U9_dq

Does Capital Structure Matter? Revisiting Modigliani

and Miller’s Empirical Work Using Latin American

and North American Data

Lucas Ayres B. de C. Barros Alexandre Di Miceli da Silveira

Rubens Famá

ABSTRACT. Modigliani and Miller (M-M) performed the first empiri- cal test of their capital structure irrelevance propositions in 1958. Their results confirmed the irrelevance hypothesis. In this paper, the M-M tests are replicated with contemporary data from Latin American and North American firms, but with two fundamental differences from M-M’s original work: (a) the CAPM is used for estimating the cost of a firm’s equity and (b) fuller specifications for the empirical models are adopted in order to improve their statistical quality. The results do not lend support to M-M’s 1958 model, but are in line with their corrected 1963 propositions.

Lucas Ayres B. de C. Barros, MSc, is a Phd Student, Faculdade de Economia, Administração e Contabilidade da Universidade de São Paulo (FEA/USP) Brazil (E-mail: [email protected]).

Alexandre Di Miceli da Silveira, MSc, is a PhD Student, Faculdade de Economia, Administração e Contabilidade da Universidade de São Paulo (FEA/USP) Brazil (E-mail: [email protected]).

Rubens Famá, PhD, MSc, is a Professor, Faculdade de Economia, Administração e Contabilidade da Universidade de São Paulo (FEA/USP) Brazil (E-mail: rfama@ usp.br).

Latin American Business Review, Vol. 5(3) 2004 http://www.haworthpress.com/web/LABR

© 2004 by The Haworth Press, Inc. All rights reserved. Digital Object Identifier: 10.1300/J140v05n03_03 43

RESUMEN. En 1958, Modigliani y Miller (M-M) realizaron el primer ensayo empírico sobre la irrelevancia de las propuestas sobre la estructura de capital. Sus resultados confirman la hipótesis de su poca importancia. En este estudio, los ensayos M-M se reprodujeron utilizando los datos con- temporáneos obtenidos de empresas latinoamericanas y norteamericanas usando, sin embargo, dos diferencias esenciales del trabajo original realizado por M-M: (a) el CAPM se utiliza para estimar el costo del capital social de una empresa y (b) para mejorar la calidad estadística se han adoptado especificaciones más amplias para los modelos empíricos. Los resultados no respaldan el modelo M-M de 1958, pero se encuentran alineados con las propuestas corregidas en 1963.

RESUMO: Modigliani e Miller (M-M) apresentaram o primeiro ensaio empírico de suas proposições sobre a irrelevância da estrutura do capital em 1958. Os resultados confirmaram a irrelevância da hipótese. Neste estudo, os ensaios de M-M são reproduzidos com dados contem- porâneos das empresas da América Latina e da América do Norte, mas com duas diferenças fundamentais do seu trabalho original: (a) o CAPM é utilizado para estimar o custo do capital social de uma empresa e (b) especificações mais completas dos modelos empíricos são adotadas para melhorar a sua qualidade estatística. Os resultados não sustentam o modelo de M-M de 1958, mas estão de acordo com as suas proposições revistas de 1963. [Article copies available for a fee from The Haworth Document Delivery Service: 1-800-HAWORTH. E-mail address: <docdelivery@haworthpress. com> Website: <http://www.HaworthPress.com> © 2004 by The Haworth Press, Inc. All rights reserved.]

KEYWORDS. Capital structure, irrelevance proposition, Modigliani and Miller, Latin America DataIntroduction

INTRODUCTION

Does the way in which a firm is financed somehow influence its value? Should the answer be negative, then why do different organiza- tions choose different capital structures? If the answer is positive, then how, in what direction and to what extent do the proportions of debt and equity financing affect the value of a firm? Furthermore, how can the existence of such varied types of available financing instruments be jus- tified?

44 LATIN AMERICAN BUSINESS REVIEW

It is difficult to establish when the scientific controversies addressing these and other issues related to the problem of capital structure began. Possibly, the first effort to deal with the subject in a more analytically rigorous manner may be attributed to Durand (1952). Nevertheless, the foremost work in the field was presented six years later by Modigliani and Miller (1958), henceforth referred to as M-M. In it the authors pro- posed that, under certain conditions, the degree of a firm’s leverage1 was irrelevant in determining its market value. In other words, a firm should continue to have the same average cost of capital whether it maintained a structure of high leverage, or one of relatively low lever- age. Since the release of their seminal work, M-M’s propositions have motivated abundant academic debate, contributing to further interest on the subject.

The empirical and theoretic works that have followed the original article by M-M have approached the subject from various different an- gles and their conclusions and propositions are somewhat heteroge- neous. Some more or less confirm the hypothesis of indifference of capital structure. Others reject it fairly strongly. Among those who re- ject the hypothesis of indifference, there is no consensus as to the spe- cific direction of the relationship between capital structure and the value of firms. Harris and Raviv (1991) presented a survey of the main theoretical trends that have emerged since the early works, and of the empirical evidence produced on the subject, showing the panorama of its diversity.

The first test of M-M’s propositions was presented by the authors themselves in their work of 1958, using data from American electric util- ity and oil companies. Their results showed that, at the time, there was no evidence of a significant relation between the average cost of a firm’s capital and its degree of leverage, confirming the indifference of capital structure hypothesis. They also confirmed that the cost of a firm’s equity increased linearly as leverage increased, another prediction of their model.

However, the M-M tests were affected by limitations of various types. Most of them, as the authors themselves acknowledge (Modigliani and Miller, 1958, pg. 282), were possibly related to the measurement of the cost of equity and the average cost of a firm’s capital. In 1958, there was no model from which theoretically acceptable estimates of the cost of equity could de drawn. A theory of market equilibrium that permitted calculating such a variable only emerged years later with Sharpe (1964) and Linter (1965) and their Capital Asset Pricing Model (CAPM). Although also fraught with controversy, the CAPM remains the model most utilized on

Barros, Silveira, and Famá 45

capital markets for calculating the return demanded by the shareholders of a firm in order to compensate them for the risk of their investment.

Would the results reported by M-M be the same if advancements in the theory of finance, such as the CAPM, had been available in 1958? Would the authors’ conclusions be the same if, moreover, the specifi- cations of the statistical models they used had been different? Above all, would their findings have been corroborated by current data? Un- fortunately, the first two questions cannot be answered. One of the rea- sons why the CAPM is today considered one of the most suitable models for estimating the cost of firms’ equity is precisely the fact that it is broadly accepted and used by investors. Thus, one should not ex- pect that the CAPM would bring about the best assessment of the re- turn demanded by the shareholders of a firm, prior to its formulation. It is possible, however, to answer the last question, and that is the pur- pose of this work. Specifically, an empirical study similar to that pre- sented by M-M in 1958 is conducted on firms of the electric and petroleum industries in the USA and in Latin America, using data of 2000. With regard to the original work, two basic differences are note- worthy: (a) the CAPM is used to calculate the cost of each firm’s eq- uity and (b) fuller specifications are adopted for the empirical model, aiming at improving its statistical quality and therefore reaching more precise conclusions.

EARLY APPROACHES

“Traditionalists” and M-M

According to the point of view sometimes referred to as “tradition- alist,” firms can and must seek to achieve an optimal combination of debt and equity capital with a view to maximizing their market value. Value maximization takes place through the minimization of the over- all cost of the capital used by a firm to finance its activities. Durand (1952) was one of the pioneers to investigate these possibilities. Ac- cording to the author, if investors agree on a method for pricing the firm based on its expected future cash flow discounted to present value, it should be possible, keeping the expected cash flow constant, to increase the firm’s value through the reduction of the discount rate; that is to say, the opportunity cost of the capital employed. Durand (1952), however, admits that it will not necessarily be possible to re-

46 LATIN AMERICAN BUSINESS REVIEW

duce the cost of capital by altering the proportions of equity and debt in the firms’ liability.

As a rule, debt should be cheaper than equity capital because the for- mer entails a contractual obligation for payments by a firm, whereas the latter represents a residual right on its cash flow. However, an increase of relative indebtedness (leverage) will not always bring about a reduc- tion of the weighted average cost of capital (WACC).2 If the firm is al- ready highly leveraged, an increase in its ratio of debt to total capital may significantly increase the risk of insolvency. Consequently, inter- est rates for new loans may rise. The risk associated with possible finan- cial difficulties will also affect the shareholders, contributing to an increase in the cost of equity. These two effects combined may cause the WACC to remain constant or even to increase, despite the relative increment in debt financing. Nevertheless, the “traditionalist” position states that the cost of debt will remain approximately unchanged for “moderate” levels of leverage, understanding “moderate” as a level of indebtedness that does not jeopardize a firm’s ability to honor its con- tractual obligations. Likewise, the cost of equity should also be insensi- tive to “reasonable” variations in financial leverage. Thus, a firm should seek more debt financing up to the point at which its WACC reaches its minimum. These and other issues are discussed by Durand (1952).

Modigliani and Miller (1958) contest the “traditionalist” outlook. In a work that became the hallmark for the study of the subject, the authors proposed that capital structure is irrelevant for determining the value of a firm if certain restrictions are met. Although this possibility had al- ready been presented by Durand (1952), M-M were the first to formally describe the mechanism through which the indifference was assured in a context of partial market equilibrium. Since their original work, an im- pressive amount of research in the field of capital structure has contrib- uted to a better understanding of the related phenomena. Controversies involving disagreement between the “traditionalists” and M-M have lasted for many years and are reflected in works such as those of Durand (1959), Modigliani and Miller (1959), Weston (1963), Solomon (1963), Boness (1964), Brewer and Michaelsen (1965) and Modigliani and Miller (1965). These and other related works may be found in Archer and D’ambrosio (1967).

The Propositions of Modigliani and Miller

The 1958 work by M-M is based on the formulation and demonstra- tion of three propositions regarding the relationship between capital

Barros, Silveira, and Famá 47

structure and a firm’s value, as well as between capital structure and a firm’s investment decisions. Their Proposition I was formulated as fol- lows (Modigliani and Miller, 1958, pg. 268):

. . . the market value of any firm is independent of its capital struc- ture and is given by capitalizing its expected return at the rate r

k

appropriate to its class.

In other words, Proposition I states that the proportions of debt and equity as sources of financing are completely irrelevant for determining a firm’s market value. That happens because different combinations of the distinct types of financing instruments will not alter the total or aver- age cost of the capital used by the firm. Using M-M’s notation, the total market value of the jth firm (Vj) belonging to class k will be given by (1), where X j corresponds to the expected return on the assets owned by the company, and rk is the appropriate discount rate. Thus, rk corresponds to the average cost of capital and is the rate that compensates all inves- tors for the risks to which they are exposed. The consideration of risk is implicit in the concept of class formulated by the authors; that is to say, all firms belonging to a given class, k, have the same level of risk. Finally, rk is equivalent to the return demanded by the shareholders of a firm that is not leveraged (not financed with any debt).

V X

j j

k

= r

(1)

The equation shown in (1) calculates the present value of the ex- pected, perpetual cash flow produced by a firm with zero growth. The supposition of no growth is a simplification used to facilitate the exposi- tion and is not a binding restriction.

Proposition I by M-M is demonstrated using a no arbitrage argument in markets where:

• all debts are risk free, • individuals can borrow and lend at the risk free rate and • there are no transaction costs.

Under such conditions, the authors demonstrate that any investor can reproduce the leverage of a firm by making a personal loan. He could also undo it by purchasing debt securities. One way or another, the in-

48 LATIN AMERICAN BUSINESS REVIEW

vestor will have an opportunity for arbitrage, meaning immediate and risk free profits, whenever a non-leveraged firm presents a different market value from that of a leveraged one, as long as both have the same expected cash flow X and belong to the same class. In an efficient mar- ket, the possibility of arbitrage should force the parity of values, render- ing the proportions of equity and debt irrelevant for determining the total value of any firm within a given class k. That is justified because the WACC will be constant and equal to rk (which is also the cost of eq- uity of a non-leveraged firm) independently of the relative amount of debt and number of shares issued by the firm.

Implicitly or explicitly, M-M utilized in their work various other as- sumptions, some of them quite restrictive. Among these, the model as- sumes that (Copeland and Weston, 1988):

• there are no bankruptcy costs, • only two types of claims are issued by firms: risk free debt and

(at-risk) equity, • all firms belong to the same risk class, • there is no information asymmetry between individuals inside and

outside of the firm (investors and managers, for example), • managers always seek to maximize shareholders’ wealth (there are

no agency costs), and • there are no taxes.

Certainly, most of these assumptions are unrealistic. However, some of them may be relaxed without materially changing the main findings. Rubinstein (1973), for instance, shows that the presence of risky debt leaves the original results totally unchanged.

Stiglitz (1969) demonstrates the M-M theorems using other argu- ments. The author works in a state-preference context and develops a general equilibrium analysis to show that the M-M results are more stal- wart than was previously thought. According to Stiglitz (1969), validity of the original results does not, for example, rest upon the existence of “risk classes,” nor on the degree of competition in capital markets. It is also independent from the agreement of all individuals about the proba- bility distribution of future returns (homogeneous expectations).

Other assumptions, explicit or implicit in the original formulation, nevertheless remain intact as significant limitations of M-M’s model. An example is the assumption of no taxation of a firm’s profits. Inclusion of a corporate income tax was, however, carried out in the original M-M work

Barros, Silveira, and Famá 49

of 1958. In it, the authors assert that the same results suggesting the irrele- vance of capital structure are attained. Later, however (Modigliani and Miller, 1963), they corrected their reasoning, pointing to an error in the original work and proposing a new formulation when a corporate tax rate greater than zero exists. That correction is outlined below.

If VU corresponds to the value of a non leveraged firm (with no debt in its capital structure), VL is the value of a leveraged firm, tc is the cor- porate income tax rate, and D is the market value of the firm’s perma- nent debts, then it can be shown that (Modigliani and Miller, 1963).

V V DL U c= + τ (2)

The relation in (2) indicates that the value of a leveraged firm is equal to the value of a non leveraged firm, plus the present value of the fiscal benefit, or tax shield, provided by debt and represented by tc D (Cope- land and Weston, 1988). It may be noted that VL = VU if tc= 0.

Stated differently, in the absence of a corporate income tax, Proposi- tion I remains valid and the financing structure of the firm will still be ir- relevant for determining its value. The importance of tc in the analysis comes from the fiscal deductibility of the interest paid as a service of the firm’s debts. Thus, the greater the leverage, the smaller will be the amount of income tax to be paid for the same profit before taxes, evi- dencing a considerable benefit provided by debt financing.

As previously mentioned, Proposition I implies that WACC = r(from hereon the underwriting k will no longer be used) in the absence of taxes. However, in the presence of t

c π0, this formulation must be

changed to:

WACC i D

D S c d= − − +

r rτ ( ) (3)

It must be noted that in (3), given that r is greater than id (this should be true because equity is riskier than debt), the weighted average cost of capital will drop with the increase of leverage because of the tax shield associated with debt.

M-M’s Proposition II, also formulated in their work of 1958, makes explicit the mechanism by which the weighted average cost of capital remains constant independently of the proportions of D and S. If, as it is reasonable to admit, id < is, then, at first sight, the WACC should

50 LATIN AMERICAN BUSINESS REVIEW

drop with the use of relatively more debt financing, at least within “moderate” levels of leverage that would not jeopardize the firm’s ability to remain solvent. This is precisely the “traditionalist” point of view. In opposition to this idea, M-M show that any increase in finan- cial leverage translates into a higher risk for the firm’s shareholders. The perception of increased riskiness will eventually drive up the cost of equity (is). Therefore, two forces will act simultaneously whenever D/(D + S) increases: on the one hand, the WACC will decrease be- cause id < is; on the other hand, the WACC will increase because is will rise due to the added risk. According to M-M, the interaction of these two opposite movements results in a null effect upon the WACC, which will remain equal to r. The expression in (4) shows how is var- ies along with D/S (at different times the authors use D/(D + S) or D/S as the measure of leverage).

i i D

S s d= + −r r( ) (4)

Looking at (4), one can note that in the absence of financial lever- age (D = 0), is = r. is will thus grow linearly with an increase of D/S, and the WACC will remain constant. Such relations are shown in Figure 1.

In their article of 1963, M-M show how the cost of equity relates to fi- nancial leverage in the presence of a corporate income tax rate greater than zero. In this case, the increase in is resulting from the increase in

Barros, Silveira, and Famá 51

%

i = + ( – i )s dr r D S

r id

WACC = r D S

FIGURE 1. WACC and Cost of Equity in the Absence of Taxes

Source: Adapted from Copeland and Weston (1988, pg. 250).

D/S will be smaller than in the previous case, shown in (4), due to the tax shield generated by debt. This relation is presented in (5).

i i D

S s c d= + − −r r( )( )1 τ (5)

Variation of the WACC and of is as leverage increases is shown in Figure 2.

To summarize, the original propositions of M-M set forth the irrele- vance of capital structure for determining the value of a firm, as long as certain restrictive conditions defined by them or implicit in their formu- lation apply, at least in approximate terms. In 1963, correcting their first work, the same authors showed that in the presence of a corporate in- come tax, the proposition of irrelevance was no longer valid. Because of the fiscal benefit provided by debt, an increase in leverage will lead to a decrease in the weighted average cost of capital, consequently increas- ing a firm’s value ceteris paribus.

Figure 2 reveals that the WACC will always decay with an increase of leverage until the limiting value of r+tc(id-r) when D – •. In this

52 LATIN AMERICAN BUSINESS REVIEW

%

= + – –i (1 )( i )s c dr t r D S

p WACC ( i )= – –r t rc d D

D + S (i )r t+ –c d r

i (1 )d c– t

D

S

FIGURE 2. WACC and Cost of Equity When tc > 0

Source: Adapted from Copeland and Weston (1988, pg. 250).

case, the apparent recommendation for a manager is: put the firm con- tinually into more debt. However, intuitively it does not seem logical that any firm would increase leverage up to the point where all of its fi- nancing came from creditors. Thus, the question is how to conciliate in- tuition with the propositions of M-M.

OTHER APPROACHES AND NEW DEVELOPMENTS

Some answers to the enigma of capital structure appeared as theoreti- cal models in which important restrictive assumptions of the M-M model were relaxed. Miller (1977) himself proposed an alternative model taking into account not only the corporate income tax, but also the income tax levied upon individual investors (creditors and share- holders). His results indicate that the tax shield effect induced by debt may be significantly smaller than the one found by M-M in 1963. An- other line of research focuses on expected bankruptcy costs, which will increase following any increase in leverage, thereby restricting the ben- efit associated with debt financing and permitting the existence of an optimal combination of equity and debt. The works of Baxter (1967), Warner (1977), Altman (1984) and Weiss (1990), among others, follow this trade-off approach.

Harris and Raviv (1991) classify the more recent contributions on the subject of capital structure into four major categories. The first lists the works based upon so-called “agency costs,” focusing on the conflicts of interest between creditors and shareholders/managers and between shareholders and managers of a firm. Noteworthy in this field are the works of Myers (1977), Jensen (1986) and Stulz (1990). The second category includes the studies inspired by the notion of informational asymmetry among agents who are inside in contrast with those who are outside the organization, and encompasses studies on signaling. The works of Ross (1977), Myers and Majluf (1984) and Myers (1984) are good examples of that approach. The third major category includes the- ories based on product/input market interactions, approaching the choice of a firm’s capital structure as part of its marketing and competitive strategies, or as a decision that intrinsically depends on the characteris- tics of its products/inputs. Important contributions to this stream of analysis were offered by Titman (1984), Balakrishnan and Fox (1993) and Maksimovic (1986), among others. Finally, the authors survey the theories driven by corporate control considerations, involving the im- plications for a capital structure of voting rights and the possibility of

Barros, Silveira, and Famá 53

hostile takeovers. Along this line, two of the more important works are attributed to Stulz (1988) and Harris and Raviv (1990).

Other approaches, besides those mentioned by Harris and Raviv (1991), continue to contribute to the understanding of the subject. For instance, one recent and promising line of research explores the tendency of managers to time the market by, for example, issuing shares when prices are high and repurchasing when prices are low, exploiting what is sometimes referred to as “windows of opportu- nity” in capital markets. One of the milestones in this field is the work of Baker and Wurgler (2002). Another promising approach models capital structure decisions by assuming that managers are excessively optimistic by nature, which leads them to overvalue the risky securities issued by the firm, resulting in underinvestment or overinvestment problems (see Heaton, 2002).

EMPIRICAL STUDY

The M-M test

In performing an empirical test of their initial propositions, Modigliani and Miller (1958) used American data for the years 1947 and 19483 from 43 electric utilities and data for the year 1953 from 42 oil companies.

First, their procedure consisted in constructing a simple regression model with the firm’s average cost of capital as the dependent variable and its degree of leverage as the independent variable. The model was estimated by ordinary least squares (OLS). Based on (1), M-M define the average cost of capital as

rk X

V =

τ

(6)

Where V is the total market value of a firm belonging to class4 k. V corresponds to the sum of the market value of the firm’s debts (D) with the market value of its common shares (S). For the oil compa- nies, D included all long term as well as some short term liabilities. For the electric utilities, only long term debt was considered. Ac- cording to the original definition, X τ represented the expected total income net of taxes generated by the firm. However, this information was not directly available. As a proxy, M-M used the average value

54 LATIN AMERICAN BUSINESS REVIEW

of actual returns net of taxes from 1947 and 1948 for the electric util- ities. Actual net returns from 1953 were used for the oil companies. Net returns was defined as the sum of interest, preferred dividends and stockholders’ income after the deduction of corporate income taxes. Leverage was operationally defined as the ratio D/(D + S) = D/V. Designating the average cost of capital as x and leverage as d, both in percentages, M-M found the following results (standard er- rors of the coefficients in parentheses):

Electric Utilities x = 5. 3 + .006d (.008)

Oil Companies x = 8. 5 + .006d (.024)

These results lend support to Proposition I. In both cases, it is not pos- sible to reject the null hypothesis that the coefficient of the independent variable is zero at usual significance levels (the t-statistics for the electric utilities and oil companies are respectively 0.75 and 0.25, and their asso- ciated p-values are 0.46 and 0.8). In short, the data does not show any im- portant relationship between leverage and the average cost of the firm’s capital. Furthermore, it can be noted that the signs of the slope coeffi- cients are positive, contrary to what would be expected by the “tradition- alist” hypothesis or even by M-M’s corrected model of 1963.

To test Proposition II, M-M estimated a regression with the cost of the firm’s equity as the dependent variable and leverage as the inde- pendent variable. According to (4), leverage should now be defined as the ratio D/S. The cost of equity, represented by is in (4), was defined as

i S

s = π τ

(7)

where π τ is the expected net income accruing to the shareholders, and S is the market value of the shares. Just as X τ τπ, is not directly observ- able. As an approximation, the authors used arithmetic averages of ac- tual net income available to shareholders reported in 1947 and 1948 for the electric utilities, and in 1952 and 1953 for the oil companies.5

Letting z represent is, as defined in (7), and h represent D/S, both in percentages, M-M obtained (standard error of the coefficients in paren- theses):

Barros, Silveira, and Famá 55

Electric Utilities z = 6.6 + .017h (.004)

Oil Companies z = 8.9 + .051h (.012)

The t-statistics for the electric utilities and oil companies both have the same value of 4.25, with a p-value of approximately zero. Such results point towards the existence of a highly significant relationship between the cost of equity and leverage. Just as was predicted in their model, M-M veri- fied that an increase in leverage would be followed by an increase in the cost of equity. The linear relationships found by M-M are similar to those graphically portrayed in Figure 1, adequately supporting their 1958 model.

As shown in Figure 1, the propositions of M-M stated that no curvi- linear relationship between the average cost of capital and the degree of leverage should exist, contrary to the “traditionalist” approach. To ver- ify that, the authors specified a third empirical model, adding a qua- dratic term to the first regression equation. Its coefficient, however, proved insignificant.

Test of the M-M Propositions with Contemporary Data

One of the most severe limitations of the test procedure used by M-M refers to the operational definition of the variables. Notably question- able are the proxies used for the average cost of capital defined in (6) and the cost of equity defined in (7). In 1958, there was no theoretical model upon which one could acceptably base estimate rk and is. An equilibrium model that was easily applicable would only emerge with the works of Sharpe (1964) and Lintner (1965), introducing the Capital Asset Pricing Model–CAPM. In its original formulation, the CAPM de- fines the cost of equity as

i i i i ßs f M f= [ – ] (8)

where is is the equilibrium cost of equity, or the return expected by the firm’s shareholders, if is the return provided by the risk free bond, im corre- sponds to the expected return of the market portfolio and b measures the systematic risk associated with the firm. Comparing (8) to (7), it might be questioned if the M-M model and the CAPM are compatible, given that in the original M-M formulation there was no consideration whatsoever about

56 LATIN AMERICAN BUSINESS REVIEW

systematic risk and its role in estimating is. Rubinstein (1973) shows that the two theories are perfectly compatible. The author demonstrates, for in- stance, the mechanism through which b increases with leverage and thus leads to a higher cost of equity. Therefore, Proposition II, as stated by M-M, is still valid within the CAPM framework.

If the M-M and CAPM formulations for is are equivalent, it is possible to replace the definition in (7) by the one presented in (8). Likewise, the definition of rk shown in (6) is replaced by the weighted average cost of capital–WACC (see endnote number 2). In calculating the WACC, we estimate is using the CAPM and id is given by the observed yield to matu- rity of the firms’ long term debt securities.

Data and Empirical Model

We use data from 68 American and 33 Latin-American electric utili- ties, and 93 American and 16 Latin-American oil and gas producers for the year 2000. The data is available in the BLOOMBERG database and the analysis follows the procedures described by M-M in their 1958 work, based on OLS regression estimates.

Initially, to test Proposition I, we construct an empirical model with ex- actly the same specification as the one built by M-M and shown in section The M-M Test, Thus, the average cost of capital appears as the dependent variable and the firm’s leverage is the only independent variable. However, differently from M-M, the dependent variable is defined as the WACC and the cost of the equity is estimated using the CAPM. We make use of the WACC and the is values for each firm provided by BLOOMBERG, as these estimates are widely used by investors. We define leverage as D/V, including in D all long term debt as well as preferred shares.

With this initial specification and using the same notation as M-M, the following results appear (t-statistics in parentheses):

American Electric Utilities Latin-American Electric

Utilities

.059+.006x = d =.094–.027

(.666

x d

) (–2.44)

American Oil and Gas Producers Latin-American Oil and Gas

Producers

=.077–.002x d =.091–.05x d

(–.43) (–2.42)

Barros, Silveira, and Famá 57

The above results still support M-M’s 1958 indifference hypothesis for the American firms of both industries. The p-values of the slope coefficients are 0.5 and 0.67 for the electric utilities and oil and gas producers, respectively. However, for the Latin-American firms the M-M results are not maintained. The slope coefficients are statisti- cally different from zero at the 5% significance level in both sectors. The signs of the coefficients are also reversed and a strong negative re- lation between financial leverage and average cost of capital is re- vealed, thus opposing Proposition I. On the other hand, considering that these firms are subject to high corporate income tax rates, the esti- mates seem to corroborate the conclusions of the M-M work of 1963, confirming the importance of the tax shield induced by debt. How- ever, with the exception of the estimates for the Latin-American oil and gas producers, all regressions exhibit poor statistical quality, as the rejection of the hypotheses of normality and homoscedasticity of the error terms suggest.6 This was especially true for the models in- volving American firms. One could also strongly suspect that the error terms are not uncorrelated with the regressor in these models, intro- ducing possible biases in the estimates. This last issue is addressed be- low.

The fact that, in their original work, M-M used such a simple specifi- cation, with one single independent variable, is one of the major sources of criticism of their procedures. Weston (1963), for instance, argues that variables such as firm size may, in practice, influence leverage as well as the average cost of capital. Their absence in the empirical formula- tion could then introduce an omitted variables bias in the estimate of the coefficient of interest. In addressing this problem and in trying to im- prove the statistical quality of the estimated models, we use a few con- trol variables. Among the most natural candidates are the firm’s size, measured by its total capital employed (TC), systematic risk, measured by the firm’s b, the corporate income tax rate to which the firm is effec- tively submitted (T), the cost of its debt financing (id) and the relative amount of short term liabilities (STL) in its financial structure. Control- ling for systematic risk seems to be of particular importance, given the model’s assumption that all firms are in the same risk class. Various al- ternative specifications were attempted, and the ones reported are those with greatest statistical adequacy.7 Some of the results are shown below (t-statistics in parentheses).

58 LATIN AMERICAN BUSINESS REVIEW

American Electric Utilities Latin-American Electric

Utilities

x =.082–.0014d+.017 –.056 –.024 =.066–.022ß T STL x d+.034 .029T

(–4.34) ß–

(–5.32) American Oil and Gas Producers Latin-American Oil and Gas

Producers

039x =. –.009 0024 .38 =.098–d+. b+ i xd .035 .053

(–4.35) d– T

(–3.60)

After introducing fuller specifications and substantially improving the statistical quality of the models, it is noteworthy that all results are opposed to those reported by M-M in 1958. The coefficients of d are now statistically significant at the 5% level in the four regressions, indi- cating a clearly negative association between leverage and the average cost of capital. Alternative specifications of these models yield similar conclusions. It is also noted that the estimated coefficient for the firm’s size was negligible in all specifications.

In the test of Proposition II, the only difference from the original M-M procedure is the operational definition of the cost of the equity, which is estimated with the CAPM. Results are reported below (t-statis- tics in parentheses):

American Electric Utilities Latin-American Electric

Utilities

=.076+.0005z h =.095–.0002

(.40)

z h

(–.12)

American Oil and Gas Producers Latin-American Oil and Gas

Producers

=.082+.0003z h =.09–.0003z h

(1.37) (–1.32)

Once again, all results contradict the predictions of the original 1958 model, as well as the conclusions based on M-M’s empirical tests. The null hypothesis that the slope coefficient is zero cannot be rejected, at the 5% significance level, in any case. In other words, the data shows no significant relationship between the cost of equity financing and the de- gree of leverage, contradicting Proposition II. However, it must again be stressed that such results are not incompatible with the M-M model

Barros, Silveira, and Famá 59

of 1963. As shown by (5), the tax shield effect provided by debt may substantially reduce the impact of leverage on is and could easily ac- count for the insignificance of the estimated regression coefficients.

The regressions reported above, estimated with Latin-American data, show satisfactory statistical quality, but that was not the case with the American firms. Several alternative specifications were then tested. Also, outlying observations were removed in some cases. In any cir- cumstance, though, the results shown above remained practically unal- tered, proving to be quite stalwart. In fact, adding control variables, including b, only produced even less significant coefficients for h.

According to the “traditionalist” point of view, there should be a curvilinear relationship between the average cost of capital and a firm’s leverage. For moderate levels of debt financing, the WACC should drop with an increase of leverage because debt is cheaper and interest pay- ments can be deducted from taxable income. At some point, however, the risk of bankruptcy caused by excessive leverage would become a se- rious concern and investors would start driving the WACC upwards if the firm continued to replace equity with debt financing. If that were the case, we should be able to find a point at which the WACC was mini- mal, thus maximizing the firm’s market value. The empirical equation used by Modigliani and Miller (1958) to test that possibility is shown below.

x d d

d = + +α α α1 2 3

2

1( – ) (9)

a1, a2 and a3 are regression coefficients. The reduced form described in (9) was estimated for the American and Latin-American firms. In general, results showed no evidence of any non linear relationship between the vari- ables; that is to say, a3 was not statistically significant in nearly all cases. Nei- ther was any clear curvilinear pattern detected in the regressions of z in h.

SUMMARY AND CONCLUDING REMARKS

Although the academic controversies concerning the issue of capital structure are more than 40 years old, they are still far from over. Since Durand (1952), and especially since Modigliani and Miller’s pioneering works of 1958 and 1963, innumerous alternative approaches to the sub-

60 LATIN AMERICAN BUSINESS REVIEW

ject have proliferated. Remarkably, though, M-M’s models remain as some of the most influential in this field of research.

The first sections of this work sought to outline a rather summarized panorama of the historical evolution of research on capital structure, fo- cusing on the seminal ideas of M-M and on their opposition to the point of view often referred to as “traditionalist.”

In the empirical study, the empirical test developed by M-M them- selves in 1958 was described. Their results confirmed their original propositions, according to which the capital structure of a firm was ir- relevant for determining its market value, even in the presence of a posi- tive corporate income tax rate. It was then argued that the testing procedures used by the authors suffered from severe limitations, espe- cially concerning the operational definition of the average cost of capi- tal and of the cost of a firm’s equity. Also, the specifications of the empirical models were too simplified and inferences based on them seem questionable.

We then replicated M-M’s tests with contemporary data, introduc- ing two basic distinguishing features from their original work: (a) the CAPM was used for estimating the firms’ cost of equity and (b) fuller specifications for the empirical models were adopted in order to en- hance their statistical quality and produce more appropriate infer- ences.

The results of the empirical research do not corroborate with the M-M model of 1958 and are similar to those reported by Weston (1963). In general, it was found that more leverage is associated with a smaller average cost of capital for both American and Latin-American electric utilities and oil and gas producers. Also, the data did not show evidence of any significant relationship between leverage and the cost of a firms’ equity. Finally, no curvilinear relationships between the average cost of capital and leverage, or between the cost of equity and leverage were found. Thus, although capital structure seems to be relevant in deter- mining a firm’s value, it was not possible to identify any optimal mix of debt and equity financing.

One possible justification for the results found is the tax shield feature associated with debt financing. The deductibility of interest payments is an advantage of debt in the presence of a corporate income tax. That was examined by M-M in their 1963 work. In this context, our evidence can- not be qualified as surprising. Much to the contrary, these should pre- cisely be the expected results if the original model (Modigliani and Miller, 1958) had been correctly extended to a world where there is a cor- porate income tax, which was finally accomplished in 1963. Of course,

Barros, Silveira, and Famá 61

some other theories, among the many that have followed since M-M’s early works, could also play significant roles in explaining the empirical patterns identified in this research.

NOTES

1. In this paper the term “leverage” is to be understood as the “financial leverage” resulting from the existence of a fixed expense, in this case the interests charged to a firm as a periodic service of its debt. Therefore, an increase in “leverage” results from an increase of the firm’s indebtedness.

2. The WACC may be represented as

WACC = id D

D S+ is

S

D S+ ,

where id is the cost of debt, is is the cost of equity, D is the market value of the firm’s debts, S is the market value of its shares and therefore, D + S represents the firm’s total value. If other forms of financing are used, they may be directly aggregated to the above equation, according to their cost and to the proportions in which they are used.

3. The average values of the two years were used. 4. The authors approximate the concept of “class” by using firms from the same indus-

try. In that sense, it is assumed that all oil companies, for example, belong to the same class. 5. For the oil companies the authors further made a minor adjustment for the varia-

tion in the size of the company from one year to the other. 6. Non-normality and heteroscedasticity were verified by standard testing proce-

dures and are particularly problematic considering that most of the samples available are considerably small, especially in the Latin-American case.

7. Diagnostic analyses were carried out based on testing procedures such as the Jarque-Bera normality test and the White test for heteroscedasticity and correct linear specification. For model selection purposes, the adjusted coefficient of determina- tion and the Schwarz Criterion were used. In some cases, when the re-specification of the model was not sufficient to ensure an adequate statistical quality, a few outlying observations were removed. The removal of outliers, though, did not materially af- fect the magnitude and sign of the coefficients of interest.

REFERENCES

Altman, E. (1984). “A Further Empirical Investigation of the Bankruptcy Cost Ques- tion,” Journal of Finance, Vol. 39, No. 6, pp. 1067-1089.

Archer, S. H. and D’ambrosio, C. H. (1967). “The Theory of Business Finance: A Book of Readings,” New York: The Macmillan Company.

Balakrishnan, S. and Fox, I. (1993). “Asset Specificity, Firm Heterogeneity and Capi- tal Structure,” Strategic Management Journal, Vol. 14, No.1, pp. 3-16.

Baker, M. and Wurgler, J. “Market Timing and Capital Structure,” Journal of Finance, Vol. 57, No.1, pp. 32.

62 LATIN AMERICAN BUSINESS REVIEW

Baxter, N. (1967). “Leverage, Risk of Ruin and the Cost of Capital,” Journal of Fi- nance, Vol. 22, No. 3, pp. 395-403.

Boness, A. J. (1964). “A Pedagogic Note on the Cost of Capital,” Journal of Finance, Vol. 19, No.1, pp. 99-106.

Brewer, D. E. and Michaelsen, J. B.(1965). “The Cost of Capital, Corporation Finance, and the Theory of Investment: Comment,” American Economic Review, Vol. 55, No. 3, pp. 261-297.

Copeland, T. E. and Weston, J. F. (1988). “Financial Theory and Corporate Policy,” 3rd edition. Reading: Addison Wesley Publishing Company.

Durand, D. (1952). “Cost of Debt and Equity Funds for Business: Trends and Problems of Measurement” in Conference on Research on Business Finance. National Bu- reau of Economic Research: New York.

______(1959). “The Cost of Capital, Corporation Finance, and the Theory of Invest- ment: Comment,” American Economic Review, Vol. XLIX, No. 4, pp. 261-297.

Harris, M. and Raviv, A. (1990). “Capital Structure and the Informational Role of Debt,” Journal of Finance, Vol. 45, No. 2, pp. 321-349.

______(1991). “The Theory of Capital Structure,” Journal of Finance, Vol. 46, No. 1, pp. 197-355.

Jensen, M. C. (1986). “Agency Costs of Free Cash Flow, Corporate Finance and Take- overs,” American Economic Review, Vol. 76, No. 2, pp. 323-329.

Lintner, J. (1965). “The Valuation of Risk Assets and the Selection of Risky Invest- ments in Stock Portfolios and Capital Budgets,” Review of Economics and Statis- tics, Vol. XLVII, No. 1, pp. 13-37.

Maksimovic, V. (1986). “Optimal Capital Structure in Oligopolies.” Doctoral Disser- tation. Harvard University.

Miller, M. H. (1977). “Debt and Taxes,” Journal of Finance, Vol. 32, No. 2, pp. 261-275. Modigliani, F. and Miller, M. H. (1958). “The Cost of Capital, Corporation Finance,

and the Theory of Investment,” American Economic Review, Vol. 48, No. 3, June, pp. 261-297.

Modigliani, F. and Miller, M. H. (1959). “The Cost of Capital, Corporation Finance, and the Theory of Investment: Reply,” American Economic Review, Vol. 49, No. 4, pp. 655-669.

Modigliani, F. and Miller, M. H. (1963). “Corporate Income Taxes and the Cost of Capital: A Correction,” American Economic Review, Vol. 53, No. 3, pp. 433-443.

Modigliani, F. and Miller, M. H. (1965). “The Cost of Capital, Corporation Finance, and the Theory of Investment: Reply,” American Economic Review, Vol. 55, No. 3, pp. 261-297.

Myers, S. C. and Majluf, N. S. (1984). “Corporate Financing and Investment Decisions When Firms Have Information that Investors do not Have,” Journal of Financial Economics, Vol. 13, No. 2, pp. 187-221.

Myers, S. C. “The Capital Structure Puzzle,” Journal of Finance, Vol. 39, No. 3, pp. 575-592.

Myers, S. C. (1977). “Determinants of Corporate Borrowing,” Journal of Financial Economics, Vol. 5, No. 39, pp. 147-176.

Ross, S. A. (1977). “The Determination of Financial Structure: The Incentive Signal- ing Approach,” Bell Journal of Economics, Vol. 8, No. 1 , pp. 23-40.

Barros, Silveira, and Famá 63

Rubinstein, M. E. (1973). “A Mean-Variance Synthesis of Corporate Financial The- ory,” Journal of Finance, Vol. 28, No. 1, pp. 167-181.

Sharpe, W. F. (1964). “Capital Asset Market Prices: A Theory of Market Equilibrium Under Conditions of Risk,” Journal of Finance, Vol. 19, No. 3, pp. 425-442.

Solomon, E. (1963). “Leverage and the Cost of Capital,” American Economic Review, Vol. 18, No. 2.

Stiglitz, J. E. “A Re-Examination of the Modigliani-Miller Theorem,” American Eco- nomic Review, Vol. 59, No. 5, pp. 784-793.

Stulz, R. (1988). “Managerial Control of Voting Rights: Financing Policies and the Market for Corporate Control,” Journal of Financial Economics, Vol. 20, No.1/2 pp. 25-54.

______(1990). “Managerial Discretion and Optimal Financing Policies,” Journal of Financial Economics, Vol. 26, No. 1 , pp. 3-28.

Titman, S. (1984). “The Effect of Capital Structure on the Firm’s Liquidation Deci- sion,” Journal of Financial Economics, Vol. 13, No. 1, pp. 137-152.

Warner, J. B. (1977). “Bankruptcy Costs: Some Evidence,” Journal of Finance, Vol. 32, No. 2, pp. 337-347.

Weiss, L. A. (1990). “Bankruptcy Resolution: Direct Costs and Violation of Priority of Claims,” Journal of Financial Economics, Vol. 27, No. 2, pp. 285-314.

Weston, J. F. (1963). “A Test of Cost of Capital Propositions,” Southern Economic Journal, Vol. 30, No. 2.

64 LATIN AMERICAN BUSINESS REVIEW