Download Relevance of Payout Policy in Frictionless Markets: A Modigliani & Miller Reassessment and more Study Guides, Projects, Research French Philology in PDF only on Docsity!
Journal of Financial Economics 79 (2006) 293–
The irrelevance of the MM dividend
irrelevance theorem
Harry DeAngelo
, Linda DeAngelo
University of Southern California, Marshall School of Business, Los Angeles, CA 90089, USA Received 20 March 2004; received in revised form 10 January 2005; accepted 7 March 2005 Available online 19 September 2005
Abstract
Contrary to Miller and Modigliani [1961. Dividend policy, growth, and the valuation of shares. Journal of Business 34, 411–433], payout policy is not irrelevant and investment policy is not the sole determinant of value, even in frictionless markets. MM ask ‘‘Do companies with generous distribution policies consistently sell at a premium above those with niggardly payouts?’’ But MM’s analysis does not address this question because the joint effect of their assumptions is to mandate 100% free cash flow payout in every period, thereby rendering ‘‘niggardly payouts’’ infeasible and forcing distributions to a global optimum. Irrelevance obtains, but in an economically vacuous sense because the firm’s opportunity set is artificially constrained to payout policies that fully distribute free cash flow. When MM’s assumptions are relaxed to allow retention, payout policy matters in exactly the same sense that investment policy does. Moreover (i) the standard Fisherian model is empirically refutable, predicting that firms will make large payouts in present value terms, (ii) only when payout policy is optimized will the present value of distributions equal the PV of project cash flows, (iii) the NPV rule for investments is not sufficient to ensure value maximization, rather an analogous rule for payout policy is also necessary, and (iv) Black’s [1976. The dividend puzzle. Journal of Portfolio
www.elsevier.com/locate/jfec
0304-405X/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jfineco.2005.03.
$This research was supported by the Charles E. Cook/Community Bank and Kenneth King Stonier
chairs at USC. For their useful comments on earlier drafts of this paper, we thank Michael Brennan, Kenneth French, Jonathan Karpoff, Ronald Masulis, David Mayers, Edward Rice, Jay Ritter, Rene´ Stulz, Mark Westerfield, and especially Eugene Fama and Clifford Smith. We also gratefully acknowledge the constructive comments of three referees. We would particularly like to thank Michael Jensen for his invaluable suggestions and advice on this project. Corresponding author. Tel.: +1 213 740 6541; fax: +1 213 740 6650. E-mail address: hdeangelo@marshall.usc.edu (H. DeAngelo).
Management 2, 5–8] ‘‘dividend puzzle’’ is a non-puzzle because it is rooted in the mistaken idea that MM’s irrelevance theorem applies to payout/retention decisions, which it does not. r 2005 Elsevier B.V. All rights reserved.
JEL classification: G35; G32; H
Keywords: Dividends; Payout policy; Dividend puzzle
- Introduction
Miller and Modigliani’s (1958, 1961) irrelevance theorems form the foundational bedrock of modern corporate finance theory. The MM theorems indicate that, in frictionless markets with investment policy fixed, all feasible capital structure and dividend policies are optimal because all imply identical stockholder wealth, and so the choice among them is irrelevant. The central lesson commonly drawn from MM is that investment policy alone determines stockholder wealth in frictionless markets, and that leverage and payout decisions have no impact on firm value, given a value- maximizing investment program (see, e.g., Allen and Michaely, 2003, p. 339). Specifically, when a firm considers different leverage and payout decisions, it is simply slicing a fixed pie (of cash flows from investment) into different pieces, whose individual values in frictionless markets must inevitably sum to the value generated by the underlying investment policy. This paper shows that payout policy, like investment policy, has first-order value consequences in frictionless markets, and cannot be reduced to a ‘‘pie-slicing’’ exercise as in Modigliani and Miller (1958). By definition, irrelevance requires a one- to-one correspondence between feasible and optimal policies—i.e., throw a dart at the feasible set and, no matter where it hits, stockholders are equally well off. Irrelevance is hard-wired into MM (1961) by assumptions that shrink the feasible set to optimal policies by forcing 100% distribution of free cash flow (FCF) in every period. In effect, MM assume away the value-relevant payout/retention decision to focus on a decision that can be reduced to pie-slicing: pay out 100% of FCF or pay out 100% and simultaneously act as an intermediary between new investors and stockholders who want to sell shares. Since portfolio trades are costless in frictionless markets, intermediation adds no value, and the firm’s payout ‘‘choice’’ is no choice at all because MM mandate full payout in all cases. When MM’s assumptions are modified to allow retention with the NPV of investment policy fixed, a firm can reduce its value by paying out less than the full present value of FCF, and so payout policy matters and investment policy is not the sole determinant of value. With retention allowed, a firm is no longer constrained to an optimal payout policy as an automatic by-product of its investment decision, and irrelevance fails because some feasible payout policies do not distribute the full present value of FCF to currently outstanding shares. Because irrelevance is a property of the opportunity set (‘‘all feasible decisions are optimal’’), payout policy (like investment policy) is relevant in the standard Fisherian model, even though that
294 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
Readers of prior drafts have raised several counter-arguments to our claim that investment policy is not the sole determinant of value in frictionless markets. We summarize and address these counter-arguments in ‘‘objection/rejoinder’’ format at the points in our argument where they typically arise. Section 2 dissects MM’s (1961) irrelevance proof. Section 3 shows why payout policy matters when MM’s assumptions are relaxed to allow retention. Section 4 discusses applications of our analysis, including its relation to Jensen’s (1986) free cash flow theory, the issue of what the market ‘‘really’’ capitalizes, and why payout policy matters in a stock bubble. Section 5 analyzes Fischer Black’s (1976) ‘‘dividend puzzle.’’ Section 6 summarizes our findings and discusses their implications.
- MM (1961) allow no payout/retention decisions
Irrelevance means that all feasible payout policies are optimal, so that any policy managers could choose yields identical stockholder wealth. MM’s (1961) irrelevance proof shows that, in frictionless markets, stockholder wealth is unchanged when all aspects of investment policy are fixed and any increase in the current payout is financed by fairly priced stock sales. In this section, we show that the reason why payout policy is irrelevant is that MM’s assumptions require firms to pay out 100% of free cash flow (FCF) in every period. By ruling out retention, MM restrict the feasible set to optimal policies and thereby ensure irrelevance. MM’s irrelevance result, however, comes at the cost of side-stepping the fundamental question they pose in their opening paragraph: ‘‘Do companies with generous distribution policies consistently sell at a premium over those with niggardly payouts?’’ Since ‘‘niggardly payouts’’ are impossible in a model that mandates 100% FCF payout every period, MM have nothing to say about the central question of payout policy they pose; thus, their irrelevance theorem is of trivial import. We maintain all of MM’s (1961) assumptions, except in Section 3 where we allow retention (while holding the NPV of investment policy fixed), and show that payout policy matters. We use the term ‘‘frictionless markets’’ as shorthand for MM’s economic setting in which there are no taxes, no security trading or flotation costs, rational expectations-enforced fair pricing of securities, and price-taking behavior by individuals and firms. For simplicity, we work in a certainty framework as do MM, but all conclusions generalize to uncertainty using the Arrow–Debreu approach. Like MM, we assume that firms use only equity financing but all findings readily translate to scenarios with debt financing. Although MM use an infinite horizon model, we use a three-date model to graphically illustrate why investment policy is not the sole determinant of value when MM’s assumptions are relaxed to allow retention. (There is no loss of generality since our conclusion that payout policy matters holds for both infinite and finite horizon formulations, as we discuss in Section 3.) Since MM’s irrelevance theorem is a statement about the firm’s opportunity set (all feasible payout policies are equally valuable), their proof makes no assumption about managerial objectives and we follow suit here. No specification of managerial
296 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
objectives is necessary because irrelevance means that, no matter how poorly motivated or self-interested managers might be, they cannot damage stockholders by their payout decisions. We use the term ‘‘standard Fisherian model’’ to refer to a frictionless market setting in which managers are assumed to select value-maximizing policies. For precision in describing a firm’s decisions at date t, let X (^) t ¼ cash flow from prior operating decisions, I (^) t ¼ investment outlays, and X (^) t I (^) t ¼ FCFt ¼ net-of- investment (free) cash flow, where the FCF label indicates the firm has chosen an optimal investment program. 2 Let D (^) t ¼ the gross distribution, which equals the sum of dividends plus repurchases, and which we pool because we wish to address whether payout versus retention decisions matter and are not concerned with how the firm splits a given distribution between dividends and repurchases, or between these two and interest/principal on debt. Finally, let S (^) t ¼ cash raised from stock sales, so that D (^) t S (^) t ¼ the net distribution (payouts minus stock sale proceeds), where D (^) t and S (^) t are nonnegative by definition. The MM irrelevance result is driven by their requirement that the firm distribute 100% of FCF in every period. This requirement is an unappreciated implication of MM’s assumption that all aspects of investment policy are fixed, coupled with their treatment of the condition that the date t distribution to stockholders cannot exceed the sum of contemporaneous FCF and stock sale proceeds. MM treat the latter condition as a strict equality so that the firm’s payout at date t is
Dt ¼ X (^) t I (^) t þ S (^) t ¼ FCFt þ S (^) t. (1)
With X (^) t and I (^) t assumed constant for all t, FCFt is also parametric for all t. Since stock sale proceeds, S (^) t, are nonnegative, the firm’s payout, Dt, is constrained to be at least as large as FCFt ¼ X (^) t I (^) t, and any distributions above the level of current free cash flow are funded by fairly priced stock sales. The irony here is that, although MM sought to avoid confounding investment and payout policy, their assumptions actually induce an interdependence between the two by mandating 100% FCF payout every period. In effect, MM force the payout decision to be a by-product of the investment decision, so that once the latter decision is made, the firm automatically distributes all FCF in every period. Obviously, stockholders cannot do better than that. When FCF retention is allowed, the firm can choose D (^) toFCFt, so that policies that pay out less than the full present value of the FCF stream become feasible and irrelevance fails, as we show in Section 3 below. Fig. 1 illustrates MM’s (1961) theorem for a three-date economy in which a given firm raises capital and invests at t ¼ 0 to generate free cash flow of FCF 1 at t ¼ 1 and FCF 2 at t ¼ 2. The x-axis represents the date t ¼ 1 distributions to all shares
(^2) In MM’s (1961) analysis, an investment program is selected arbitrarily and held fixed, with no presumption that it is optimal. Thus, one could interpret X (^) t I (^) t as ‘‘net cash flow from investment policy’’ rather than ‘‘free cash flow,’’ since the latter term is commonly used to indicate the amount of cash left after selecting investments with maximal overall NPV. With X (^) t I (^) t representing free cash flow, there can be no misconception that our payout relevance conclusion depends on a hidden assumption that the firm has adopted a strictly suboptimal investment program.
H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–315 297
initially. The firm distributes the same total FCF 1 plus the stock sale proceeds to the t ¼ 0 stockholders, who now receive more cash at t ¼ 1 and less at t ¼ 2, with an unchanged present value. At t ¼ 2, the firm distributes FCF 2 , which is now split between shares outstanding at t ¼ 0 and those issued at t ¼ 1. In substance, movements along AW 1 are simply trades between investors at t ¼ 1, with no variation in the firm’s level of retention. If the incremental ‘‘distribution’’ at t ¼ 1 comes as a repurchase, it is exactly as if old stockholders sold their shares directly to outside investors. If it comes as a dividend, it is exactly as if new investors paid cash to old stockholders for a portion of the firm’s equity (and then the firm split its stock to increase the number of shares outstanding). The firm is merely a financial intermediary in these transactions and, since trading is costless in frictionless markets, intermediation adds no value for stockholders. But the issue of concern here is payout policy, not intermediation. And these transactions represent trivial changes in payout/retention decisions for firms whose retention levels never change because they are forced to distribute all FCF under any allowed ‘‘payout choice.’’ Bottom line, in MM (1961), the only policies the firm can choose entail 100% FCF payout, and that is why the payout choices examined by MM are all equally valuable to stockholders and why investment policy is the sole determinant of value. Although their proof assumes 100% FCF payout, MM (1961, footnote 12) indicate elsewhere in the paper that stockholder wealth is invariant to all payout/retention decisions except those with exactly zero payouts in every period, thereby creating the impression that, except for one pathological and economically trivial case, all feasible policies yield identical stockholder wealth. 3 This cannot be true, however, since rational investors will set a near-zero value on the equity of firms whose payout policies entail near-zero distributions every period. In fact, MM (1961) does not apply to payout/retention decisions, since their assumptions prohibit retention.
(^3) MM’s argument follows (with (14) being the ‘‘discounted dividend’’ formula for equity value, (12) the
‘‘investment opportunities’’ formula, and (9) the discounted ‘‘net cash flow’’ formula):
‘‘The statement that Eqs. (9), (12), and (14) are equivalent must be qualified to allow for certain pathological extreme cases, fortunately of no real economic significance. An obvious example of such a case is the legendary company that is expected never to pay a dividend. If this were literally true then the value of the firm by (14) would be zero; by (9) it would be zero (or possibly negative since zero dividends rule out X ðtÞ 4 IðtÞ but not X ðtÞoIðtÞ; while by (12) the value might still be positive. What is involved here, of course, is nothing more than a discontinuity at zero since the value under (14) and (9) would be positive and the equivalence of both with (12) would hold if that value were positive as long as there was some period T, however far in the future, beyond which the firm would pay out 40 percent of its earnings, however small the value of .’’ (MM, 1961, footnote 12, emphasis added in final sentence).
The closing sentence of the passage states that the ‘‘dividend discount’’ valuation formula yields the same value as the ‘‘investment opportunities’’ formula as long as there is an arbitrarily small, but positive, stream of payouts beginning in some future period. The term ‘‘discontinuity at zero’’ indicates that equity value under the discounted dividend formula is identical for all payout vectors except for the singular point at which the vector of time-dated dividend payments has every element exactly equal to zero. These statements are not correct.
H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–315 299
Objection #1: I always thought that MM showed that, although firms must pay out the full present value of free cash flow, the timing of those payouts is irrelevant. Rejoinder: MM allow no variation in the timing of FCF payout. MM mandate payout of all FCF every period, and if the firm wants to pay a yet-higher dividend or buy back more stock today, it sells shares to outsiders and immediately hands over the cash to current stockholders. It is no different than if the current stockholders had sold some (ex-dividend) shares to other investors. The fact that current stockholders get more cash today and less tomorrow is the result of portfolio trades, and has nothing to do with resources received from the firm since the firm’s retention level never changes. Objection #2: It is unfair to criticize MM (1961) for requiring 100% FCF payout, since that assumption is perfectly appropriate for proving that the choice of how to divide a given cash payout between dividends and repurchases is a matter of indifference in frictionless markets. Rejoinder: Indifference to the dividend/repurchase mix is not sufficient to establish that investment policy is the sole determinant of value, a conclusion that is incorrect once retention is allowed. Moreover, MM’s (1961) concern is the level of distributions and not the dividend/repurchase mix, as is evident from the following. (1) MM’s opening paragraph asks: ‘‘Do companies with generous distribution policies consistently sell at a premium over those with niggardly payouts?’’ (2) MM do not mention stock repurchases anywhere in their article, not even in the closing section where they discuss the effect of taxes on payout policy. (3) If MM had repurchases in mind, they would have (should have) stated that the discounted ‘‘stream of dividends approach’’ understates equity value by the present value of buyback proceeds. (4) The body of MM’s irrelevance proof asks: ‘‘Which is the better strategy for the firm in financing the investment: to reduce dividends and rely on retained earnings or to raise dividends and float new shares?’’ (emphasis added). It is not surprising that repurchases are mentioned nowhere in MM (1961) since dividends were the only empirically meaningful equity payout at that time, and so the issue of the dividend/repurchase mix was simply not on the profession’s radar screen.
- Why payout policy matters with retention allowed
Because their assumptions ensure payout policy optimality by forcing 100% FCF distribution to be an automatic by-product of the investment choice, MM (1961) confound investment and payout policy and mistakenly attribute the value impact of payout policy optimization to investment policy. When MM’s assumptions are relaxed to allow retention, payout policy optimization is not an automatic by- product of the investment choice. Rather, to maximize stockholder wealth, managers must make a separate decision to adopt a payout policy that distributes the full PV of FCF to currently outstanding shares. One can resurrect irrelevance by assuming that costless contracting restricts managers to payout policies that fully distribute FCF, but costless contracting renders both investment and payout policy irrelevant
300 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
are indifferent to the timing of the payout. This logic establishes that no single payout policy is uniquely optimal, i.e., optimal payout policy is indeterminate. But payout policy is not irrelevant, as we next establish. Fig. 2 describes feasible and optimal payout policies with FCF retention allowed. The feasible payout set is the shaded area bounded by V 2 V 1 W 1 W 2 , with the upper boundary W 2 W 1 determined by the PV of the FCF stream generated by optimal investment policy. Rational expectations dictate a lower bound, V 2 V 1 , on the set of feasible payout policies that will elicit the firm’s desired equity infusion at t ¼ 0. This
$ at t = 2 W 2 Feasible payout policies with retention and rational expectations Infeasible payout policies given rational expectations V (^2)
A FCF (^2) Feasible payout policies under MM where retention is ruled out
O FCF 1 V 1 W 1 $ at t = 1
Fig. 2. Feasible and optimal payout policies for a firm operating in frictionless markets: MM (1961) relaxed to allow retention of FCF, holding the NPV of investment policy fixed. The x-axis plots consumption claims (dollars) at t ¼ 1 and the y-axis plots consumption claims at t ¼ 2 to all shares outstanding at t ¼ 0. The slope of W 2 W 1 is dictated by the market interest rate for transforming consumption claims from t ¼ 1 to t ¼ 2. FCF 1 and FCF 2 are free cash flow at t ¼ 1 and t ¼ 2 from investment at t ¼ 0. In MM (1961), all aspects of investment policy are fixed and FCF must be fully distributed each period (retention is not allowed). AW 1 is the feasible set of distributions to shares outstanding as of t ¼ 0 in MM, and it is also the optimal set since all these payout policies imply identical stockholder wealth, as shown in Fig. 1. Fig. 2 holds the NPV of investment policy fixed by allowing the firm unlimited access to zero-NPV projects. Optimal investment policy is indeterminate, with W 2 A showing the set of equally valuable investment policies that are reached by retaining cash at t ¼ 1 and investing it in zero-NPV projects. Optimal payout policy is also indeterminate, with value-maximizing policies plotting along W 2 W 1. The set of feasible payout policies is the full shaded region. V 1 and V 2 are derived by treating the rational expectations constraint on capital supply (Eq. (2) in the text) as a strict equality. Hence V 1 ¼ ð 1 þ r 01 ÞðI 0 =yÞ and V 2 ¼ ð 1 þ r 12 ÞV 1 where I 0 is the total equity capital raised from outsiders at t ¼ 0 in exchange for fraction y of the equity, r 01 is the market interest rate for transforming dollars from t ¼ 0 to t ¼ 1, and r 12 is the rate between t ¼ 1 and t ¼ 2. Rational expectations rule out payout policies with permanently low or near-zero distributions in OV 2 V 1 because these distributions are too low to induce outside investors to fund the t ¼ 0 outlay that generates FCF 1 and FCF 2.
302 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
bound is determined by I 0 , the amount of external capital the firm seeks at t ¼ 0, and y, the fraction of stock issued to outsiders. Since the figure plots total distributions at t ¼ 1 and t ¼ 2 to all shares outstanding at t ¼ 0 (not just to shares issued for the t ¼ 0 capital contribution I 0 ), feasible payout policies are those with distributions D^01 and D^02 whose present value, V 0 , ensures that the value of future payouts to shares sold at t ¼ 0, yV 0 , is at least as great as the capital contribution I 0. Let r 01 and r 12 , respectively, denote market interest rates for transforming dollars from t ¼ 0 to t ¼ 1 and from t ¼ 1 to t ¼ 2. Since market interest rates are investors’ opportunity cost of capital, only payout policies that satisfy the following condition enable the firm to raise sufficient funds at t ¼ 0 to generate FCF 1 and FCF 2 :
V 0 ¼
D^01
ð 1 þ r 01 Þ
þ
D^02
ð 1 þ r 01 Þð 1 þ r 12 Þ
XI 0 =y. (2)
In Fig. 2, the ‘‘efficient frontier’’ W 2 W 1 is the set of optimal policies whose distributions to shares outstanding as of t ¼ 0 have present value equal to that of the FCF stream. Hence, the set of optimal payout policies expands beyond AW 1 , the optimal policies in MM, to the full line segment W 2 W 1. Stockholders are no longer indifferent among all feasible policies because that set is now the entire shaded region V 2 V 1 W 1 W 2 , while optimal payout policies lie along W 2 W 1. With retention admissible, firms are no longer forced to distribute the full present value of FCF as an automatic by- product of selecting an optimal investment policy, and that is why payout policy matters. Objection #3: I always thought that MM’s payout policy irrelevance result holds when retention is allowed, as long as the NPV of investment policy is fixed. Rejoinder: This seemingly innocuous extrapolation from the MM model to one that allows retention is actually a major logical error because it fails to recognize that MM’s payout irrelevance conclusion relies on mandated 100% FCF payout. Allowing retention gives managers the previously unavailable opportunity to choose suboptimal payout policies, i.e., policies from within the shaded region in Fig. 2. Payout policy irrelevance is therefore out the window once retention is allowed. Objection #4: In the standard Fisherian model, managers are assumed to act in the interests of stockholders, and so they will always choose a payout policy along the W 2 W 1 frontier, i.e., they will distribute the full present value of FCF to currently outstanding shares and ignore the suboptimal payout policies that fall in the shaded region of the feasible set. And so payout policy is irrelevant. Rejoinder: The fact that value-maximizing managers will always choose optimal payout policies is completely beside the point because irrelevance is a property of the opportunity set: all choices that could be made are equally valuable. The fact that more than one payout policy satisfies the optimality condition does not mean that all feasible policies do so. Irrelevance fails because some feasible payout policies are better than others in exactly the same way that some investment projects are better than others. (See also Objection #5 below.) Managers can choose any payout policy on the W 2 W 1 frontier without affecting stockholder wealth. And so it is correct to say that, provided that managers distribute the full present value of FCF, the timing of those payouts is a matter of indifference to stockholders. But this is not ‘‘payout policy irrelevance,’’ since managers can also
H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–315 303
are constrained to choose only among policies that yield the same (maximized value of) stockholder wealth. However, the costless contracting assumption that eliminates suboptimal payout policies also precludes suboptimal investment policies, and so investment policy is irrelevant in the same conditional sense. No one has ever argued that investment policy is irrelevant because of costless contracting. Why not? Because it is obvious that some investment programs are better than others, and it is precisely because project choice matters that stockholders can benefit from disciplinary mechanisms that constrain managerial choice. It is equally obvious from Fig. 2 that some payout policies are better than others and, as with investment policy, it is precisely because payout policy matters that disciplinary constraints on managers can increase stockholder wealth. And so, if we apply the same criterion universally applied to investment policy, payout policy matters in exactly the same sense that investment policy matters when costless contracting is assumed. In general, if takeover pressure, incentive contracts, or other disciplinary mechanisms are required to force particular choices and rule out others, then the decision under analysis cannot possibly be irrelevant. Consistent with this principle, MM’s dividend and leverage irrelevance proofs invoke no such disciplinary mechanisms. In MM (1958), the ‘‘pie-slicing’’ nature of the leverage decision they analyze ensures irrelevance: because no feasible change in the mix of dividends versus interest/principal payments alters the total payout delivered to investors, the present value of that total payout is invariant to the debt-equity mix. Wealth redistributions aside, any decision that legitimately can be reduced to a pie-slicing exercise is irrelevant, such as the choice between dividends and stock repurchases in frictionless markets, which is the equity payout analog to the leverage choice analyzed in MM (1958). However, the choice between ‘‘generous’’ and ‘‘niggardly’’ distributions posed by MM (1961) is by nature not reducible to a pie-slicing exercise. The reason is that in rational markets stockholder wealth equals the discounted value of payouts and, with retention allowed, the size of the pie delivered to stockholders varies with alternative payout/retention policies that could be chosen. MM’s (1961) irrelevance theorem has led to the mistaken belief that payout policy is automatically optimized as long as the firm chooses a value-maximizing set of investment projects. Automatic optimization of payout policy does occur in MM (1961), but only because they mandate 100% FCF payout. With retention allowed, payout of the full PV of FCF no longer happens automatically and so optimizing payout policy requires an extra step beyond selecting an optimal investment program. Two optimality conditions are necessary for stockholder wealth maximization: managers must both (i) select projects that generate an overall cash flow stream with maximal attainable NPV, and (ii) distribute the full present value so generated (over the life of the enterprise) to currently outstanding shares. 7 With retention admissible, condition (ii) is not satisfied automatically when (i) is satisfied
(^7) These conditions follow MM (1961) and assume the firm is unlevered. For a levered firm, optimal payout policy continues to require distribution of the full PV of FCF, but (ii) must be modified to stipulate that part of the value of FCF flows to debtholders and that the full remainder flows to currently outstanding shares.
H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–315 305
because there is nothing inherent in project choice that forces the firm to distribute the full PV of FCF generated by that choice. Payout policy matters in infinite horizon models of the payout/retention decision because, with no final date for possible ‘‘settling up,’’ many feasible policies fail to distribute the full PV of FCF. It also matters in finite horizon models because arrival at the final date does not automatically trigger full payout. (If finite horizon models did necessitate payout of the full PV of FCF, they would be inappropriate for analyzing agency costs, which they obviously are not since the principal/agent literature is dominated by finite horizon models.) Full payout requires an action on the part of managers beyond any decisions they make about project choice, and that action is not automatic at the last date in a finite horizon model (or at any other time). We are not arguing that managers will fail to distribute full value, only that they could do so and therefore the choice of payout policy matters. Objection #6: If managers select a payout policy that fails to distribute the full PV of FCF to currently outstanding shares, they have essentially changed the firm’s investment decisions, and so the associated wealth loss for stockholders is attributable to selection of a suboptimal investment policy, and not to payout policy. Therefore, investment policy alone determines value and payout policy is irrelevant. Rejoinder: This is a semantic trick to resuscitate the conclusion that investment policy alone determines value by defining payout policy-related changes in value as elements of investment policy, e.g., by defining the value loss from a failure to distribute project-generated cash as due to a sub-optimal investment policy. If all value-relevant actions are defined as investment choices, then investment policy is tautologically the sole determinant of value and there was no need for MM to provide a formal theorem and detailed analytical proof to establish that ‘‘result.’’ Irrelevance follows as a meaningless tautology when payout policy is defined to remove the choice of distributing less than full value to stockholders. Mandatory payout of the full PV of FCF, whether in the last period of a finite horizon model or at any other date, simply restricts the set of feasible policies to those that are optimal, and is the retention analog to MM’s mandated 100% FCF payout every period. In such cases, stockholders are certainly not indifferent to receiving less than full value, rather the model employed simply defines such suboptimal outcomes as impossible. Similarly, if one defines any failure to distribute full value to be the selection of a suboptimal investment policy, the principle that ‘‘only investment policy counts’’ becomes tautological because the expanded definition of investment policy includes all decisions that affect stockholder wealth. Objection #7: You claim that payout policy and investment policy are both first- order determinants of value, i.e., that investment policy is not the only important value driver. Yet, there can be no distributions without investment returns (ignoring the return of capital contributions). Therefore, investment policy is the fundamental value driver, and payout policy is at best of second-order concern. Rejoinder: It is certainly true that there can be no economically meaningful distributions without investment returns. It is also true that the NPV rule for investment policy is specified without reference to payout policy, whereas the payout
306 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
investment program for the firm. The larger those rents, the bigger is the shaded area and the greater is the potential for agency problems that generate substantial value losses.^8 This suggests that agency pressures for cash disbursement are especially important for firms that earn substantial economic rents. Obviously, pursuing this point would take us too far afield, but it does illustrate that useful new ideas can emerge from recognizing that the incentives for cash payouts in the standard Fisherian model are intimately related to the incentives for payouts in free cash flow theory. Fig. 3 shows that MM’s (1961) irrelevance theorem does not, as commonly believed, overturn the conventional pre-MM view that increasing a firm’s payout increases stockholder wealth. The figure plots the relation between stockholder wealth at t ¼ 0 and the firm’s capitalized payout ratio (CPR), defined as the ratio of the PV of the distributions to shares outstanding at t ¼ 0 to the PV of FCF. Only those payout policies that lie on W 2 W 1 in Fig. 2 entail full payout of FCF to shares outstanding as of t ¼ 0 and thus have CPR^ ¼ 1. All other feasible policies imply lower current stockholder wealth and CPRo1. As long as the firm is constrained to payout policies with CPR^ ¼ 1, i.e., constrained to distribute the full PV of FCF as in MM’s irrelevance proof, payout policy has no impact on stockholder wealth. Payout policies with CPRo1 are strictly suboptimal in the same sense that leverage policies are suboptimal when they do not plot at the top of the value- leverage curve in ‘‘trade-off’’ theories of capital structure—i.e., they are feasible policies that will not be adopted by value-maximizing managers. When CPRo1, any variation in payout policy that increases the firm’s CPR increases stockholder wealth (because it delivers a distribution stream with larger PV). Thus, MM’s analysis does not universally refute the traditional practitioner intuition that increased payout means increased stockholder wealth. Rather, MM simply shrink the payout policy choice to a tiny region of the feasible set for which conventional wisdom does not apply because their assumptions force firms to pay out 100% of FCF. The modern version of the traditional view is agency theory, which holds that most publicly traded firms operate at CPRo1, so that a payout increase does increase stockholder wealth. MM (1961, Section II) conclude that the long-standing controversy over what the stock market ‘‘really’’ capitalizes is essentially empty because the discounted value of cash flows from investment policy (grouped a variety of different ways) must equal the discounted value of distributions to currently outstanding shares. Although MM’s ‘‘equivalence principle’’ is widely believed to hold universally for all payout policies, it actually holds only for those that distribute full value, as is easily seen from Fig. 3. Specifically, when CPRo1, as it does in the agency equilibrium, distribution value (the PV of payouts to currently outstanding shares) falls strictly below investment value (the PV of FCF), and the equivalence principle fails. When
(^8) If a firm has only zero-NPV projects (zero economic rents), it cannot raise sufficient equity capital to fund its full desired investment outlay unless investors believe that payout policy will distribute the full PV of FCF. In this case, the capital market will constrain the firm to policies along W 2 W 1 , since any other policy will fail to elicit the necessary capital, and both investment and payout policy are irrelevant.
308 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
CPR^ ¼ 1, distribution and investment values are equal and the equivalence principle holds because the chosen payout policy distributes the full value of FCF. With rational expectations, the stock market ‘‘really’’ capitalizes distributions because investors value securities only for the payouts they are expected to provide. Earnings matter, of course, since you can’t create distributions out of thin air (see the Rejoinder to Objection #7 above), but distribution value can fall short of investment value due, e.g., to managerial appropriation of FCF and, when it does, the stock market value equals the capitalized value of expected payouts. And so, at the most fundamental level, stockholder wealth is determined by payout policy, with investment policy relevant because it determines the capacity to distribute cash. Since value is generated for investors only to the extent that this capacity is transformed into actual payouts, selection of an optimal payout policy is necessary to ensure that the discounted value of distributions equals the discounted value of investment cash flows.
Stockholder wealth at t = 0
PV(FCF) All optimal policies entail full payout of the FCF stream (CPR = 1) Feasible but suboptimal CPRs*
I 0 / θ Infeasible CPRs under rational expectations
γ 1 Capitalized payout ratio (CPR)
Fig. 3. Stockholder wealth and a firm’s capitalized payout ratio. The capitalized payout ratio, CPR, is bounded between 0 and 1 and is defined as the ratio of the present value of the stream of distributions paid to the shares outstanding at t ¼ 0 divided by the present value of the firm’s free cash flow stream. All payout policies on the efficient frontier W 2 W 1 in Fig. 2 provide distributions with a present value equal to that of the free cash flow stream (denoted PV(FCF)) so that these policies map into Fig. 3 at CPR^ ¼ 1. Payout policies in Fig. 2 that plot on any line parallel to W 2 W 1 but strictly within the region OW 2 W 1 have equal present values that imply CPRo1, since they entail distributions to currently outstanding shares whose present value falls below that of the free cash flow stream. As these parallel ‘‘iso-value’’ lines in Fig. 2 move closer to the origin, CPR falls closer and closer to 0. When CPR falls below the critical level, g, that corresponds to payout policies along V 2 V 1 in Fig. 2, rational investors will not supply the level of outside equity that the firm seeks to generate the desired FCF stream. If, as in MM (1961), the firm must distribute 100% of FCF in every period, all feasible payout policies have CPR^ ¼ 1 and therefore yield the maximum feasible stockholder wealth. When distributions of less than 100% of FCF are feasible, the choice of payout policy matters, and policies that have CPRo1 are strictly suboptimal, while all policies that plot on W 2 W 1 in Fig. 2 have CPR^ ¼ 1. Stockholder wealth increases monotonically as the capitalized payout ratio increases from g to 1.
H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–315 309
destroying stockholder wealth. If firms respond to payout taxes in the manner Black recommends, the value of equities will collapse to near-zero levels and firms will only be able to raise trivial amounts of capital. When investors are taxed at rates tp1 and tp2 on payouts received at t ¼ 1 and t ¼ 2, the rational expectations condition that governs minimum feasible payout policies applies on an after-tax basis:
V 0 ¼
ð 1 tp1ÞD^01 ð 1 þ r 01 Þ
þ
ð 1 tp2ÞD^02 ð 1 þ r 01 Þð 1 þ r 12 Þ
XI 0 =y. (3)
Condition (3) dictates that payout policies with low or near-zero future distributions are infeasible if the firm seeks to raise substantial external equity at t ¼ 0. Only if the level of taxation is confiscatory, with tp1 and tp2 in the neighborhood of 1, will distributions be eliminated and, in this case, equity-financed corporations will disappear because shares that yield trivial after-tax payouts are essentially worthless. With payouts taxed, rational investors will not purchase shares whose expected after-tax distributions have a present value below their initial cost. With low or near- zero payout policies thus infeasible, Black’s argument that payout taxes should largely eliminate cash distributions is incorrect. The conditions for optimal payout policy mirror those in Section 3, but with distributions to stockholders now specified on an after-tax basis: the firm should adopt investment and payout policies that maximize the current market value of the after-tax distributions to currently outstanding shares. Although payouts must be large in present value terms, is it possible that the tax- efficient optimal policy for firms is to defer payouts for as long as possible, making one or a few massive distributions far in the future? Yes, it is possible, but it is unlikely unless firms can avoid immediate payout taxes while satisfying aggregate consumption demand in other ways. DeAngelo (1991, Section I) considers an airtight tax code through which all current payouts to stockholders are taxed without fail. He applies Miller’s (1977) logic to show that market prices adjust to encourage firms to distribute cash to meet immediate aggregate consumption demand, even when retention implies both that no current taxes are due (because nothing is paid out today) and that future payouts escape taxation. DeAngelo’s argument is analogous to the standard price theory analysis of a unit tax on production under perfect competition. The tax raises the marginal cost of an immediate payout, which reduces the market-clearing quantity, but not to near-zero levels because in the aggregate investors typically demand substantial consumption in each period (since consumption claims at different dates are by nature imperfect substitutes). If consumption demand can be satisfied in other ways by firms that circumvent the payout tax, the equilibrium unravels and firms will make massive payouts to stockholders at some point in the far distant future. For example, suppose the tax authorities allow firms to make unlimited zero-interest rate (non-taxable) loans to current stockholders at each date until some distant horizon, T. Firms will use such loans to satisfy consumption demand at each date prior to T. At date T, stockholders will repay the loans and immediately receive liquidating distributions from firms, which owe no corporate tax at this or any other date, since the interest
H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–315 311
rate on the loans is zero. Of course, the IRS is unlikely to sit idly by if such tax- evasion strategies enable firms and investors to avoid billions of dollars in taxes. And so the most plausible tax-based model lies somewhere between the airtight tax code analysis of DeAngelo (1991) and unfettered tax deferral. In the middle ground, firms make some taxable payouts in periods prior to T, but also engage to some degree in retention strategies that reduce the overall tax bite on the unequivocally large (in present value terms) distributions they must provide to maximize stockholder wealth. But Fischer Black was not puzzled about the extent to which firms adopt reasonable strategies to temporarily defer some taxable payouts to stockholders. Rather, he was puzzled that firms make any taxable payouts at all given the MM (1961) result that, in the absence of taxes, firms can largely avoid doing so forever without reducing stockholder wealth (see our footnote 1). Black began his article with the question ‘‘Why do corporations pay dividends?’’ and ended with the answer ‘‘We don’t know.’’ Had he followed his opening question not with a statement that dividend policy is irrelevant in frictionless markets, but with recognition that ‘‘value maximization requires firms to distribute the full PV of FCF,’’ he surely would not have found large taxable payouts puzzling, although he (and the rest of us) would probably have continued to wonder why tax-advantaged repurchases don’t constitute a larger portion of the massive taxable payouts that firms must deliver in order to maximize stockholder wealth.
- Summary and implications
Contrary to Miller and Modigliani (1961), payout policy is not irrelevant and investment policy is not the sole determinant of value in frictionless markets. MM’s assumptions force 100% FCF payout, thereby restricting the feasible set of payout policies to those that are optimal and eliminating the value-relevant payout/retention decision from consideration. When MM’s assumptions are modified to allow retention with the NPV of investment policy fixed, payout policy matters and investment policy is not the sole determinant of value because some now-feasible payout policies distribute less than the full PV of FCF. Because irrelevance is a property of the opportunity set (‘‘all feasible decisions are optimal’’), payout policy (like investment policy) remains relevant in the standard Fisherian model, even though that model’s value-maximization assumption ensures that managers will never make suboptimal payouts (or take negative-NPV projects). In short, payout policy inherently affects stockholder wealth, and not only when it affects project choice or because of market imperfections such as personal taxes. Although MM (1958, 1961) deserve enormous credit for providing the foundational framework for modern corporate finance theory, more than 40 years later a needless disconnect still exists between the perceived implications of the Fisherian model and the beliefs of corporate managers, investors, and students. From day one, the MM principle that ‘‘only investment policy counts’’ met resistance from practitioners who believed that payout policy also matters. Miller
312 H. DeAngelo, L. DeAngelo / Journal of Financial Economics 79 (2006) 293–
