Financing decisions: who issues stock? $

Transcription

1 Journal of Financial Economics 76 (2005) Financing decisions: who issues stock? $ Eugene F. Fama a, Kenneth R. French b, a Graduate School of Business, University of Chicago, Chicago, IL 60637, USA b Amos Tuck School of Business, Dartmouth College, Hanover, NH 03755, USA Available online 30 December 2004 Abstract Financing decisions seem to violate the central predictions of the pecking order model about how often and under what circumstances firms issue equity. Specifically, most firms issue or retire equity each year, and the issues are on average large and not typically done by firms under duress. We estimate that during , the year-by-year equity decisions of more than half of our sample firms violate the pecking order. r 2004 Elsevier B.V. All rights reserved. JEL classification: G132 Keywords: Financing decisions; Pecking order; Asymmetric information; Equity issues; Repurchases 1. Introduction The modern corporate finance literature focuses on two competing models to explain the financing decisions of firms. In the tradeoff model, firms identify optimal leverage by weighing the costs and benefits of an additional dollar of debt. The benefits of debt include, for example, the tax deductibility of interest and the $ We acknowledge the helpful comments of Elroy Dimson, Michael Katz, Stewart Myers, Richard Roll, Richard Sansing René Stulz, workshop participants at Dartmouth, the University of Chicago, Yale, and the NBER, and especially Clyde Stickney and two anonymous referees. We thank Michael McWilliams and especially Savina Rizova for excellentresearch assistance. Corresponding author. address: kfrench@dartmouth.edu (K.R. French) X/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi: /j.jfineco

2 550 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) reduction of free-cash-flow agency problems. The costs of debt include potential bankruptcy costs and agency conflicts between stockholders and bondholders. At the leverage optimum, the benefit of the last dollar of debt just offsets the cost. Myers (1984) advocates an alternative theory, the pecking order model. The pecking order arises if the costs of issuing risky securities transactions costs and especially the costs created by management s superior information about the value of the firm s risky securities overwhelm the costs and benefits proposed by the tradeoff model. The costs of issuing risky securities spawn the pecking order: firms finance new investments first with retained earnings, then with safe debt, then risky debt, and finally, but only under duress, with outside equity. The pecking order sequence for financing decisions leads to a prediction about capital structures. Specifically, variation in a firm s leverage is driven not by the tradeoff model s costs and benefits of debt or equity, but more simply, by the firm s financing deficit (dividends plus investment outlays minus earnings). Quoting Myers (1984), The crucial difference between this and the static tradeoff story is that, in the modified pecking order story, observed debt ratios will reflect the cumulative requirement for external financing a requirement cumulated over an extended period. In short, Myers (1984) presents the pecking order model as a theory both about how firms finance themselves and about the capital structures that result from pecking order financing. Subsequent tests of the model follow these two routes. For example, Shyam-Sunder and Myers (1999), Fama and French (2002), and Frank and Goyal (2003) test the model s predictions about the securities firms issue to cover financing deficits, while Titman and Wessels (1988), Rajan and Zingales (1995), Shyam-Sunder and Myers (1999), Fama and French (2002), and Huang and Song (2003) test the model s predictions about capital structures. This earlier work mainly uses cross-section regressions to test the pecking order model. Cross-section regressions measure average responses of financing decisions and capital structures to variables such as growth and profitability (the ingredients of the financing deficit). But average responses may conceal important details relevant for judging the model. We take a more direct approach. We test pecking order predictions about financing decisions by examining how often and under what circumstances firms issue and repurchase equity. We uncover what seem to be pervasive contradictions of the model. The first important result is striking evidence against the pecking order prediction that firms rarely issue stock. As motivation for the pecking order, Myers (1984) emphasizes that aggregate net new issues of equity are small relative to net new debt. It is also well-known that seasoned equity offerings (SEOs) are rare. But the aggregate level of equity financing and the scarcity of SEOs are misleading. In addition to SEOs, firms issue equity in mergers and through private placements, convertible debt, warrants, direct purchase plans, rights issues, and employee options, grants, and benefit plans. During 1973 to 1982, on average 67% of our sample firms issue some equity each year, and the proportion rises to 74% for 1983 to 1992 and 86% for 1993 to During much of the sample period, however, repurchases by some firms offset the equity issues of others, and aggregate annual

3 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) net new equity is small. This result, along with the low frequency of SEOs, leads to the misleading impression that new issues of stock are rare. In fact, most firms issue, repurchase, or do both every year. And our examination of the firms that issue or retire equity shows that equity decisions often violate the pecking order. Thus, equity issuers are not typically under duress; net issues are common among firms with moderate leverage and financing surpluses (earnings exceed the sum of dividends and investment). Also in violation of the pecking order, repurchases are not limited to firms with low demand for outside financing; many firms with financing deficits repurchase stock. We estimate that during 1973 to 2002, the year-by-year equity decisions of more than half of our sample firms contradict the pecking order. Moreover, annual net issues of equity are material. For example, on average 61.5% of small firms (total assets below the NYSE median) make net issues of stock each year from 1983 to 1992, and the average rises to 73.7% for 1993 to These annual net stock issues average 6.0% of assets during 1983 to 1992 and 12.6% for 1993 to 2002, both larger than the annual net new debt of these firms, which averages 5.2% and 6.4% of assets for the two periods. On average 66.5% of big firms make net stock issues each year of the 1993 to 2002 period, and their net equity issues are about the same magnitude, 7.5% of assets, as their net issues of debt, 7.9%. The fact that equity issues and repurchases are commonplace and commonly not in line with the pecking order seems like a telling blow to the argument of Myers (1984) and Myers and Majluf (1984) that asymmetric information problems drive the capital structures of firms. Myers (1984) and Myers and Majluf (1984) do notallow for equity issues that do not have an asymmetric information problem. One story for our results is that there are important ways to issue equity that avoid this problem. If so, the pecking order, as the stand-alone model of capital structure proposed by Myers (1984), is dead: financing with equity is not a last resort, and asymmetric information problems are not the sole (or perhaps even an important) determinant of capital structures. This does not mean the asymmetric information problem disappears. But its implications become quite limited: firms do not follow the pecking order in financing decisions; they simply avoid issuing equity in ways that involve asymmetric information problems. Our measure of equity issues is all encompassing, including any transaction that increases the number of (split-adjusted) shares outstanding. This leads some readers to argue that our results say nothing about the pecking order because we include stock issues that do not have asymmetric information problems. But again, the pecking order is proposed by Myers (1984) and Myers and Majluf (1984) as a complete model of capital structures, so our broad measure of equity issues is relevant for analyzing its predictions about how capital structures are determined. For example, issues of stock to employees via options and grants play a big role in our results on the frequency of equity issues. Stock issues to employees may not have an asymmetric information problem. A firm nevertheless alters its capital structure when it compensates employees with stock instead of cash and chooses not to offset the stock issues with repurchases. In short, if the pecking order model can only

4 552 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) handle ways of issuing equity that involve large asymmetric information problems, it is not a model of capital structure. The breakdown of the pecking order does not require that equity can be issued with minor asymmetric information problems. Anything that produces the result that equity is not a last resort will do. For example, agency problems may sometimes lead managers to ignore the costs of issuing equity (Jung etal., 1996). Itis also likely that some equity issues have benefits that outweigh their costs, i.e., tradeoff effects. For example, getting stock in a merger can have tax benefits for shareholders of the acquired firm that lead them to accept a lower price for their shares. Similarly, stock issued to employees may have motivation benefits that outweigh any asymmetric information costs. The important point is that any forces that cause firms to deviate systematically from pecking order financing (retained earnings, then debt, and equity only as a last resort) imply that the pecking order model, on its own, cannot explain capital structures. Our story proceeds as follows. Section 2 outlines the pecking order model. Section 3 documents how the profitability and growth characteristics of firms, which are important for understanding financing decisions, change through time. Our main empirical results are in Sections 4 6. The theme of these sections is that violations of the pecking order become more evident when financing decisions are examined at more disaggregated levels. Section 4 examines financing decisions at the level of the market, where we find little evidence against the pecking order model, except later in the sample period. Section 5, the paper s centerpiece, disaggregates firms into 12 groups formed on size, profitability, and growth, and examines in detail how equity issuers in the 12 groups differ from repurchasers. Here we find widespread evidence of pecking order violations. Section 6 provides evidence on the mechanisms firms use to issue equity. The concluding section discusses the implications of our findings. 2. The pecking order model Myers (1984) uses Myers and Majluf (1984) to motivate the pecking order. In Myers and Majluf (1984), managers use private information to issue risky securities when they are overpriced. Investors are aware of this asymmetric information problem, and the prices of risky securities fall when new issues are announced. Managers anticipate the price declines, and may forego profitable investments if they must be financed with risky securities. To avoid this distortion of investment decisions, managers follow what Myers (1984) calls the pecking order. They finance projects first with retained earnings, which have no asymmetric information problem, then with low-risk debt, for which the problem is negligible, then with risky debt. Equity is issued only under duress, or when investment so far exceeds earnings that financing with debt would produce excessive leverage. Myers (1984) also posits that in the short term, dividends are (for unspecified reasons) sticky, leaving variation in net cashflows to be absorbed mainly by debt. Myers (1984) largely ignores share repurchases. Shyam-Sunder and Myers (1999) address the issue, and argue that the asymmetric information problem of new

5 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) common stock applies to repurchases. If a firm announces a repurchase, investors assume managers have positive information not reflected in the stock price, causing the price to rise. This can deter the repurchase if the price rises above what managers consider the equilibrium level. More important, debt capacity is a valuable option for future financing. Thus, when firms use financing surpluses to retire securities, they first retire debt. They retire equity only when leverage is low or when poor investment opportunities (relative to earnings) lower the value of debt capacity. In short, repurchases should be limited to firms with little or no leverage, few investment opportunities, or both. Two additional points about the pecking order are pertinent for interpreting our empirical results. First, Myers (1984) emphasizes asymmetric information problems, but he recognizes that transactions costs alone can produce pecking order financing if they are higher for debt than for retained earnings and higher yet for equity. In other words, asymmetric information may be unnecessary. Second, in Myers (1984) and Myers and Majluf (1984), the pecking order arises through an implicit assumption that there is no way to issue equity that avoids asymmetric information problems. If firms find ways to issue equity without such problems, asymmetric information may not constrain equity issues. As a result, pecking order financing can disappear; that is, financing with equity is not a last resort, the incentive to avoid repurchases to maintain debt capacity is gone, and asymmetric information problems do not drive capital structures. This does not mean asymmetric information is irrelevant. But its implications become quite limited. Firms do avoid issuing risky securities in ways that involve asymmetric information problems, but financing decisions do not follow the pecking order. The capital structure literature focuses on SEOs as the source of outside equity. There are at least seven other ways firms issue equity: (1) mergers via an exchange of stock, (2) employee stock options, grants, and other employee benefit plans, (3) subscription rights issued to stockholders, (4) warrants attached to other securities, (5) convertible bonds, (6) dividend reinvestment and other direct purchase plans, and (7) private placements. Do some of the mechanisms for issuing equity involve low transactions costs and minor asymmetric information problems? Consider transactions costs. SEOs, warrants, and convertible bonds have large underwriting costs. The large price concessions in private placements of equity (Wruck, 1989; Hertzel et al., 2002) are also a high transaction cost. In contrast, rights offerings are notcostly (Smith, 1977). Issuing stock to employees via grants, options, and other benefit plans also probably involves low transactions costs. The same is true for direct purchase plans. Negotiating mergers is costly, but the marginal cost of carrying out an exchange of stock may not be high. In short, four of the alternatives for issuing stock seem to involve low transactions costs. Asymmetric information is a stickier issue. Myers (1984) posits that stock price declines in response to announcements of equity issues reflect asymmetric information problems. If so, the problems are severe for SEOs (Masulis and Korwar, 1986), and presentbutweaker for convertible bonds (Mikkelson and Partch, 1986). Wruck (1989) and Hertzel et al. (2002) find that the stock price

6 554 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) response to private placements is positive. Though their samples are small, Smith (1977) and Eckbo and Masulis (1992) find no evidence of a reliable negative price response to announcements of rights issues. The price responses to initiations of dividend reinvestment plans (Peterson et al., 1987; Allen etal., 1995) and ESOP plans (Chaplinsky and Niehaus, 1994) also seem to be small. Moeller etal. (2004) find that the negative price responses to stock-financed mergers are limited to acquisitions of publicly traded firms by big publicly traded firms. Otherwise, mergers financed with stock do not seem to have negative price effects. Stock issues in direct purchase plans are initiated by the purchasers (mostly existing stockholders), not by managers, so asymmetric information problems may be absent. Asymmetric information problems may be minor in stock-financed mergers because mergers are negotiated between informed parties. Likewise, asymmetric information may not be important in grants of stock and options to employees because employees are informed or the issues are a matter of routine. And stock issues in mergers and to employees are important in our later evidence on the frequency and magnitude of issues. But asymmetric information problems can never be ruled out. Firms can always use repurchases to offset stock issues to employees, in mergers, and in direct purchase plans. The decision not to repurchase may be interpreted as evidence that managers think the stock is overvalued an asymmetric information problem. As always, the information conveyed by any action depends on the market s expectations. Unfortunately, event studies cannot settle the matter unambiguously since price responses to announcements may not be due solely to asymmetric information problems. We are not arguing that asymmetric information problems are unavoidable. Indeed, it is likely that the asymmetric information problems that arise with some ways of issuing stock (most notably SEOs) do not lead to pecking order financing (equity is not a last resort and asymmetric information problems do not drive capital structures) because firms can issue equity in other ways that largely avoid asymmetric information problems. The discussion above suggests that the prime candidates for breaking the grip of the pecking order are stock issues to employees, with perhaps a minor assist from rights offerings and direct purchase plans, and a big assistfrom stock-financed mergers. Breaking the grip of the pecking order does not require that equity can be issued with minor asymmetric information problems and low transactions costs. Any forces that produce the result that equity is not a last resort will do. One possibility is that agency problems sometimes lead managers to ignore equity issuing costs (Jung etal., 1996). A potentially important alternative is that equity issues can have benefits that outweigh their costs, i.e., tradeoff effects. For example, an exchange of stock in a merger can create a tax benefit that offsets asymmetric information costs. By paying with stock, the acquiring firm allows target shareholders to postpone capital gains taxes, which should lead the shareholders to accept a lower price for their shares. Similarly, the motivation benefits of stock issued to employees may outweigh asymmetric information costs if there are any. The important point is that any forces that cause firms to deviate systematically from pecking order financing (retained

7 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) earnings, then debt, and equity only as a last resort) imply that the pecking order model cannot alone explain capital structures. Finally, in developing the pecking order model, Myers (1984) takes dividend decisions as given and outside the purview of the model. This is a patch on the model, a tacit admission that tradeoff forces or other factors are important in dividend decisions since the asymmetric information story in itself would predict that dividends (like repurchases) are rare. There is a second important patch. The concluding sections of Myers (1984) and Myers and Majluf (1984) argue that firms want to avoid debt that may result in distress that prevents them from exercising future investment options. As a result, firms with large current and expected investments relative to earnings may issue equity even though leverage is moderate. This patch on the pecking order model allows firms with large current and expected financing deficits to opt out of simple pecking order behavior, causing such firms to look more like those predicted by the tradeoff model (they incur current asymmetric information costs because of expected future benefits). In evaluating the empirical results that follow, we largely let the pecking order retain these two patches, to give the model its best shot at explaining the evidence. 3. The characteristics of sample firms The profitability and growth characteristics of firms are central in evaluating their financing decisions. Thus, we begin the empirical work with a brief description of the changing characteristics of listed firms. The data are from CRSP and Compustat, the period is 1973 to 2002 (when Compustat s coverage of listed firms is fairly complete), and the sample includes NYSE, AMEX, and Nasdaq firms. We exclude financial firms and utilities. Financial intermediaries do not seem relevant for testing models of financing decisions, and during much of the sample period, regulation is important in the capital structure decisions of utilities. The number of listed non-financial, non-utility firms with the necessary data for later tests rises from an average of 2,951 for 1973 to 1982, to 4,417 for 1993 to 2002 (Table 1). The increase is primarily among small firms, defined as NYSE, AMEX, and Nasdaq firms with total assets below the NYSE median. The number of big firms only rises from an average of 617 for 1973 to 1982, to 712 for 1993 to To illustrate the changing characteristics of firms, Table 1 separately allocates small and big firms to portfolios formed as the intersections of three profitability and two growth groups. The breakpoints for the profitability and growth groups are averages of NYSE annual medians for 1973 to The three profitability groups for each year include firms with (i) negative earnings before interest (E), (ii) low profitability (earnings relative to total assets, E/A, below the average median for NYSE firms with positive E), and (iii) high profitability (E/A above the average NYSE median). The two growth groups include firms with (i) low asset growth for year t (da/a=(a t A t 1 )/A t below the average median for NYSE firms) and (ii) high assetgrowth (da/a above the NYSE average median). The appendix contains precise definitions of E/A, da/a, and our other variables.

8 556 Table 1 Average number of firms in size-profitability-growth groups Each year s sample of all firms includes all non-financial (SIC codes between 6000 and 6999), non-utility ( ) firms with the necessary CRSP and Compustat data for that year (see the appendix). Firms are assigned to the Small and Big portfolios and to the 12 size-profitability-growth portfolios in year t based on their characteristics at the end of their fiscal year in calendar year t. The assets of small and big firms in year t are below or above those of the median NYSE firm in t. The breakpoints for growth and positive profitability are the averages of the annual NYSE medians for 1973 to A firm is assigned to the low- or high-growth group for year t if its growth in assets, da/a=(a t A t 1 )/A t, is below or above the average median for 1973 to A profitable firm is assigned to the low- or high-profitability group for t if its earnings relative to assets, E/A, is below or above the 1973 to 2002 average median of NYSE firms with positive profitability. All firms Small Big Small Big E/A: Negative E Low E/A High E/A Negative E Low E/A High E/A da/a: Low High Low High Low High Low High Low High Low High E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) ARTICLE IN PRESS

9 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) There are fairly general increases in the number of big firms in different profitability and growth groups. The group of big firms with high profitability and high growth is the only one that shrinks, from an average of 246 for 1973 to 1982, to 142 for 1993 to Thus, the population of profitable, rapidly growing big firms contracts through time, but all other big groups expand. The number of small firms (mostly on Nasdaq) explodes during the sample period, but the increases are not uniform across groups. The notable surges are for small unprofitable firms and small firms with positive but low profitability. The number of small highly profitable firms actually declines, from an average of 1,227 for 1973 to 1982, to 1,005 for 1993 to In short, the number of listed firms increases from 1973 to 2002, largely due to an influx of unprofitable and low-profitability firms. The number of highly profitable firms actually declines. These patterns are common to both small and big firms, but for groups with increasing numbers of firms, the increases among small firms dwarf those for big firms. Lemmon and Zender (2002) argue that equity issues by fast-growing firms with low profitability are rational opt-outs allowed by the pecking order (via the patch discussed above). They argue that the increased frequency of such firms, obvious in Table 1, explains the apparent evidence against the pecking order of Frank and Goyal (2003). When we later examine equity issuers and repurchasers in the 12 sizeprofitability-growth groups in detail, however, we find that violations of the pecking order are pervasive. The violations are not limited to fast-growing low-profitability firms. 4. The market Table 2 summarizes the aggregate financing decisions of all sample firms, the market. The variables in Table 2 (and in later tables) are ratios, and the denominator is total assets. Specifically, the variables are book leverage (total liabilities over total assets, L/A), the market-to-book ratio for total assets (V/A, a proxy for Tobin s Q), profitability (E/A), assetgrowth (da/a), netdebt issues (dl/a), the change in balance sheet retained earnings (dre/a), and two measures of netequity issues (dsb/a and dsm/a). The ratios are calculated as the aggregate value of the numerator divided by aggregate assets, which in effect treats the entire sample as a single firm. Equivalently, the ratios are size-weighted averages of the ratios for firms, where a firm s ratio is weighted by its assets relative to aggregate assets. Finally, the ratios are multiplied by 100, and thus are expressed as percents of assets. Our results center on pecking order predictions about which firms issue equity. We show two measures of annual equity issues. The first, dsb, is the change in the book value of stockholders equity in excess of the change in Compustat s adjusted balance sheetretained earnings, dsb ¼ dse dre: (1)

10 558 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) Table 2 Average characteristics of all firms Each year s sample includes all non-financial, non-utility firms with the necessary CRSP and Compustat data for that year (see the appendix). We report the average ratio of the annual aggregate value of the numerator for a portfolio divided by the aggregate value of assets, in percent. Thus, leverage, L/A(t), is the sum of total liabilities in year t (Compustat item 181) divided by the sum of assets in year t (6); L/A(t 1) is L/A for the previous fiscal year, t 1; dl/a is the aggregate change in liabilities from t 1 tot divided by the sum of assets at t; da/a is the aggregate growth in assets from t 1 tot divided by the sum of year t assets; profitability, E/A, is the aggregate of the sum of income before extraordinary items (18), interest expense (15), and extraordinary income (48) if available, divided by aggregate assets; dre/a is the sum of the change in Compustat s adjusted retained earnings (36) from t 1 tot divided by aggregate assets at t; the financing deficit, Def/A, is the difference between the growth in assets and the change in retained earnings, da/a dre/a; the book measure of net equity issued, dsb/a, is the financing deficit minus the change in liabilities, Def/A dl/a; and the market-based measure of net stock issued, dsm, is the splitadjusted change in Compustat shares outstanding (25) during the fiscal year times the average of the beginning and ending split-adjusted Compustat stock prices (199). The market-to-book ratio for total assets, V/A, is the aggregate of the sum of book value of debt and the market value of equity divided by aggregate assets. Firms dsm/a dsb/a dl/a dre/a da/a V/A L/A E/A Def/A t t The change in stockholders equity, dse, combines (i) issues and repurchases of equity, (ii) the change in retained earnings, and (iii) dirty surplus transactions such as foreign currency translation adjustments. Since dirty surplus transactions do not flow through the income statement, they do not affect a firm s reported value of retained earnings. These transactions are typically incorporated, however, in Compustat s adjusted value of retained earnings (data item 36), so we use this measure to compute dsb. Both pooling of interests mergers and employee stock options cause dsb to understate stock issues. In a pooling of interests merger, stockholders equity increases by the book value of the acquired firm s stock, not by the (typically greater) market value of the shares issued to the acquired firm s shareholders. When stock options are exercised, stockholders equity rises by the strike price plus the taxes saved because of the exercise, not by the (typically greater) market value of the shares issued. 1 Stock dividends also contaminate dsb. When a firm pays a stock 1 The tax effect arises because of differences between the way options are expensed for tax and financial reporting purposes. During our sample period, few firms recognize the cost of executive options in their financial statements. When an option is exercised, however, the difference between the market value of the stock and the strike price is an expense for tax purposes. This expense reduces the firm s tax liability and drives a permanent wedge between the tax expense reported in the financial statements and the actual taxes paid. The tax savings are credited to stockholders equity without flowing through the income statement.

11 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) dividend, it transfers the market value of the shares issued from its retained earnings account to its contributed capital account. As a result, dsb misclassifies stock dividends as new equity issued. Our second measure of equity issued during a fiscal year, dsm, is the (splitadjusted) change in the number of shares outstanding over the fiscal year times the average of the (split-adjusted) stock prices at the beginning and end of the fiscal year. (See the appendix for details.) Since the market-based dsm does not suffer from the accounting problems of dsb, it is a more accurate measure of equity financing, and thus is the central variable in the tests below. But we also show dsb because it balances financing sources and investment, dl=a þ dsb=a þ dre=a ¼ da=a (2) and itproduces the financing deficitin Table 2, Def=A ¼ da=a dre=a ¼ dl=a þ dsb=a: (3) Both dsb/a and dsm/a are superior to another measure of outside equity financing, net cash from stock issued the difference between inflows from new issues and outflows for repurchases, from the statement of cash flows. This measure, used by Frank and Goyal (2003) and Lemmon and Zender (2002), understates equity issued because the statement of cash flows does not show stock issued in mergers or outright grants of stock to employees because such issues produce no cash flows. The change in Treasury stock used by Fama and French (2001) has similar problems. The aggregate profitability and growth of sample firms change through time. Profitability, E/A, falls from an average of 7.7% for 1973 to 1982, to 5.0% for 1993 to 2002 (Table 2). Earnings in excess of dividends (dre/a) also fall, averaging 3.0% of assets for 1973 to 1982 and 0.5% for 1993 to Asset growth declines, but less, from 10.3% per year to 9.1%. Since asset growth declines less than the change in retained earnings, outside financing increases. The increase is not met with new debt, which falls a bit; dl/a averages 6.5% per year for 1973 to 1982, and 6.2% for 1993 to New issues of equity fill the void. During 1973 to 1982 and 1983 to 1992, annual aggregate net new equity averages 0.8% and 0.5% of assets, and most outside financing is debt. Myers (1984) uses the relatively low level of aggregate equity financing to motivate the pecking order. Confirming Frank and Goyal (2003), however, during 1993 to 2002 equity financing is important, even at the aggregate level; annual net new equity, dsm/a, averages a substantial 4.0% of assets, which is about two-thirds as large as net new debt (6.2% per year). Aggregate results are, however, largely irrelevant for judging the predictions of the pecking order. The pecking order is a model of financing decisions by individual firms. Because aggregate results hide much variation in the financing of individual firms, aggregate results can be misleading. When we next examine financing decisions at a more disaggregated level, we find lots of behavior inconsistent with the pecking order throughout the 1973 to 2002 period.

12 560 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) Who issues stock? In the pecking order model, transactions costs and asymmetric information problems lead firms to finance first with internal funds and then with debt. Equity is issued only under duress or when investment so far exceeds earnings that financing with debt would produce excessive leverage. Since pecking order predictions about when firms issue equity are the lynchpin of the model, this section examines the evidence in detail Incidence and materiality of equity issues The pecking order model predicts that few firms issue or repurchase shares. In fact, most firms issue, repurchase, or do both every year. Table 3 shows average annual percents of sample firms with net repurchases (dsmo0), netissues (dsm40), or neither (dsm=0) mutually exclusive categories. The table also shows percents of firms with gross issues and gross repurchases. These groups are not mutually exclusive; a firm can be in one, both, or neither of the groups. In addition to results for all sample firms, Table 3 shows separate results for small and big firms. Raising immediate suspicions about the extent to which financing decisions follow the pecking order (equity as a last resort), the average fraction of sample firms with no net equity issues (dsm=0) falls from a rather low 24% per year for 1973 to 1982, to 8% per year for 1993 to Big firms almost always make annual net issues or net repurchases. The fraction of big firms with no net issues is 11% per year for 1973 to 1982 and 2% for 1993 to Despite the literature s emphasis on repurchases, the fraction of all sample firms with net repurchases hovers around a modest 20% per year. During the last 20 years of the sample (1983 to 2002), a more impressive 30% per year of big firms make netrepurchases. Most firms make annual net issues of equity. The average annual fraction of sample firms with net issues (dsm40) rises from more than half (54% per year) for 1973 to 1982, to 62% for 1983 to 1992, and an impressive 72% for 1993 to The increase in netissuers among all sample firms is driven by small firms. The fraction of big firms with net issues of equity is always high, averaging 72%, 66%, and 67% per year for the three ten-year periods of Table 3. The pecking order model is silent on why firms pay dividends, but issuing stock to pay dividends seems like a problem. Specifically, if asymmetric information means stock must be issued at below equilibrium value, issuing stock to pay dividends decreases the wealth of current shareholders. In contrast, reducing dividends does not cause permanent wealth losses for long-term shareholders even if asymmetric information problems cause the stock price to fall temporarily below its equilibrium value. Dividend payers are more likely to repurchase than non-payers, especially during 1983 to For example, though not shown in Table 3, during 1993 to 2002, on average 37% of each year s dividend payers make netrepurchases, versus 15% for non-payers. And firms that do not pay dividends are more likely to issue equity than

13 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) Table 3 Average percents of firms that issue and repurchase equity and average issues and repurchases as percents of total assets Each year s sample (All) includes all non-financial, non-utility firms with the necessary CRSP and Compustat data for that year (see the appendix). Firms are assigned to the Small or Big category in year t if their assets are below or above those of the median NYSE firm in t. The market-based measure of net stock issued, dsm, is the split-adjusted change in Compustat shares outstanding during the fiscal year times the average of the beginning and ending split-adjusted Compustat stock prices. Thus, net stock issuers have a positive split-adjusted growth in shares, dsm40, and net stock repurchasers have a negative growth in shares, dsmo0. Gross stock repurchases, GR, is Compustat annual item 115. Gross stock issues, GI, is netissues, dsm, plus gross repurchases. % in dsm group % with GI40 % wit h GR40 % with GI/A41% All Small Big All Small Big All Small Big All Small Big Part A: percent of firms in dsm categories that issue and repurchase All dsm groups dsmo dsm= dsm dsm/a GI/A GR/A All Small Big All Small Big All Small Big Part B: issues and repurchases as percents of assets All dsm groups dsmo dsm= dsm

14 562 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) dividend payers. Nevertheless, in apparent contradiction of the pecking order, during 1973 to 2002 on average about 58% of each year s dividend payers make net issues of equity. The fractions of firms with net equity issues or net repurchases each year are high, but they understate the frequency of issues and repurchases. Many firms that are net issuers in a given year repurchase in the same year, and most firms that make net repurchases also issue. On average, 19% of the firms that make net issues of equity during 1973 to 1982 repurchase in the same year, and the average rises to 33% for 1993 to 2002 (Table 3). As a result, the fraction of all sample firms making gross repurchases (30% per year, rising to 42%) is much higher than the fraction making netrepurchases. More impressive, of each year s sample of firms with net repurchases (dsmo0), 56% issue equity in the same year during 1973 to 1982 and the fraction rises to 64% during 1993 to 2002 (Table 3). As a result, most sample firms (67% per year for 1973 to 1982 and 86% for 1993 to 2002) issue some equity each year. 2 Big firms are especially active. The fractions with gross issues of equity, gross repurchases, and both issues and repurchases are higher for big firms than for small firms (Table 3). Some equity issues are tiny, so it is interesting to examine the frequency of issues above a non-trivial threshold, say 1% of assets. The fraction of all sample firms with gross issues that exceed 1% of assets rises from 24% per year for 1973 to 1982, to 40% for 1983 to 1992, and 57% for 1993 to 2002 (Table 3). More impressive, among the high fractions of firms that make net issues of equity, in a typical year 38% of the 1973 to 1982 sample make equity issues that exceed 1% of assets, and the fraction rises to 58% for 1983 to 1992 and 71% for 1993 to A caveat is in order. We see later that for individual firms, the equity issuing process is lumpy, with smaller issues during most years but large issues during some years, the result of relatively infrequent SEOs and mergers. The distribution of equity issues is thus skewed right, and tabulating annual percents of firms that make issues above some threshold (1% of assets) gives a misleadingly low impression of the cumulative effects of issues. A better picture of cumulative effects is obtained (next) by comparing average annual equity issues with average debt issues. Consider the firms that are net issuers of equity in a given year, dsm40. Their annual net equity issues average 1.4% of assets during 1973 to 1982, 2.3% for 1983 to 1992, and an impressive 8.2% for 1993 to 2002 (Part B of Table 3). Their gross issues are larger, averaging 1.5%, 2.6%, and 8.8% of assets. The annual net new debt of these firms averages about 7.0% of assets (Table 4). Thus, for firms that make net issues of equity, the issues are substantial relative to new issues of debt, especially during the last 20 years of the sample period. For the seven of ten sample firms that 2 Note that Compustat reports zero gross repurchases for some firms making net repurchases (dsmo0). Compustat s gross repurchases are from the Statement of Cashflows. A cashless repurchase can occur in a variety of ways. For example, a shareholder (typically an employee) can deliver shares as payment on a loan from the company, the firm can exchange other property for outstanding shares, or an employee can forfeit non-vested shares when leaving the firm. Some cases, however, are probably Compustat errors (reporting zero cash repurchases when they are in fact positive).

15 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) make annual net equity issues during 1993 to 2002, aggregate net new equity exceeds netnew debt. The aggregate issues of all net issuers quoted above are dominated by big firms, so it is interesting to examine small firms. The net equity issues of small firms that are net issuers tend to be even more substantial, averaging 2.7% of assets for 1973 to 1982, 6.1% for 1983 to 1992, and 12.6% for 1993 to 2002 (Table 3). In contrast, the netnew debtfor these firms averages about6.0% of assets, and netnew debtis less than net new equity during both 1983 to 1992 and 1993 to 2002 (Table 4). In short, during the last 20 years of the sample, on average about two-thirds of each year s firms make net equity issues. For big firms that are net issuers, net new issues of equity are on average about one-third the size of net new debt during 1983 to 1992, and they are about as large as new debt during 1993 to 2002 (Table 4). For small firms that make net equity issues, the issues are on average larger than net new debt for the entire 1983 to 2002 period. And it is worth adding that the evidence for small firms cannot be dismissed as unimportant for judging the pecking order model since the asymmetric information problems that generate its predictions about financing decisions are probably more serious for smaller firms. Though net issuers outnumber repurchasers about three to one, firms that retire equity tend to be larger. As a result, for the sample as a whole, aggregate net new equity averages just 0.8% of assets for the first ten years of the sample (1973 to 1982) and 0.5% for 1983 to 1992 (Table 3). These aggregates are clearly misleading, however. They are not due to the paucity of equity issues predicted by the pecking order. On the contrary, they are the result of equity issues by a large fraction of firms that are offset by repurchases by others. During 1993 to 2002, even aggregate net new equity is no longer small, averaging 4.0% of the assets of sample firms and 8.2% of the assets of net issuers (as compared to about 6.2% for net new debt) The characteristics of firms that make net issues or repurchases Since the pecking order model leads us to expect that equity issues and repurchases are rare, the fact that almost all firms do one or the other or both every year suggests that many firms violate the model s predictions about equity decisions. The key variables in the predictions are profitability and growth, the main ingredients of the financing deficit. Thus, to examine the extent to which equity decisions conform to the model, we break the sample into groups of firms with similar profitability and growth. Specifically, as in Table 1, we allocate firms to 12 groups each year based on size (big and small), profitability (negative, low, and high E/A), and growth (low and high da/a). Table 4 summarizes the characteristics of the net equity issuers and net repurchasers in the overall sample, for small and big firms separately, and for each of the 12 size-profitability-growth groups. The market-to-book ratios, V/A, in Table 4 are normalized. Fama and French (2001) find that the upward drift in market V/A after 1980, apparent in Table 2, is not matched by drift in aggregate profitability or growth. Based on this evidence, they conclude that the drift in V/A reflects a decline in expected returns (discount rates) rather than better investment opportunities that might affect financing

16 564 Table 4 Average characteristics of firms that make net repurchases (dsmo0) or netissues (dsm40) Each year s sample (All firms) includes all non-financial, non-utility firms with the necessary CRSP and Compustat data for that year (see the appendix). Firms are assigned to the 12 size-profitability-growth groups in year t based on their characteristics at the end of their fiscal year in calendar year t. The assets of small and big firms in year t are below or above those of the median NYSE firm in t. The breakpoints for growth (da/a) and positive profitability (E/A) are the averages of the annual NYSE medians for 1973 to We report the average ratio of the annual aggregate value of the numerator for a group divided by the aggregate value of assets, in percent. Thus leverage, L/A(t), is the sum of total liabilities in year t divided by the sum of assets in year t; L/A(t 1) is L/A for the previous fiscal year, t 1; dl/a is the aggregate change in liabilities from t 1 tot divided by the sum of assets at t; da/a is the aggregate growth in assets from t 1 tot divided by the sum of year t assets; profitability, E/A, is the aggregate sum of income before extraordinary items, interest expense, and extraordinary income if available, divided by aggregate assets; dre/a is the sum of the change in Compustat s adjusted retained earnings from t 1 tot divided by aggregate assets at t; the financing deficit, Def/A, is the difference between the growth in assets and the change in retained earnings, da/a dre/a; the book measure of netequity issued, dsb/a, is the financing deficit minus the change in liabilities, Def/A dl/a; and the market-based measure of net stock issued, dsm, is the split-adjusted change in Compustat shares outstanding during the fiscal year times the average of the beginning and ending split-adjusted Compustat stock prices. The % of Group is the average of the annual percents of All firms and of firms in the 12 size-profitability-growth groups with dsmo0 (Netrepurchases) and dsm40 (Net issues). The market-to-book ratio for total assets, V/A, is the aggregate of the sum of book value of debt and the market value of equity divided by aggregate assets. We standardize V/A by dividing each year s V/A for a group by V/A for all firms. Firms % of Group Part A: average characteristics of all sample firms dsm/a dsb/a dl/a dre/a da/a V/A L/A E/A Def/A All firms: dsmo t t 1 E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) ARTICLE IN PRESS

17 All firms: dsm Small firms: dsmo Small firms: dsm Big firms: dsmo Big firms: dsm E.F. Fama, K.R. French / Journal of Financial Economics 76 (2005) ARTICLE IN PRESS