HOW STICKY ARE DIVIDENDS?

Transcription

1 The Pennsylvania State University The Graduate School Smeal College of Business HOW STICKY ARE DIVIDENDS? ANALYSIS UNDER CASH SHORTFALLS A Dissertation in Business Administration by Thomas O. Miller 2011 Thomas O. Miller Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy December 2011

2 The dissertation of Thomas O. Miller was reviewed and approved* by the following: David Haushalter Clinical Associate Professor of Finance Dissertation Adviser Chair of Committee Laura Field Associate Professor of Finance Peter Illiev Assistant Professor of Finance Ed Coulson Professor of Economics William Kracaw David Sykes Professor of Finance Department Chair *Signatures are on file in the Graduate School ii

3 Abstract: This paper analyzes dividend stickiness by examining managers decisions whether to cut dividends when facing cash flow shortfalls. I present two hypotheses regarding this decision. One, which I label tradition, suggests that managers are looking back at dividend history, and view cutting as a last resort. The other suggests managers are forward looking and their decisions primarily on expectations about future performance. Analysis of maintaining and cutting groups with poor cash flows lend support to both hypotheses, but firms that maintained their dividend see significant increases in sales growth, and in some cases cash flow, the following year, indicating that managers are forward looking. Firm leverage is particularly important for determining whether a firm cuts its dividend under cash shortfalls. iii

4 Contents List of Tables...v List of Figures...vii 1 Introduction Literature Review Overview of Dividend Policy Repurchases, the Substitution Hypothesis, and Dividend Disappearance Related Literature Hypotheses The Tradition Hypothesis The Forward Looking Hypothesis Data Who s Their Dividends? A Summary Analysis Market wide Data Dividend Decisions Under Financial Stress and Cash Shortfalls Defining Financial Distress Why Do Firms Dividends in Extreme Stress? How do firms perform after extreme stress? How Do Firms Fund Dividends During Cash Shortfalls? Regressions Analysis Market wide Regressions Regressions on Financially Distressed Firms Industry Fixed Effects Regressions Direct Tests of Hypotheses Dividend Policies of Firms with Public Debt Summary Statistics Logit Regression Are Real Dividends Sticky? A Look at Inflation adjusted Payouts Dividend Policies at Industry Level Conclusion References Appendix...47 iv

5 Tables Table 1 - Summary Stats Grouped by Decision to Dividend 47 Table 2 - Summary Stats by Decision to Dividend w/income<0 48 Table 3 - Summary Stats by Decision to Dividend w/ocf<0 49 Table 4 - Summary Stats by Decision to Dividend w/ocf<last Year's Dividend 50 Table 5 - Summary Stats by Decision to Dividend w/ocf+cash on Hand Last Year<Last Year's Dividend 51 Table 6 - Summary Stats by Decision to Dividend w/income & OCF<0 52 Table 7 - Switch Table of Repeat Distressed Firms 53 Table 8 - Future Performance After Income and OCF<0 54 Table 9 - Future Performance After OCF<Last Year's Dividend 54 Table 10 - Summary of Future Performance by Decision to Dividend with OCF<Last Year's Dividend 55 Table 11 - Fixed Effect Logit Regressions on MAINTAIN variable 56 Table 12 - Fixed Effect Logit Regressions on MAINTAIN variable w/ocf and Income<0 57 Table 13 - Fixed Effect Logit Regressions on MAINTAIN variable when OCF<Last Year's Dividend Table 14 - Fixed Effect Logit Regressions on MAINTAIN variable when OCF+Last Year's Cash<Last Year's Dividend Table 15 - Industry Fixed Effect Logit Regressions on MAINTAIN variable 60 Table 16 - Fixed Effect Univariate Regressions on MAINTAIN on Future Sales Growth 61 Table 17 - Summary Stats for Firms with Public Debt 62 Table 18 - Summary Stats for Firms with Public Debt and OCF & Income<0 63 Table 19 - Summary Stats for Firms with Public Debt and OCF<Last Year's Dividend 64 v

6 Table 20 - Summary Stats for Firms with Public Debt following a Downgrade 65 Table 21 - Logit Regressions on MAINTAIN variable following Credit Ratings Downgrades 66 Table 22 - Summary Stats Grouped by Decision to Real Dividend 67 Table 23 - Summary Stats Grouped by Decision to Real Dividend with OCF+Last Period's Cash on Hand<Last Period's Dividend Table 24 - Summary Stats Grouped by Decision to Real Dividend with OCF<0 & Income< Table 25 - Dividend Policy by Industry 70 vi

7 Figures Figure 1 - Percentage of Firms ing or Increasing Dividends by Year 71 Figure 2 Financing Methods of Dividend Paying Firms with Cash Shortfalls 72 Figure 3 - Percentage of Firms ing Dividend by Dividend Payment History 73 Figure 4 - Percentage of Firms ing or Increasing Real Dividends by Year 74 vii

8 1 Introduction Since Lintner (1956), it has been well documented that dividend payouts to shareholders are smoothed, with last period s payout having the largest impact on this period s. Lintner s findings along with Miller and Rock s (1985) documentation of a negative reaction to dividend cuts have led to the interpretation of dividends as sticky. Once a firm starts paying a level of dividends it becomes very reluctant to cut, meaning they are essentially stuck there. One explanation for stickiness is that it reflects tradition. Managers may believe that if their firm has been paying dividends for an extended period that those dividends are part of the identity and culture of the firm. Consistent with this argument, Brav, Graham, Harvey, and Michaely (2005) report in their survey of managers that most consider maintaining dividends of equal importance to determining investment levels, which is in stark contrast to Miller and Modigliani s (1961) dividend irrelevancy hypothesis. Skinner (2008), building on Brav et. al (2005), also suggests that firms pay dividends out of historical obligation. This represents a view that stickiness signals the past; firms that have always paid dividends will strive to continue that policy, and perhaps incur serious costs to do so, all to avoid going against long-standing tradition. An alternative explanation is that firms use dividends as a signal of their future expectations. Benartzi, Michaely, and Thaler (1997) examine whether dividend changes signal the future or the past. They find that although increases do not lead to future increases in cash flow, firms that raise dividend payouts are much less likely to experience declines in cash flow in future periods. Cash flows fluctuate even for the most stable firms, but a fluctuating dividend is less attractive to income-seeking investors. To address this disconnect between cash flow 1

9 stability and investor preferences, firms maintain dividends to retain consistency in shareholder payouts, a phenomenon Lintner refers to as dividend smoothing. Under this model, short-term shocks should not affect dividend payouts, unless they also indicate changes to a firm s future earning potential. To further investigate explanations for dividend stickiness, I focus on the dividend decisions of firms facing cash flow shortfalls. Following severe negative shocks to cash flow and/or earnings, dividend-paying firms are faced with a decision between two presumably costly alternatives. The firm can continue to pay dividends at current levels, either by drawing down internal cash reserves or engaging in external financing, or cut dividends and risk potentially giving a negative signal to the market and reduction in share price. Capturing this second concern is a 2008 quote from then BP CEO Tony Hayward. Defending a decision to maintain dividends as oil prices plummeted and BP took significant losses, he said, I pay taxes, so I don't go to jail. I pay dividends, so I don't get fired. 1 Similarly, Leary and Michaely (2010) note that dividend-paying stocks are often held by income-seeking institutional investors very concerned about payout levels. Although there is clear pressure to maintain dividends, how firms do so when facing cash shortfalls is less clear. Anecdotal evidence indicates that sometimes firms are raising debt levels to fund dividend payments. The idea of borrowing to pay dividends became popular during the early 2000s when interest rates were at historical lows. Some extreme cases, such as aircraftparts manufacturer TransDigm Group Inc., involved the issue of $425 million in bonds to fund a $360 million dividend payout to shareholders. 2 1 BP Takes Heat from Two Sides on Dividend, Wall Street Journal Online, Herron, J., June, 9, Borrowing for Dividends Raises Worries, Wall Street Journal Online, Rappaport, L., Oct. 6,

10 The first goal of this paper is to provide a market-wide analysis of firms dividend decisions, i.e. the decision to maintain or increase dividends from last period versus cut the dividend. The two hypotheses I examine, outlined in detail later, relate to whether this decision is primarily influenced by the past (the firm s history of paying and maintaining dividends), or the future (the firm s future potential to support payouts). The analysis focuses on dividendpaying firms and begins with summary statistical analysis of firm characteristics, including indicators of financial condition, historic dividend decisions, and growth trends. For most variables there are statistically different means and medians between the maintaining and cutting groups. Two exceptions that do not differ significantly are cash-to-total asset ratio and capital expenditure. As would be expected, the maintaining firms are performing much better than the cutting firms with higher returns, cash flows, asset growth rates, sales growth rates and lower leverage. In the next step of the analysis I identify firms that have faced severe negative shocks in both income and operating cash flow and are therefore most likely to be facing pressure to cut their dividends. The results show nearly half of firms maintain dividends even when experiencing both negative cash flows and accounting income. For firms that maintain dividends when cash flow is less than the established dividend levels, there is a significantly higher increase in operating cash flow compared to firms that cut their dividend. Asset growth is also significantly higher for these firm two years following the decision. To better understand the probability a firm will maintain its dividend in a given period, I conduct a series of logit regressions. These regressions are run on both the whole dividendpaying sample and on three sub-samples of financially distressed firms. The results of these regressions show that leverage and whether a firm cut its dividend last period are key factors in 3

11 determining whether the firm maintains its dividend. Industry-fixed models also produce similar results, along with significant coefficients for asset growth, profitability, and firm size. Also in regressions on a sub-sample of firms with negative cash flow and income, a firm s streak of years maintaining its dividend entering the period does not load as significant. This suggests that a firm s long-term dividend history (i.e. its tradition) may not be a key factor in the decision to maintain or cut. This section concludes with a series of univariate cross-sectional regressions of future sales growth with the decision to maintain or cut this period as the independent variable. The results suggest that maintaining firms see higher sales growth in subsequent years than cutting firms, supporting the forward-looking hypothesis. The next section of the paper focuses on the dividend policies of firms with publicly traded and rated (by Standard & Poor s) debt. This section follows the framework of the previous sections of the paper: summary statistical analysis of all firms and distressed subsamples, and a logit regression of firms with ratings downgrades. Unique to this sample are repeated instances of cutting firms being older than maintaining firms, which was not the case earlier. Results from the logit regression are consistent with the previous section where leverage and maintaining last period seem to be the most significant variables. Stickiness is clearly observed in nominal dividends, but the behavior of firm-level real (i.e. inflation-adjusted) dividends is largely unexplored. The next section which examines decisions involving real dividends finds that fewer firms (about half) are able to maintain their real dividends from year to year. Firms that maintain real dividends tend to be consistently younger than cutting firms across all sub-samples. Under conditions of financial distress, the maintain rate of real dividends drops to below 20% (17.8% for firms with negative operating cash flow and income). 4

12 The final section of the paper examines the subset of firms that never cuts their dividend during the sample period. This is done by comparing the number of firms by industry in three categories: total firms, dividend payers, and firms who never cut their dividends. There does not seem to be a clear relationship between the number of dividend payers in an industry and the number of those payers who never cut dividends. The overall take-away from the analysis is that firms in even extreme financial distress continue to pay and maintain their dividends despite their dire condition. While the firm s history of paying dividends does seem to have some influence on a firm s decision to maintain dividends, it seems that the most recent year matters far more than any long-standing tradition. Leverage seems to have a strong effect on the decision to cut dividends, with cutting firms having much higher leverage than firms that maintained across all sub-samples. The paper is outlined as follows. Section 2 reviews the relevant literature, Section 3 outlines the hypotheses, Section 4 describes the data, Section 5 provides summary statistical analysis, Section 6 examines funding dividends, Section 7 provides regression analysis, Section 8 focuses on firms with public debt, Section 9 looks at inflation-adjusted dividend policy, Section 10 addresses dividend policy at the industry level and Section 11 concludes. 2 Literature Review 2.1 Overview of Dividend Policy Frankfurter and Wood (1997) note that the evolution of corporate dividends has been closely intertwined with the evolution of the corporation itself. Dating back to the 1600s, European ships would raise funds before venturing to the New World and then disperse all capital and profits back to investors. As the industry developed, ships stopped liquidating and became 5

13 going concerns, distributing only their profits back to investors and retaining the capital. These entities would eventually grow into the corporations we know today. In the early days of the modern corporation, financial information was often scarce and unreliable. Because dividend payments were clearly observable, investors would value firms using dividend information more so than reported earnings data. Changes in dividend policy would in turn be reflected in the stock price. However, over time as markets became more efficient and asymmetric information reduced, investors need of dividends to value firms was mitigated. The continued importance of dividends in corporate finance has puzzled many researchers, notably Black (1976), and has lead to a number of theories aiming to explain why they are paid. Among the various theories, the effects of dividends on firm value are argued in every direction: positive, neutral, and negative. Some predict that dividend policy is irrelevant (Modigliani and Miller, 1961), others say that dividends increase firm value, and still others attest that dividends are a detriment to value. Some of the most popular theories regarding dividends fall into the following major camps: the bird-in-hand argument, the tax-preference argument, dividend clienteles, information content and agency cost hypothesis. The bird-in-hand argument was the prevailing theory on dividends in the early part of the 20th century (Al-Malkawi, Rafferty, and Pillai, 2010). The prospects of future capital gains was viewed as risky, and investors preferred the bird in the hand of cash dividends rather than the two in the bush of future capital gains, in reference to the common idiom. Dividends increase firm value by reducing uncertainty regarding future cash flows, which, in turn, lowers the firms the cost of capital. A number of papers by Gordon (Gordon and Shapiro, 1956; Gordon 1959, 1963) provide support for the bird-in-hand hypothesis. 6

14 One of the first papers to challenge the bird-in-hand hypothesis and form the foundation of current dividend research was Modigliani and Miller (1961). Given certain assumptions about perfect capital markets (i.e. no taxes, transaction costs), they construct a proof that if firms invest at their optimal level, dividend policy is irrelevant to firm value. M&M argue that investment decisions drive firm value, not how income is distributed. Various empirical tests of the irrelevancy hypothesis have produced mixed results. Also M&M s assumptions do not hold in the real world, and the authors themselves even make note that differences in taxation between dividends and capital gains could lead to effects on firm value. 2.2 Repurchases, the Substitution Hypothesis, and Dividend Disappearance An increasingly popular alternative form of payout to investors is share repurchases. Researchers come down on opposing sides of whether repurchases can serve the function of dividends and thus be used as a substitute for them. If the costs of maintaining dividends are a hindrance to firms, the existence of a more flexible substitute could provide great benefits and cause a substantial shift in payout policy. Grullon and Michaely (2002) provide evidence that, since regulatory changes in 1983, repurchases have slowly been substituting for dividend payments. They show that younger firms are much more likely to initiate a payout policy with a repurchase, and that growth in repurchases among all firms comes from funds that would have previously been devoted to dividends. They also observe, seemingly in conflict with their argument, that large established firms are cutting their dividends. The conclusion of their work is that repurchases are a viable substitute for dividends and an increasing number of firms will continue to gravitate toward the more flexible payout method. In contrast to the substitution hypothesis, Guay and Harford (2000) 7

15 and Jagannathan, Stephens, and Weisbach (2000) show that firms repurchase shares with temporary cash flows and pay dividends with permanent operating cash flows. Guay and Harford also find a more positive market reaction to dividend increases than repurchases. This positive reaction may be a motivating factor for a firm to maintain or even increase their dividends even when doing so is costly. Stemming from the substitution hypothesis is a debate as to whether or not dividends are destined to disappear as repurchases become a larger slice of the payout pie. Fama and French (2001) provide support for the disappearance by documenting that the percentage of U.S. industrial firms paying dividends declined from 66.5% to 20.8% between 1978 and DeAngelo, DeAngelo, and Skinner (2004) take the opposite viewpoint, showing that from a dollar value standpoint, real dividend payouts from industrial firms have grown over the same period from $31.3 billion in 1978 to $38.4 billion in They argue that the results in Fama and French (2001) are driven by a change in the earnings concentration over the time period. Firms that used to pay small dividends did stop paying, but concentration of dividend payouts closely resembled the concentration of earnings (the top 25 dividend payers accounted for 54.9% of the dividends paid and 51.4% of earnings in 2000, respectively). A recent work on this topic is Skinner (2008). Building on the evidence provided in both Fama and French (2001) and DeAngelo, DeAngelo, and Skinner (2004), Skinner concludes that repurchases are becoming the dominant form of payout at the expense of dividends. A contribution of the paper is its categorization of firms with a payout policy based on whether they pay dividends and their frequency of repurchases. Firms that pay dividends and don t repurchase shares regularly are dwindling from both a firm count and percentage of total payouts. According to Skinner, however, a small group of firms (345) who pay annual dividends 8

16 and repurchase regularly account for well over half of aggregate payouts in recent years. This concentration is consistent with DeAngelo, DeAngelo, and Skinner (2004). 2.3 Related Literature An important concept to the framework of this paper is the costs of external financing that could determine a firm s willingness to raise funds in order to maintain its dividend. As described by Myers and Majluf (1984), firms would be likely to forgo increasing dividends to build financial slack to ensure optimal future investment. If firms pay dividends and maintain them through difficult times, they must bear the costs of external financing to do so. Therefore, firms with high costs of external financing should then be more likely to cut dividends if short-term cash flows cannot support them. These costs of external financing can be measured by firm leverage and, more directly, credit ratings for those firms where data is available. Related to this issue is Easterbrook (1984), which provides two agency cost explanations for why firms pay dividends. The first is that it is costly to monitor management and that paying dividends reduces the amount of cash at managers disposal. The second rationale is with reduced cash, managers must now go to the capital markets to finance new ventures. The markets can then monitor managers by only financing value-creating projects. In this situation, firms would simultaneously raise cash from external sources while also paying dividends. Because of stickiness, firms using this practice may transition from raising cash while paying dividends to raising cash specially to pay dividends. This concept of dividend financing is explored further in a later section of the paper. A more recent paper studying dividend stickiness is Aivazian, Booth, and Cleary (2006), which focuses on the type of debt a firm has and how these different debt types affect the firm s 9

17 dividend decision. Aivazian et al. find that firms with bank debt exhibit a strong correlation between dividends and current earnings, which essentially equates to non-sticky dividends. Conversely, they also find that firms with publicly traded debt, and thus a credit rating, are more likely to engage in dividend smoothing, as predicted by the Lintner model. The paper also finds empirically that the same fundamental factors that influence paying a dividend also influence access to public debt, particularly size and asset tangibility. Aivazian et al. indicate a clear link between a firm s outside financing options and its dividend behavior. Grullon, Michaely, and Swaminathan (2002) build on Benartzi, Michaely, and Thaler (1997) by arguing that dividend increases are a sign of firm maturity. They find that earnings do not increase following dividend increases, but, rather, increase leading up to the dividend change. They do find, however, that earnings are much less likely to decrease following a dividend increase, suggesting the real information in dividend changes is that previous increases in earnings are permanent. Grullon, Michaely, and Swaminathan use these results to suggest that dividend increases convey information about a firm s discount rate. They find empirical support that as firms age, become less risky and offer less return, they will increase dividend payments. This provides support for the forward-looking hypothesis to be discussed in the next sections. This suggests that when firms notice a change in their future prospects, such as decreased risk, there will be an impact on the firm s dividend policy. 10

18 3 Hypotheses 3.1 The Tradition Hypothesis The first hypothesis I consider for how firms make their dividend decision is the tradition hypothesis. The tradition hypothesis is based on the work of Brav et al. (2005) and Skinner (2008). Brav et al. find in a survey of top management (CEOs and CFOs) that many managers view dividends as a burden and felt significant pressure to maintain them. When questioned further, many said if they were to run a company from inception, they would never introduce a dividend, instead distributing cash through repurchases. Skinner (2008) builds on this result by observing that a vast majority of the dividends in the market are paid by a small number of firms (approx. 300) with long dividend histories. He asserts that these firms are continuing to pay dividends simply because they always have; essentially dividends are a tradition for these firms. This hypothesis is also consistent with a line of behavior research that suggests dividends have become a social norm (Frankfurter and Lane, 1992). The rationale is that dividends once served a purpose in mitigating information asymmetry problems, but over the course of time, however, dividend paying evolved into a custom that is difficult to question and hard to resist. Baskin (1988) reviews the historical development of corporations in the United States and the UK and observes that investor pressure turned dividend paying into a normative behavior that is difficult to evade. The tradition hypothesis says that firms are looking backward when deciding whether to maintain or cut dividends this period. This hypothesis predicts that the longer a firm s history of maintaining dividends, the greater the likelihood it will continue to maintain them. This hypothesis is consistent with Lintner s model, as last period s dividend is the primary factor influencing this period s. The tradition hypothesis also implies that dividend cuts would be a 11

19 method of last resort. Firms have little incentive to cut dividends until they absolutely must. This leads to a prediction that dividend policy choices (to cut or not) should have little effect on future performance. The difficulty in testing this hypothesis is that a tradition is a nebulous concept. Surely thirty years of paying a dividend would constitute a tradition, but what about ten or five? Is a firm with twenty years of maintaining dividends any less set in its policy than a firm with thirtyfive years of maintaining? Providing clarity in this area will be an important step in testing this hypothesis. 3.2 The Forward Looking Hypothesis If the tradition hypothesis is backward looking, then its alternative is forward looking. The forward-looking hypothesis predicts that future firm prospects are the primary influence on a firm s dividend policy. Grullon, Michaely, and Swaminathan (2002) provide some of the strongest support for this argument. They show a firm s understanding of its future discount rate is the driving factor behind dividend increases. The difficulty in testing this hypothesis is determining the best way to measure future firm prospects. Easterbrook (1984) suggests that the costs of and access to external financing would be a good way to capture such prospects. This hypothesis is also consistent with Lintner (1956). Dividend smoothing is a firm s attempt to separate earnings volatility from payout volatility. Under this model, a choice to cut dividends would indicate a change in the future outlook of the company. The forward-looking hypothesis suggests that maintaining dividends signals that while current circumstances for the firm may be poor, managers outlook on the future of the company has not changed. A 12

20 prediction of the forward-looking hypothesis is that, following poor performance, firms that maintain dividends should improve, returning to established levels. 4 Data To examine these arguments, I gathered annual data on all firms in the Compustat database over the period of 1977 to 2007, excluding financials and utilities (SIC codes and ). Also, to be included in the sample, firms must have information on the following variables: DIV (sum total of all firm dividends declared in the given year) and EARN (the total income before extraordinary items). The data ranging from 1977 through 1986 are then used to create historic dividend policy and other backward-looking variables for the early years of my sample. The final sample includes all dividend-paying firms during the years 1987 to 2007, resulting in a total of 25,773 firm-year observations. The year 1987 was chosen to begin the sample as it is the first year in which statement of cash flow variables are widely reported in Compustat. The most important variable for my analysis is whether a firm cuts, maintains, or increases its dividend from one year to the next. I define this variable, MAINTAIN, to be 1 if dividend payouts are equal to or greater than those in the previous year. I define MAINTAIN to be 0 whenever dividends are less than the previous years. Firms with MAINTAIN=0 are considered to have cut their dividends. Looking at the entire sample, dividend-paying firms maintained or increased their payouts in 85.3% of firm-years. Figure 1 examines this variable s trend on a yearly basis. For the first twelve years of the sample, this maintain rate was consistently between 80-90% before falling into the 70% range for the first time following the 13

21 burst of the dotcom bubble. In the most recent years of the sample, the rate rebounded to crack 90% for the first time. The rarity of cuts is consistent with dividend stickiness. Whether this is primarily due to tradition or signaling of future prospects is not clear. 5 Who s Their Dividends? A Summary Analysis 5.1 Market wide Data Table 1 provides summary statistics for firm-years grouped by the MAINTAIN variable, along with a t-test of the difference in means between the two categories and the Wilcoxon rank sum test. The goal of Table 1 is to gain a broad understanding of the type of firm that cuts their dividend. By looking at data across three sections (financials, historical dividend policy, and growth rating), this table provides insights to characteristics shared by dividend cutters and maintainers. Because the decision to cut or, more appropriately, not cut dividend payouts creates dividend stickiness, understanding that decision is vital to understanding stickiness. Due to extreme outliers, many of the variables were winsorized at the 1% level. The difference in means test shows noticeable differences between the two groups, with all but five t-stats significant at 1%. The results are even stronger for the rank sum test with only a ratio of cash total assets not significantly different at 10%. Table 1 is broken into four sections: financial data, dividend history data, growth data and credit rating data. The financial data section aims to give a snapshot of the fiscal state of firms when they decide to maintain or cut their dividends. Although firms who maintain their dividends are older and larger than the firms who cut, the cutting firms are by no means young and small, with a mean age of and mean assets totaling $2.3 billion. The maintaining firms, as one would expect, are more profitable by both accounting (return on assets) and cash flow (operating cash flow/total assets) metrics. However, not all the cutting firms appear financially stressed, with more than half of the 14

22 firms posting positive return on assets and operating cash flows. ting firms appear to have higher leverage (as defined by debt-to-asset ratio) than firms maintaining or increasing their dividend payouts. Looking at the change in leverage and lagged changed in leverage variable, this seems to be the result of increases in leverage over the previous two years 3. The information in this first panel of Table 1 can be interpreted in a number of ways. There does seem to be a clearly distinguished difference in the performance between firms maintaining dividends and firms cutting their dividends. This is consistent with the notion that stickiness is driven by a firm s desire to maintain dividends whenever possible, cutting them only in times of serious financial stress. However, when looking only at the data of the cutting firms, there is no evidence of widespread cases of such stress. Most of these firms have adequate cash on hand to pay their dividends and the median firm remains profitable. Panel B of Table 1 focuses on firms previous decisions regarding its dividend policy. For maintaining firms in the current period, dividends scaled by assets have a significantly higher mean and median. In the previous period, however, the cutting firms had higher values in both measures. This suggests that cutting firms had a more burdensome dividend in the previous period, which has been greatly reduced in the current period. The next variable is the dummy for whether or not a firm maintained its dividends in the previous period. This variable captures the fraction of total firms maintaining their dividend, and is displayed as a percentage. The results show a much greater percentage of firms maintaining their dividend in the current period also maintained their dividend last period. Focusing on the cutting side, nearly one third of firms cutting their dividend this period also cut in the period prior. The final two variables in the panel indicate how long the firm had maintained (consecutive years without cutting) and paid (consecutive years with any payout) dividends prior to the current period. The difference is much greater for the maintaining streak 3 Change in leverage is defined by the absolute change in the leverage ratio, e.g. a firm whose debt to asset ratio increased from 50% to 53% would be listed at 3%, not 3%/50% = 6%. 15

23 variable, although this effect is being driven by one third of the cutting group having a maintaining streak of zero. With the paying streak variable the difference is much smaller, with less than half a year separating the groups means. Perhaps the most interesting take-away from this second panel of Table 1 is in the dividend decision from the prior period. Over one third of all dividend cuts from were by firms who had cut their dividend in the previous year as well. This seems to be consistent with the tradition hypothesis, in that firms that have not recently cut dividends are much more reluctant to do so than those firms who cut in the last year. The results from the paying streak variable, however, seem to lean slightly against the tradition hypothesis, as the maintaining firms do not have an appreciably longer history of paying out their dividend than cutting firms. In Panel C of Table 1 an important difference surfaces between maintaining and cutting firms. ing firms have a much higher asset and sales growth rates than cutting firms. In fact, the median cutting firm sees its assets shrink by -1.4% and its sales by -0.1%. This difference persists when looking back to the prior period with mean asset growth for a maintaining firm three times that of a cutting firm, which again has a negative median as well. Only when looking back two periods does the difference in asset growth between the groups become insignificant. Sales growth is significantly higher in maintaining firms both one and two years prior, although firms appear much more similar in year T-2. Another growth variable with a clear gap between the groups is the operating cash flow growth. ing firms cash flow grew by 8.8% on average while the mean cutting firm fell -13.3%, and the median cutting firm fell even further at -16.9%. ting firms performed very poorly compared to the previous period, which is more indicative of the financial stress expected to cause dividend reductions. No statistical difference is found between the groups capital expenditure growth. The results of Panel C are important to the overall question of this paper. Understanding trends in performance is a key factor in assessing the chances of future success. Managers are highly 16

24 likely to incorporate this data into any expectations they have about the firm s future. If these trends suggest a decline for cutting firms, but not for maintaining firms, that would support the forwardlooking hypothesis. Panel C does give some indication that this is the case, but there are a few other possible take-aways. As noted earlier, the poor, and often negative, growth rates for cutting firms are consistent with the notion that only extreme financial stress motivates firms to cut their dividends. As the next section of this paper will examine, financial distress, however, can come in a variety of degrees. The fact that firms that cut their dividends are performing poorly is not surprising. Given the documented adverse market reaction to dividend cuts it would be counterintuitive for firms doing well to reduce payouts and risk stock price reduction. This supports the last resort prediction of the tradition hypothesis. Only firms doing extremely poorly even consider cutting their dividends. The analysis raises a number of interesting questions. When a firm faces potential stress, what factors determine if it is the time to cut their dividend? Do they consider only their tradition of paying dividends, or do they consider not only their ability to payout in this period but future periods as well? The next section of this paper will examine firm behavior under very difficult financial conditions. From the analysis of Table 1, there are indications that both the tradition and forwardlooking hypotheses are viable. The next step of the analysis will focus on those firms facing difficult financial conditions as these firm are the most likely to be considering dividend cuts, regardless of whether they execute those cuts. 5.2 Dividend Decisions Under Financial Stress and Cash Shortfalls Defining Financial Distress The incentive for firms to maintain dividends is strong. Miller and Rock (1985) document negative stock price reaction to dividend cuts and nearly 90% of managers surveyed in Brav et al. (2005) 17

25 believe there will be negative consequences if they cut dividends. Firms performing well are therefore very unlikely to even consider cutting dividends. Table 1 captures the entire market, making analysis of a decision to cut dividends difficult; hence a large number of successful firms aren t really facing a decision at all. Alternately, firms facing financial distress are forced into making these decisions. By studying these firms in particular, the factors that influence dividend cuts should become clearer. Isolating and studying these firms will therefore become the primary focus of my analysis. Financial distress can come in many forms. The variables chosen to measure such stress are income before extraordinary items, a measure of accounting performance, operating cash flow, a metric to capture actual cash generated, and the firm s actual cash level. For the purpose of dividends, cash shortfalls would be expected to be of more importance since dividends are cash payouts to shareholders. However, because accounting return is a common measure of firm performance and financial health, it was included as well. Changes in both accounting profit and cash flows were also considered as proxies but were not chosen due to difficulties in defining what level of reduction would constitute stress and the possibility of small increases occurring in poor performing firms (i.e. a reduction in losses from $1.1M to $1M). To measure various types of financial stress, a number of constraints using these three proxies are created and then grouped sample statistics (in the style of Table 1) are generated for these conditions. A main motivation for this sub-sample selection is to understand the sustainability of dividend policies. Firms doing well can maintain dividend policies for decades even if they are relics that don t provide firm value. Only in financial stress are firms forced to reconsider their dividend policy. Firms performing poorly will have difficulty paying their dividend, and must evaluate whether paying dividends at the current level is sustainable into the future. These tables represent sub-samples of Table 1 and aim to understand the characteristics of maintaining and cutting firms just as Table 1 did, only now under a condition of financial stress. The 18

26 first and most straightforward is whether a firm had a year with negative accounting profit. These results are reported in Table 2. Under this condition approximately 54% of firms maintain their dividends and, as expected, firms that maintain are generating more cash than cutting firms. The results of difference in mean and rank sum tests for Table 2 are very similar to those in Table 1, suggesting that using negative income as the only criteria does not produce a unique sub-sample of firms. Next, Table 3 uses negative cash flow for its stress condition and finds, somewhat surprisingly, that nearly 70% of firms in this sample maintained their dividend. In this case, the opposite of the previous table occurs; the maintaining firms in this sample have much higher return on assets than the cutters, even though the cash flow generated by each group is similar. The results for the difference in means tests change considerably from Tables 1 and 2. Most notably, there is no longer a significant difference between the size and age of the maintaining and cutting groups. The rank sum test does continue to find significantly higher assets in maintaining firms, but age is no longer significant in that test either. Because generating positive, but relatively small, cash flows may also lead to financial stress, Table 4 provides statistics under the constraint that operating cash flows are less than the previous period s cash dividends. The logic behind this condition is that if cash flow is less than a firm s established dividend level, the firm must choose to either cut dividends or payout more cash to shareholders than it generated. In this sample, slightly more than 60% of firms maintained their dividends. The results of Table 4 are consistent with those in Tables 2 and 3. Table 5 accounts for not only the firm s performance, but also the level of cash on hand. Firms in this sample had a sum total of operating cash flow and cash on hand last period less than the previous year s cash dividends. In this sample 55.3% of firms maintain their dividend. In Tables 2-5, the variable that has the most statistical difference between the two groups is the leverage variable, where maintaining firms have lower leverage than cutting firms in all three cases. This difference is 19

27 significant at the 1% in both the t-test and rank sum test. In Tables 4 and 5 maintaining firms actually have less cash on hand than cutting firms. This suggests that cutting firms are recognizing the need to keep cash on hand and thus cut dividends. This would contradict the last resort prediction of the tradition hypothesis, as these firms have cash reserves but are not using them to maintain dividends. Because each of the tables uses only one variable to measure stress, it is difficult to ensure that all the firms in the sample are experiencing stress, or if they simply operate in that environment (i.e., profitable with low cash flows). Table 6 uses both accounting profit and cash flow to create a condition likely to capture only those firms in the most dire straits. To be included in Table 6, sample firms must have both a negative income before extraordinary items and a negative cash flow from operations. This creates a sample of 777 firm-years, with 355 (45.7%) maintaining their dividend Why Do Firms Dividends in Extreme Stress? Table 6 provides a look at firms that are clearly in a grave financial situation, with both negative accounting profits and cash flows in the current period. One would expect a very large portion of these firms to cut dividends, but nearly half (45.7%) choose to maintain their payouts. The interesting question in this case shifts from why do firms cut dividends (which is obvious in this case) to why do firms maintain dividends when faced with such grim financial performance. It is this decision to maintain, when the situation would seem to clearly dictate otherwise, that is at the heart of dividend stickiness. Understanding the motivations of these firms should in turn yield a much greater understanding of stickiness and what it indicates. If these maintaining firms are clinging to a long history of dividends, that would give clear support to the tradition hypothesis. This would predict longer maintaining and paying streaks (entering the period) for maintaining firms. Alternatively, if there are variables indicating potential improvement, that would support the future 20

28 performance hypothesis. In line with this prediction would be significantly higher performance and growth for maintaining firms. Now that all firms in the sub-sample are facing similar financial conditions, many of the variables are no longer statistically different between the two groups. The difference in means test for age, size, cash flow/assets, and return on assets are no longer significant at even the 10% level. Of these variables, only the difference in return on assets remains significant (at 1%) in the rank sum test. Although maintaining firms hold a higher percentage of their assets in cash, cutting firms do not appear to be cash poor, with a mean cash-to-assets ratio of 13.2% and median of 4.7%. The main difference in the financials again appears to be in leverage. The firms who maintained their dividend have both a mean and median leverage over 10% less than cutting firms, a difference significant at the 1% in both the t-test and rank sum test. The difference in leverage seen in Panel A of Table 6 potentially indicates a forward-looking dividend policy. Obviously firms in this sample, if they choose to do so, must pay dividends out of some other source besides current cash income. Firms with more debt capacity are willing to payout cash from reserves knowing that if cash shortfalls occur in the future, they can still take on more debt. Firms with little to no excess debt capacity have very little choice but to hold on to as much cash as possible to sustain their business, thus forcing managers to make the difficult decision to cut dividends. If leverage is a factor in determining whether firms maintain dividends, then managers are concerned with their firm s future access to capital which is consistent with a forward-looking approach. Panel B of Table 6 shows the difference in dividend history variables between the two sets of firms. Going against the tradition hypothesis, cutting firms actually came into the period with a longer streak of paying dividends than maintaining firms, and the median maintaining streak for both groups is 1. The main difference between the two groups, as was the case in the overall sample, is maintaining firms had maintained in the prior year much more often than cutting firms. Total 21

29 dividends scaled by assets are higher for maintaining firms in the current period but were significantly higher for cutting firms in the previous period. This suggests that firms with a higher dividend relative to their size are more likely to cut their dividends. There is also no statistical difference found in either test between the total instances of previous dividend payouts for the two groups. Interpreting the dividend history data in Table 6 is not easily done. Neither set of the firms has a particularly long streak of maintaining its dividend, and there is not much distinguishing the other variables, save the difference in whether the firms maintained the prior year. One possible case for the tradition hypothesis last resort prediction could be made based on the higher scaled dividends for cutting firms in the previous year. Firms with lower dividends relative to their size are likely to have an easier time maintaining them and would not need to turn to a last resort. Panel C of Table 6 examines the growth data across the two groups of distressed firms. This table includes data on how a firm s leverage has changed over the prior two periods. The change in leverage variable is a raw change in percentage leverage, i.e. a firm with a change in leverage equal to 10% would have went from 50% leverage in the prior period to 60% leverage in the current period. This number is statistically higher in the current period for cutting firms, although the median change of 5.7% is very close to the maintaining firms median of 5.3%. In the prior period, however, the change is much greater for cutting firms, with maintaining firms actually reducing leverage on average. Also in this panel is three periods of asset and sales growth for the firms. As was the case in Table 1, asset growth is noticeably higher in the maintaining firms than the cutting firms in both the current and previous period. There is no statistical difference between the groups when looking back two years. Sales growth has significantly higher medians for maintain firms in both the current and previous year, but the means are not statistically different. Operating cash flow growth is highly negative for both groups with very large standard deviations resulting in no significant difference in 22

30 means. The rank sum test does find lower operating cash flow growth for maintaining firms to be significant at the 10% level. Panel C highlights some possible indicators for determinants of the dividend decision. The cutting group shows a trend of consecutive periods of both low asset growth as well as rising leverage. Conversely, firms maintaining their dividend while having low asset growth and increased leverage in the current period have stronger asset growth and stable leverage in the prior period. If a firm has one bad year they more than likely will continue to payout dividends at their current levels. When it s two bad years, however, the environment seems to change. What remains unclear is whether these firms are cutting their dividend because they are lacking in cash and have no other choice, or if they are updating their expectations about their future prospects and thus reduce or stop dividend payouts accordingly. The role dividend history and tradition plays are also unclear from these difference in means tests. Trying to isolate these effects will be the main goal of the duration of this study. One potential drawback of using panel data is many of the firms in the cutting group of Table 6 could simply be older versions of the maintaining firms in the other group. This is also consistent with the findings in Table 6 that firms maintain dividends through the first signs of trouble, but cut when that trouble persists over time. Table 7 is a small switch table designed to capture the overlap between the two groups. The table shows 147 firms are present in the sample over two consecutive years. Twenty-two firms were able to maintain their dividend through two years of negative cash flow and income while 54 firms who maintained once had to cut in the second period. Of 71 firms who cut their dividend in the first period, only four maintained their new level in the second, with 67 firms cutting again. Although there is clearly some overlap present, only 54 of the 422 firm-years in the cutting group are T+1 firm year from the maintain group. This suggests the result is not simply the effect of the panel data set. It is also a possible indication that dividend cuts are often rolling. 23