THE UNIVERSITY OF CHICAGO MANAGING FINANCIAL POLICY: EVIDENCE FROM THE FINANCING OF EXTRAORDINARY INVESTMENTS A DISSERTATION SUBMITTED TO

Transcription

1 THE UNIVERSITY OF CHICAGO MANAGING FINANCIAL POLICY: EVIDENCE FROM THE FINANCING OF EXTRAORDINARY INVESTMENTS A DISSERTATION SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF BUSINESS IN CANDIDACY FOR THE DEGREE OF DOCTOR OF PHILOSOPHY BY ERIK STAFFORD CHICAGO, ILLINOIS AUGUST 1999

2 Copyright 1999 by Erik Stafford All rights reserved

3 TABLE OF CONTENTS LIST OF TABLES... v LIST OF FIGURES... vii ACKNOWLEDGEMENTS... viii ABSTRACT... ix MANAGING FINANCIAL POLICY: EVIDENCE FROM THE FINANCING OF EXTRAORDINARY INVESTMENTS... 1 APPENDIX A. TABLES B. DETAILS ON SAMPLE AND VARIABLE DEFINITIONS C. EXAMPLES OF EXTRAORDINARY INVESTMENTS D. SIMULATION EVIDENCE ON EVENT-TIME METHODOLOGY E. AGGREGATE FINANCIAL POLICY FOR VALUE LINE AND EVENT FIRMS F. EVENT-TIME BEHAVIOR OF SHORT-TERM ASSETS & LIABILITIES G. FIGURES BIBLIOGRAPHY iv

4 LIST OF TABLES A.1.A EXTRAORDINARY INVESTMENTS BY YEAR A.1.B EXTRAORDINARY INVESTMENTS BY INDUSTRY A.2 ABNORMAL INVESTMENT ACTIVITY BY FIRMS MAKING EXTRAORDINARY INVESTMENTS A.3 ABNORMAL CASH FLOWS OF FIRMS MAKING EXTRAORDINARY INVESTMENTS A.4 VARIOUS DIVIDEND POLICY STATISTICS FOR FIRMS MAKING EXTRAORDINARY INVESTMENTS A.5 ABNORMAL DIVIDEND PAYMENTS BY FIRMS MAKING EXTRAORDINARY INVESTMENTS A.6 DEBT AND EQUITY ISSUE FREQUENCY FOR FIRMS MAKING EXTRAORDINARY INVESTMENTS A.7 ABNORMAL EXTERNAL FINANCING BY FIRMS MAKING EXTRAORDINARY INVESTMENTS A.8 ABNORMAL CAPITAL STRUCTURES OF FIRMS MAKING EXTRAORDINARY INVESTMENTS A.9 TARGET-ADJUSTMENT AND PECKING ORDER FORECAST ERRORS FOR FIRMS MAKING EXTRAORDINARY INVESTMENTS A.10 POST-EVENT CASH-TO-SALES RANKINGS RELATIVE TO INDUSTRY FOR FIRMS MAKING EXTRAORDINARY INVESTMENTS & RAISING EXTERNAL FUNDS A.11 REGRESSION ANALYSIS ON THE DETERMINANTS OF CASH HOLDINGS FOR VALUE LINE FIRMS B.1 VARIABLE DEFINITIONS B.2 CAPITALIZED ADVERTISING AND R&D CAPITAL BY INDUSTRY D.1 SIMULATED SPECIFICATION LEVELS (SIZE) OF VARIOUS TEST STATISTICS FOR ABNORMAL FINANCIAL POLICY VARIABLES E.1 AGGREGATE CAPITAL STRUCTURES FOR VALUE LINE FIRMS E.2 AGGREGATE INVESTING ACTIVITIES AS A PERCENT OF LONG-TERM ASSETS FOR VALUE LINE FIRMS v

5 E.3 AGGREGATE PAYMENTS TO STAKEHOLDERS AS A PERCENT OF TOTAL ASSETS FOR VALUE LINE FIRMS E.4 AGGREGATE INVESTMENT AND FINANCING ACTIVITIES FOR VALUE LINE FIRMS. 89 E.5 AGGREGATE INVESTMENT AND FINANCING ACTIVITY FOR FIRMS MAKING EXTRAORDINARY INVESTMENTS F.1 ABNORMAL WORKING CAPITAL OF FIRMS MAKING EXTRAORDINARY INVESTMENTS vi

6 LIST OF FIGURES G.1 MARGINAL COSTS OF EXTERNAL FINANCING CONDITIONAL ON RAISING EXTERNAL FUNDS vii

7 ACKNOWLEDGEMENTS I thank Harry DeAngelo, Doug Diamond, J. B. Heaton, Peter Hecht, Raghu Rajan, Matthew Rothman, Per Strömberg, Francis Yared, and seminar participants at Columbia, Cornell, Duke, Harvard, NYU, University of Rochester, University of Chicago, and USC for helpful comments and discussions. I especially thank Gregor Andrade, Eugene Fama, Steven Kaplan, Mark Mitchell, and Luigi Zingales for their valuable guidance, support, and nurturing on this project and throughout my years in the Ph.D. program. I gratefully acknowledge financial support from the Oscar Mayer Foundation and the Sanford Grossman Fellowship. viii

8 ABSTRACT How do managers set financial policy? Most popular theories of financial policy are developed in terms of a marginal analysis for a new project available to the firm that is extraordinary in spirit. However, extraordinary investment observations have not been previously isolated for study. This paper uses a sample of firms making extraordinary investments to probe deeper into the driving forces behind financial policy theories. Event-time analysis reveals that the financial policies of the sample firms can reasonably be characterized as pecking order behavior as described by Donaldson (1961) and Myers (1984): (1) internal funds are the dominant source; (2) equity issues are relatively unimportant; and (3) debt issues are the residual financing variable. There is no evidence that dividend paying firms adjust dividend policy to accommodate the extraordinary investment. To address whether this observed pecking order behavior in financial policy is necessarily supportive of the pecking order theory [Myers and Majluf (1984)], I develop a test designed to distinguish between target-adjustment and pecking order models. A competing prediction of the two models concerns what happens to the cash holdings of firms raising external financing. The empirical results suggest (a) transaction costs appear to be an important determinant of financial policies and (b) pecking order behavior does not necessarily provide strong support for the pecking order theory. ix

9 MANAGING FINANCIAL POLICY: EVIDENCE FROM THE FINANCING OF EXTRAORDINARY INVESTMENTS How do managers set financial policy? In a world of perfect and complete capital markets, capital structure and dividend policy are a matter of irrelevance [Modigliani and Miller (1958, 1961)]. However, once the Modigliani-Miller assumptions are relaxed, financial policy becomes a major consideration. Most popular theories of financial policy perform a marginal analysis for a firm with a new investment project that is extraordinary in spirit, rather than typical or routine. This paper uses a sample of firms making extraordinary investments to probe deeper into the driving forces behind theories of financial policy. For the most part, what we know about financial policy can be described as a large set of stylized facts, but little that distinguishes between competing views, and little that reliably identifies the driving forces behind these policies. In some sense, after Modigliani and Miller (1958, 1961) there are two points of departure for subsequent research on financial policy. One branch uses transaction costs, financial distress costs, agency conflicts, and taxes to produce optimal capital structures and dividend payouts. In this view, a tradeoff of the costs and benefits of borrowing determine a firm s financial policy. The firm borrows to the point where the tax and incentive benefits of increased leverage are just offset by the value lost to increased expected costs of financial distress is determined by a tradeoff between the benefits of dividend payments [see Easterbrook (1984)] and the present value of the tax disadvantage of dividend payments. The firm sets target debt-to-value and dividend payout ratios, but because of transaction costs, does 1

10 not adjust continuously or immediately. Instead, firms only gradually move towards their targets. Another branch of the financial policy literature, developed by Myers (1984, 1991) and Myers and Majluf (1984), focuses on information asymmetries between managers and investors as the driving force of financial policy. In this view, managers prefer to finance investment with internally generated funds to avoid having to choose between issuing undervalued securities and forgoing positive net present value projects. Since managers know more about the true value of the firm s assets and the expected profitability of the firm s investment opportunities, investors are skeptical when the firm seeks external financing. If internally generated funds are insufficient to finance investment, such that external funds are necessary, securities that are relatively insensitive to asymmetric information, such as debt, are issued before riskier securities such as equity. Since there is no optimal capital structure, profitable firms accumulate financial slack and lower their leverage, while unprofitable firms are forced to issue debt and increase their leverage. Each of these branches of the literature has been filled out with considerable research in recent years, generally extending and fine tuning the model. In addition, each view receives considerable empirical support. However, this research has generally made little progress towards distinguishing between specific alternatives. Since each view of financial policy has evolved to describe the stylized facts, each view is generally consistent with the data. Most empirical research has focused on what happens to leverage ratios and the debt-equity choice, but the two views make similar predictions in most situations, making it very difficult to decide which view is receiving more support from the data. This paper attempts to determine which of these two views is a better firstorder description of how firms manage their financial policies. 2

11 Firms making extraordinary investments comprise a unique sample for learning about intended financial policy. In general, it is difficult to learn about planned financial policies unless firms take actions that in some way reveal their intent. Learning is further complicated by there being few unique implications of most competing theories. For example, when a firm merely replaces depreciated capital or makes a routine investment, both the pecking order and the static tradeoff theory with transaction costs predict that the firm will finance this investment with retained earnings if possible. To the extent that the transaction costs for debt issues are smaller than are those for equity issues, both theories predict that small requirements of external funds will be satisfied with debt financing. However, when a firm makes an extraordinary investment, this need not be the case. The extraordinary investment presents the firm with an opportune time to adjust its financial policy, returning to an optimum, if one in fact exists. At the same time, the extraordinary investment exerts substantial pressure on the firm to reduce accumulated cash holdings, and perhaps reduce dividend payouts, to limit external financing requirements, if minimizing external financing requirements is the first-order consideration when making financing decisions. Thus, examination of financial policy surrounding extraordinary investments may yield more powerful tests, and increase the likelihood of distinguishing between competing theories. To address how firms manage their financial policy when making extraordinary investments, I examine various financial policy variables in event-time relative to an industry based comparison group. This methodology provides a natural way to explore the dynamic aspects of financial policies by examining how these policies accommodate extraordinary investment in the years before, during, and after the event. The event-time analysis reveals that the financial policies of the sample firms can be reasonably characterized as pecking order behavior, much like those making ordinary investments. 3

12 In particular, (1) internal funds are the dominant source of financing, accounting for over 65% of investing activities in the event year, as compared to 86% for Value Line firms; (2) the relative importance of equity financing is unchanged for the firms making extraordinary investments versus ordinary investments; and (3) debt issues are the residual financing variable. Interestingly, dividend paying firms do not reduce or slow dividend growth before or during the event, but instead are more likely to increase their dividends per share, including firms that also raise external financing. Both the target-adjustment and pecking order theories have to some extent emerged from the survey studies of Lintner (1956) and Donaldson (1961) that find managers are reluctant to reduce dividends and prefer to finance investment with internal funds. As such, both of these theories can potentially explain the observed pecking order behavior in these data. The question is whether this observed pecking order behavior in financial policies is more consistent with the pecking order theory [Myers and Majluf (1984) and Myers (1984)] or a target-adjustment alternative. In order to address whether this observed pecking order behavior in financial policies is necessarily supportive of the pecking order theory described by Myers (1984, 1991) and Myers and Majluf (1984), I develop a test designed to distinguish between target-adjustment and pecking order models. At the heart of all theories of financial policy, are two basic questions: (1) how much financing should be raised? and (2) how should the financing be raised? Most empirical research on financial policy is devoted to addressing the second question. However, insights into which theory best characterizes how firms manage their financial policy can also come from addressing the first question. I argue that the target-adjustment model is associated with concave costs of external finance, while the traditional pecking order theory described by Myers (1984, 1991) and Myers and Majluf (1984) has costs of external finance that are convex in the amount 4

13 raised. Consequently, one competing hypothesis of the target-adjustment and pecking order models concerns what happens to cash holdings when firms raise external financing. The target-adjustment model predicts that firms use the event as an opportunity to replenish their cash holdings, while the pecking order predicts that firms try to minimize external financing requirements by reducing cash reserves. A direct test of these competing hypotheses produces empirical results more supportive of the targetadjustment view of financial policy. Overall, the empirical evidence suggests that (1) the existence of at least some transaction costs is necessary to explain the financial policies of the sample firms and (2) the mere presence of pecking order behavior does not provide strong support for the pecking order theory. 5 I. EXTRAORDINARY INVESTMENTS: MAJOR AND NON-ROUTINE A. Sample Selection Overview The dataset for this paper consists of major and non-routine (extraordinary) investments, which are defined as significant additions to the capital stock of the firm, and significant deviations from the firm s typical investment policy, respectively. Extraordinary investment events are used as an indicator of informative observations, exploiting the notion that these firms are relatively more likely to reveal their intended financial policy. The very nature of extraordinary investments requires that firms carefully manage their financial policy in the years surrounding the event, as the righthand side of the balance sheet must accommodate a major change on the left-hand side. Moreover, to the extent that extraordinary investments are characteristic of the projects described in most formal theories of financial policy, but not the projects that provide the empirical basis or motivation for the development of the theories, these events allow for out-of-sample tests.

14 The amount and intensity of thought associated with investment and financing decisions is much greater when these projects are major and non-routine. These decisions are not only carefully scrutinized by management, but also by the board of directors, as board approval is much more likely to be required for these expenditures than for ordinary investment outlays. In addition, these decisions are likely to be highly visible, inducing a manager with career concerns (valuing his reputation) to take additional care in his decision making. To the extent that extraordinary investments are material events, these actions require public disclosure and prominent placement in the annual report, informing investors and the labor market alike. Therefore, financial policy observations surrounding extraordinary investments should constitute a sample with a higher signal-tonoise ratio than a random sample or the population as a whole, and may reveal more about the intended financial policies of firms than typical or ordinary observations. In the end, this sample should lead to more efficient and powerful tests. 1 In addition, most popular theories of dividend and financing policies are developed in terms of a marginal analysis for a new project available to the firm [examples include Myers (1977), Leland and Pyle (1977), Myers and Majluf (1984), Bhattacharya (1979), Miller and Rock (1985)]. The new projects in these models are extraordinary in spirit, rather than typical or routine. Since these theories describe how firms manage financial policy when making extraordinary investments, their implications are likely to be strongest in this situation, and often, only in this situation. To the extent that extraordinary investments are characteristic of the events described by financial 6 1 One concern that arises is that there may be a sample selection bias. If investment and financial policies are jointly determined, then the use of an endogenous variable to select the sample can be problematic for the generality of the findings if a bias results. This potential bias in no way undermines the logic of why it is interesting to examine the financial policy of firms in the years surrounding extraordinary investment decisions, merely requires that appropriate care be taken in the sample selection and the interpretation and generalization of the findings.

15 policy theories, it is natural (at least reasonable) to examine the financial policies of firms surrounding these events. Moreover, these theories have largely evolved from casual observation or actual surveys of how firms typically manage financial policy. To the extent that extraordinary investments do not provide the empirical motivation for the development of the financial policy theories, these observations represent an unexplored and essentially independent data source for testing. Finally, I am going to try to distinguish between competing views of financial policy and the empirical tests will have some assumptions. The sample is selected to be reasonably consistent with these assumptions. In particular, I am going to argue that one of the main distinctions between competing views relates to the costs of external financing convex in the amount raised in the pecking order theory and concave in the target-adjustment view. Since some epsilon transaction costs are required in the pecking order theory to explain why firms do not continuously issue securities, there may be an S shape cost function, concave at first and then convex. The extraordinary investment sample is likely to consist of firms that are at the convex portion of the function, if the convex portion in fact exists, because these firms are raising a substantial amount of funds. At the same time, the test is going to assume that the costs of external financing, whatever their shape, are constant through time. This is another one of the motivations for the extraordinary investment sample, because to directly select the sample based on large security issues would increase the likelihood of identifying firms that have experienced a shock to their cost function. In other words, this procedure would identify firms that have issued securities because the costs of external financing are temporarily low. Finally, the test assumes that the firm s marginal benefit of slack is the same under both views of financial policy. In other words, the marginal benefits of financial slack due to information asymmetry about the value of assets and potential 7

16 investment opportunities are assumed to be negligible immediately following the extraordinary investment. There can be increased uncertainty about whether subsequent investment is necessary, but the test assumes that there is no increased information asymmetry about the value of assets and prospects. This seems reasonable for a sample of firms that have just completed an extraordinary investment and thus just revealed what private information they may have had. In addition, the events are extraordinary with respect to both the past and the future. This is to rule out permanent shifts in investment policy where the level of information asymmetries may also be changing dramatically. B. Sample Selection Details The extraordinary investment sample is selected by identifying firms making major investments that are also non-routine when compared to the firm s own investment patterns. The universe of firms for the purpose of this study are those covered by Value Line from 1970 to 1996, excluding firms in financial industries. There are 2,728 unique 8 firms, representing 35,613 firm-years in the Value Line sample. 2 All of the accounting data are from Compustat. The CRSP Events database is used to compile a list of Value Line firms completing mergers so that these observations can be excluded from the extraordinary investment sample. Details on the Value Line sample, the industry classifications, and the variable definitions are in Appendix B. In order to compare annual investment expenditures across both firms and time, some type of transformation is required. It is natural to scale investment by a measure of assets in place, such as book assets. However, book assets do not reflect current or historical advertising and R&D expenditures. Therefore, book assets are adjusted to reflect the accounting treatment of advertising and R&D investment (details on this paper. 2 Andrade and Stafford (1999) develop the Value Line based industry classifications used in this

17 procedure can be found in Appendix B). Annual investment intensities are calculated as investment (advertising + CAPX + R&D) in year t divided by long-term book assets (book assets plus R&D capital (RDC) plus advertising capital (ADC) minus cash & marketable securities minus working capital (WC)) at the end of year t-1 (see Appendix B for details on variable definitions). Advertising + CAPX + R&D i, t i, t i, t Investment Intensity, = (1) i t Assetsi, t 1 + ADCi, t 1 + RDCi, t 1 Cashi, t 1 WCi, t 1 All annual investment intensities greater than 0.25 (approximately the 90 th 9 percentile across all firm-years) are considered major investments. 3 In order to identify non-routine investments, annual percentage deviations from each firm s mean investment intensity are calculated for all firms that have at least four annual observations. All annual percentage deviations greater than 50% are considered nonroutine investments. Extraordinary investment events are defined as the intersection of the major and non-routine investments. The final sample satisfies the following additional criteria: 1. Must have been in Value Line for at least two years before the event to avoid sample selection bias. Firms entering the Value Line sample are likely to have recently experienced unusually high growth, and may not be directly comparable to the other firms, which are generally more established. 2. Must not be part of a consecutive series of investment events that is longer than three years. It is not clear whether or not unusually long-lived extraordinary investment events were originally planned by management. 3. Firms with multiple extraordinary investment events must have at least four years between them. This is to allow for meaningful analysis of pre- and post-event financial policy. 3 Annual investment intensities greater than 1.0 are dropped from the sample (20 observations).

18 4. The firm does not make an acquisition or merger in the year of the extraordinary investment. In particular, if the firm makes an acquisition greater than 5% of beginning of period long-term book assets, or is identified as an acquirer in the CRSP Events database, the observation is excluded. Mergers are excluded because there may be interactions between the financing choice and the type of the deal, which are not necessarily related to the firm s intended financial policy. The final sample includes 794 extraordinary investment event-years by 538 different firms. Table 1 displays the number of events by year and by industry, as well as the number of Value Line firms that otherwise meet the sample criteria, and Appendix C describes a few of these events. 10 II. METHODOLOGY FOR ANALYZING EVENT-TIME FINANCIAL POLICY To address how firms manage financial policy when making extraordinary investments, I examine various financial policy variables in event-time relative to an industry based comparison group. There is a dynamic aspect to financial policy that is generally difficult to capture unless some type of event-time analysis is performed. The event-time analysis provides a natural way to explore how financial policy accommodates extraordinary investment in the years before, during, and after the event. In order for the output to be useful, a meaningful comparison group and adjustment procedure are required to determine the level of, or change in, the variables expected in the absence of an extraordinary investment. There are two important considerations in determining the abnormal financial policy variables: (1) mean reversion in accounting variables; and (2) cross-sectional dependence of abnormal variables. First, a wide variety of accounting variables exhibit mean reversion, which may induce spurious correlations if event firms tend to have either high or low levels of the variable before the event. Second, major corporate actions are not random events,

19 and thus may not represent independent observations. The very nature of an event sample is that all of the sample firms have chosen to participate in the event, while other firms have chosen not to. This may lead to positive cross-correlations in abnormal financial policy variables, leading to overstated test-statistics if not accounted for. A. Calculating Abnormal Financial Policy Variables The comparison group consists of all firms in the same industry, with similar levels of the variable in the year prior. Specifically, for each variable examined, three groupings of the Value Line sample firms are formed (independent of industry) based on whether the level of the firm s financial policy variable falls below the 33 rd percentile, between the 33 rd and 67 th percentiles, or above the 67 th percentile of the distribution of the variable across all firm-years in the year before the event. These groupings are then intersected with the industry classifications to form the comparison groups. This method of grouping assumes that firms in the same industry are subject to similar shocks that may affect the cross-sectional variation in the variable, but which are unrelated to the actual investment event. Moreover, by matching on the prior-year level of the variable, spurious correlations related to mean reversion in accounting variables are less likely. This comparison group definition is similar in spirit to the matching procedure advocated by Barber and Lyon (1996), which controls for both the pre-event level of the variable and the industry of the sample firms. Once the comparison group is identified, an abnormal financial policy variable is calculated for each sample firm by subtracting the mean of the variable associated with the comparison group, as in equation (2). This adjustment should roughly account for the typical behavior of firms in the comparison group, and thus provide a meaningful benchmark against which to evaluate the sample firms financial policy variables. 11 AbnX i, t = X i, t mean( X j, t ) where i indexes all event firms and j indexes all firms in the comparison group for firm i. (2)

20 Finally, the mean and standard error are calculated using a variant of the time series of cross-sections approach developed by Fama and MacBeth (1973). In this approach, the time series mean of annual cross-sectional means of the individual abnormal financial policy variables is calculated, and a t-statistic is calculated using the time series standard error. In other words, each year in calendar-time, I calculate the mean abnormal financial policy variable for all firms completing an extraordinary 12 investment that year, AbnX t. The mean and standard deviation of the time series of annual means is then used to calculate the t-statistic, ( T mean( AbnX t ) / std( AbnX t ) ). This time series of cross-sections approach produces standard errors that are robust to cross-sectional dependence because the variation in the annual means includes the effects of cross-correlated abnormal financial policy variables. B. Simulation Evidence One concern with any type of adjustment is that the properties of the resulting statistic may not be known, particularly when the underlying data are not well behaved. This problem is common to most corporate event-time studies, and Barber and Lyon (1996) document how sensitive results can be to different methodologies. To further complicate the matter, there are several differences between the financial policy variables in this study and the operating performance variables examined by Barber and Lyon, which may be problematic. The most severe is that many financial policy variables are bounded below at zero, and often have a large mass at zero, whereas profitability measures, such as return on assets, are close to normally distributed. Moreover, the sample size of 50 firms that Barber and Lyon examine is many times smaller than the extraordinary investment sample in this study. In order to determine whether the adjustment procedure described above is reasonable for the variables of interest in this

21 study, I perform simulations to test the size of the test statistics. The details of the simulation procedure and the results are in Appendix D. The simulation results on the specification of the test statistics are presented in Table D.1. In particular, the fraction of random samples that reject the null hypothesis are reported. For virtually all of the variables studied, the adjustment procedure produces well-specified t-statistics. The only exceptions are that the test statistics for the cash-tosales, dividend-to-income, and long-term debt-to-sales ratios tend to reject too often when negative. Overall, the abnormal financial policy methodology appears to produce wellspecified t-statistics, which should be robust to cross-correlated observations. C. Investment Activity in Event-Time To demonstrate the effectiveness of the event-time approach, abnormal investment activity results are displayed in Table 2. The sample consists of firms making extraordinary investments, and as such, it is important to examine how the investment activity of these firms evolves in event-time. Beginning of period long-term book assets scales all the various components of investment activity. For the most part, this table simply confirms that these are indeed extraordinary investments, as the event firms on average have investment intensities over 14 percentage points higher than comparable 13 firms, during the event year (t-statistic = 21.7). 4 This is more than double the average aggregate investment intensity for Value Line firms from , which is 12.7% (reported in Appendix E). On average, it appears that there is an investment run-up the year before the event, culminating in year t with an extraordinary investment. The investment event is immediately followed by a sharp decline in investment activity in years t+1 and t+2. This 4 Equivalently, the average investment intensity for the event firms is 14.2 percentage points higher than what is expected, given the firm s industry and the amount invested in the year prior.

22 same general pattern persists when sales, rather then long-term assets, scale investment activity. Finally, most of the abnormal investment comes from CAPX. Advertising expenditures are marginally higher during the event year, and R&D expenditures are significantly larger, but the vast majority of the abnormal investment intensity relates to CAPX. 14 III. FINANCIAL POLICY SURROUNDING EXTRAORDINARY INVESTMENTS The estimates presented in this section provide insights into how financial policies adjust to accommodate the extraordinary investment in year t, and capture the average financial position of firms making extraordinary investments in the years surrounding the event. Emphasis is on cash flows, dividends, external financing, and the resulting capital structures. The reported coefficients are averages of adjusted financial policy variables for firms making extraordinary investments in year t. The variables are adjusted by subtracting the average of the comparison group firms in the same industry, with similar levels of the variable in year t-1. In other words, if a sample firm from industry j, has high leverage (above the 67 th percentile) in the year prior to the extraordinary investment, the average leverage of all firms with leverage ratios above the 67 th percentile in year t-1, from industry j, serves as the benchmark. After each sample firm s leverage has been adjusted by subtracting the corresponding benchmark, the average of the adjusted leverage ratios is calculated each year in calendar time. The time series mean and standard deviation of annual averages are used to calculate the reported statistics. The average adjusted variables are reported in event-time from three years before to two years after the event.

23 15 A. Cash Flows Several different abnormal cash flow measures for firms making extraordinary investments are presented in Table 3. In order to facilitate event-time comparisons, all cash flow measures are scaled by sales rather than total book assets, as sales should be less sensitive to the extraordinary investment event. 5 Thus, the variables are actually cash flow margins, which are typically interpreted as measures of profitability. The particular cash flow variables examined are EBITDA, operating cash flows, pre-tax cash flows, and after-tax cash flows. Details of the variable definitions are in Appendix B. Firms that make extraordinary investments have slightly higher cash flows than comparable firms do in the years before the investment, consistently peaking in year t-1. For example, in the three years prior to the extraordinary investment, average abnormal operating margins range from to (corresponding t-statistics are 1.34 and 2.34). The aggregate average cash flow margin for all Value Line firms is 0.13 (results not reported), implying that firms making extraordinary investments have cash flow margins roughly 4% to 7% higher than non-event firms prior to the event. During the event year, and following the extraordinary investment, cash flow margins are indistinguishable from those of the comparable firms. The average abnormal operating margins during the event and in the two years after the event are , , and (t-statistics equal -0.44, 0.06, and 0.48, respectively). The results from examining changes in the cash flow-to-sales ratio are consistent with the level results. The three-years before the event are characterized by marginally positive abnormal increases in cash flow margins, and then changes are generally flat following the event. 5 The extraordinary investment event is essentially tantamount to a significant increase in total book assets. However, there is no such mechanical relation between the investment event and sales.

24 Cash flow margins and return on assets are often used as measures of operating performance. The assumption is that if the profitability of a firm is at least as high as that of comparable firms, then the firm is performing efficiently. However, the denominators in each of these measures are also related to performance. For example, examining the cash flow margin for a firm experiencing a large decrease in sales, with no change in profitability, may disguise the fact that the firm is enduring serious operating problems. To ensure that the previous cash flow inferences are consistent with strong operating performance, sales growth is also examined (results not reported). The average abnormal sales growth of the event firms is invariably non-negative and reliably positive in the year before the event and during the event year. This suggests that the operating performance of the sample firms is indeed slightly higher than that of comparable firms in the years before the extraordinary investment, and certainly no worse than average subsequently. B. Dividend Policy Do managers reduce dividends or slow dividend growth to accommodate extraordinary investment? While there is a large literature devoted to dividend policy, there is little empirical evidence on the interaction between dividend policy and extraordinary investments [see the review by Allen and Michaely (1995)]. Firms making extraordinary investments provide a unique sample for studying this aspect of financial policy, as the investment event exerts strong pressure on the firm to alter its dividend policy. A dividend reduction may allow the firm to avoid the direct transaction costs of issuing securities, as well as the potential costs associated with asymmetric information. On the other hand, dividends may serve an important signaling role, such that a dividend cut is not actually a feasible source of financing. 16

25 Table 4 reports the fraction of Value Line and event firms satisfying various dividend policy criteria and test statistics denoting whether the fractions are significantly different across the two samples. There are several interesting observations. First, dividend increases defined as a 25% increase in dividends per share, adjusting for stock splits and stock dividends are significantly more likely for event firms in the years before, and the year of, the extraordinary investment. In particular, during any given year, 11.6% of the Value Line firms increase their dividends per share at least 25%, whereas in both the year before the event and the year of the event, roughly 20% of the firms making extraordinary investments increase their dividends by this amount. At the same time, dividend decreases are significantly less likely for event firms in the years before and during the investment event. Firms making extraordinary investments are only half as likely to reduce their dividends per share by at least 25% as other Value Line firms in the 17 two years before the event, and during the event year. 6 Moreover, these general findings hold for the subset of event firms that raise external financing in year t. 7 This suggests that dividends do not accommodate extraordinary investment in the sense of providing a source of funds, but rather dividend payments exert additional pressure on an already tight sources and uses constraint. Second, it appears that firms making extraordinary investments are significantly less likely to pay dividends in the years before the event than other Value Line firms. 6 These general findings (event firms are significantly more likely to increase dividends in year t, and marginally less likely to reduce dividends in years t-1 and t) are also true for the group of firms that make substantial dividend payments in year t-3, defined as those above the 75 th percentile of dividends-tosales. The notion being that firms that actually pay substantial dividends will have a greater source of funding from a dividend cut, than firms with low dividend payout, and thus may be more likely to use this financing option. 7 Firms raising external financing are identified as those with debt issues plus equity issues in year t larger than 5% of beginning of period long-term book assets. The results are virtually identical if these firms are instead identified as those with debt issues plus equity issues greater than 25% of investment in year t.

26 However, by the time the event has occurred, the payment frequency of event firms is indistinguishable from that of the rest of the Value Line sample. Consistent with this evidence, the event firms are nearly twice as likely to initiate dividends in the years before the event, and just as likely in the years afterwards. Dividend omissions by the event firms are not reliably different from other Value Line firms, except during the event year and again two years after the extraordinary investment. During the event year, firms are less likely to omit dividends, whereas, two years after the extraordinary investment firms are twice as likely to omit dividends. There is some hint that dividend policy may accommodate investment at longer horizons, much the way it does for small, rapidly growing firms. In other words, firms that expect to invest more than their cash flows allow are less likely to pay dividends at all. Abnormal event-time dividend policy variables are presented in Table 5. Emphasis is on dividends-to-sales and growth in dividends per share. Again, scaling by sales should dominate total book assets or equity income, both of which have a more 18 direct relationship to the event. 8 For the most part, the event firms have dividend policies that are remarkably similar to comparable firms. As suggested in Table 4, growth in dividends-per-share is higher for event firms before the event, and lower immediately after. This confirms the previous evidence that dividend-paying firms do not reduce dividends, or even slow dividend growth, to accommodate their extraordinary investments. Overall, it appears that once dividends are paid, which describes the vast majority of the sample, firms are extremely reluctant to reduce payouts, or even to slow their 8 Moreover, simulation evidence suggests that the specification (size) of the test statistics for the traditional payout ratio (dividends divided by equity income) is sensitive to sample size. The results for dividends-to-equity income, reported in Table B.1, for a sample size of 500 are somewhat misleading, in that all other sample sizes are poorly specified.

27 growth unless they hit bad times. However, dividend policy may accommodate investment for the event firms much the same way it does for small, growth firms both are less likely to pay dividends at all. C. External Financing Firms making extraordinary investments are selected for study largely because these firms are likely to rely more heavily on external financing. Table E.5 in the appendix reports that this is indeed the case. On average, external financing accounts for 32.9% of investing activity for firms making extraordinary investments, while external funds account for only 14% of investing activities for firms making ordinary investments. In some sense, this justifies the approach of examining the financial policies of firms making extraordinary investments because it will be difficult to distinguish between competing theories of financial policy with ordinary investment observations. Even for this sample, internal funds are by far the dominant source of financing. Also interesting is the finding that the relative importance of debt and equity financing for ordinary and extraordinary investment is similar. Debt issues account for 89.2% of external financing for ordinary investments and 93.8% for extraordinary investments, on average. Table 6 reports the event-time frequency of debt and equity issues for firms making extraordinary investments and for the full Value Line sample. Again, it is clear that debt issues are the dominant source of external financing. Over 56% of the event firms issue debt worth over 5% of beginning of period long-term book assets in the year of the event, compared to 25% of all Value Line firms. This is also dramatically larger than the 11.7% of event firms issuing equivalent amounts of equity in the event year. The vast majority of security issues occur during the actual event year. Nonetheless, debt issues also appear to be significantly more likely in the year after the event, suggesting that there is some follow-up financing that is required. In particular, 19

28 25% of all Value Line firms issue debt over 5% of beginning of period long-term book assets in any given year, whereas, over 32% of the firms making extraordinary investments issue an equivalent amount of debt in year t+1 (t-statistic of difference is 2.17). The event-time frequency of equity issues is not as dramatic, as they are more evenly spaced over years t-1 through t+1, peaking in the year before the event. Table 7 presents abnormal external financing variables, which largely confirm the results presented so far. Net debt issues are by far the primary source of external funding for firms making extraordinary investments, with an abnormal measure more than 7 times larger than that for net equity issues. In the years before the event, net debt issues are somewhat lower for event firms than for comparable firms. Virtually all debt financing occurs in the year of the event. In addition, most of the net debt financing comes from long-term debt issues, as long-term debt reductions for event firms are indistinguishable from those made by comparable firms. Moreover, short-term debt issues are not particularly important, although they are marginally higher than expected in the year of the event. Again, equity issues are more evenly spaced over years t-1 through t+1. Equity repurchases by firms making extraordinary investments are not reliably different from comparable firms, which is interesting in light of the dividend policy evidence. In other words, not only are firms reluctant to reduce dividends in the years before and during extraordinary investments, but equity repurchase programs appear to continue unaltered as well. In the year after the investment event, the average abnormal equity repurchase is significantly lower for the sample firms than for the comparable firms. This is sufficient to make abnormal net equity issues in year t+1 marginally positive, despite the average amount raised from equity issues being similar for event firms and comparable firms. 20

29 The financing of extraordinary investments appears to be quite similar to that for ordinary investment, with internal funds being the dominant source, equity issues remaining relatively unimportant, and debt being the residual variable. Although equity issues are significantly more likely for event firms than for comparable firms, debt financing overwhelms both the amount of equity financing and the frequency of equity issues. D. Leverage Most empirical studies of financial policy focus on changes in leverage. However, there are conflicting explanations for the determinants of leverage. For example, in some theories leverage is the outcome of financing decisions, while in others, leverage is one of the primary choice variables, such that financing decisions are merely a means to an end. Regardless of which view is taken, it is interesting to examine how leverage ratios evolve in event-time for firms making extraordinary investments. 9 In the three years before the event, long-term debt to total book assets is significantly lower for event firms than for comparable firms, with average abnormal long-term debt ratios ranging from to (corresponding t-statistics are and -4.74). As indicated in the previous table, most of the financing during the event year is provided by debt issues, resulting in a large increase in leverage. At the end of the event year, the average abnormal long-term debt ratio is 2.41 percentage points higher than expected, with a t-statistic of The average aggregate long-term debt ratio for the Value Line firms is (reported in Appendix Table E.1), implying an average increase in long-term debt ratios for firms making extraordinary investments over 10%. In the years following the event, long-term debt tends towards the average of the 21 9 Most research focuses on the capital structure of long-term claims, but it is interesting to examine short-term claims as well. Appendix E reports the event-time behavior of short-term assets and liabilities.

30 comparison firms, and is in fact indistinguishable in years t+1 and t+2. Examining total liabilities produces similar results, as short-term and long-term debt ratios behave similarly, although by year t+2, abnormal total liabilities are significantly negative. On the whole, firms making extraordinary investments have significantly lower levels of debt than comparable firms prior to the event, but similar to the comparison group afterwards. As such, common equity as a fraction of total assets behaves just the opposite higher than average prior to the event, over correction the year of the event, and then drifting upwards afterwards. E. Pecking Order Behavior in Financial Policy The event-time analysis suggests that the pecking order descriptions of Donaldson (1961) and Myers (1984, 1991) can reasonably characterize the financial policies of the firms making extraordinary investments. Internal funds are the dominant source of financing, the relative importance of equity issues is the same for firms making extraordinary investments as for firms making ordinary investments, and debt issues are the residual financing variable. Interestingly, dividend-paying firms do not reduce dividends or slow dividend growth in the year of the extraordinary investment or in the prior years. This is consistent with the findings of Fama and French (1997), that dividends do not accommodate ordinary investment. 22 IV. DISTINGUISHING BETWEEN PECKING ORDER BEHAVIOR AND THEORY The results from the previous section are very suggestive of pecking order behavior, but there has been no direct test of the pecking order theory. In order to determine whether the data are necessarily supportive of the pecking order theory of Myers (1984) and Myers and Majluf (1984), the predictions of the pecking order theory are compared to those of the target-adjustment model. To some extent, both of these

31 theories have evolved to explain existing empirical regularities, including the surveybased findings of Lintner (1956) and Donaldson (1961). Consequently, it is generally difficult to distinguish between these two theories. This section interprets the existing stylized facts and attempts to devise a test that can distinguish between the targetadjustment and pecking order models of financial policy. A. Interpreting the Existing Evidence Empirical research has been instrumental in the development of many popular theories of financial policy, and has produced a wealth of information about how firms typically manage these policies. For the most part, these studies take investment as given, and focus on how investment is financed by analyzing the cross-section of leverage ratios, the time-series of leverage ratios, the announcement period stock-price reaction to financing decisions, and the determinants of the debt-equity choice. Cross-sectional regressions generally find that leverage ratios are positively related to the tangibility of the firm s assets and investments, and negatively related to the firm s market-to-book ratio, R&D investment, and profitability [see for example: Bradley, Jarrell, and Kim (1984), Long and Malitz (1985), Titman and Wessels (1988), Barclay, Smith, and Watts (1995), Rajan and Zingales (1995), and Fama and French (1997)]. Generally, these findings are consistent with the existence of optimal capital structures. More direct evidence on how financial policy is set comes from time-series models that find mean reversion in leverage and dividend ratios, consistent with firms gradually adjusting towards targets [see Fama and Babiak (1968), Taggart (1977), Marsh (1982), Jalivand and Harris (1984), Auerbach (1985), Opler and Titman (1997), and Fama and French (1997)]. Underlying most of these empirical tests is the notion that an optimal capital structure indeed exists, and these findings are generally viewed as supporting this notion. 23