NOTE ON THE TAXINDUCED CLIENTELE EFFECT AND TAX REFORM KENNETH R
Presentations text content in NOTE ON THE TAXINDUCED CLIENTELE EFFECT AND TAX REFORM KENNETH R
Page 1 NOTE ON THE TAX-INDUCED CLIENTELE EFFECT AND TAX REFORM KENNETH R. FERRIS* AND WILLIAM R. REICHENSTEIN** HE relationship between corporate linear association between expected re- dividend policy and security prices is turn and dividend yield following TRA. of interest to investors and corporate Finally, an after-tax CAPM is also esti- managers alike. It was traditionally mated. thought that, ceteris paribus, investors Our findings consistently indicate that preferred high dividend yield securities the tax deferral of unrealized capital gains relative to low yield securities (Gordon, and the ability to time the recognition of 1963); however, Miller and Modigliani gains in lower tax rate years are suffi- (1961) argued that, in world without ciently beneficial as to maintain these tax- taxes and transaction costs, corporate induced clienteles. Our estimates indicate dividend policy and security prices were that the association between before-tax independent. Further, where dividends return and yield (for given level of risk) were taxed at higher rate than capital is approximately one-half as large as the gains and tax rates varied across inves- relationship that existed prior to TRA. tors, they observed that corporations would adjust their dividend policies to attain an equilibrium with the income desires of 1. Tax Models shareholders. Hence, corporations would Prior to January 1, 1987, long-term attract "clientele" of investors whose tax capital gains enjoyed three preferential tax situation fit the corporation's dividend features relative to ordinary dividend in- policy. come: (1) an exclusion of 60 percent of the Given the higher taxation of dividends versus capital gains that existed prior to gain value; (2) the deferral of taxes on any 1987, high tax bracket investors would unrealized gains; and (3) the ability to time thus tend to prefer low dividend yield se- the realization, and thus taxation, of gains curities; and conversely, investors with low for low tax rate years. While TRA elimi- tax rates (e.g., pension funds and IRA ac- nated the first feature, it left intact the counts) would prefer high yield securities. other two. The Tax Reform Act of 1986 (TRA), how- necessary condition for the existence ever, eliminated the 60 percent long-term of tax-induced clientele effect is that capital gain exclusion, and hence presum- higher tax bracket investors receive higher ably this positive but nonlinear associa- after-tax returns on low yield securities. tion between dividend yield and common To investigate this, we utilize two simu- stock returns. lation models; both assume an percent The present study utilizes simulated annual pre-tax return on the underlying data and ex ante data from The Value Line investment, tax rate in all years be- Investment Survey to investigate the as- fore withdrawal, tax rate tn in the year sociation between yield and return fol- of withdrawal, and an year investment lowing TRA. In Section 1, two tax models horizon. The investment and withdrawal are used to illustrate that the existing of funds are assumed to occur on Jan- capital gains treatment can still produce uary 1. substantially higher after-tax returns for Model assumes that all of the per- high tax bracket investors. In Section 2, cent return is current income and is fully we estimate tax-clientele Capital Asset taxed annually at percent. This is rep- Pricing Model (CAPM) to illustrate the resentative of many debt and equity in continued presence of positive but non- vestments in portfolios with rapid turn- over; it also represents an equity *Southern Methodist University. investment with all expected return in the **North Texas State University. form of dividends. $1 investment in this 131 Page 2 132 NATIONAL TAX JOURNAL [Vol. XLI model accumulates at the after-tax rate investor that should benefit most from the of i(l t); and the after-tax value of favorable capital gains tax features.' The Model investment after years is: predominant values for and are as- sumed to be 13.2 percent and 0.8, which L, i(l t)ln. (1) are based on The Value Line Investment Survey forecasts of median expected re- Model is based on Doyle's (1984) model turn and the proportion of return ex- o@ the after-tax value (adjusted for the pected to be realized as capital gains for elimination of the 60 percent long-term the securities that investment service fol- capital gains exclusion) of an investment, lows as of the time of this study.' where percent of the percent annual The predominant value of p, or the pro- return is in the form of capital gains and portion of accumulated capital gains re- is the proportion of accumulated capital alized annually, is assumed to be 0.33, gains realized annually. The after-tax which appears reasonable for many high value after n-years of an investment sub- bracket investors under TRA. The 60 per- ject to this tax structure is: cent exclusion rule previously encouraged rapid turnover of equity portfolios for (1 r)n tn(l P) these investors; however, today's remain- ing capital gains tax advantages encour- -gi[(l r)n 1/rj (2) age slower realization of gains. Thus, most high tax bracket investors should where respond to the elimination of the long-term capital gains exclusion by lengthening the (1 tp)gi (1 t)(1 9)i. average time before the realization of such gains. This model assumes that an initial $1 in- The values reported in rows and in vestment grows at percent per year for Table are representative of the impor- years at which time taxes are paid at tance of the favorable capital gains fea- the rate tn on the accumulation of un- tures for some higher tax bracket inves- realized capital gains (the second part of tors in equity securities. All of the annual equation 2). If all capital gains are real- return in row is in the form of current ized annually (p 1), the model reduces income, while 80 percent of row Bs re- to equation 1. Setting 1.0 and 0.0 turn is in the form of tax-favored capital makes the model representative of the re- gains. The tax rate in the year of with- turn prospects for art, coins, gold, and drawal (tn) is assumed to be 28 percent, many other real assets. which is only slightly lower than the 33 should be apparent that any differ- percent tax rate in earlier years. The val- ence in the values produced by models ues in column show the average annual and is due to the existence of prefer- geometric after-tax rates of return from ential tax features associated with long- to 20 .3 The average after-tax term capital gains. Further, it should be rate of return in row is 10.50 percent, noted that the difference in values will or 1.70 percent above the rate in row A, increase as the proportion of return in the which ignores the advantageous capital form of capital gains (g) increases and as gains benefits. The 1.70 percent reported the proportion of gains realized annually in column AR is measure of the capital- (p) decreases. gains tax advantage. To illustrate, Table presents the af- Rows through show the importance ter-tax values for an initial investment of of the tax timing advantage. The lower $1 after 5, 10, 15, and 20 years for both the tax rate in the year of withdrawal, the models and 2, for various combinations greater is the capital gains advantage. of i, g, p, and tn. marginal tax rate in Many investors, especially those with vol- the years before withdrawal of 33 percent atile income, will experience larger tax is used to illustrate the higher tax bracket timing benefit than the percent differ- Page 3 No. 11 TAX-INDUCED CLIENTELE EFFECT 133 Table Comparison of the After Tax Values and Growth Rates for an Investment of $1 for Models and After Tax Values in Model tn 'g Yr -- 10 Yrs 15 Yrs. 20 Yrs. &R Yrs 13.2 0.33 1.528 2.334 3.565 5.446 0.088 13 0.8 0.33 0.28 1.575 2.551 4.210 208 0.105 0.017 13.2 0.8 33 0.33 1.553 2.492 4.087 6.796 0.103 0.015 13.2 0.8 0.33 0.15 632 2.706 4.531 631 0.108 0.020 13.2 0.8 67 0.28 1.551 2.437 3.864 6.162 0.096 0.008 13 0.00 0.28 599 671 588 ois 113 0.025 13.2 33 0.28 1.563 2.494 4.034 6.583 0.101 012 13.2 00 0.28 618 2.768 904 875 0.121 0.323 18.0 0.33 7.767 3.123 5.518 750 0.121 18.0 0.8 33 0.28 850 577 7.088 14 226 0.146 0.025 Any value of would produce the same outcome. annual before tax return on investment. proportion of in the form of capital gains. proportion of accumulated capital gains realized annually. tn marginal tax rate in year n. average annual geometric after tax growth rate between years and 20. AR difference between for models and for given i. ential, assumed in most of the analysis. of the after-tax advantage for higher tax Row C, with tn 0.33, assumes no tax rate investors in passive investment timing benefit, and the AR value of 1.50 fund, such as an index fund. percent shows the steady-state advantage Reducing g, the proportion of annual of tax deferral of unrealized capital gains. returns in the form of capital gains, re- Rows B, E, and reveal the importance duces the capital gains advantage as of the proportion of capital gains realized shown by comparing values in rows and annually (p). There is no capital gains ad- G. The tax advantage of precious metals vantage for very active portfolios that re- and many other real assets is illustrated alize all capital gains annually, and the in row with 1.0 and 0.0. The value of the advantage (AR) increases from capital gains advantage for given is 0.80 to 1.70 to 2.50 percent as decreases maximized under these conditions. Fl- from 0.67 to 0.33 to 0.0. The 2.5 percent nally, rows and show that the capii annual after-tax advantage is indicative gains advantage is sensitive to the in- Page 4 134 NATIONAL TAX JOURNAL [VOI. XLI Table Comparison of the Annual After Tax Growth Rates Due to Capital Gains Treatment Before and After January 1987 Row __In AR ARI AR/ARI 13.2 0.8 0.33 0.28 0.017 0.028 0.596 13.2 0.8 0.33 0.33 0.015 0.027 0.552 13.2 0.8 0.33 0.15 0.020 0.029 0.689 13.2 0.8 0.67 0.28 0.008 0.024 0.326 13.2 0.8 0.00 0.28 0.025 0.031 0.799 13.2 0.6 0.33 0.28 0.012 0.021 0.589 13.2 1.0 0.00 0.28 0.322 -- 18.0 0.8 0.33 0.28 0.025 0.039 0.647 AR capital gains tax advantage of Model versus Model 1, or the annual after-tax growth rate advantage due to capital gains after January, 1987. ARI annual after-tax return advantage attributable to favorable capital gains treatment before January, 1987. Definitions of i, g, p, tn, and appear in Table 1. vestment's rate of return (i). AR increases degree. reasonable estimate appears to to 2.50 percent when equals 18 percent be, however, that capital gains retain on (row J) compared with the 1.70 percent average 50 percent of their previous tax advantage when equals 13.2 percent (row advantage. B). Table compares the capital gains tax advantage (AR) under TRA with the cap- 2. Tax-Clientele CAPM ital gains tax advantage (AR') that ex- isted before TRA. AR' was computed as- Another approach to investigate the suming that 60 percent of the realized continued existence of tax-induced capital gains are excludable from taxable clientele effect is to estimate an extended income. The column AR/AR' reveals that tax-clientele CAPM for the period after capital gains usually retain over one-half January, 1987. Brennan (1970) was the of their prior tax advantage. It should be first to derive an after-tax CAPM that ac- noted that AR' values assume that all counted for both progressive tax scheme capital gains are long-term, which exag- and the differential taxation of dividends gerates the &R' values and underesti- and capital gains. The model was subse- mates the AR/AR' values by an unknown quently extended by Litzenberger and Page 5 No. 11 TAX-INDUCED CLIENTELE EFFECT 135 Ramaswamy (1979, 1980) and Ang and as historical data on variety of financial Peterson (1985). variables. Complete data sets were avail- The model we estimate is able for 314 industrial firms; financial companies and utilities were not consid- RPi ao a, P, Ot2 LEVI ered due to their unique regulatory en- vironment and evidence that investors may interpret dividend policy differently CL3 S, a, 0, Y@ ei (3) for such firms (Baker, Edelman and Far- relly (1985)). where RP, =- ex ante risk premium, Following Litzenberger and Ramas- defined as the before-tax wamy (1979) and Ang and Peterson (1985), expected return on se- the sample of dividend-paying securities curity less the risk-free were sorted on the basis of forecasted div- rate of return on U.S. idend yield into one of groups (i.e., 04 government securities ---, 08) with approximately equal markei (i.e., three-month U.S. values in each group. Multiple linear Treasury bill rates) ;4 regression analysis (ACLRA) was then used pi the market-adjusted beta to estimate the coefficient values. com- for security i; mon problem encountered in cross-sec- LEV, financial leverage, de- tional research using MLRA is the pres- fined as the ratio of cur- ence of heteroskedasticity, or non-constant rent debt, long-term residual variance. The Golffeld-Quandt debt, preferred stock, and (1965) test indicated the existence of this unfunded vested pen- condition in our data, and to correct for sion liability to the total this, the t-statisties reported herein were market value of security based on the heteroskedasticity-consis- i; tent covariance matrix estimator devel- S, firm size, defined as the oped by White (1980). natural logarithm of the The empirical results of the regression total market value of se- analysis are summarized in Table 3. The curity i; yield coefficients, U4 through ot8, follow Y, expected dividend yield descending pattern which implies the ex- stence of tax-induced clientele effect. less the risk-free rate; and The implied tax rate in this study for 0, series of binary vari- investors in low yield stocks (04) is 42 per- ables based on rank- cent, and descends almost monotonically ing of security accord- to an implied rate of 15 percent for high ing to the size of the yield stocks (08). The observed maximum expected dividend yield tax rate is in line with the maximum fed- (i.e., 04 (low yield), eral tax rate for 1987 of 38 1/2 percent. (high yield)). 08 Ang and Peterson (1985) report the range of implied tax rates for the period 1973 to The coefficients of the binary variables 1983 to be 0.72 to 0.07, whereas Litz- represent estimates of the marginal tax enberger and Ramaswamy (1979) found rates for investors in low yield to high yield the implied tax rates to vary from 0.63 to securities. If tax clientele effect exists 0.05 for the period 1940 to 1980. The 42 after January, 1987, the coefficients (X4 percent implied tax rate on low yield se- through cL8 should be positive in sign, but curities from this study is lower than the descending in value. corresponding rates observed in earlier Data to estimate equation (3) were ob- studies, which accords with the sharply tained from The Value Line Investment lower maximum tax rates under TRA. Survey data base, which contains current As final step in our analysis, we es- forecasts (i.e., as of April, 1987) of divi- timated an after-tax CAPM which pre- dends, earnings, and stock prices, as well dicted linear relationship between risk Page 6 136 NATIONAL TAX JOURNAL [VOI. XLI Table Regression Results of Tax Induced Clientele CAPM* RPi 0.057 0.007 Pi 0.005 LEVI 0.004 Si 0.420 04yi (0.91) (2.55) (-3.78) (4.05) 0.340 95yi 0.190 86Yi 0.040 87yi 0.150 ()8yi (2.76) (1.20) 0.17) (0.82) Adjusted R2 0.21 Degrees of freedom 304 t-statistics are presented in parentheses below the regression coefficients. premium and dividend yield. This model higher before-tax returns, for given level takes the same form as equation (3) ex- of risk, than low yield securities. cept that Y, replaces the binary yield The analyses in this study indicate that variables, 04yi through OsYi. The coeffi- these capital gains tax advantages are cient of the yield variable, using April sufficiently strong as to produce mean- 1987 Value Line Investment Survey fore- ingfully higher before-tax returns on high casts was found to be 0.29, which is sig- current yield securities. An analysis of nificant at the .02 level.' Ang and Pe- simulated data and ex ante data from The terson, who also used ex ante Value Line Value Line Investment Survey suggests data, found the average yield coefficient that the relationship between the before- to be 0.57 for the period 1973 to 1983. tax return and yield is approximately one- These coefficients suggest that high yield half as large as the relationship which securities continue to offer higher before- existed prior to TRA. Further, coefficient tax returns, but the strength of the clien- estimates from the extended tax-clientele tele effect is approximately one-half of its CAPM revealed the continued existence previous size. This result is simlilar to our of positive but nonlinear association be- earlier estimate that the strength of the tween security risk premium and divi- before-tax-yield-expected return relation- dend yield. ship is perhaps one-half as strong as it was before TRA and the elimination of the 60 percent long-term capital gains exclusion. FOOTNOTES 'The maximum Federal marginal tax rate begin 3. Conclusion ning in 1988 is currently scheduled to decrease to 33 percent; however, there is considerable Congressional Today's Tax Code grants capital gains discussion about raising the maximum Federal rate, two advantages compared to the tax and state taxes also raise individual rates. Thus, the 33 percent rate may actually underestimate the ex- treatment of current income (dividend or pected marginal rates for high tax bracket investors, interest): the tax deferral of unrealized in which case our analysis underestimates the iTn- capital gains and the ability to time gains portance of the capital gains tax advantages. for low tax rate year. This implies that 2The Value Line forecasts on June 26, 1987, called for median dividend yield of 2.5 percent and median tax clienteles should continue to exist and annual capital gains of 10.7 percent based on pro- that high yield securities should offer jected median 3-to-5 year price appreciation of 50 Page 7 No. 11 TAX-INDUCED CLIENTELE EFFECT 137 percent, interpreted to occur in four years (or 'The signs and significance levels of 0,, LEV,, and (1.50)**.25 0.107). 8, are similar to those reported in Table 3. 'The models assume an initial $1 investment and no initial unrealized capital gains in the portfblio. The average annual after-tax return after year is con- REFERENCES sidered to allow some accumulation of unrealized cap- Ang, J. S. and D. R. Peterson, "Return, Risk, and Yield: ital gains. This average is more representative of Evidence from Ex Ante Data," Journal of Finance, steady-state after-tax return. (June 1985), 537-648. 4Expected return for security was measured using Baker, H., G. Farrelly, and R. Edelman, "A Survey normalized earnings-price model, where normalized on Management Views of Dividend Policy," Fi earnings (NE,) was weighted average of near-term nancial Management, (Autumn 1985), 78 84. and distant earnings-per-share projections: Brennan, M. J., "Taxes, Market Valuation and Cor- porate Financial Policy," National Tax Journal, NE, Wi, EPSI, (1 W,) EPS, (December 1970), 417-427. Doyle, R. J., "IRAs and the Capital Gains Tax Effect," The near-term (EPSI) and distant (EPS3 5) earnings Financial Analysts Journal, (May-June 1984), 60- projections are Value Line projections of EPS for the 66. year ending months ahead and 3-to-5 year earn- Goldfeld, S. M. and R. F. Quandt, "Some Tests for ings, respectively. The weighting factors (i.e., W, and Homoskedasticity," Journal of the American Sta- (1 Wl)) were based on Value Line's ranking of the tistical Association, (September 1965), 539-547. difficulty of predicting earnings; firms whose earn- Gordon, M. J., "Optimal Investment and Financing ings are relatively easy (hard) to predict receive most Policy," Journal of Finance, 18 (1963), 264 272. of the weight on the distant (near-term) earnings; Litzenberger, R. H. and K. Ramaswamy, "The Effect hence, of Personal Taxes and Dividends on Capital Asset WI, (2 RANK,)/10 Prices: Theory and Empirical Evidence," Journal of Financial Economics, (1979), 163-195. to Litzenberger, R. H. and K. Ramaswamy, "Dividends, where RANK, is the Value Line ranking (I [easy] Short Selling Restrictions, Tax-Induced Investor [difficult)) of the difficulty of forecasting firm's Clienteles and Market Equilibrium," Journal Fi- earnings. The corresponding model of expected return nance, (May 1980), 469-482. Of (EIR,) was: Miller, M. and F. Modighani, 'Mvidend Policy Growth NEI and the Valuation of Shares," Journal of Business, El@ (1961), 411-433. PO@ White, H., "A Heteroskedasticity-Corisistent Covari ance Estimator and Direct Test for Heteroske- where Po is the current per share price of security i. dasticity," Econormtrica, (May 1980), 817-838.