6

The Capital Structure Puzzle

Another Look at the Evidence

Michael J. Barclay and Clifford W. Smith, Jr.

A perennial debate in corporate finance concerns the question of optimal capital structure: Given a level of total capital necessary to support a company's activities, is there a way of dividing up that capital into debt and equity that maximizes current firm value? And, if so, what are the critical factors in setting the leverage ratio for a given company?

Although corporate finance has been taught in business schools for almost a century, the academic finance profession has found it remarkably difficult to provide definitive answers to these questions – answers that can guide practicing corporate executives in making their financing decisions. Part of the difficulty stems from how the discipline of finance has evolved. For much of this century, both the teaching of finance and the supporting research were dominated by the case-study method. In effect, finance education was a glorified apprenticeship system designed to convey to students the accepted wisdom – often codified in the form of rules of thumb – of successful practitioners. Such rules of thumb may have been quite effective in a given set of circumstances, but as those circumstances change over time such rules tend to degenerate into dogma. An example was Eastman Kodak's long-standing decision to shun debt financing – a policy stemming from George Eastman's brush with insolvency at the turn of the century that was not reversed until the 1980s.

But this “anecdotal” approach to the study of finance is changing. In the past few decades, financial economists have worked to transform corporate finance into a more scientific undertaking, with a body of formal theories that can be tested by empirical studies of market and corporate behavior. The ultimate basis for judging the usefulness of a theory is, of course, its consistency with the facts – and thus its ability to predict actual behavior. But this brings us to the most important obstacle to developing a definitive theory of capital structure: namely, the difficulty of designing empirical tests that are powerful enough to distinguish among the competing theories.

What makes the capital structure debate especially intriguing is that the different theories represent such different, and in some ways almost diametrically opposed, decision-making processes. For example, some finance scholars have followed Miller and Modigliani by arguing that both capital structure and dividend policy are largely “irrelevant” in the sense that they have no significant, predictable effects on corporate market values. Another school of thought holds that corporate financing choices reflect an attempt by corporate managers to balance the tax shields of greater debt against the increased probability and costs of financial distress, including those arising from corporate underinvestment. But if too much debt can destroy value by causing financial distress and underinvestment, others have argued that too little debt – at least in large, mature companies – can lead to overinvestment and low returns on capital.

Still others argue that corporate managers making financing decisions are concerned primarily with the “signaling” effects of such decisions – for example, the tendency of stock prices to fall significantly in response to common stock offerings (which can make such offerings very expensive for existing shareholders) and to rise in response to leverage-increasing recapitalizations. Building on this signaling argument, MIT professor Stewart Myers has suggested that corporate capital structures are simply the cumulative result of individual financing decisions in which managers follow a financial pecking order – one in which retained earnings are preferred to outside financing, and debt is preferred to equity when outside funding is required. According to Myers, corporate managers making financing decisions are not really thinking about an optimal capital structure – that is, a long-run targeted debt-to-equity ratio they eventually want to achieve. Instead, they simply take the “path of least resistance” and choose what then appears to be the low-cost financing vehicle – generally debt – with little thought about the future consequences of these choices.

In his 1984 speech to the American Finance Association in which he first presented the pecking order theory, Professor Myers referred to this conflict among the different theories as the “capital structure puzzle.” As we already suggested, the greatest barrier to progress in solving the puzzle has been the difficulty of devising conclusive tests of the competing theories. Over 30 years ago, researchers in the capital markets branch of finance, with its focus on portfolio theory and asset pricing, began to develop models in the form of precise mathematical formulas that predict the values of traded financial assets as a function of a handful of (mainly) observable variables. The predictions generated by such models, after continuous testing and refinement, have turned out to be remarkably accurate and useful to practitioners. For example, the Black-Scholes option pricing model – variations of which have long been widely used on options exchanges – has enabled traders to calculate the value of traded options of all kinds as a function of just six variables (all but one of which can be directly observed).

The key to financial economists' success in capital markets is this: Armed with specific hypotheses, they have been able to develop sophisticated and powerful empirical tests. The evidence from such tests has in turn allowed theorists to increase the “realism” of their models to the point where they have been used, and in some cases further refined, by practitioners. And while no one would argue that all major asset pricing issues have been resolved, the continuing interaction between theory and testing has yielded a richer understanding of risk-return tradeoffs than anyone might have imagined decades ago.

Empirical methods in corporate finance have lagged behind those in capital markets for several reasons. First, our models of capital structure decisions are less precise than asset pricing models. The major theories focus on the ways that capital structure choices are likely to affect firm value. But rather than being reducible, like the option pricing model, to a precise mathematical formula, the existing theories of capital structure provide at best qualitative or directional predictions. They generally identify major factors like taxes or bankruptcy costs that would lead to an association between particular firm characteristics and higher or lower leverage. For example, the tax-based theory of capital structure suggests that firms with more non-interest tax shields (like investment tax credits) should have less debt in their capital structures; but the theory does not tell us how much less.

Second, most of the competing theories of optimal capital structure are not mutually exclusive. Evidence consistent with one theory – say, the tax-based explanation – generally does not allow us to conclude that another factor – the value of debt in reducing overinvestment by mature companies – is unimportant. In fact, it seems clear that taxes, bankruptcy costs (including incentives for underinvestment), and information costs all play some role in determining a firm's optimal capital structure. With our current tests, it is generally not possible to reject one theory in favor of another.

Third, many of the variables that we think affect optimal capital structure are difficult to measure. For example, signaling theory suggests that the managers' “private” information about the company's prospects plays an important role in their financing choices. But, since there is no obvious way to identify when managers have such proprietary information, it is hard to test this proposition.

For all of these reasons and others, the state of the art in corporate finance is less developed than in asset pricing. Thus it is important for the academic community to continue to develop the theory to yield more precise predictions, and to devise more powerful empirical tests as well as better proxies for the key firm characteristics that are likely to drive corporate financing decisions.

In this paper, we offer our assessment of the current state of the academic finance profession's understanding of these issues and suggest some new directions for further exploration. We also offer in closing what we feel is a promising approach to reconciling the different theories of capital structure.

THE THEORIES¹

Current explanations of corporate financial policy can be grouped into three broad categories: (1) taxes, (2) contracting costs, and (3) information costs. Before discussing these theories, it is important to keep in mind that they are not mutually exclusive and that each is likely to help us understand at least particular facets of corporate financing. Our aim is to determine the relative importance of the different theories and to identify those aspects of financial policy that each theory is most helpful in explaining.

Taxes

The basic corporate profits tax allows the deduction of interest payments but not dividends in the calculation of taxable income. For this reason, adding debt to a company's capital structure lowers its expected tax liability and increases its after-tax cash flow. If there were only a corporate profits tax and no individual taxes on corporate securities, the value of a levered firm would equal that of an identical all-equity firm plus the present value of its interest tax shields. That present value, which represents the contribution of debt financing to the market value of the firm, could be estimated simply by multiplying the company's marginal tax rate (34% plus state and local rates) times the principal amount of outstanding debt (assuming the firm expects to maintain its current debt level).

The problem with this analysis, however, is that it overstates the tax advantage of debt by considering only the corporate profits tax. Many investors who receive interest income must pay taxes on that income. But those same investors who receive equity income in the form of capital gains are taxed at a lower rate and can defer any tax by choosing not to realize those gains. Thus, although higher leverage lowers the firm's corporate taxes, it increases the taxes paid by investors. And, because investors care about their after-tax returns, they require compensation for these increased taxes in the form of higher yields on corporate debt – higher than the yields on, say, comparably risky tax-exempt municipal bonds.

The higher yields on corporate debt that reflect investors' taxes effectively reduce the tax advantage of debt over equity. In this sense, the company's shareholders ultimately bear all of the tax consequences of its operations, whether the company pays those taxes directly in the form of corporate income tax or indirectly in the form of higher required rates of return on the securities it sells. For this reason alone,² the tax advantage of corporate debt is almost certainly not 34 cents for every dollar of debt. Nor is it likely to be zero, however, and so a consistently profitable company that volunteers to pay more taxes by having substantial unused debt capacity is likely to be leaving considerable value on the table.

Contracting Costs

Conventional capital structure analysis holds that financial managers set leverage targets by balancing the tax benefits of higher leverage against the greater probability, and thus higher expected costs, of financial distress. In this view, the optimal capital structure is the one in which the next dollar of debt is expected to provide an additional tax subsidy that just offsets the resulting increase in expected costs of financial distress.

Costs of Financial Distress (or the Underinvestment Problem). Although the direct expenses associated with the administration of the bankruptcy process appear to be quite small relative to the market values of companies,³ the indirect costs can be substantial. In thinking about optimal capital structure, the most important indirect costs are likely to be the reductions in firm value that result from cutbacks in promising investment that tend to be made when companies get into financial difficulty.

When a company files for bankruptcy, the bankruptcy judge effectively assumes control of corporate investment policy – and it's not hard to imagine circumstances in which judges do not maximize firm value. But even in conditions less extreme than bankruptcy, highly leveraged companies are more likely than their low-debt counterparts to pass up valuable investment opportunities, especially when faced with the prospect of default. In such cases, corporate managers are likely not only to postpone major capital projects, but to make cutbacks in R&D, maintenance, advertising, or training that end up reducing future profits.

This tendency of companies to underinvest when facing financial difficulty is accentuated by conflicts that can arise among the firm's different claimholders. To illustrate this conflict, consider what might happen to a high-growth company that had trouble servicing its debt. Since the value of such a firm will depend heavily on its ability to carry out its long-term investment plan, what the company needs is an infusion of equity. But there is a problem. As Stewart Myers pointed out in his classic 1977 paper entitled “Determinants of Corporate Borrowing,”⁴ the investors who would be asked to provide the new equity in such cases recognize that much of the value created (or preserved) by their investment would go to restoring the creditors' position. In this situation, the cost of the new equity could be so high that managers acting on their shareholders' behalf might rationally forgo both the capital and the investment opportunities.

Myers referred to this as “the underinvestment problem.” And, as he went on to argue, companies whose value consists primarily of intangible investment opportunities – or “growth options,” as he called them – will choose low-debt capital structures because such firms are likely to suffer the greatest loss in value from this underinvestment problem. By contrast, mature companies with few profitable investment opportunities where most of their value reflects the cash flows from tangible “assets in place” incur lower expected costs associated with financial distress. Such mature companies, all else equal, should have significantly higher leverage ratios than high-growth firms.

The Benefits of Debt in Controlling Overinvestment. If too much debt financing can create an underinvestment problem for growth companies, too little debt can lead to an overinvestment problem in the case of mature companies. As Michael Jensen has argued,⁵ large, mature public companies generate substantial “free cash flow” – that is, operating cash flow that cannot be profitably reinvested inside the firm. The natural inclination of corporate managers is to use such free cash flow to sustain growth at the expense of profitability, either by overinvesting in their core businesses or, perhaps worse, by diversifying through acquisition into unfamiliar ones.

Because both of these strategies tend to reduce value, companies that aim to maximize firm value must distribute their free cash flow to investors. Raising the dividend is one way of promising to distribute excess capital. But major substitutions of debt for equity (for example, in the form of leveraged stock repurchases) offer a more reliable solution because contractually obligated payments of interest and principal are more effective than discretionary dividend payments in squeezing out excess capital. Thus, in industries generating substantial cash flow but facing few growth opportunities, debt financing can add value simply by forcing managers to be more critical in evaluating capital spending plans.⁶

Information Costs

Corporate executives often have better information about the value of their companies than outside investors. Recognition of this information disparity between managers and investors has led to two distinct, but related theories of financing decisions – one known as “signaling,” the other as the “pecking order.”

Signaling. With better information about the value of their companies than outside investors, managers of undervalued firms would like to raise their share prices by communicating this information to the market. Unfortunately, this task is not as easy as it sounds; simply announcing that the companies are undervalued generally isn't enough. The challenge for managers is to find a credible signaling mechanism.

Economic theory suggests that information disclosed by an obviously biased source (like management, in this case) will be credible only if the costs of communicating falsely are large enough to constrain managers to reveal the truth. Increasing leverage has been suggested as one potentially effective signaling device. Debt contracts oblige the firm to make a fixed set of cash payments over the life of the loan; if these payments are missed, there are potentially serious consequences, including bankruptcy. Equity is more forgiving. Although stockholders also typically expect cash payouts, managers have more discretion over these payments and can cut or omit them in times of financial distress.

For this reason, adding more debt to the firm's capital structure can serve as a credible signal of higher future cash flows.⁷ By committing the firm to make future interest payments to bondholders, managers communicate their confidence that the firm will have sufficient cash flows to meet these obligations.

Debt and equity also differ with respect to their sensitivity to changes in firm value. Since the promised payments to bondholders are fixed, and stockholders are entitled to the residual (or what's left over after the fixed payments), stock prices are much more sensitive than bond prices to any proprietary information about future prospects. If management is in possession of good news that has yet to be reflected in market prices, the release of such news will cause a larger increase in stock prices than in bond prices; and hence current stock prices (prior to release of the new information) will appear more undervalued to managers than current bond prices. For this reason, signaling theory suggests that managers of companies that believe their assets are undervalued will generally choose to issue debt – and to use equity only as a last resort.

To illustrate this with a simple example, let's suppose that the market price of a stock is $25.00. Investors understand that its “real” value – that is, the value they would assign if they had access to the same information as the firm's managers – might be as high as $27.00 or as low as $23.00; but given investors' information $25.00 is a fair price. Now let's suppose that the managers want to raise external funds and they could either sell equity or debt. If the managers think the stock is really worth only $23.00, selling shares for $25.00 would be attractive – especially if their compensation is tied to stock appreciation. But if the managers think the stock is really worth $27.00, equity would be expensive at $25.00 and debt would be more attractive.

Investors understand this – and so if the company announces an equity offer, investors reassess the current price in light of this new information. Since it is more likely that the stock is worth $23.00 than $27.00, the market price declines. Such a rapid adjustment in valuation associated with the announcement thus eliminates much of any potential gain from attempting to exploit the manager's superior information.

Consistent with this example, economists have documented that the market responds in systematically negative fashion to announcements of equity offerings, marking down the share prices of issuing firms by about 3% on average. By contrast, the average market reaction to new debt offerings is not significantly different from zero.⁸ The important thing to recognize is that most companies issuing new equity – those that are undervalued as well as those that are overvalued – can expect a drop in stock prices when they announce the offering. For those firms that are fairly valued or undervalued prior to the announcement of the offering, this expected drop in value represents an economic dilution of the existing shareholders' interest. Throughout the rest of this paper, we refer to this dilution as part of the “information costs” of raising outside capital.

The Pecking Order. Signaling theory, then, says that financing decisions are based, at least in part, on management's perception of the “fairness” of the market's current valuation of the stock. Stated as simply as possible, the theory suggests that, in order to minimize the information costs of issuing securities, a company is more likely to issue debt than equity if the firm appears undervalued, and to issue stock rather than debt if the firm seems overvalued.

The pecking order theory takes this argument one step farther, suggesting that the information costs associated with issuing securities are so large that they dominate all other considerations. According to this theory, companies maximize value by systematically choosing to finance new investments with the “cheapest available” source of funds. Specifically, they prefer internally generated funds (retained earnings) to external funding and, if outside funds are necessary, they prefer debt to equity because of the lower information costs associated with debt issues. Companies issue equity only as a last resort, when their debt capacity has been exhausted.⁹

The pecking order theory would thus suggest that companies with few investment opportunities and substantial free cash flow will have low debt ratios – and that high-growth firms with lower operating cash flows will have high debt ratios. In this sense, the theory not only suggests that interest tax shields and the costs of financial distress are at most a second-order concern; the logic of the pecking order actually leads to a set of predictions that are precisely the opposite of those offered by the tax and contracting cost arguments presented above.

THE EVIDENCE

Having discussed the different theories for observed capital structure, we now review the available empirical evidence to assess the relative “explanatory power” of each.

Evidence on Contracting Costs

Leverage Ratios. Much of the previous evidence on capital structure supports the conclusion that there is an optimal capital structure and that firms make financing decisions and adjust their capital structures to move closer to this optimum. For example, a 1967 study by Eli Schwartz and Richard Aronson showed clear differences in the average debt to (book) asset ratios of companies in different industries, as well as a tendency for companies in the same industry to cluster around these averages.¹⁰ Moreover, such industry debt ratios seem to align with R&D spending and other proxies for corporate growth opportunities that the theory suggests are likely to be important in determining an optimal capital structure. In a 1985 study, Michael Long and Ileen Malitz showed that the five most highly leveraged industries – cement, blast furnaces and steel, paper and allied products, textiles, and petroleum refining – were all mature and asset-intensive. At the other extreme, the five industries with the lowest debt ratios – cosmetics, drugs, photographic equipment, aircraft, and radio and TV receiving – were all growth industries with high advertising and R&D.¹¹

Other studies have used “cross-sectional” regression techniques to test whether the theoretical determinants of an optimal capital structure actually affect financing decisions. For example, in their 1984 study, Michael Bradley, Greg Jarrell, and Han Kim found that the debt to (book) asset ratio was negatively related to both the volatility of annual operating earnings and to advertising and R&D expenses. Both of these findings are consistent with high costs of financial distress for growth companies, which tend to have more volatile earnings as well as higher spending on R&D.¹²

Several studies have also reported finding that the debt ratios of individual companies seem to revert toward optimal targets. For example, a 1982 study by Paul Marsh estimated a company's target ratio as the average ratio observed over the prior ten years. He then found that the probability that a firm issues equity is significantly higher if the firm is above its target debt ratio, and significantly lower if below the target.¹³

As described in a 1995 article in this journal, we (together with colleague Ross Watts) attempted to add to this body of empirical work on capital structure by examining a much larger sample of companies that we tracked for over three decades.¹⁴ For some 6,700 companies covered by Compustat, we calculated “market” leverage ratios (measured as the book value of total debt divided by the book value of debt and preferred stock plus the market value of equity) over the period 1963–1993. Not surprisingly, we found considerable differences in leverage ratios, both across companies in any given year and, in some cases, for the same firm over time. Although the average leverage ratio for the 6700 companies over the 30-year period was 25%, one fourth of the cases had market leverage ratios that were higher than 37.5% and another one fourth had leverage ratios less than 10.3%.

To test the contracting cost theory described earlier in this paper, we attempted to determine the extent to which corporate leverage choices can be explained by differences in companies' investment opportunities. As suggested earlier, the contracting cost hypothesis predicts that the greater these investment opportunities (relative to the size of the company), the greater the potential underinvestment problem associated with debt financing and, hence, the lower the company's target leverage ratio. Conversely, the more limited a company's growth opportunities, the greater the potential overinvestment problem and, hence, the higher should be the company's leverage.

To test this prediction, we needed a measure of investment opportunities. Because stock prices reflect intangible assets such as growth opportunities but corporate balance sheets do not, we reasoned that the larger a company's “growth options” relative to its “assets in place,” the higher on average will be its market value in relation to its book value. We accordingly used a company's market-to-book ratio as our proxy for its investment opportunity set.

The results of our regressions provide strong support for the contracting cost hypothesis. Companies with high market-to-book ratios had significantly lower leverage ratios than companies with low market-to-book ratios. (The t-statistic on the market-to-book ratio in the leverage regression was about 130.) To make these findings a little more concrete, our results suggest that, as one moves form companies at the bottom 10th percentile of market-to-book ratios (0.77) to the 90th percentile (2.59), the predicted leverage market ratio falls by 14.3 percentage points – which is a large fraction of the average ratio of 25%. (For further discussion of these results, see the box on the next page.)

Moreover, such a negative relation between corporate leverage and market-to-book ratios appears to hold outside the U.S. as well. In a 1995 study, Raghuram Rajan and Luigi Zingales examined capital structure using data from Japan, Germany, France, Italy, the U.K. and Canada, as well as the U.S. They found that, in each of these seven countries, leverage is lower for firms with higher market-to-book ratios and higher for firms with higher ratios of fixed assets to total assets.¹⁵

The above evidence on leverage ratios, it should be pointed out, is also generally consistent with the tax hypothesis in the following sense: The same low-growth companies that face low financial distress costs and high free-cash-flow benefits from heavy debt financing are also likely to have greater use for interest tax shields than high-growth companies. At the same time, the above evidence is inconsistent with the predictions of the pecking order theory – which, again, suggests that low-growth firms with high free cash flow will have relatively low debt ratios.

Debt Maturity and Priority. Like this article up to this point, most academic discussions of capital structure focus just on the leverage ratio. In so doing, they effectively assume that all debt financing is the same. In practice, of course, debt differs in several important respects, including maturity, covenant restrictions, security, convertibility and call provisions, and whether the debt is privately placed or held by widely dispersed public investors. Each of these features is potentially important in determining the extent to which debt financing can cause, or exacerbate, a potential under-investment problem. For example, debt-financed companies with more investment opportunities would prefer to have debt with shorter maturities (or at least with call provisions, to ensure greater financing flexibility), more convertibility provisions (which reduce the required coupon payments), less restrictive covenants, and a smaller group of private investors rather than public bondholders (which makes it easier to reorganize in the event of trouble). By recognizing this array of financing choices, we can broaden the scope of our examination and raise the potential power of our tests, while at the same time increasing the relevance of the analysis for managers who must choose the design of their debt securities.

As described in our 1996 article in this journal,¹⁶ we designed an empirical test of the suggestion – offered by Stewart Myers in his 1977 article – that one way for companies with lots of growth options to control the underinvestment problem is to issue debt with shorter maturities. The argument is basically this: A firm whose value consists mainly of growth opportunities could severely reduce its future financing and strategic flexibility – and in the process destroy much of its value – by issuing long-term debt. Not only would the interest rate have to be high to compensate lenders for their greater risk, but the burden of servicing the debt could cause the company to defer strategic investments if their operating cash flow turns down. By contrast, shorter-term debt, besides carrying lower interest rates in such cases, would also be less of a threat to future strategic investment because, as the firm's current investments begin to pay off, it will be able over time to raise capital on more favorable terms.¹⁷

When we tested this prediction (again using market-to-book as a measure of growth options), we found that growth companies tended to have significantly less debt with a maturity greater than three years than companies with limited investment opportunities. More specifically, our regressions suggest that moving from companies at the 10th to the 90th percentile of market-to-book ratios (that is from 0.77 to 2.59) reduces the ratio of long-term debt to total debt by 18 percentage points (a significant reduction, given our sample average ratio of 46%).

Moreover, we also found in the same study that the debt issued by growth firms is significantly more concentrated among high-priority classes. Consistent with our results indicating that firms with more growth options tend to have lower leverage ratios, we find that changing the market-to-book ratio from the 10th to the 90th percentile is associated with reductions in leasing of 89%, in secured debt of 71%, in ordinary debt of 78%, and in subordinated debt of almost 250%. Our explanation for this is as follows: When firms get into financial difficulty, complicated capital structures with claims of different priorities can generate serious conflicts among creditors, thus exacerbating the underinvestment problem described earlier. And because such conflicts and the resulting underinvestment have the greatest potential to destroy value in growth firms, those growth firms that do issue fixed claims are likely to choose mainly high-priority fixed claims.

The Evidence on Information Costs

Leverage. Signaling theory says that companies are more likely to issue debt than equity when they are undervalued because of the large information costs (in the form of dilution) associated with an equity offering. The pecking order model goes even farther, suggesting that the information costs associated with riskier securities are so large that most companies will not issue equity until they have completely exhausted their debt capacity. Neither the signaling nor the pecking order theory offers any clear prediction about what optimal capital structure would be for a given firm. The signaling theory seems to suggest that a firm's actual capital structure will be influenced by whether the company is perceived by management to be undervalued or overvalued. The pecking order model is more extreme; it implies that a company will not have a target capital structure, and that its leverage ratio will be determined by the gap between its operating cash flow and its investment requirements over time. Thus, the pecking order predicts that companies with consistently high profits or modest financing requirements are likely to have low debt ratios – mainly because they don't need outside capital. Less profitable companies, and those with large financing requirements, will end up with high leverage ratios because of managers' reluctance to issue equity.

A number of studies have provided support for the pecking order theory in the form of evidence of a strong negative relation between past profitability and leverage. That is, the lower are a company's profits and operating cash flows in a given year, the higher is its leverage ratio (measured either in terms of book or market values).¹⁸ Moreover, in an article published in 1998, Stewart Myers and Lakshmi Shyam-Sunder added to this series of studies by showing that this relation explains more of the time-series variance of debt ratios than a simple target-adjustment model of capital structure that is consistent with the contracting cost hypothesis.¹⁹

Such findings have generally been interpreted as confirmation that managers do not set target leverage ratios – or at least do not work very hard to achieve them. But this is not the only interpretation that fits these data. Even if companies have target leverage ratios, there will be an optimal deviation from those targets – one that will depend on the transactions costs associated with adjusting back to the target relative to the costs of deviating from the target. To the extent there are fixed costs and scale economies in issuing securities, companies with capital structure targets – particularly smaller firms – will make infrequent adjustments and often will deliberately overshoot their targets. (And, as we argue in the closing section of this paper, a complete theory of capital structure must take account of these adjustment costs and how they affect expected deviations from the target.)

In our 1995 paper with Ross Watts, we attempted to devise our own test of how information costs affect corporate financing behavior. According to the signaling explanation, undervalued companies will have higher leverage than overvalued firms. One major challenge in testing this signaling argument is coming up with a reliable proxy for undervaluation that can be readily observed. In devising such a measure, we began with the assumption that corporate earnings follow a random walk, and that the best predictor of a company's next year's earnings is thus its current year's earnings. We then classified firms as undervalued in any given year in which their earnings (excluding extraordinary items and discontinued operation and adjusted for any changes in shares outstanding) increased in the following year. We designated as overvalued all firms whose ordinary earnings decreased in the next year.

Our regressions showed a very small (but statistically significant) positive relation between a company's leverage ratio and its unexpected earnings, thus suggesting that this undervaluation variable has a trivial effect on corporate capital structure. For example, moving from the 10th percentile of abnormal earnings in our sample (those firms whose earnings decreased by 12%) to the 90th percentile (those whose earnings increased by 13%) raised the predicted leverage ratio by only 0.5 percentage points. Moreover, in our 1996 study (which also uses Compustat data, although for a somewhat different time period), we again found a small relation between leverage and unexpected earnings. In this regression, however, the relation was negative.

Debt Maturity and Priority. Signaling theory implies that undervalued firms will have more short-term debt and more senior debt than overvalued firms because such instruments are less sensitive to the market's assessment of firm value and thus will be less undervalued when issued. The findings of our 1996 study are inconsistent with the predictions of the signaling hypothesis with respect to debt maturity. Companies whose earnings were about to increase the following year in fact issued less short-term debt and more long-term debt than firms whose earnings were about to decrease. And, whereas the theory predicts more senior debt for firms about to experience earnings increases, the ratio of senior debt to total debt is lower for overvalued than for undervalued firms.

In sum, the results of our tests of managers' use of financing choices to signal their superior information to the market are not robust, and the economic effect of any such signaling on corporate decision-making seems minimal.

According to the pecking order theory, the firm should issue as much of the security with the lowest information costs as it can. Only after this capacity is exhausted should it move on to issue a security with higher information costs. Thus, for example, firms should issue as much secured debt or capitalized leases as possible before issuing any unsecured debt, and they should exhaust their capacity for issuing short-term debt before issuing any long-term debt. But these predictions are clearly rejected by the data. For example, when we examined the capital structures of over 7,000 companies between 1980 and 1997 (representing almost 57,000 firm-year observations), we found that 23% of these observations had no secured debt, 54% had no capital leases, and 50% had no debt that was originally issued with less than one year to maturity.

To explain these more detailed aspects of capital structure, proponents of the pecking order theory must go outside their theory and argue that other costs and benefits determine these choices. But once you allow for these other costs and benefits to have a material impact on corporate financing choices, you are back in the more traditional domain of optimal capital structure theories.

The Evidence on Taxes

Theoretical models of optimal capital structure predict that firms with more taxable income and fewer non-debt tax shields should have higher leverage ratios. But the evidence on the relation between leverage ratios and tax-related variables is mixed at best. For example, studies that examine the effect of non-debt tax shields on companies' leverage ratios find that this effect is either insignificant, or that it enters with the wrong sign. That is, in contrast to the prediction of the tax hypothesis, these studies suggest that firms with more non-debt tax shields such as depreciation, net operating loss carryforwards and investment tax credits have, if anything, more not less debt in their capital structures.²⁰

But before we conclude that taxes are unimportant in the capital structure decision, it is critical to recognize that the findings of these studies are hard to interpret because the tax variables are crude proxies for a company's effective marginal tax rate. In fact, these proxies are often correlated with other variables that influence the capital structure choice. For example, companies with investment tax credits, high levels of depreciation, and other non-debt tax shields also tend to have mainly tangible fixed assets. And, since fixed assets provide good collateral, the non-debt tax shields may in fact be a proxy not for limited tax benefits, but rather for low contracting costs associated with debt financing. The evidence from the studies just cited is generally consistent with this interpretation.

Similarly, firms with net operating loss carryforwards are often in financial distress; and, since equity values typically decline in such circumstances, financial distress itself causes leverage ratios to increase. Thus, again, it is not clear whether net operating losses proxy for low tax benefits of debt or for financial distress.

More recently, several authors have succeeded in detecting tax effects in financing decisions by focusing on incremental financing choices (that is, changes in the amount of debt or equity) rather than on the levels of debt and equity. For example, a 1990 study by Jeffrey Mackie-Mason examined registered security offerings by public U.S. corporations and found that firms were more likely to issue debt if they had a high marginal tax rate and to issue equity if they had a low tax rate.²¹ In another attempt to avoid the difficulties with crude proxy variables, a 1996 study by John Graham used a sophisticated simulation method to provide a more accurate measure of companies' marginal tax rates.²² Using such tax rates, Graham also found a positive association between changes in debt ratios and the firm's marginal tax rate.

On balance, then, the evidence appears to suggest that taxes play at least a modest role in corporate financing and capital structure decisions. Moreover, as mentioned earlier, the results of our tests of contracting costs reported above can also be interpreted as evidence in support of the tax explanation.

TOWARD A UNIFIED THEORY OF CORPORATE FINANCIAL POLICY

In addition to explaining the basic leverage (or debt vs. equity) decision, a useful theory of capital structure should also help explain other capital structure choices, such as debt maturity, priority, the use of callability and convertibility provisions, and the choice between public and private financing. As discussed above, the contracting-cost theory provides a unified framework for analyzing the entire range of capital structure choices while most other theories, such as the signaling and pecking order theories, are at best silent about – and more often inconsistent with – the empirical evidence on these issues.

We now take this argument one step further by suggesting that a productive capital structure theory should also help explain an even broader array of corporate financial policy choices, including dividend, compensation, hedging, and leasing policies. The empirical evidence suggests that companies choose coherent packages of these financial policies. For example, small high-growth firms tend to have not only low leverage ratios and simple capital structures (with predominantly short-maturity, senior bank debt), but also low dividend payouts as well as considerable stock-based incentive compensation for senior executives. By contrast, large mature companies tend to have high leverage, more long-term debt, more complicated capital structures with a broader range of debt priorities, higher dividends, and less incentive compensation (with greater reliance on earnings-based bonuses rather than stock-based compensation plans).²³ Thus, corporate financing, dividend, and compensation policies, besides being highly correlated with each other, all appear to be driven by the same fundamental firm characteristics: investment opportunities and (to a lesser extent) firm size. And this consistent pattern of corporate decision-making suggests that we now have the rudiments of a unified framework for explaining most, if not all, financial policy choices.

As mentioned earlier, proponents of the pecking order theory argue that the information costs associated with issuing new securities dominate all other costs in determining capital structure. But, as we also noted, the logic and predictions of the pecking order theory are at odds with, and thus incapable of explaining, most other financial policy choices. For example, in suggesting that firms will always use the cheapest source of funds, the model implies that companies will not simultaneously pay dividends and access external capital markets. But this prediction can, of course, be rejected simply by glancing at the business section of most daily newspapers. With the exception of a few extraordinarily successful high tech companies like Microsoft and Amgen, most large, publicly traded companies pay dividends while at the same time regularly rolling over existing debt with new public issues. And, as already discussed, although the pecking order predicts that mature firms that generate lots of free cash flow should eventually become all equity financed, they are among the most highly levered firms in our sample. Conversely, the pecking order theory implies that high-tech startup firms will have high leverage ratios because they often have negative free cash flow and incur the largest information costs when issuing equity. But, in fact, such firms are financed almost entirely with equity.

Thus, as we saw in the case of debt maturity and priority, proponents of the pecking order must go outside of their theory to explain corporate behavior at both ends of the corporate growth spectrum. In so doing, they implicitly limit the size and importance of information costs; they concede that, at least for the most mature and the highest-growth sectors, information costs are less important than other considerations in corporate financing decisions.

Integration of Stocks and Flows

Although the pecking order theory is incapable of explaining the full array of financial policy choices, this does not mean that information costs are unimportant in corporate decision-making. On the contrary, such costs will influence corporate financing choices and, along with other costs and benefits, must be part of a unified theory of corporate financial policy.

In our view, the key to reconciling the different theories – and thus to solving the capital structure puzzle – lies in achieving a better understanding of the relation between corporate financing stocks and flows. The theories of capital structure discussed in this paper generally focus either on the stocks (that is, on the levels of debt and equity in relation to the target) or on the flows (the decision of which security to issue at a particular time). For example, the primary focus of the contracting-cost theories has been leverage ratios, which are measures of the stocks of debt and equity. By contrast, information-based theories like the pecking order model generally focus on flows – for example, on the information costs associated with a new issue of debt or equity. But, since both stocks and flows are likely to play important roles in such decisions, neither of these theoretical approaches taken alone is likely to offer a reliable guide to optimal capital structure.

In developing a sensible approach to capital structure strategy, the CFO should start by thinking about the firm's target capital structure in terms of stock measures – that is, a ratio of debt to total capital that can be expected to minimize taxes and contracting costs (although information costs may also be given some consideration here). That target ratio should take into consideration factors such as the company's projected investment requirements; the level and stability of its operating cash flows; its tax status; the expected loss in value from being forced to defer investment because of financial distress; and the firm's ability to raise capital on short notice (without excessive dilution).

If the company is not currently at or near its optimal capital structure, the CFO should come up with a plan to achieve the target debt ratio. For example, if the firm has “too much” equity (or too much capital in general), it can increase leverage by borrowing (or using excess cash) to buy back shares – a possibility that the pecking order generally ignores. (And the fact that U.S. corporate stock repurchases have been growing at almost 30% per year for most of this decade is by itself perhaps the single most compelling piece of evidence that corporate managers are thinking in terms of optimal capital structure.) But, if the company needs more capital, then managers choosing between equity and various forms of debt must consider not only the benefits of moving toward the target, but also the associated adjustment costs. For example, a company with “too much” debt may choose to delay an equity offering – or issue convertibles or PERCS instead – in order to reduce the cost of issuing securities that it perceives to be undervalued.

As a more general principle, the CFO should adjust the firm's capital structure whenever the costs of adjustment – including information costs as well as out-of-pocket transactions costs – are less than the costs of deviating from the target. Based on the existing research, what can we say about such adjustment costs? The available evidence on the size and variation of such costs suggests that there is a material fixed component – one that again includes information costs as well as out-of-pocket costs.²⁴ And, since average adjustment costs fall with increases in transaction size, there are scale economies in issuing new securities that suggest that small firms, all else equal, are likely to deviate farther from their capital structure targets than larger companies.

Although the different kinds of external financing all exhibit scale economies, the structure of the costs varies among different types of securities. Equity issues have both the largest out-of-pocket transactions costs and the largest information costs. Long-term public debt issues, particularly for below-investment-grade companies, are less costly.²⁵ Short-term private debt or bank loans are the least costly. And, because CFOs are likely to weigh these adjustment costs against the expected benefits from moving closer to their leverage target, it is not surprising that seasoned equity offerings are rare events, that long-term debt issues are more common, and that private debt offerings or bank loans occur with almost predictable regularity. Moreover, because of such adjustment costs, most companies – particularly smaller firms – are also likely to spend considerable time away from their target capital structures. Other things equal, larger adjustment costs will lead to larger deviations from the target before the firm readjusts.

In sum, to make a sensible decision about capital structure, CFOs must understand both the costs associated with deviating from the target capital structure and the costs of adjusting back toward the target. The next major step forward in solving the capital structure puzzle is almost certain to involve a more formal weighing of these two sets of costs.

NOTES

1. This section draws on the discussion of capital structure theory in Michael J. Barclay, Clifford W. Smith, Jr. and Ross L. Watts “The Determinants of Corporate Leverage and Dividend Policies,” Journal of Applied Corporate Finance, Vol. 7 No. 4 (Winter 1995).

2. The extent to which a company benefits from interest tax shields also depends on whether it has other tax shields. For example, holding all else equal, companies with more investment tax credits or tax loss carryforwards should have lower leverage ratios to reflect the lower value of their debt tax shields. See Harry DeAngelo and Ronald Masulis, “Optimal Capital Structure Under Corporate and Personal Taxation,” Journal of Financial Economics, Vol. 8 No. 1 (1980), pp. 3–29.

3. Perhaps the best evidence to date on the size of direct bankruptcy costs comes from Jerry Warner's study of 11 railroads that declared bankruptcy over the period 1930–1955. (Jerold B. Warner, “Bankruptcy Costs: Some Evidence,” Journal of Finance, Vol. 32 (1977), pp. 337–347.) The study reported that out-of-pocket expenses associated with the administration of the bankruptcy process were quite small relative to the market value of the firm – less than 1% for the larger railroads in the sample. For smaller companies, it's true, direct bankruptcy costs are a considerably larger fraction of firm value (about five times larger in Warner's sample). Thus there are “scale economies” with respect to direct bankruptcy costs that imply that larger companies should have higher leverage ratios, all else equal, than smaller firms. But, even these higher estimates of direct bankruptcy costs, when weighted by the probability of getting into bankruptcy in the first place, produce expected costs that appear far too low to make them an important factor in corporate financing decisions.

4. Stewart C. Myers, “Determinants of Corporate Borrowing,” Journal of Financial Economics, Vol. 5 (1977), pp. 147–175.

5. See Michael C. Jensen, “Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers,” American Economic Review 76 (1986), pp. 323–329.

6. More generally, the use of debt rather than equity reduces what economists call the agency costs of equity – loosely speaking, the reduction in firm value that arises from the separation of ownership from control in large, public companies with widely dispersed shareholders. In high-growth firms, the risk-sharing benefits of the corporate form are likely to outweight these agency costs. But, in mature industries with limited capital requirements, heavy debt financing has the added benefit of facilitating the concentration of equity ownership. To illustrate this potential role of debt, assume that the new owner of an all-equity company with $100 million of assets discovers that the assets can support $90 million of debt. Reducing the firm's equity from $100 million to $10 million greatly increases the ability of small investor groups (including management) to control large asset holdings.

The concentration of ownership made possible by leverage appears to have been a major part of the value gains achieved by the LBO movement of the ′80s, and which has been resurrected in the 1990s. And, to the extent there are gains from having more concentrated ownership (and, again, these are likely to be greatest for mature industries with assets in place), companies should have higher leverage ratios.

7. Stephen Ross, “The Determinination of Financial Structure: The Incentive Signaling Approach,” Bell Journal of Economics, Vol. 8 (1977), pp. 23–40.

8. More generally, the evidence suggests that leverage-increasing transactions are associated with positive stock price reactions while leverage-reducing transactions are associated with negative reactions. In reaction to large debt-for-stock exchanges, for example, stock prices go up by 14% on average. The market also reacts in a predictably negative way to leverage-reducing transactions, with prices falling by 9.9% in response to common-for-debt exchanges and by 7.7% in preferred-for-debt exchanges. For a review of this evidence, see Clifford Smith, “Investment Banking and the Capital Acquisition Process,” Journal of Financial Economics, Vol. 15 (1986), pp. 3–29.

9. See Stewart Myers, “The Capital Structure Puzzle,” Journal of Finance, 39 (1984), pp. 575–592.

10. Eli Schwartz and J. Richard Aronson, “Some Surrogate Evidence in Support of Optimal Financial Structure,” Journal of Finance Vol. 22 No. 1 (1967).

11. Michael Long and Ileen Malitz, “The Investment-Financing Nexus: Some Empirical Evidence,” Midland Corporate Finance Journal, Vol. 3 No. 3 (1985).

12. Michael Bradley, Greg Jarrell, and E. Han Kim, “The Existence of an Optimal Capital Structure: Theory and Evidence,” Journal of Finance, Vol. 39 No. 3 (1984).

13. Paul Marsh, “The Choice Between Equity and Debt,” Journal of Finance, 37, (1982), pp. 121–144.

14. Barclay, Smith, and Watts (1995), cited above.

15. See Raghuram Rajan and Luigi Zingales, “What Do We Know About Capital Structure? Some Evidence From International Data,” Journal of Finance, Vol. 50 No. 5 (1995). These relations are statistically significant for each country for the coefficient on growth options and for every country but France and Canada for the coefficient on assets in place.

16. Michael J. Barclay and Clifford W. Smith, Jr., “On Financial Architecture: Leverage, Maturity, and Priority,” Journal of Applied Corporate Finance, Vol. 8 No. 4 (1996).

17. If the firm's debt matures before a company's growth options must be exercised, the investment distortions created by the debt are eliminated. Since these investment distortions are most severe, and most costly, for firms with significant growth options, high-growth firms should use more short-term debt.

18. See, for example, Carl Kester, “Capital and Ownership Structure: A Comparison of United States and Japanese Manufacturing Corporations,” Financial Management, Vol. 15 (1986); Rajan and Zingales (1995); and Sheridan Titman and Roberto Wessels, “The Determinants of Capital Structure Choice,” Journal of Finance, 43 (1988), pp. 1–19.

19. Lakshmi Shyam-Sunder and Stewart Myers, “Testing Static Tradeoff Against Pecking Order Models of Capital Structure,” Journal of Financial Economics, Vol. 51 No. 2.

20. See, for example, Bradley, Jarrell, and Kim (1984); Titman and Wessels (1988); and Barclay, Smith, and Watts (1995), all of which are cited above.

21. Jeffrey Mackie-Mason, “Do Taxes Affect Corporate Financing Decisions?”, Journal of Finance, 45 (1990), pp. 1471–1494.

22. John Graham, “Debt and the Marginal Tax Rate”, Journal of Financial Economics, 41 (1996), pp. 41–73.

23. See Clifford W. Smith and Ross L. Watts, “The Investment Opportunity Set and Corporate Financing, Dividend and Compensation Policies,” Journal of Financial Economics, 32 (1992), pp. 263–292.

24. See, for example, David Blackwell and David Kidwell, “An Investigation of Cost Differences Between Private Placements and Public Sales of Debt,” Journal of Financial Economics, 22 (1988), pp. 253–278; and Clifford Smith, “Alternative Methods for Raising Capital: Rights vs. Underwritten Offerings,” Journal of Financial Economics, 5 (1977), pp. 273–307.

25. See, Sudip Datta, Mai Iskandar-Datta, and Ajay Patel, “The Pricing of Debt IPOs,” Journal of Applied Corporate Finance, Vol. 12 No. 1 (Spring 1999).