UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-C ;i>AIGN BOOKSTACK§ ec . -2-3 Digitized by the Internet Archive in 2011 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/analyticalempiri953leec b.u \ \ ■a £A ^y\£ FACULTY WORKING PAPER NO. 953 An Analytical and Empirical Comparison of Alternative Cost of Equity Capital Estimation Methods Cheng F. Lee Charles M. Linke J. Kenton Zumwalt College cf Commerce and Business ^dminis Bureau of Economic and Business Research University cf Illinois. Urbana-Chamoaign BEBR FACULTY WORKING PAPER NO. 953 College of Commerce and Business Administration University of Illinois at Urbana-Champaign April 1983 An Analytical and Empirical Comparison of Alternative Cost of Equity Capital Estimation Methods Cheng F. Lee, Professor Department of Finance Charles M. Linke, Professor Department of Finance J. Kenton Zumwalt, Associate Professor Department of Finance Abstract Cost of equity capital estimation is an important issue for both regulated and unregulated firms. There are several methods for estimating the cost of equity capital, including: (1) discounted cash-flow methods; (2) capital asset pricing model methods; (3) M&M's cross-sectional method; and (4) integrated methods. This paper reviews the alternative estimation methods and empirically compares three of the most frequently used methods. The results indicate there are no consistent relationships among the three methods analyzed and the results vary across different industries. I. INTRODUCTION Cost of equity capital estimation is one of the most important issues in public utility regulation and in capital budgeting decisions. The concept of the cost of capital presents no particular difficulty under conditions of perfect certainty — the assumption on which most of classical theory has been developed. It is simply the market rate of interest. Under conditions of uncertainty, the cost of capital is no longer directly observable and several methods have been developed to estimate the cost of equity capital. These include: (1) the earnings yield method, (2) the discounted cash-flow (DCF) model, (3) the capital asset pricing model (CAPM) method; (4) M&M's [13] cross-sectional esti- mation method; and (5) integrated methods. The integrated method com- bines the CAPM with M&M's proposition III in estimating the cost of equity capital. Most recently Glenn and Litzenberger [3] have integrated M&M's cross-sectional method and the CAPM method to propose "an interindustry approach for cost of capital estimation." The main purposes of this paper are to review the alternative cost of equity capital estimation methods and to empirically compare three of the most frequently used methods. Possible biases related to the application of the CAPM for different industries will also be discussed. In the second section, alternative cost of equity capital estimation methods will be reviewed. In the third section empirical data from both the utility industry and three non-utility industries will be used to estimate empirically the cost of capital using the three most popular alternative methods. Finally, some guidelines for choosing amonp, the alternative cost of capital estimation methods are established in accor- dance with a compromise of theory and practice. -2- II. ALTERNATIVE COST OF EQUITY CAPITAL ESTIMATION METHODS: A REVIEW There are two basic approaches for estimating the cost of equity capital — the discounted cash flow method which is based on the firm's operating and financial characteristics, and the capital asset pricing model method which is based on the relationship between a firm's expected returns and the returns of a market index. There are several alternative specifications under each of these two basic approaches, and each par- ticular specification has its own set of underlying assumptions. This section develops the different models, summarizes the assumptions of each model, and describes the estimation problems associated with each model. Discounted Cash Flow Methods The discounted cash flow (DCF) method assumes the market price of an asset, here a common stock, is equal to the present value of expected future cash flows to the suppliers of equity capital, the stockholders. Since cash flows to stockholders come in the form of dividends, the DCF model is D D D P n = — + 1 —~ + . . . + = (1) u i + k a + k r (1 + K )" e e e or 00 d P n = E (2) t u t=i a + k r e where P n is the current observable market price of the stock, D , ..., D ot are the expected future (unobservable) dividends, and k is the rate of return necessary to equate expected dividends to the share price and, hence, is the required rate of return or cost of equity capital. Obviously, the crucial problem is the estimation of future dividends and several -3- alternative specifications have been developed in attempting to simplify the estimation problem. The earnings-price ratio or earnings yield model simplifies the dividend estimation problem by assuming a 100% dividend payout and no growth in earnings. In this model dividends equal earnings and the earnings constitute a perpetuity. That is: E E E °° 1 + k (1 + k ) z (1 + k ) e e e which may be simplified to: e or k e = f- . " (5) Here, k is the return required on a perpetuity and is the inverse of the familiar price-earnings ratio reported in the financial press. The major problems associated with using the earnings yield as a cost of equity capital measure are the omission of possible earnings growth and dividend policy considerations. It should also be noted that the P/E ratio reported in the financial press is generally based on historical earnings, not expected earnings. A second DCF specification is the dividend growth model . In this model the dividends in equation (1) are assumed to follow a particular pattern. A pattern frequently used by institutional investors involves dividing the expected dividend stream into three periods based on dif- ferent growth expectations. This results in: -4- m D (1 + g ) t n D(l + g ) t_m » Dd + g-)*" 11 P Q = E-2 i- + i JS L_ + E _S L_ (6) t-1 (1 + k ) L t=m+l (1 + k ) t=n+l (1 + k ) c e e e where g. is the expected growth in dividends for the first m years, g_ is the expected growth for the second n years and g,, is the growth rate assumed to be constant in perpetuity. Using this specification, the dividends can easily be estimated for the first two terms and the D n+1 last term approaches. r With this information and the k e" 8 3 current market price, the expected rate of return can be calculated. A similar growth model developed by the Wells Fargo Bank includes estimation of dividends and earnings for the next five years, the ex- pected earnings growth rate and payout ratio for the sixth year, the time when a constant growth rate is expected, the amount of the constant growth rate, and the expected payout ratio when the firm reaches its constant growth. Using this information, dividends are estimated up to the time the firm is expected to achieve its constant growth; the value of the constant growth term is determined; then the expected dividends and constant growth value is used with the current price to determine the stock's expected return. Probably the most widely used specification of equation (1) is: °° D n (l + g) t P = I -2 . (7) u t=i (i + k r e The crucial assumptions for this model are that the dividend growth rate, g, will be constant forever and that k is greater than g. This model also e generally assumes a firm's earnings are growing at the rate g and that the payout ratio and P/E ratio are constant. This allows the model to be simplified to: -5- D l P o = IT^T (8) or D l k - jr + g (9) where D is the expected dividend, P_ is the current stock price, and g is the expected growth in earnings /dividends. The obvious difficulty faced in applying any DCF model involves the estimation of future dividends or the expected growth rate(s) in future dividends. Whether the analyst is using a specification such as the Wells Fargo model which requires actual dividend estimates for a number of years or a specification such as the constant growth model which re- quires an estimate of the constant growth term, assumptions must be made regarding the general economic outlook and the outlook for the markets in which the firm competes. M&M's Cross-Sectional Model M&M's proposition 2 deals with the cost of capital determination. In their 1966 AER article, they extended their proposition 1 with taxes and growth potential and proposed a cross-sectional cost of capital model Following M&M [14], proposition 1 (with taxes but without growth) can be defined as V = V + xD (10) J-j u where V T and V are market values for the levered and the unlevered L u firm, respectively; x is the marginal tax rate and D is the total debt of the firm. By allowing for growth potential, M&M [13] suggest a cross- sectional model to estimate the cost of capital as -6- V L " TD 1 X(l - t) AA , ,,„ N — IT" =a 0l +a l A - + Q 2 "A + U (11) where X is the expected earnings, A is total assets, and AA/A is the average growth rate. X is generally not observable and the current earnings are used to replace the expected earnings. M&M suggest the proxy error problem arising from the use of X instead of X may be alleviated by using, m Y = E r.Z. + W (12) i-1 X X where Y = , and the Z.'s refer to a firm's size, growth, debt, A 1 preferred stock, and dividends. If the estimated Y is used to replace xYl - t) : , then the inverse of the estimated a n is the estimated cost of A 1 capital. This model assumes all firms used in estimating equation 11 belong to a single business risk class. The cost of capital estimated by this procedure is an industry cost of capital estimate. This concept and method were extended by Litzenberger and Rao [10, 11] and Higgins [6]. Capital Asset Pricing Model The Sharpe [19]-Lintner [8] Capital Pricing Model CAPM predicts the relationship between betas and risk premiums to be: E(r.) = E(r )6., (13) J m 3 where E(r.) and E(r ) are the expected excess rates of return above the risk- J m less rate of interest for the j-th security and the market portfolio, respec- tively, and 6. is the beta of the j-th security as measured against the true market portfolio of all assets. The model assumes risk averse investors with homogeneous expectations, existence of a riskless asset, marketability of all assets and the absence of transactions costs and taxes. -7- When short selling of securities is limited, the relationship be- tween risk premiums and betas becomes: E(r ) = E(r m )6 j + E(r z )(l - S ), (14) where E(r ) is the risk premium associated with the minimum variance z zero beta portfolio. Recently, Litzenberger, Ramaswamy and Sosin [9] [LRS] discussed the possible estimation bias associated with the specification of the CAPM. One important source of bias results from using the traditional CAPM instead of the after-tax version of the CAPM. The after-tax version of the CAPM may be written as : E(r.) = E(r )6. + E(r')(l - 6.) + E(r.)(d. - S.d ), (15) 3 ™ 3 z 3 n Jjm where E(r f ) is the risk premium of a zero beta portfolio with a zero dividend yield, E(r, ) is the expected return on a hedged portfolio with a zero beta and a dividend yield of unity, and d. is the dividend vield. 3 Comparing this more generalized cost of equity capital model with the traditional model of equation (13) shows that the traditional CAPM has three possible biases; (1) a risk free rate bias and (2) a dividend yield bias and, (3) a beta bias. The alternative CAPM specifications as in equations (13) , (14) and (15) do not explicitly consider the impact of growth on the cost of equity capital estimation. Fewings [1975] and Gordon and Gould [1978] indicate that systematic risk is a positive function of the growth rate of cor- porate earnings. However, Myers and Turnbuli [1977] suggest systematic risk is negatively related to beta. Senbet and Thompson [1982] review -8- the controversy and conclude that the relationship between growth and systematic risk "depends on the way in which the response of cash flows to unanticipated changes in the economy changes with g." Myers [15], using real option theory, argued that the estimated beta in terms of the traditional CAPM is a positive function of the proportion of the stock's value accounted for by growth in the M&M sense, That is, if the market value of a growth firm can be decomposed into a perpetual component and a growth component, the hurdle rate obtained from the traditional CAPM will be an overestimate of the correct rate for any firm having valuable growth opportunities. Integrated Models The DCF and CAPM methods of cost of equity capital estimation have been used in determining the required rates of return in the electric utility industry. The DCF model of share valuation in partial equilib- rium emphasizes corporate characteristics such as growth while risk is consigned to a black box. The CAPM, on the other hand, provides a gen- eral equilibrium risk and return model in which everything else is con- signed to a black box, hence, some integrations of the alternative cost of capital estimation methods have been made. One method of combining M&M's formulation with the CAPM involves using the CAPM to allow for different levels of risk instead of assuming equivalent risk classes. As shown in Copeland and Weston [1, p. 293], M&M's after-tax cost of equity model is; k = p + (p - k,)(l - t )(B/S) (16) e d c -9- where p is the unlevered cost of equity, k, is the cost of debt, t d c is the corporate tax rate, and B/S is the ratio of the market value of debt to the market value of equity. Substituting E (R ) — R P-E(R)-R f +— H. £