginal” shareholder does not pay the dividend income tax, neither stock market valuations nor value-maximizing investment decisions should be influenced by dividend tax rates. A related argument has been made in the context of small open economies by Boadway and Bruce (1992). In their model, the rate of return on domestic investment is determined either by the rate demanded by foreign shareholders (for capital importers), or the rate available on outbound investment (for capital-exporters). Both of these rates are determined on the world market and are unaffected by domestic taxes on dividends. Boadway and Bruce do not explicitly consider the market value of equity, since they assume that the domestic firm maximizes the utility of the domestic shareholder, rather than the market value. However, where the firm is partly owned by foreign investors, these foreign investors are effectively the “marginal” shareholders, and the firm’s market value is independent of domestic taxes on dividends. Fuest and Huber (2000) also assume that foreign investors are the marginal shareholders.

The assumption that share prices depend only on the tax treatment of one class of investors is difficult to reconcile with asset pricing theories based on optimal portfolio allocation by risk-averse investors. However the finance literature provides a model that—at least in a small open economy context—leads to broadly similar predictions as the model of Boadway and Bruce (1992). This is the tax-adjusted capital asset pricing model (CAPM), first set out by Brennan (1970), in which different investors may face different tax rates on dividend income. This approach makes no assumption about the identity of “the” marginal shareholder. In contrast, all investors can be “marginal” in the sense of being just willing to hold the equity at the ruling market price, even though tax rates may differ both across investors and across assets. This is possible because they are also concerned about the risk they bear by holding risky assets. In equilibrium, an investor with a lower tax rate on a specific asset will tend to hold more of that asset, but at the cost of holding a less diversified, and hence more risky, portfolio.

In this model, the effect of dividend taxes on share prices depends on an average of tax rates across all investors. It does not matter that shares in a specific asset may be held predominantly by one group of shareholders; the relative ownership of a particular asset across different types of investors is irrelevant. What does matter is the relative total wealth of different investors. In the context of a small open economy, this implies that dividend income taxes on domestic investors are likely to be irrelevant to the tax capitalization effect. If firms choose investment to maximize their stock market valuations, this again implies that domestic dividend taxation will have little or no effect on investment.5

The 1997 dividend tax reform in the U.K. provides an interesting opportunity to test these predictions. As we explain in more detail in section II, before this reform, U.K. pension funds and insurance companies managing pension-related assets could reclaim part of the corporate income tax paid by firms on the underlying profits, when they received dividends paid by U.K. companies. These rebates cost the U.K. government around £5bn per annum, equivalent to about 20 percent of U.K. corporation tax revenue. Before 1997, U.K. pension funds alone owned around 30 percent of all U.K. equities, and held approximately three quarters of their equity holdings in U.K. companies. These rebates were abolished by the new Labour government in its first Budget in July 1997. The post-tax value of a given cash dividend paid by a U.K. firm to a U.K. pension fund fell by 20 percent, while other shareholders—both U.K. and foreign—were largely unaffected by this reform.

Extending the argument of Miller and Scholes (1978)—that share prices are largely determined by the valuation of domestic financial institutions that own the largest shares of these assets—to this U.K. setting, Bell and Jenkinson (2002) suggested that this reform of U.K. dividend taxation caused a significant fall in the value of the U.K. stock market. To support this claim, they studied the behavior of share prices on ex-dividend days—the day on which the owner of the share ceases to be entitled to a recently announced dividend payment. If U.K. pension funds were “the” marginal investors, then the ex-dividend day fall in the market value of a U.K. company paying £100 of cash dividends should have been £125 (including the rebate of underlying corporation tax) before the 1997 reform and £100 after the reform. Equivalently the “drop-off ratio” (the change in the market value expressed as a proportion of the cash dividend) should have fallen from 1.25 to 1, a fall of 20 percent. In contrast, the tax-adjusted CAPM predicts that this reform of U.K. dividend taxation should have had essentially no effect on either share prices or drop-off ratios. U.K. pension funds would be expected to reduce their holdings of U.K. equities, as the tax advantage that induced them to bear more U.K.-specific risk was eliminated. But since U.K. pension funds are small relative to the world capital market, this portfolio reallocation should have little or no effect on U.K. share prices.

As a first step towards considering more detailed empirical evidence, it is useful to review movements in the U.K. stock market index on and after the announcement of the 1997 tax reform. The FTSE 100 index is presented in Figure 1, between 1995 and 2002. As is well known, there was considerable volatility in the stock market during this period, with the index almost doubling between 1995 and 1999 before falling back. The date of the 1997 tax reform is marked by the vertical line. If equity valuations had followed the Bell-Jenkinson prediction, then, ceteris paribus, there should have been a fall of around 20 percent in the value of the index on the announcement of the reform. Clearly, this did not happen. Instead the index continued to rise. It is conceivable that other announcements in Gordon Brown’s first Budget may have overshadowed the dividend tax reform. However, it is difficult to think of precedents for announcements of changes in economic policy producing the required 20 percent rise in the stock market, at least in advanced economies.

Figure 1.
Figure 1.

FTSE 100 Index, July 2, 1997 Indicated by Vertical Line.

Citation: IMF Working Papers 2007, 204; 10.5089/9781451867688.001.A001

Source: Thomson Financial Datastream

Nevertheless, Bell and Jenkinson (2002) find that the mean drop-off ratio for listed U.K. companies fell significantly after this tax reform, although their central estimate is from around 1.05 to 0.85. We confirm their finding in our empirical analysis; however we do not share their conclusion that U.K. share prices are determined by the valuations of U.K. pension funds.

Our doubts about their conclusion are based both on the theoretical claim that one type of shareholder should be “the” marginal investor in a market where the same assets are held simultaneously by different investors facing different tax rates; and the methodological concerns about inferring the impact of dividend taxes on share prices from fluctuations in the mean drop-off ratio, that have been highlighted recently by Chetty, Rosenberg and Saez (2007). To support this, we present a more detailed empirical study of the behavior of ex-dividend day drop-off ratios in the U.K., both around the 1997 tax reform and over a longer period. While the mean drop-off ratio did fall significantly in the second half of the 1990s, further analysis reveals that this fall was associated with a sharp increase in the proportion of observations where the share price rose on the ex-dividend day, generating a negative value for the drop-off ratio. It is not clear how this development could be related to the dividend tax reform. A similar pattern was observed in the late 1980s, with the mean drop-off ratio being low and the proportion of observations with negative values being high in the period after the 1987 stock market crash. There were no changes to the tax treatment of U.K. pension funds that could explain these patterns in the late 1980s. In line with Chetty, Rosenberg and Saez (2007), we conclude that the mean drop-off ratio in the U.K. is too volatile for short term fluctuations around tax reforms to provide reliable evidence on the effects of dividend taxation on the stock market valuation of firms.6

In a related paper, we investigated more directly whether the 1997 U.K. dividend tax reform affected the investment behavior of U.K. companies, and found no significant effect.7 Our findings in this paper are consistent with that result, but also provide an explanation for why there should be no effect of U.K. dividend taxation on U.K. corporate investment. At least in the U.K. context, this does not seem to be explained by the standard “new view” argument, which would imply a significant tax capitalization effect on share prices. Rather, it appears to reflect the size of the U.K. and the openness of its capital market. Consistent with the tax-adjusted CAPM in a small open economy setting, there is no significant effect of domestic dividend taxes on U.K. share prices, and hence no impact on investment by U.K. companies.

In the next section we present a brief summary of the U.K. dividend tax regime before and after the 1997 reform. Following that, in Section III we outline a simple version of the Brennan (1970) tax-adjusted CAPM model, which serves to highlight the features of the market which are important in determining prices. In Section IV, we summarize the empirical predictions of the model and set out how we implement empirical tests. Section V presents the data, and Section VI the results. We conclude in Section VII.

II. The Taxation of Dividends in the U.K.

We briefly summarize the main elements of the dividend tax regime both before and after 1997.8 From the early 1970s until 1999, the U.K. operated a partial imputation system. On paying a cash dividend, U.K. firms were obliged to pay a proportion of the dividend in tax: Advance Corporation Tax (ACT). Subject to restrictions (principally that the dividend did not exceed U.K. taxable profit), the ACT could be credited against the main corporation tax charge, and thus generally only affected the timing of corporation tax payments. In addition, however, U.K. shareholders could also claim a credit against the U.K. income tax due on the receipt of the dividend. In general, ACT was charged at the basic rate of income tax (20 percent for dividend income in 1997) on the grossed-up dividend (i.e. the cash dividend plus the ACT). Hence basic rate shareholders were deemed to have paid tax in full on any dividends received, and consequently did not have to pay any further tax. Higher rate taxpayers, whose marginal tax rate was 40 percent, had to pay additional tax. For a £100 cash dividend, they had to pay tax on the grossed-up value of £125, i.e. a total of £50, but they could offset against that the £25 tax credit, leaving them with another £25 to pay.

The crucial element of the tax regime for our purposes is that tax-exempt U.K. shareholders were entitled to claim a tax rebate equal to the ACT paid by the firm. Just before the tax reform in 1997, this was worth 25 percent of the cash dividend (equivalent to 20 percent of the grossed-up dividend). As noted earlier, the cost of paying this rebate prior to 1997 was around £5 billion per year, around 20 percent of U.K. corporation tax revenue.

The 1997 tax reform abolished this cash rebate for U.K. pension funds, and the pension-related assets of U.K. insurance companies. Other tax-exempt shareholders—charities, non-tax-paying individuals, and holders of tax-advantaged personal equity plans—were unaffected. Some tax treaties also provided for non-U.K. shareholders to receive part of the tax credit—worth approximately 6percent of the cash dividend. They too were unaffected by the 1997 reform.

In 1999 the system was further reformed. The cash rebate was now abolished for most other non-tax-paying individuals,9 including those foreign shareholders that used to receive some benefit.10 The credit rate was halved to 10 percent, but U.K. tax-paying shareholders were unaffected, as income tax rates on dividend income were also reduced. At this time, ACT was also abolished, and new payment arrangements were introduced for companies paying corporation tax.

III. A Simple Portfolio Model

We present a simple version of the one period tax-adjusted CAPM model of Brennan (1970) which has been widely used to study the case of shareholders with heterogeneous tax rates.11 The aim here is to identify the effects of differences in tax rates, not only across investors, but also across assets for an individual investor.

There are a large number, N, of investors. Investor i has an endowment of Xi, which is divided between two risky assets, H and W, and a risk-free asset. Investor i holds Hi shares at price p in asset H, Wi shares at price q in W, and the remainder, B = Xi-pHi-qWi, in the risk-free asset. Dividends from H and W, denoted DH and DW, are taxed at rates mHi and mWi respectively, net of any dividend tax credits. Capital gains are taxed at rate zi for both assets. Interest income from the risk-free asset is taxed at rate mi. Dividends are assumed to be known, but the prices of the risky assets at the end of the period, denoted P˜Hand P˜W, are stochastic. Random variables are denoted with a tilde—their expected values at the start of the period are shown without the tilde.

The end-of-period wealth of investor i is Z˜i, where

Z˜i=(1+r(1mi))Bi+(P˜H+(1miH)DHzi(P˜Hp))Hi+(P˜W+(1miW)DWzi(P˜Wq))Wi=Xi+(1zi)[ρiBi+(G˜H+γiHDH)Hi+(G˜W+γiWDW)Wi](1)

where G˜i is the stochastic capital gain on asset i eg. G˜H=P˜Hp, where r is the risk- free interest rate,

ρi=(1mi)r1zi(2)

is the tax-adjusted discount rate of investor i, and

γiH=(1miH)1ziand γiW=(1miW)1zi(3)

are the tax discrimination variables of investor i for assets H and W respectively.

Investors choose H and W to maximize

Vi=Ziφi2var(Z˜i)(4)

where φi is a risk aversion parameter. The form of φi is important: we discuss two special cases below.

The expected value of Z˜i, denoted Zi, is equal to the expression in (1), with the stochastic capital gains terms replaced by their expected values. The variance of Z˜i is

var(Z˜i)=(1zi)2(Hi2σH2+Wi2σW2+2HiWiσHW)(5)

where σH2,and σW2 are, respectively, the variances of P˜Hand P˜W and, and σHW is the covariance.

Assuming an interior solution in which the investor simultaneously holds all three assets, the investor’s demand for each asset can be derived from the first order conditions for Hi and Wi, which are:

Hi=GH+γiHDHpρiφi(1zi)σH2WiσHWσH2andWi=GW+γiWDWqρiφi(1zi)σW2HiσHWσW2(6)

We can use these demand equations to solve for the equilibrium prices, and rates of return. Suppose there are, in aggregate, H and W shares in the two risky assets respectively. Define λi = 1/(1-zi)φi so that a higher λi implies either lower risk aversion or a higher capital gains tax rate. Aggregating the first expression for Hi over N investors and rearranging implies

H=i=1NHi=GHλi+DHλiγiHpρiλiσH2σHWWσH2=λi[GH+γ¯HDHpρ¯]σH2σHWWσH2(7)

where γ¯Hand ρ¯ are weighted averages:

γ¯H=γiHλiλi(8)
and ρ¯=ρiλiλi(9)

An equivalent expression holds for asset W. Alternatively, we can express the equilibrium expected return to purchasing a share in H, as

GH+γ¯HDHp=ρ¯+σH2H+σHWWpλi(10)

This takes a familiar form: the expected return is equal to the weighted average return on the risk-free asset, plus an adjustment for risk. The definition of the weighted average return is discussed below. The risk adjustment depends on the variance of the end-of-period price of the asset itself and the covariance with the end-of-period price of the other risky asset, where the weights on these two terms depend on their relative size in the overall market. If asset H is sufficiently small relative to W, then only the covariance term matters. This expression is consistent with Brenan’s (1970) model of the CAPM with personal taxes and has been the subject of extensive empirical testing.12

The portfolio choice of investor i depends on his own tax rates relative to that of other investors. Specifically,

Hi=λiHλi+λiσW2σH2σHW2[(γiHγ¯H)σW2DH(γiWγ¯W)σHWDW(ρiρ¯)(pσW2qσHW) ](11)

Clearly, investor i will tend to hold more or less of H, depending on whether his tax parameter, γiH is above or below the weighted average, γ¯H. If, for example, γiH=1.25 and γ¯H= 1, as was broadly the case for U.K. pension funds holding U.K. equities before 1997, then pension funds would hold more of this asset. How much more depends on the risk of the two assets and the investor’s risk aversion, λi. For example, the more risk averse is the investor (the lower λi), the less would be the tendency to have additional holdings of this asset in response to favorable tax treatment. Of course, the tax treatment of the other assets also affect the holdings of H. Advantageous tax treatment of the return from W or the risk-free asset relative to a weighted average of other investors (i.e. γiW>γ¯Wor ρi>ρ¯) would reduce holdings of H by investor i.

To examine the effects of differential taxation further, it is necessary to examine the weighted average tax rates. It is useful to simplify by assuming that all investors face the same rate of capital gains tax on all assets, in which case the weighted averages depend only on the risk aversion parameter, φi:

γ¯H=γi/φi1/φiand ρ¯=ρi/φi1/φi.(12)

Now consider two special cases:

(i) All investors have the same degree of risk aversion: φi = φ for all i. In this case, ∑λi=N/(1−z)φ and  γ¯Hand ρ¯ reduce to unweighted averages across all investors. One implication of this is that individual holdings of the risky assets do not depend on the initial endowment. Consider (11), but setting the tax rates faced by all investors on each asset to be the same. Then the second term is zero and Hi = H/N: all investors hold the same number of shares in H. Any difference in endowments is reflected only in the holding of the risk-free asset. Of course, holdings of the risky assets are affected by tax rates; but the fact that holdings differ across investors is not reflected in the construction of the average tax rates, which are unweighted. This is because each investor is at a margin and is equally likely to trade part of the holding.

A simple alternative to this is:

(ii) Risk aversion differs only across endowments: φi = φ/Xi.13 In this case, the weights for γ¯and ρ¯ are initial endowments:

γ¯H=γiHXiXiand ρ¯=ρi/XiXi(13)

This is more intuitive; abstracting from differences in taxes again, holdings of risky assets are exactly proportional to the endowment since λi/∑ λi = Xi/∑ Xi. Note though that again the weighted tax rates do not depend on the holdings of each asset: indeed, the weights for H and W are the same. Suppose investor i has a tax advantage from H and hence holds a greater proportion of his investment in H compared to other investors. It is not the case that the weighted average tax rate for H disproportionately reflects i’s tax rate. As in the previous case, all investors are at the margin; the difference from the previous case is that since holdings are proportional to the endowment, a wealthier investor would trade more in response to a change in, say, the expected end-of-period price. As a result, his tax rate is weighted more.

It is interesting to note the consequences of taxes varying only across investors, so that mi=miH=miW and hence γiHγ¯H=γiWγ¯W=ρiρ¯ for all investors. In this case, investor i would hold more or less than the weighted average (λiH/∑λi) holding, depending on whether he faced a relatively high tax rate (that is, whether γiH>/<γ¯H and on the sign of σW2DHσHWDW(pσW2qσHW). Even in this case, it is therefore generally not true that all investors would divide their portfolio across assets in the same way. Hence the weights for constructing the average tax rates would still not be equal to relative holdings of the individual assets.

However, finally note that from (10), the market valuation of each asset depends only on the tax rates applied to that asset. An implication of this is that expression (10) is equally valid in considering the price implications of the recent US dividend tax reform, even though that tax reform applied to dividends from all equities. The US tax reform reduced the dividend tax rate for US personal investors. The effect of this on US equity prices depends on how the average tax rate across all investors in US equities was affected. If the wealth of the group of US taxpayers affected was sufficiently small, relative to the wealth of tax-exempt US investors and non-US investors, then again as a first approximation, there would be little or no impact on US equity prices.14

We note three qualifications to this simple model. First, we have assumed an internal solution in which all investors hold all assets. Consider the introduction of a subsidy to a group of investors on the income from asset H. This will induce those investors to switch their holdings in favor of H. They will continue to do so either up to the point at which, at the margin, the gain from the subsidy is exactly offset by the additional risk they bear by moving away from an optimally diversified portfolio—this is characterized by (6)—or until those investors have switched all their holdings to asset H. In practice we do not observe investors holding only one form of asset, and so (6) seems the most likely equilibrium.

Second, as argued by Miller and Scholes (1978), it may be the case that trading costs deter some investors responding to small changes in the prices or expected returns from particular assets. If only a subset of all investors respond to new information, then at the margin, it is only the tax rates of those “marginal” investors which will be reflected in the weighted average tax rates. The relevance of this observation for examining the U.K. tax reform depends on whether the weight of U.K. pension funds should be higher than if all investors were taken into account. This is a key empirical issue which we address below.

Third, this model ignores trading around the ex-dividend day. To prevent a tax-favored investor holding only asset H cum-dividend and then diversifying ex-dividend, it is necessary to introduce some cost to this trading strategy. For example, there may be transaction costs, or a risk of unfavorable underlying price movements around the ex-dividend day. Michaely and Villa (1995) develop a theoretical model in which ex-dividend day trading is allowed but is endogenously limited. Lasfer (1995) presents empirical evidence that ex-dividend day returns in the U.K. are not significantly affected by short-term trading. We follow Bell and Jenkinson (2002) in assuming that an analysis of U.K. ex-dividend day returns can in principle identify the impact of dividend taxation.

IV. Empirical Implications

This model suggests two empirical hypotheses about the impact of dividend taxes.

First, the overall effect of dividend taxes on share prices reflects the weighted average tax rate of all investors, γ¯H. What does this suggest about the impact of the 1997 U.K. dividend tax reform on U.K. share prices? Even taking the second special case above, U.K. pension funds control only small proportion of the total wealth invested in all markets. Any change in their tax rate is therefore likely to have a negligible effect on U.K. equity prices. Thus:

The 1997 tax reform should have little or no effect on the prices of U.K. equities.

We test this proposition below using the standard technique of analyzing drop-off ratios. When a share goes ex-dividend, marginal shareholders are indifferent between either selling the share at the cum-dividend price, thus forgoing the dividend, or keeping the share and thus receiving the dividend. Denote the cum-dividend price by Pc, the ex-dividend price by Pex, and the dividend by D. Then following Elton and Gruber (1970), and using (10), we have

PcPex=D(14)
or PcPex/D=γ¯(15)

The term on the left hand side of this expression is the drop-off ratio: the fall in the price expressed as a proportion of the dividend. The term on the right hand side is the tax discrimination variable, described above. The drop-off ratio can therefore be used to estimate the average value of the tax discrimination parameter—which determines the share price.

In practice, we measure Pc at the end of trading on the last day the share trades cum-dividend, and Pex at the end of trading on the first day the share trades ex-dividend. Clearly, the difference between these two prices will reflect not only the dividend payment, but all other news about the value of the firm that emerges on the ex-dividend day. Averaging across a large number of independent observations on drop-off ratios is therefore required to obtain a useful estimate of γ¯. Adjustments can also be made for market movements on ex-dividend days (see below).

For comparison with Bell and Jenkinson (2002), we follow the same approach as they in estimating γ¯. Briefly, assume, following Lakonishok and Vermaelen (1983), that the price changes are random variables which can be written as

PcPex=θD+ε(16)

where the ε are independently distributed with

E(ε)=0 and var(ε)=Pc2σ2;(17)

that is the standard deviation of the unexplained price change is assumed to be proportional to the share price. As proposed by Boyd and Jagannathan (1994), an efficient estimate of γ¯, before and after the tax reform, can then be found by estimating

PcPexPc=θ1DPc+θ2FDPc+e(18)

where

e=εPcand hence var(e)=σ2 .(19)

In (18), θ1 provides an estimate of γ¯ prior to the tax reform. F is a dummy variable which takes the value of 0 for observations before the tax reform and 1 for observations after the tax reform; hence θ2 is an estimate of the change in γ¯ following the tax reform. Based on microstructure models developed by Boyd and Jagannathan (1994) and Frank and Jagannathan (1998), which suggest a negative intercept in such a regression, a constant term may also be included.

A further common adjustment is to account for market movements on the ex-dividend day multiplied by a historic estimate of the correlation between the return on the share and the return on the market. That is, we replace Pex with Pex*=PexPcβRm where Rm is the return on the market on the ex-dividend day, and β is the CAPM measure of risk of that equity.

The second empirical prediction concerns holdings of U.K. equities. Expression (11) makes clear that, ceteris paribus, any investor will tend to hold more of a given asset— say H—when his tax discrimination variable for that asset (γiH) is above the average of all investors (γ¯H). As is clear from the discussion above, until 1997 the value of γ for U.K. pension funds holding U.K. equities was 1.25 and therefore significantly above the average value across all investors. However, after 1997, this value fell to 1. This implies that:

U.K. pension funds should hold a disproportionately high share of U.K. equities before 1997, but this share should fall after 1997. By contrast, other investors should hold a disproportionately low share of U.K. equities before 1997, but should increase their share after 1997.

To investigate this proposition, we report evidence on the composition of equity portfolios before and after 1997 for U.K. pension funds and other institutional investors, and we report evidence on the share of U.K. equities held by different classes of investors.

V. Data

We set up our data to mirror as closely as possible the data used by Bell and Jenkinson (2002), henceforth BJ, to ensure that any differences we encounter are not caused by the samples.

Specifically, we use data from Thomson Financial Datastream on dividend payments of quoted U.K. companies. This data set contains one observation per payment, i.e. typically two observations per firm per year, as most U.K. firms pay an interim and a final dividend in each accounting year. We merge daily data on share prices and return indices into this data set, keeping in each case the observation on the day when the share first trades ex-dividend and on the day before, i.e. the ex-dividend and cum-dividend prices.

Before running regressions, we clean the resulting data sets as follows. We drop any observations where core data are missing, such as the payment date, the ex-dividend date, the (cum- or ex-dividend) share price or the value of the dividend. We also drop observations where the last cum-dividend observation predates the ex-dividend observation by more than 5 trading days. We drop a few observations for which we cannot work out the accounting year end date, because we need this in order to match the dividend payment data with information from company accounts. After matching the data with company accounts, we drop all firms for which the sum of individual dividend payments over the year does not match up with the total dividend payment reported in the accounts. Then we drop all dividend payments that were designated as Foreign Income Dividends, as the tax treatment for this form of dividends was different. We also drop any observation for which the share price did not move on the ex-dividend date, which suggests that there was no trading. Finally we drop outliers, which we define as drop-off ratios in excess of 5.

As explained above, we adjust returns for general market movements using the CAPM. To allow comparisons with BJ, we follow their approach in estimating the correlation of each share’s monthly returns with market returns (β). We thus run separate regressions of each share’s monthly return (including capital gains and dividends) on the monthly return of the FTSE All-Share index during the 5 years preceding the tax reform. We only keep shares with at least 36 historical observations.

The cleaning procedure used by BJ is virtually the same as ours, except that they did not delete data where the sum of dividend payments differed from the figure reported in company accounts, and they did not drop outliers as defined above. Hence our sample is slightly smaller than theirs, with data on 7966 dividend payments by 1275 firms.

VI. Results

This section first presents empirical evidence on the behavior of drop-off ratios in the U.K. It then briefly considers evidence on U.K. equity ownership. BJ use the 1997 reform to test whether taxes affect the valuation of dividends and to attempt to find the identity of what they refer to as “the” marginal shareholder. We first replicate their main findings using our sample, confirming that the mean drop-off ratio did fall significantly in the U.K. in the late 1990s. We then look in more detail at the nature and timing of this 17 change in the distribution of drop-off ratios, and consider fluctuations in the mean drop-off ratio over a longer horizon.

A. Bell and Jenkinson (2002) Replication

Table 1 presents the results obtained from estimating mean drop-off ratios for pre-reform and post-reform periods in a similar way to BJ, based on OLS estimation of equation (18). Columns 1 and 2 reproduce the results from BJ; columns 3 and 4 present our replications. Like BJ, we compare the 30 month period before the 1997 tax reform with the 30 month period after the reform. Following BJ, we report results for the sample of all firms and for the sub-sample of the largest 250 firms. Our results are very similar to those obtained by BJ. While we estimate a smaller fall than BJ, we confirm that there was a significant fall in the mean drop-off ratio in the U.K. after July 1997, particularly for larger firms.

Table 1.

Regression Results Obtained by BJ and Replication

article image
Source: Authors’ calculations.Notes: Heteroskedasticity-robust standard errors in parentheses. Results obtained by regressing drop-off ratios multiplied by D/Pc on D/Pc, and a post-reform dummy multiplied by D/Pc. Stars indicate the level of significance (*: 10percent, **: 5percent).

We implemented a number of robustness checks, which suggested that these results are robust. Specifically, we considered the following alternative specifications. (a) Including a constant term to allow for certain ex-dividend day trading behavior as suggested by microstructure models in Boyd and Jagannathan (1994) and Frank and Jagannathan (1998): this does not affect the estimated coefficients. (b) Not correcting the ex-dividend price (Pex) for market movements: this hardly affects the coefficients and leads to slightly more significant falls in the mean drop-off ratios. (c) Not dealing with heteroskedasticity, i.e. just regressing the drop-off ratio on a constant and a post reform dummy: this does not affect the results for the sample of large firms. For the full sample, this reduces the estimated fall in the mean drop-off ratio by half. The estimated fall in this case is only significant at the 13percent level.

Before extending the investigation, it is worth discussing the interpretation of these results. It is true that the estimated change in the mean drop-off ratio, at least for the larger companies, is close to the theoretical drop in the value of for U.K. pension funds: 20percent (BJ) or 17percent (our results) as against 20percent in theory. However, if pension funds were “the” marginal shareholders, then the estimated levels of these mean drop-off ratios are not as expected. If pension funds were the marginal shareholders, the mean drop-off ratio should be around 1.25 before the reform and 1 after the reform. 18 For the largest firms, the empirical results suggest a mean drop-off ratio of around 1 before the reform and around 0.8 after the reform.

Of course, based on the asset pricing model set out in Section III, we would not expect the mean drop-off ratio to reflect only the tax rates of U.K. pension funds, but rather an average across all investors in U.K. equities. While the levels of the mean drop-off ratios before and after the reform could reflect an average across investors, from this perspective, the significant fall in the mean drop-off ratio is more surprising.

B. Drop-off Ratios and Dividend Yields

BJ also consider changes in the mean drop-off ratio for sub-samples divided by dividend yields. The 1997 tax reform affected those shareholders with the highest valuation of U.K. company dividends. In the presence of clientele effects, highly taxed investors would be expected to hold shares in low-dividend-paying firms, and lightly taxed (or subsidized) investors would be expected to hold shares in high-dividend-paying firms. This suggests that, before 1997, U.K. pension funds were more likely to be “the” marginal shareholders for U.K. firms with relatively high dividend yields. If this were the case, then the 1997 reform is expected to have most impact on the mean drop-off ratio for high-dividend-paying firms.

BJ report results that appear to support such clientele effects. Specifically they use annual data on dividend yields to divide their observations in the pre-reform and post-reform periods, separately, into quintiles. They then compare the mean drop-o¤ ratio for each quintile in the pre-reform period with the mean drop-o¤ ratio for the corresponding quintile in the post-reform period. We replicate these results in Table 2.15 Like BJ, we find that the mean drop-off ratio fell significantly only when comparing observations with relatively high dividend yields, although it is not the case that observations in the top quintile had the largest or most significant drop.

Table 2.

Regression Results by Dividend Yield Quintiles

article image
Source: Authors’ calculations.Notes: Heteroskedasticity-robust standard errors in parentheses. Results obtained by regressing drop off ratios multiplied by D/Pc on D/Pc, and a post-reform dummy multiplied by D/Pc. Sample split by dividend yield quintiles before and after reform, where (5) is the top quintile. Stars indicate the level of significance (*: 10percent, **: 5percent).

Given that the rationale for splitting the sample by dividend yields is based on the tax preference of U.K. pension funds for a high dividend yield in the pre-reform period, it would seem more appropriate to divide the full sample into quintiles based on dividend yields in the pre-reform period only. We use data on average dividend yields in the pre-reform period to divide our sample of firms into quintiles. We then compare the mean drop-off ratio for each quintile in the pre-reform period with the mean drop-o¤ ratio for the same sub-sample in the post-reform period. Unlike the procedure used by BJ, this ensures that we are comparing mean drop-off ratios for the same firms in the two sub-periods.

Table 3 presents these results. When the samples are classified in this way, it is notable that the fall in the mean drop-off ratio becomes small and statistically insignificant for the sub-sample with the highest dividend yields in the pre-reform period. The clear pattern in the behavior of drop-off ratios by dividend yields reported by BJ is thus quite sensitive to the precise way in which their sub-samples were chosen. Moreover, and regardless of the method used to select the sub-samples, we can note that the pattern of estimated mean drop-off ratios in the pre-reform period provides little support for the view that the tax treatment of U.K. pension funds was particularly important for the stock market valuation of U.K. firms with relatively high dividend yields.16

Table 3.

Regression Results by Pre-reform Dividend Yield Quintiles

article image
Source: Authors’ calculations.Notes: Heteroskedasticity-robust standard errors in parentheses. Results obtained by regressing drop-off ratios multiplied by D/Pc on D/Pc, and a post-reform dummy multiplied by D/Pc. Sample split by pre reform dividend yield quintiles, where (5) is the top quintile. Stars indicate the level of significance (*: 10percent, **: 5percent).

C. Evidence on the Distribution of Drop-off Ratios

To investigate the behavior of drop-off ratios further, we now consider the distribution of drop-off ratios. The 1997 tax reform reduced the tax discrimination parameter i for the class of shareholders which previously had the highest valuation of dividends. If U.K. pension funds were indeed “the” marginal investors for certain types of U.K. firms, the fall in the mean drop-off ratio reported in Table 1 should be associated with compression in the upper part of the distribution of drop-off ratios. Essentially, the highest values of γi were eliminated by the tax reform, while lower values of γi were unaffected.

To examine this prediction, Figure 2 plots various quantiles of the distribution of drop-off ratios over the same sample period used in Table 1. In fact we see the opposite pattern, with the fall in the mean drop-off ratio after 1997 being associated with a fall in drop-off ratios at the bottom end of the distribution. The upper quartile increases from 1.3 in 1995 to 1.4 in 1999, with no sign of any reduction following the 1997 tax reform. In contrast, the bottom decile falls steadily throughout this period, from 0.1 in 1995 to -0.6 in 1999. This indicates that there was a considerable increase in the proportion of observations with negative drop-off ratios. A negative drop-off ratio is found when the firm’s share price increases (relative to the market) on the ex-dividend day, notwithstanding the loss of the entitlement to the dividend payment. We discuss this development further below, but note that shifts at the bottom end of the distribution of drop-off ratios are not easily explained by the change in the tax treatment of U.K. pension funds. More generally, Figure 3 shows that the distribution of drop-off ratios widened after 1997, while differences in the tax treatment of different classes of investors were reduced. This suggests that developments other than the 1997 tax reform may have been the dominant influence on the behavior of U.K. drop-off ratios during this period.

Figure 2.
Figure 2.

Drop-off Ratios, 1988–2000

Citation: IMF Working Papers 2007, 204; 10.5089/9781451867688.001.A001

Notes: Results obtained by regressing drop-off ratios multiplied by D/Pc on D/Pc for each year and half-year using the full sample of firms.
Figure 3.
Figure 3.

The Distribution of Estimated Drop-off Ratios

Citation: IMF Working Papers 2007, 204; 10.5089/9781451867688.001.A001

D. Further Evidence on Timing

We now extend the analysis to consider more carefully the timing of these changes in the mean drop-off ratio, and the longer term evidence. Following BJ, our regression analysis in Tables 1-3 neglected precise timing issues, as there was just one post-reform dummy: the test compared a 30 month period before the reform with a 30 month period after the reform. In order to see more precisely when the fall in the mean drop-off ratio occurred, we can estimate the mean drop-off ratio for 6 and 12 month periods. To maintain comparability with the previous results, we again use the GLS estimation procedure explained in Section IV. Table 4 presents these estimates of mean drop-off ratios for each year and half-year from 1995 to 1999.

Table 4.

Share of Negative Drop-off Ratios

article image
Source: Authors’ calculations.Notes: Calculated for full sample of firms.

The annual estimates suggest that the mean drop-off ratio did not fall significantly until 1999, although the tax reform was implemented in July 1997. The six monthly estimates suggest that there was a marked fall in the second half of 1997. However they also show that there was a larger increase in the mean drop-off ratio in the second half of 1998, which stops this showing up in the annual estimate for 1998. Indeed the mean drop-off ratio in the second half of 1998 is the highest found for any of these 21 six-month periods. This indicates that there are substantial fluctuations in these estimates of mean drop-off ratios, which may have more to do with general stock market dynamics rather than tax changes.

To explore this further, we consider longer term evidence. Figure 2 plots annual and six-monthly estimates of mean drop-off ratios between 1988 and 2000.17 Our original sample period is marked here by the two vertical bars, with the tax reform occurring in the middle of that period.

This evidence confirms that the behavior of the mean drop-off ratio in the U.K. is indeed erratic. There is a sharp increase from 1988 to 1991, which is not explained by any change in the tax treatment of U.K. pension funds. Both the fraction of equity owned by tax-exempt institutions (see below) and their tax treatment were stable over this period. The tax discrimination parameter for U.K. pension funds fell from 1.33 in 1992 to 1.25 in 1994, when the rate of the refundable dividend tax credit was reduced from 25percent to 20percent. However we see that there was no fall in the mean drop-off ratio for U.K. companies over this period. The period studied by BJ is thus unique in showing an association between a significant fall in the mean drop-off ratio and an increase in dividend taxation for U.K. pension funds. Furthermore, the mean drop-off ratio at the end of this period, in the second half of 2000, is very similar to that at the start of the period, in the first half of 1988, although the relevant tax discrimination parameter for U.K. pension funds had fallen from 1.33 to 1.

These fluctuations in the mean drop-off ratio appear to be associated with changes at the bottom end of the distribution, and in particular with the fraction of observations where the drop-o¤ ratio is negative. Table 5 reports annual figures for the share of observations with negative drop-off ratios. This fraction falls sharply from 1988 to 1991 and increases sharply towards the end of the 1990s, mirroring the fluctuations in the mean drop-off ratio shown in Figure 2. Negative values for the drop-off ratio—observations where the share price increases despite the share going ex-dividend—seem to be most common in periods when the stock market is both volatile and rising. This was the case immediately after the 1987 stock market crash, and again during the dot com bubble period of the late 1990s. These developments in the tail of the distribution of drop-off ratios appear to exert a strong influence on the behavior of the estimated mean.

Table 5.

Estimated Drop-off Ratios by Year/half-year

article image
Source: Authors’ calculations.Notes: Heteroskedasticity-robust standard errors in parentheses. Results obtained by regressing drop-off ratios multiplied by D/Pc on D/Pc for each year and half-year using the full sample of firms.

The behavior of the mean drop-off ratio in the US provides further grounds for doubting whether the fall in the U.K. emphasized by BJ was related to the 1997 U.K. dividend tax reform. Estimates presented in Chetty, Rosenberg and Saez (2007) show that the mean drop-off ratio in the US also fell sharply over the period studied by BJ, from around 0.8 in 1994 to around 0.4 in 2000, with the sharpest fall also occurring from 1999 to 2000. Using data over the period 1963-2004, Chetty, Rosenberg and Saez show that such fluctuations in the mean drop-off ratio in the US are not uncommon, and display little relationship with changes in dividend taxation. Our evidence for the U.K. supports their conclusion: estimates of mean drop-off ratios are too volatile to provide reliable evidence about the impact of dividend taxes on the stock market valuation of firms.

Equity ownership

The second proposition discussed in Section IV concerned the share of U.K. equities in the portfolios of U.K. pension funds. This share is expected to fall after the July 1997 tax reform eliminated a major tax advantage for U.K. pension funds of dividends from U.K. companies. This prediction also applies to the holdings of U.K. insurance companies insofar as they relate to the provision of pension plans, although not to the provision of life insurance.

Table 6 reports the proportion of U.K. equities in the total equity holdings of U.K. pension funds, insurance companies and unit trusts between 1990 and 2001. The U.K. share of pension fund equity portfolios did indeed fall sharply, from around three quarters at the end of 1996 to around two thirds by the end of 2001, having been quite stable during the first half of the 1990s. The U.K. share of insurance company equity portfolios also fell, from around 80percent to around 75percent, having also been stable in the period before this tax reform. In contrast, for unit trusts—mutual funds whose tax treatment did not change at all in 1997—there was a temporary increase in the U.K. share of their equity holdings immediately after the tax reform, although this has since returned to its pre-reform level. This indicates that the change in the composition of equity portfolios observed for U.K. pension 23 providers was not common to all U.K. institutional investors during this period. The smaller fall for insurance companies relative to pension funds is consistent with the tax change affecting only part of insurance company portfolios.

Table 6.

Share of U.K. Equities in Total Equity Holdings of Pension Funds, Long-term Insurance Companies and Unit Trusts

article image
Source: Financial Statistics (U.K.), tables 5.1A and 5.1B.

Table 7 reports estimates of the fraction of total U.K.-quoted equity owned by different types of shareholders, for the years between 1990 and 2000 for which these estimates are available. The last observation before the July 1997 tax reform refers to the end of 1994, when U.K. pension funds owned 28 percent of equity quoted on the London stock exchange. This share fell to 18 percent by the end of 2000, although it is not possible to date precisely the timing of this fall. The share owned by U.K. insurance companies has been much more stable. This suggests that the fall in U.K. equity as a share of insurance company total equity holdings, shown in Table 6, reflected an increase in the total size of their portfolios rather than a fall in their holdings of U.K. equity. The fall in the share of U.K. equity owned by U.K. pension funds was accompanied by an increase in the share of U.K. equity owned by foreign shareholders, which rose from 16 percent at the end of 1994 to 32 percent by the end of 2000. While this was partly a continuation of a longer term trend, it seems likely that this increase in foreign ownership of U.K. equity was given further impetus by the reduced attractiveness of U.K. equity to U.K. pension funds after the abolition of refundable dividend tax credits in July 1997.

Table 7.

Beneficial Ownership of U.K. Equities

article image
Source: National Statistics (2001), Table A. End of year figures, no data available for 1995 and 1996.

As we discussed in Section III, the effect on the valuation of U.K. equities would depend on the size of the premium required by foreign investors to increase their exposure to U.K.-specific risks. A negligible change in this risk premium is consistent with the absence of any noticeable effect on the level of the U.K. stock market (Figure 1). At first sight this is inconsistent with the change in the mean drop-off ratio after 1997 emphasized by BJ, but as discussed earlier in this section, there are reasonable grounds for doubting whether this fall in the mean drop-off ratio was driven by the change in the tax treatment of U.K. pension providers.

VII. Conclusions

This paper explores the extent to which dividend taxes are capitalized into share prices, using information derived from a significant U.K. tax reform. When different investors are subject to different tax rates, the Capital Asset Pricing Model suggests that the market valuation of equity should depend on an average of tax rates across all investors. In the context of the U.K., the model suggests that all investors that hold both U.K. equities and other assets are marginal, with differences in tax treatments just offset by differences in exposure to risks at the optimal portfolio allocations. Moreover, since the wealth invested by U.K. pension providers is small relative to the size of the world capital market, as a first approximation we would expect a change in the tax treatment of U.K. pension funds to have little or no effect on the pricing of U.K. equities. The first-order effects of this tax reform should instead be seen in a shift in the composition of pension providers’ portfolios away from U.K. equities. This affects equity prices only to the extent that other investors—notably foreign investors—require a higher risk premium to take up additional holdings of U.K. equity.

The abolition of refundable dividend tax credits in July 1997 represented a substantial increase in the taxation of dividends paid by U.K. companies to U.K. pension funds and U.K. insurance companies providing pension plans. Bell and Jenkinson (2002) argued that this tax reform had a substantial impact on the stock market valuation of dividends paid by U.K. firms, implying that U.K. pension providers were “the” marginal investors in U.K. equity, at least in the period before the tax reform. However, we question this conclusion, which is not consistent with the CAPM. We confirm that there was a fall in the mean drop-off rate after 1997, but question whether it can confidently be attributed to the abolition of refundable dividend tax credits for U.K. pension funds. The largest fall occurred in 1999, some eighteen months after the tax reform. While the tax change affected investors with the highest valuation of U.K. dividends, the main change occurred in the lower tail of the distribution of drop-off ratios. As in the US, the mean drop-off ratio in the U.K. is shown to fluctuate erratically over a longer time period. Indeed the period studied by Bell and Jenkinson (2002) appears to be unique in showing an association with changes to the tax treatment of U.K. pension funds. Interestingly, Chetty, Rosenberg and Saez (2007) have shown that there was also a sharp fall in the mean drop-off ratio in the US in late 1990s. Presumably this 25 fall in the US was not driven by the tax treatment of U.K. pension funds; though it may have been driven by factors that were common to the US and the U.K. stock markets during this dot com bubble period.

We share the skepticism of Chetty, Rosenberg and Saez (2007) about the use of mean drop-off ratios to test hypotheses about the impact of dividend taxation. Mean drop-off ratios are simply too volatile to attach causal significance to short- run fluctuations around tax reforms.

We report evidence on the ownership of U.K. equities that are consistent with the predicted changes to the composition of U.K. pension fund portfolios after this tax reform in 1997. U.K. pension funds have reduced their exposure to U.K.-specific risks following the abolition of a uniquely favorable tax treatment of U.K. dividends. Foreign shareholders have increased their holdings of U.K. equities over the same period. Noting the relative size of U.K. and foreign investors, asset pricing theory does not suggest that this should have had a major impact on the market valuation of U.K. equity. This is consistent with the absence of a crash in the U.K. stock market on the announcement of this tax reform; and, in our view, it is also consistent with a closer examination of the available empirical evidence.

References

  • Auerbach, A.J., 1979, “Wealth Maximization and the Cost of Capital”, Quarterly Journal of Economics 94, pp. 433-436.

  • Auerbach, A.J., 1983, “Stockholder Tax Rates and Firm Attributes”, Journal of Public Economics 21, pp. 107-27.

  • Auerbach, A. J. and K. A. Hassett, 2002, “On the Marginal Source of Investment Funds”, Journal of Public Economics 87, pp. 205-32.

  • Bell, L. and T. Jenkinson, 2002, “New Evidence of the Impact of Dividend Taxation and on the Identity of the Marginal Investor”, Journal of Finance 57, pp. 1321-1346.

    • Search Google Scholar
    • Export Citation
  • Black, F. and M. Scholes, 1974, “The Effects of Dividend Yield and Dividend Policy on Common Stock Prices and Returns”, Journal of Financial Economics 1, pp. 1-22.

    • Search Google Scholar
    • Export Citation
  • Boadway, R. and N. Bruce, 1992, “Problems with Integrating Corporate and Personal Income Taxes in an Open Economy”, Journal of Public Economics 48, pp. 39-66.

    • Search Google Scholar
    • Export Citation
  • Bond, S. R., M. P. Devereux and A. Klemm, 2007, “Dissecting Dividend Decisions: Some Clues About the Effect of Dividend Taxation from Recent UK Reforms”, in A. Auerbach, J. Hines and J. Slemrod, eds.,Taxing Corporate Income in the 21st Century, Cambridge University Press, pp. 46-75.

    • Search Google Scholar
    • Export Citation
  • Boyd, J. H. and R. Jagannathan, 1994, “Ex-dividend Price Behavior of Common Stocks: Fitting Some Pieces of the Puzzle”, Review of Financial Studies 7, pp. 711-741.

    • Search Google Scholar
    • Export Citation
  • Brennan, M. J., 1970, “Taxes, Market Valuation and Financial Policy”, National Tax Journal 23, pp. 417-429.

  • Chetty, R. and E. Saez, 2005, “Dividend Taxes and Corporate Behavior: Evidence from the 2003 Dividend Tax Cut”, Quarterly Journal of Economics 120, pp. 791-833.

    • Search Google Scholar
    • Export Citation
  • Chetty, R., J. Rosenberg and E. Saez, 2007, “The Effects of Taxes on Market Responses to Dividend Announcements and Payments: What Can we Learn from the 2003 Dividend Tax Cut?” in A. Auerbach, J. Hines and J. Slemrod, eds., Taxing Corporate Income in the 21st Century, Cambridge University Press.

    • Search Google Scholar
    • Export Citation
  • Edwards, J.S.S. and M. Keen, 1984, “Wealth Maximization and the Cost of Capital: A Comment”, Quarterly Journal of Economics 99, pp. 211-214.

    • Search Google Scholar
    • Export Citation
  • Elton, E. and M. Gruber, 1970, “Marginal stockholders. Tax Rates and The Clientele Effect”, Review of Economics and Statistics 52, pp. 68-74.

    • Search Google Scholar
    • Export Citation
  • Frank, M. and R. Jagannathan, 1998, “Why Do Stock Prices Drop by Less than the Value of the Dividend? Evidence from a Country Without Taxes”, Journal of Financial Economics 47, pp. 161-188.

    • Search Google Scholar
    • Export Citation
  • Fuest, C. and B. Huber, 2000, “The Optimal Taxation of Dividends in a Small Open Economy”, CESifo Working Paper 348.

  • Gordon, R.H. and D.F. Bradford, 1980, “Taxation and the Stock Market Valuation of Capital Gains and Dividends”, Journal of Public Economics 14, pp. 109-136.

    • Search Google Scholar
    • Export Citation
  • Kalay, A. and R. Michaely, 2000, “Dividends and Taxes: A Reexamination”, Financial Management 29, pp. 55-75.

  • King, M.A. Taxation and the Cost of Capital”, Review of Economic Studies 41, pp. 21-35.

  • Lakonishok, J. and T. Vermaelen, 1983, “Tax Reform and Ex-dividend Day Behavior”, Journal of Finance 38, pp. 1157-1179.

  • Lasfer, MA, 1995, “Agency Costs, Taxes and Debt: The U.K. EvidenceEuropean Financial Management 1, pp. 265-85.

  • Litzenberger, R. and K. Ramaswamy, 1979, “The Effects of Personal Taxes and Dividends on Capital Asset Prices: Theory and Empirical Evidence”, Journal of Financial Economics 7, pp. 163-195.

    • Search Google Scholar
    • Export Citation
  • Litzenberger, R. and K. Ramaswamy, 1980, “Dividends, Short Selling Restrictions, Tax Induced Investor Clientele and Market Equilibrium”, Journal of Finance 35, pp. 469-482.

    • Search Google Scholar
    • Export Citation
  • Litzenberger, R. and K. Ramaswamy, 1982, “The Effects of Dividends on Common Stock Prices: Tax Effects or Information Effects?”, Journal of Finance 37, pp. 429-443.

    • Search Google Scholar
    • Export Citation
  • Michaely, R. and J-L Vila, 1995, “Investors’ Heterogeneity, Prices and Volume Around the Ex-dividend Day”, Journal of Financial and Quantitative Analysis 30, pp. 171-198.

    • Search Google Scholar
    • Export Citation
  • Miller, M. and M. Scholes, 1978, “Dividends and Taxes”, Journal of Financial Economics 6, pp. 333-364.

  • Miller, M. and M. Scholes, 1982, “Dividends and Taxes: Empirical Evidence”, Journal of Political Economy 90, pp. 1118-1141.

  • National Statisitcs, 2001, “Share ownership – A report on ownership of shares as of 31 December 2000”, London: The Stationery Office.

    • Search Google Scholar
    • Export Citation
  • Sinn, H.-W., 1991, “The ‘Old’ View, the ‘New View’ and Another View”, in D. Bradford, ed., Tax Policy and the Economy 5, pp. 25-54.

    • Search Google Scholar
    • Export Citation
1

Stephen Bond is at Nuffield College, Oxford and the Institute for Fiscal Studies (IFS), Michael Devereux is at the Said Business School, Oxford, and IFS, and Alexander Klemm is in the European Department of the IMF. The authors thank Tim Besley, John Norregaard and Jim Poterba for helpful comments.

2

For legal reasons linked to bilateral tax treaties, the U.K. still formally operates a dividend tax credit. For domestic shareholders this is equivalent to income tax at a preferential rate, and the remaining tax credit has negligible value for foreign shareholders.

3

Edwards and Keen (1984) made more precise the result that neutrality holds as long as the marginal source of finance is the same when the investment is made and when the returns are paid to shareholders (and as long as the dividend tax rate is constant).

4

See, for example, Sinn (1991). Recent papers that test predictions of the ‘new view’ include Auerbach and Hassett (2002, 2005), Chetty and Saez (2005) and Chetty, Rosenberg and Saez (2007).

5

If most stock market wealth is controlled either by (domestic or foreign) tax-exempt institutions or by foreign investors who can avoid paying income tax on dividends paid by domestic firms, this would further imply the Miller-Scholes view that dividend taxation in general is largely irrelevant for stock market valuations. However this stronger implication is not needed for the hypotheses we investigate in this paper.

6

Chetty, Rosenberg and Saez (2007) document similar volatility over time in the behavior of the mean drop-off ratio in the US. Interestingly their estimates show that there was also a sharp fall in the mean drop-off ratio in the US in the second half of the 1990s, although there was no similar tax change in the US that would explain this development.

8

A more detailed description of the tax system is provided in Bond and others (2007), where we investigate the impact of the 1997 reform on company dividend payments and investment.

9

Charities received temporary compensation for this loss. Holders of “Individual Savings Accounts” continued to receive credits until 2004.

10

For foreign shareholders this was achieved by halving the tax credit to 10percent and applying a withholding tax of 10percent, rather than by formally abolishing the credit.

11

See, for example, Litzenberger and Ramaswamy (1979, 1980, 1982), Gordon and Bradford (1980), Auerbach (1983) and Michaely and Villa (1995).

13

Of course it is straightforward to allow for differences in preferences as well as endowments. For example, φi = θi/Xi where θi represents individual preferences.

14

Chetty, Saez and Rosenberg (2007) provide more detail on this US tax reform, and empirical evidence on its impact on US equity prices.

15

Unlike BJ, we present results based on individual dividend payments, rather than artificial portfolios made up of all dividend payments on the same day. BJ state that results were similar in both cases.

16

Consistent with this, we find that the simple correlation coefficient between the drop-off ratio and the dividend yield in the pre-reform period is less than 1 percent.

17

Unfortunately we do not have the required data on dividend payments to calculate drop-off ratios before 1988.

The Effects of Dividend Taxes on Equity Prices: A Re-examination of the 1997 U.K. Tax Reform
Author: Stephen Bond, Mr. Alexander D Klemm, and Michael B. Devereux