Do dividends matter more in declining markets?

Journal of Corporate Finance 17 (2011) 457–473

Contents lists available at ScienceDirect

Journal of Corporate Finance j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j c o r p f i n

Do dividends matter more in declining markets?☆ Kathleen P. Fuller a,1, Michael A. Goldstein b,⁎ a b

University of Mississippi, School of Business, Department of Finance, 244 Holman Hall, University, MS 38667, United States Babson College, Finance Department, 231 Forest Street, Babson Park, MA 02457-0310, United States

a r t i c l e

i n f o

Article history: Received 18 April 2007 Received in revised form 21 December 2010 Accepted 1 January 2011 Available online 12 January 2011 JEL classification: G35 Keywords: Dividend policy Asymmetry Market movements

a b s t r a c t We find dividends do matter to shareholders, but more in declining markets than advancing ones. Dividend-paying stocks outperform non-dividend-paying stocks by 1 to 2% more per month in declining markets than in advancing markets. These results are economically and statistically significant and robust to many risk adjustments and across industries. In addition, we find an asymmetric response to dividend changes based on market conditions: dividend increases matter more in declining markets than advancing ones. Tests indicate that results are not due to more profitable firms and appear not to be caused either by free cash flow or signaling explanations. We also find that it is the existence of dividends, and not the dividend yield, that drives returns' asymmetric behavior relative to market movements. © 2011 Elsevier B.V. All rights reserved.

1. Introduction Ever since the five major investigative questions (who, what, when, where, and why) were supplemented by a sixth (whether) fifty years ago, corporate dividends have been the focus of significant debate among academics and practitioners alike. Of these six questions, not all have been given equal attention. While the questions of what and where firms pay dividends have had some investigation, the questions of whether or not firms should pay dividends, who pays dividends (and to whom), and why firms pay dividends have received the most attention. However, the question of when dividends matter most has not been as studied. This paper aims to fill that gap by examining when dividends matter more, based on market conditions. Specifically, we examine whether dividend-paying firms outperform non-dividend-paying firms by more when the market declines than when the market is increasing. Using S&P 500 returns from January 1970 to December 2007 as a proxy for market conditions, we examine the returns of dividend-paying and non-dividend-paying firms in both advancing and declining markets. We find that dividend-paying firms outperform non-dividend-paying firms by more when the market declines than when the market is

☆ We thank Deepak Agarwal, Jeff Bacidore, Jennifer Bethel, Wayne Ferson, Kathleen Hevert, Paul Irvine, Jon Karpoff, Gautam Kaul, Laurie Krigman, Marc Lipson, James Mahoney, Donna Paul, Tyler Shumway, John Scruggs, Chris Stivers, and Avandihar Subrahmanyam for their comments and suggestions. We also thank seminar participants at the 2001 All Georgia Conference, the 2001 and 2004 Financial Management Association Meetings, the 2002 Eastern Finance Association Meeting, the University of Michigan Lunch Seminar series, the 2004 University of Colorado Burridge Center Conference, the 2004 Financial Management Association European Meeting (Zurich), and the 2005 Western Finance Association Meeting for their comments. Kathleen Fuller acknowledges financial support from the Terry-Sanford Research Grant. Michael Goldstein acknowledges support from the Babson College Board of Research. An earlier version of this paper was titled “Dividend Policy and Market Movements.” ⁎ Corresponding author. Tel.: +1 781 2394402; fax: +1 781 2395004. E-mail addresses: [email protected] (K.P. Fuller), [email protected] (M.A. Goldstein). 1 Tel.: + 1 662 915 5463. 0929-1199/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jcorpfin.2011.01.001

458

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

increasing. Our findings are robust to a number of controls, time periods, and risk factors, and hold both for “low” and “high” quality firms.2 We find that dividend-paying firms outperform non-dividend-paying firms by more in declining markets than they do in advancing markets, implying that dividends do matter more in declining markets. Our results show an outperformance of 1% to 2% each month and this difference is statistically and economically significant.3 Our results are robust to a variety of adjustments for risk as well as controls for size, book to market, time period, and the magnitude of the market advance or decline. In addition, we also show that even during the months when the dividend-paying firms do not pay a dividend, dividend-paying stocks outperform non-dividend-paying stocks by more in declining markets than in advancing markets and that the results are not driven by magnitude of the dividend yield. Thus, our results are not driven by the receipt of the cash dividend itself. We do not include a firm in the dividend-paying set until the month after the first dividend is paid and we continue to include a firm in the dividend-paying set until the month after they did not pay a regularly scheduled dividend. Therefore, our findings that dividend-paying firms outperform non-dividend-paying firms by more in declining markets are not due to the inclusion of dividend initiations in the dividend-paying set or dividend-omissions in the non-dividend-paying set. Our results are in line with anecdotal evidence which suggests shareholders' preferences for dividend-paying stocks over non-dividend-paying stocks vary over time, depending on the state of the market, i.e., advancing and declining markets.4 We also control for the quality of the firm to make sure that our asymmetric results are not driven by the suggestion of Allen et al. (2000) that better firms pay dividends.5 Beyond our controls for size and book-to-market (both of which can control for perceived quality), we segment firms based on their profitability (previous year's net income or return on assets) while still controlling for size and book-to-market, and continue to find that dividend-paying firms outperform non-dividend-paying firms by more in down markets for both high and low quality groups. We also run our Fama–MacBeth regressions and sequentially add return on assets (ROA), return on equity (ROE), or Tobin's Q directly to the regression in order to control for possible quality differences further; our results continue to hold, and, in fact, neither ROA nor ROE are significant. Our results also hold across tests that segment firms based on their future earnings (EBITDA or EPS), cash flow, or Tobin's Q, which also implicitly control for the quality of the firm. Overall, our results collectively do not indicate that the variation in preference for dividend-paying firms is driven by large quality differences. Our finding that dividends matter differentially based on market conditions also has implications for the question of whether dividends matter. DeAngelo and DeAngelo (2006) recently argued that the Miller and Modigliani (1961) irrelevance theory was itself incomplete, as it ignored the ability of managers to immediately destroy wealth through the dissipation of cash flow which is not then immediately or subsequently returned to capital providers.6 A key argument in DeAngelo and DeAngelo (2006) is that for dividends to be irrelevant, they must be irrelevant under all conditions. In this way, our results support and augment the findings of DeAngelo and DeAngelo (2006); if payout policy were truly irrelevant as suggested by Miller and Modigliani (1961), then the variation between dividend- and non-dividend-paying stocks should not vary across time and market conditions.7 In addition, the awareness of and cost to investors of the value destruction suggested in DeAngelo and DeAngelo (2006) could vary, particularly with perceived changes in current or future economic conditions. The effect of managers burning cash generated by optimal investments will be greater during perceived economic downturns as large compensation packages, lavish parties, and excess expenses will account for a higher percentage of revenue and gross margin and operating expenses.8 Jagannathan and Wang (2007) note that when future projections of the economy are poor or uncertain, shareholders are more likely to review their investments. Since the market is the aggregator of future expectations, negative or uncertain economic projections tend to 2 While Black (1976) argued that due to taxes non-dividend-paying firms should be worth more than firms which pay dividends, Miller and Modigliani (1961), Elton and Gruber (1970), Brennan (1970), Black and Scholes (1974), and Litzenberger and Ramaswamy (1979, 1980) suggest there may be different clienteles across investors, so that, depending on their tax bracket, some shareholders may prefer high-dividend-paying stocks while others may prefer nondividend-paying stocks. Similar to Litzenberger and Ramaswamy (1980, 1982), Blume (1980), Elton, Gruber, and Rentzler (1983), and Christie (1990) , we also find that dividend-paying firms outperform non-dividend-paying firms, but also add when this outperformance is largest (i.e., during market declines). Christie (1990) also notes that the outperformance cannot solely be due to the tax clientele effects; our findings that this outperformance varies with the direction of the market helps support this conjecture. 3 We note that by creation we are not testing a tradable strategy; instead, we are examining contemporaneous changes to examine differences in shareholder responses to different conditions. The lack of tradability may be one reason why this difference is persistent across time and stock characteristics. 4 For example, in 2000, Fidelity began advertising a mutual fund that consisted of only dividend-paying stocks. In the ad, an advisor informs his client that this fund can help diversity his portfolio and moderate losses in declining markets. Also in 2000, Standard and Poor's predicted a revived interest in dividends, stating “market weakness may boost interest in dividends as investors begin to see the value of a ‘bird in the hand.” Interestingly, this sentiment almost exactly mimics that mentioned in the third sentence of Black and Scholes (1974). 5 Allen et al. (2000) suggest that due to differential taxation better firms issue dividends to attract institutions “which have a relative advantage in detecting high quality firms and ensuring firms are well managed” (p. 2499), supporting DeAngelo and DeAngelo's (2006) suggestion that firms less likely to destroy value are more likely to pay dividends. 6 DeAngelo and DeAngelo (2006) note that managers can simply destroy value, as suggested by Smith (1776) and Jensen and Meckling (1976). Therefore, the present value of the cash flows paid out by a firm may be less than the present value of cash flows spun off by an optimal investment policy. To the extent that dividends reduce the amount of money wasted by managers, dividends will not be irrelevant. 7 Our results may also be less model specific than previous tests, such as those by Litzenberger and Ramaswamy (1980, 1982) or Christie (1990). By comparing the returns in advancing verses declining markets, our results indicate that the difference in valuation between dividend and non-dividend paying stocks is likely not due to improper model specification as, unlike previous papers, we find the outperformance varies with the direction of the market regardless of which model is used. 8 Such large cash burning expenses may either push a company towards bankruptcy or even continue during significant negative performance, as recently demonstrated by the bonuses and corporate events paid by financial firms in both the US and the UK during the current economic crisis even after requiring significant bailout money from the government.

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

459

produce declining markets. Investors may therefore become more aware of the suboptimal payout policies noted in DeAngelo and DeAngelo (2006) as their economic projections worsen and the market declines. Therefore, it is possible that investors' preferences for dividend-paying firms may vary over time, depending on the market. There are a variety of reasons why investors may prefer dividends more in declining markets than in advancing ones.9 For example, the free cash flow hypothesis developed by Easterbrook (1984) and Jensen (1986) suggests that firms may take actions, such as paying dividends, to bind managers and thereby reduce the cash flow available to managers to waste. The probability that the bonding will be binding increases as the state of the economy decreases and the market declines. Alternatively, signaling theory implicit in Lintner (1956) and developed by Bhattacharya (1979), John and Williams (1985), and Miller and Rock (1985) argues that dividends may signal managers' private information on future earnings.10 Since the probability and costs of financial distress generally rise in declining economies, even maintaining dividend payments is more likely to be difficult. Thus, the information conveyed by each payment, and not just dividend changes, is more valuable in down markets than in up markets.11 To investigate the effects of market movements on the signaling hypothesis, we examine the effects of dividend increases, decreases, and maintenance of a dividend (i.e., no change) conditional on the direction of the market. While dividend changes have been examined by many others, such as Pettit (1972), Aharony and Swary (1980), and Grullon et al. (2002), dividend changes have not previously been examined based on the direction of market movements. We find that the market reacts more positively to dividend increases in declining markets than in advancing markets, and that the market reacts more positively to the maintenance of dividend payments in declining markets than in advancing markets. We further test for signaling differences by segmenting our dividend changes into those with high/low future earnings. Our results continue to hold for all firms, with the exception of low future earnings firms that decrease their dividends. We then further test our results by re-running our Fama– MacBeth regressions but segment the sample into low and high future earnings, either defined by future EBITDA or future EPS; our results continue to hold. We also provide a variety of tests to check whether our finding that dividend-paying firms outperform non-dividend-paying firms by more in declining markets than advancing markets is driven by free cash flow differences. First, we segment firms into high and low cash flow firms based on their cash flow as defined in Lehn and Poulsen (1989) and still find that dividend-paying firms outperform non-dividend-paying firms by more in declining markets for both high and low cash flow firms. As in Lang and Litzenberger (1989), Lang et al. (1991), Denis et al. (1994), Yoon and Starks (1995), and others, we also segment our firms into high (Q N 1) and low (Q b 1) Tobin's Q and our results still hold (although the difference is not significant for low Tobin Q firms). We also segment our dividend changes analysis by cash flow and continue to find our results hold except for dividend decreases, where high cash flow firms that decrease dividends do better in down markets (but not significantly), while low cash flow firms do much more poorly if they cut their dividends in declining markets than advancing ones. The paper is organized as follows. In Section 2, we expand on how dividend payments relate to market movements and discuss some unique characteristics of dividend payments. We also provide testable empirical predictions. Section 3 presents our data and method. Section 4 describes our major empirical results. In Section 5 we examine our results as they relate to both the signaling and free cash flow theories. Section 6 provides robustness checks, and Section 7 concludes. 2. Market movements, dividends, and empirical predictions While firms have a variety of choices in what they pay their investors, we concentrate on dividend payments because dividends have a variety of special features that lend themselves nicely to studying investors' preferences in declining markets.12 While Grullon and Michaely (2002) suggest examining the total payout of a firm, dividends have a variety of unique features not shared by repurchases.13 First, Brennan and Thakor (1990) suggest that repurchases are more beneficial for informed investors while dividend payments benefit all investors equally. Second, Cook et al. (2004) note that the timing of share repurchases is uncertain, including if and when they are completed. Ng (2007) notes that repurchase programs provide only weak signals since companies may cease their repurchases due to market or business conditions. Howe et al. (1992) also indicate that repurchases are discretionary and are not always observable by investors. 9 Other papers have focused on the questions of who pays dividends. Fama and French (2001), DeAngelo et al. (2004), and Julio and Ikenberry (2004) examine the types of firms and how much dividends they pay. DeAngelo and DeAngelo (2006), Denis and Osobov (2008) and Blau and Fuller (2008) suggest that a combination of the life cycle of the firm and the needs for financial flexibility help determine who pays dividends. Baker and Wurgler (2004a,b) suggest that the firms that issue or initiate dividends do so when requested by shareholders. 10 As noted in Allen et al. (2003), despite considerable academic research, empirical tests of the signaling versus free cash flow hypothesis fail to pick an overwhelming winner. One possible reason may be found in Blau and Fuller (2009), which suggests that both the free cash flow and signaling theories may be correct, and that the use of dividends varies over the life and success of the firm. 11 Prospect theory, developed by Kahneman and Tversky (1979), the “disposition effect” of Shefrin and Statman (1985), and the “mental accounting” of Thaler (1980) may also suggest a stronger preference for the relatively certain dividend payment from dividend-paying firms when shareholders predict future uncertainty or economic downturns, particularly if they are downside risk-averse. Shefrin and Statman (1984) note reasons for investor preferences of cash dividends overall. Such responses are similar to the “flight to quality” tendency seen during market declines documented by Connolly et al. (2005). 12 Although there are several reasons why investors might prefer dividend-paying stocks in declining markets, neither Miller and Modigliani (1961) nor traditional asset pricing models such as the Sharpe (1964) – Lintner (1965) CAPM or the Fama and French (1992, 1993) models account for state-specific investor preferences. However, previous research suggests that investors may have asymmetric preferences; for example, papers as early as Markowitz (1952, 1959) noted that investors care about downside risk differentially. In addition, an increasing body of work examines asymmetric responses of returns to market movements. See for example, DeBondt and Thaler (1987), Goldstein and Nelling (1999), Ang and Chen (2002), Hong et al. (2003), and Ang et al. (2006). 13 While investors in non-dividend-paying firms could mimic this cash flow by selling stock, they would be doing so at depressed prices (given the market downturn) and incur potentially non-trivial transaction costs (a guaranteed loss).

460

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

In contrast, the timing and completion of dividend payments is transparent: dividends are announced and then paid on a certain date. It is immediately clear to shareholders whether or not this obligation has been met in part or in full. Unlike repurchase programs, dividend payments are regularly scheduled: investors know in advance when to expect the next dividend payment. Thus, the value of dividend-paying stocks should be highest when there is either increased uncertainty or the future outlook has become less rosy. In any case, to the extent that repurchases are beneficial in declining markets, not considering repurchases will only bias us against finding significant results. For example, if repurchases are beneficial, they would mitigate our results on nondividend-paying stocks, since, by definition, non-dividend-paying stocks can only increase payout through share repurchases. If repurchases are highly correlated with dividend payments, our results will still hold. To investigate when, and therefore indirectly whether, investors prefer dividend-paying stocks, we begin by examining three empirical predictions. First, we test whether dividend-paying stocks are more likely to outperform non-dividend-paying stocks in declining markets than in advancing markets. According to Miller and Modigliani (1961) and traditional asset pricing models, dividends are irrelevant and, thus, we should find no differences between dividend-paying and non-dividend-paying firms. Yet, DeAngelo and DeAngelo (2006) note, if dividends are irrelevant, they should be irrelevant at all times; irrelevance is rejected by finding even one case where dividends matter. In addition, if the value difference is solely due to taxes, as suggested by Black and Scholes (1974), Black (1976), and others, then, since tax policy is relatively constant, this value difference should not vary month to month with the movement of the S&P 500. Even more to the point, we should not expect to see larger differences between dividend and non-dividend-paying stocks the more the market declines. Second, a firm's ability to maintain a dividend should cause a favorable response during a declining market but not during an advancing market. The increase of a dividend should also matter more in declining markets than advancing markets under either signaling theory (a stronger signal) or free cash flow (more binding). Again, if dividends are irrelevant, we should find no difference in responses to dividend changes. Third, investors should prefer dividend-paying stocks to non-dividend-paying stocks even during those months between dividend payments during which no dividend is paid. Alternative explanations such as tax explanations require cash payments.14 If taxes were a primary factor in our differential results, we should not see a result for months during which no dividend was paid. Therefore, if these explanations are correct, our results would not hold in non-dividend-paying months for dividend-paying stocks. We also perform a variety of other tests that are consistent with or driven by different theories of why firms pay dividends or which ones. For example, Allen et al. (2003) suggest that “better” firms pay dividends. If our results are due to better firms, our results should vary with ROA, ROE, or future earnings. Alternatively, if our results are not due to better firms, we should continue to see a difference after controlling for quality differences. We also investigate whether the results are driven by other factors suggested in the literature. Finally, we investigate whether the results across advancing and declining markets across dividend and non-dividend-paying stocks are caused either by signaling on the part of firms or due to free cash flow considerations. If the signaling theory of dividends drives our results, we should see results vary with future earnings. If our results are driven by free cash flow, we should see our results vary with estimates of cash flow or Tobin's Q. We should also see more of a differential factor for firms of higher dividend yield. Alternatively, if something else is at work that is causing this difference in advancing and declining markets, then our results will continue to hold under both tests.

3. Data and method We use the Center for Research in Security Prices (CRSP) monthly master file to identify a sample of dividend-paying and nondividend-paying firms. We examine all NYSE, AMEX, and Nasdaq listed stocks with a price greater than $2 over the 38-year period from January 1970 to December 2007, a long period with many market up and down months. For each firm, we collect its monthly return, market capitalization, and share volume data from CRSP. We also collect data on the firm's book value of equity and SIC code from Compustat. As many of our tests use corporate data, our analyses only include firms with data on both CRSP and Compustat. We identify dividend-paying stocks by comparing the CRSP total return to the CRSP return that does not include dividends. If the returns are different, we consider that the firm has paid a dividend in that month, and consider the difference to be the dividend.15 Only quarterly-dividend-paying stocks are classified as divided-paying stocks; firms that pay dividends other than quarterly dividends are treated as non-dividend-paying firms.16 In this way, to the extent that non-quarterly dividend payers have any positive differential effect, we bias our results against finding any results for quarterly-dividend-paying firms.17 14 Previous arguments that dividends are tax-disadvantaged to selling stock may no longer hold since a great deal of stock is held in tax-deferred accounts and the 2003 Tax Law eliminated the tax benefits of capital gains over dividend payments. 15 Although this method might result in calculating negative dividends, we retain these likely errors to avoid introducing any bias by correcting errors on only one side. 16 We use the distribution code in CRSP to determine if the dividend is a special, or an annual, semi-annual, quarterly, or monthly. Choosing quarterly-dividendpaying stocks increases the frequency of the possibility of signal observations and decreases the length of time between potential signals. 17 To ensure that these rules did not affect our results, we reran our tests using alternate definitions of dividend-paying stocks. For example, we also included all regularly scheduled dividend payments (monthly, quarterly, and yearly) when classifying firms as dividend-paying. As another alternative classification method, we dropped all non-quarterly-dividend payments from the sample so non-dividend-paying firms never paid any type of cash dividend. Results are qualitatively similar.

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

461

Table 1 Summary statistics. Summary statistics for 25,238 NYSE, Amex, and Nasdaq listed firms for the 456 calendar months from January 1970 to December 2007. There are 2,777,716 firm months of which 1,419,664 are non-dividend-paying firm months and 1,358,052 are dividend-paying firm months. Advancing markets are the 268 months in the sample when the S&P 500 has a positive return; declining markets are the 188 months in the sample when the S&P 500 does not have a positive return. All data are from CRSP. Monthly Volume is the average monthly trading volume, Price is the average end-of-the-month price per share, Market Cap is the average end-of-the-month market capitalization, Dividend per share is the average quarterly dividend per share, Beta is the average end-of-the-month CRSP estimate of beta, and Number of Obs is the total number of firm-months. The number of volume observations is less than other variables since for some months no volume was reported on CRSP for some firms; in such cases, we dropped those observations when computing the average monthly volume, but retained that firmmonth for all other calculations. Non-dividend-paying

Dividend-paying

Panel A: all markets (456 months) Monthly volume Price Market cap. Dividend per share Beta Number of Obs.

33,540 $10.13 $318,016,840 None 0.862 1,419,664

35,545 $23.97 $1,527,729,870 $0.071 0.703 1,358,052

Panel B: advancing markets (268 months) Monthly volume Price Market cap. Dividend per share Beta Number of Obs.

35,335 $10.56 $330,096,950 None 0.882 836,364

37,383 $24.78 $1,614,258,510 $0.072 0.702 800,113

Panel C: declining markets (188 months) Monthly volume Price Market cap. Dividend per share Beta Number of Obs.

30,981 $9.52 $300,796,270 None 0.858 566,300

32,925 $22.82 $1,404,380,530 $0.069 0.704 557,939

For each month, we classify firms as either dividend-paying or non-dividend-paying. While Litzenberger and Ramaswamy (1979, 1980, 1982) define a dividend-paying stock-month as only the month in which the firm pays a dividend, we follow Black and Scholes (1974) and Kalay and Michaely (2000) by defining a stock as a dividend-paying stock if that a firm has paid dividends in the past and is expected to continue paying on a regular basis. We therefore classify a known regular quarterly-dividend-payer, such as General Electric or IBM, as a dividend-paying stock for all 12 months, not just for the four months of the year a dividend is paid. We also take care not to include dividend initiations in the dividend-paying set of firms. Specifically, if a firm does not pay a dividend and then begins to do so, we classify it as a non-dividend-paying firm until the month after the dividend is paid. That is, if a firm lists on January 1989 but does not pay a quarterly dividend until June 1992, we classify the firm as a non-dividend-paying firm from January 1989 through June 1992 and reclassify it as a dividend-paying firm as of July 1992. The months up to and including the first dividend payment are attributed to the non-dividend-paying set. In this way, we can attribute any positive return in the stock price that is due to the initiation of a dividend to the non-dividend-paying stock group, thus biasing our results against finding outperformance by dividend-paying stocks. At the same time, we are careful to keep a firm in the dividend-paying set until after the firm stops paying a dividend, so that any cessation of dividend payments affect the dividend-paying set and not the non-dividend-paying set. Specifically, we continue the classification as a dividend-paying stock until the firm stops paying a dividend, the firm is delisted, or the sample period ends. Similarly, if a firm pays a dividend and then stops paying a dividend, we classify it as a dividend-paying firm until the month after the scheduled quarterly-dividend payment. That is, if a firm lists in January 1989 and begins paying a quarterly-dividend as of June 1992 but does not pay the September 1994 dividend, then we classify the firm as a non-dividend-paying firm through June 1992, a dividend-paying firm from July 1992 to September 1994 (the month of the expected quarterly-dividend), and a non-dividendpaying firm from October 1994 until it is either delisted, pays a dividend, or the sample period ends. Thus, we can attribute any negative surprise due to the nonpayment of a dividend to the dividend-paying group, further biasing our results against finding outperformance by dividend-paying stocks.18

18 It is possible that our results could be driven by return drift as Michaely et al. (1995) find a positive price drift for the year after dividend initiations and negative price drift for the year after dividend omissions. To verify that our results are not primarily driven by return drift, we test whether firms that have always paid a dividend (i.e., firms that were paying a dividend as of January 1, 1969 and continue to until they are delisted, merged, etc.) outperform firms that have never paid a dividend (i.e., firms that were not paying a dividend as of January 1, 1969 or never pay a dividend from when they were listed until delisted, merged, etc). Our results, that dividend-paying firms outperform non-dividend-paying firms in down markets, still hold, indicating that return drift is not driving our findings.

462

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

Table 2 Average return for both advancing and declining markets. The table reports the average monthly return to dividend- and non-dividend-paying stocks for the 456 calendar months from January 1970 to December 2007. Advancing markets are the 268 months in our sample when the S&P 500 has a positive return; declining markets are the 188 months in our sample when the S&P 500 does not have a positive return. Difference of differences is the difference of non-dividend-paying stocks minus dividend-paying stocks in advancing markets minus the difference of non-dividend-paying stocks minus dividend-paying stocks in declining markets.

Panel A: all years All markets Advancing markets Declining markets Difference of differences Panel B: subperiods 1970s All Markets Advancing markets Declining markets Difference of differences 1980s All markets Advancing markets Declining markets Difference of differences 1990s All markets Advancing markets Declining markets Difference of differences 2000s All markets Advancing markets Declining markets Difference of differences

Non-dividend-paying (%)

Dividend-paying (%)

Difference (%) a

0.96 4.13 − 3.72

1.34 3.62 − 1.88

− 0.38 ⁎⁎, b, c 0.51 ⁎⁎, b, c − 1.84 ⁎⁎, b, c 2.35 ⁎⁎

0.87 6.47 − 3.52

1.29 5.24 − 2.40

− 0.42 ⁎⁎, b, c 1.23 ⁎⁎, b, c − 1.12 ⁎⁎, b, c 2.35 ⁎⁎

0.57 3.39 − 3.65

1.71 4.20 − 1.89

− 1.14 ⁎⁎, b, c − 0.81 ⁎⁎, b, c − 1.76 ⁎⁎, b, c 0.95 ⁎⁎

1.38 3.35 − 3.00

1.23 2.65 − 1.91

0.15 ⁎⁎, b, c 0.70 ⁎⁎, b, c − 1.09 ⁎, b, c 1.79 ⁎⁎

1.08 6.97 − 4.71

1.37 3.50 − 0.88

− 0.29 ⁎⁎, b, c 3.47 ⁎⁎, b, c − 3.83 ⁎, b, c 7.30 ⁎⁎

a

Significance was tested using only parametric tests for the Differences of differences. Indicates the Wilcoxon sign-rank test is significant at the 1% level. Indicates the Kruskal–Wallis test is significant at the 1% level. ⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level. b c

We define declining markets as a time when investors should more highly value dividend payment (i.e., declining markets are when dividend payments should be most valuable to investors). Similar to Goldstein and Nelling (1999), we collect the S&P 500 returns for each month from CRSP and classify an advancing market as a month during which the monthly return on the S&P 500 was positive, while a declining market is one where the S&P 500 posted a negative monthly return.19 Overall, our sample includes 25,238 NYSE, Amex and Nasdaq listed firms for the 456 calendar months from January 1970 to December 2007. We classify each firm as either dividend-paying or non-dividend-paying for every month of the sample period in which data are available. We develop a total of 2,777,716 firm months in our time period, of which 1,419,664 are non-dividendpaying firm months and 1,358,052 are dividend-paying firm months. Table 1 describes the dividend- and non-dividend-paying firm months in our sample. Panel A provides averages across all observations for all 456 calendar months in our sample. Panels B and C provide averages for those observations that occur in the 268 months in our sample when the S&P 500 has a positive return (“advancing markets”) and the 188 months where it does not (“declining markets”). Panel A indicates that the average market capitalization of firms during the months when they were classified as dividend-paying is almost five times that of firms classified as non-dividend-paying. This larger size is due to having twice as many shares outstanding and an average price about 2.4 times that of the non-dividend-paying firms. Trading volume is similar for dividend-paying firms and non-dividend-paying firms. Betas for non-dividend-paying firms (0.862) are slightly larger than those of dividend-paying firms (0.703). These general results in Panel A are similar for both advancing and declining markets, as indicated by Panels B and C, indicating that the relative relationships between dividend-paying and non-dividend-paying do not vary significantly with overall market movements.

19 While the analyses tabulated in this paper use this definition for advancing and declining markets, we have also examined alternative definitions of advancing and declining markets. First, we have redefined advancing markets to require that the S&P 500 return exceeds the risk-free rate, as in Ang and Chen (2002). Second, we examine bull and bear markets as defined by Ned Davis Research. Third, we use the National Bureau of Economic Research's business cycle classification to define expansions (up) and contractions (down). In each case, we find substantively similar results.

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

463

Table 3 Fama–MacBeth returns. This table contains the average coefficients of monthly ordinary least squares of dividend-paying and non-dividend-paying firms. We run the regressions cross-sectionally each month for every firm, as in Fama and MacBeth (1973). The coefficients reported below are the average coefficients for each group. The regressions take the form: rit −rFt = αit + γit βt + μ it LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividend-paying firm in month t. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less. Large Positive Movements are when the S&P 500 return for that month is in excess of + 5%; Small Positive Movements are when the S&P 500 return for that month is between 0% and + 5%. Small Negative Movements are when the S&P 500 declined and its return for that month is between 0% and − 5%; Large Negative Movements are when the S&P 500 declined by more than 5%.

Panel A: overall Declining markets Advancing markets Differences Panel B: size of movement Advancing markets Large Positive Movements Small Positive Movements Differences Declining markets Small Negative Movements Large Negative Movements Differences

Intercept

β

Ln(Mktcap)

Ln(BVEquity)

DIV

− 0.0261 ⁎⁎ 0.0092 ⁎ ⁎⁎

− 0.0370 ⁎⁎ 0.0256 ⁎⁎ ⁎⁎

0.0022 ⁎⁎ 0.0000 ⁎⁎

− 0.0140 ⁎⁎ − 0.0180 ⁎⁎ ⁎⁎

0.0076 ⁎⁎ − 0.0040 ⁎⁎ 0.0116 ⁎⁎

0.0131 ⁎ 0.0085 ⁎ ⁎⁎

0.0553 ⁎⁎ 0.0151 ⁎⁎ ⁎⁎

0.0005 ⁎ − 0.0004 ⁎ ⁎

− 0.0210 ⁎⁎ − 0.0170 ⁎⁎

− 0.0080 ⁎ − 0.0020 ⁎⁎ − 0.0060 ⁎⁎

− 0.0180⁎ − 0.0510 ⁎⁎ ⁎⁎

− 0.0260 ⁎⁎ − 0.0770 ⁎⁎ ⁎⁎

0.0014 ⁎⁎ 0.0056 ⁎⁎ ⁎⁎

− 0.0140 ⁎⁎ − 0.0110 ⁎⁎ ⁎⁎

0.0058 ⁎⁎ 0.0164 ⁎⁎ − 0.0106 ⁎⁎

⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level.

4. Empirical results 4.1. Overall To investigate how investor preferences for dividends vary across market conditions, we examine the returns of dividendpaying and non-dividend-paying stocks in advancing and declining markets separately. In addition to showing results for all markets, Table 2 presents evidence for the 268 months in our sample when the S&P 500 has a positive return (“advancing markets”) and the 188 months where it does not (“declining markets”). Panel A indicates that we find that dividend-paying firms significantly outperform non-dividend-paying firms by 0.38% per month across all the months in our sample. This difference is statistically significant at the 1% level for the Student t-test, the Wilcoxon sign-rank test, and the Kruskal–Wallis test. This difference, however, depends on the state of the market: not risk adjusted, non-dividend-paying firms significantly outperform dividend-paying firms in advancing markets by about 0.51%, while non-dividend-paying stocks underperform dividend-paying stocks by 1.84% in declining markets. Using a difference-ofdifferences test, we find that dividend-paying stocks outperform non-dividend-paying stocks by 2.35% more in declining markets than in advancing markets, and that this difference is significant at the 1% level.20 To verify that our overall result is not driven by one particular sub-period, Panel B examines four separate decade sub-periods: January 1970 to December 1979, January 1980 to December 1989, January 1990 to December 1999, and January 2000 to December 2007. Interestingly, dividend-paying stocks outperformed non-dividend-paying stocks overall in every decade except the 1990s. The results for advancing markets across sub-periods are relatively similar to the overall period, except that dividend-paying stocks outperformed non-dividend-paying stocks even in advancing markets in the 1980s. Even so, the differences-of-differences tests for all four sub-periods indicate that dividend-paying stocks significantly outperform non-dividend-paying stocks by more in declining markets than in advancing markets at the 1% level. The difference-of-difference results are consistent in sign and significance both overall and in each of the four sub-periods. This finding supports the first major empirical prediction, that investors differentially prefer dividend-paying stocks over non-dividend-paying stocks more in declining markets than in rising markets.

20 We note that the difference of differences that we use in this paper is the difference of non-dividend-paying stocks minus dividend-paying stocks in advancing markets minus the difference of non-dividend-paying stocks minus dividend-paying stocks in declining markets. A positive number for this test indicates that dividend-paying stocks outperform non-dividend-paying stocks by more in declining markets than in advancing markets. Due to the nature of the data (unequal number of observations across all four potential categories), throughout the paper we use only parametric methods to test the significance of the difference of difference test.

464

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

4.2. Fama and MacBeth (1973) Style regressions Similar to Grinblatt and Han (2005), we examine Fama and MacBeth (1973) style regressions to determine if dividend-paying stocks outperform non-dividend-paying stocks in declining markets. We run the regressions cross-sectionally each month for every firm as in Fama and MacBeth (1973). Specifically, we estimate the following: rit −rFt = αit + γit βt + μ it LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit

ð1Þ

where rit − rFt is the return on a stock in month t minus the three-month Treasury bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividend-paying firm in month t. In addition, by including both beta and size, we also control for some of the small-firm beta effects for non-dividend-paying firms noted in Keim (1985) and Christie (1990). Table 3 reviews the basic results of these regressions. The results in Panel A for the overall regressions reports that at the 1% level, the coefficient for DIV is significantly greater in declining market months (0.0076) than in advancing market months (−0.0040), indicating that in declining markets dividend-paying firms outperform non-dividend-paying firms by approximately 1.2% each month more than in advancing months. This shows that investors value dividend-paying firms more in declining markets and more so than in advancing markets.21 An interesting question is whether these results vary with the magnitude of the movement in the market. We therefore divide the sample into months with large (greater than 5%) and small (between 0% and 5%) movements in the S&P 500 for both advancing and declining markets. The results in Panels B are revealing. The magnitude of the coefficient for DIV monotonically increases as the S&P 500 return decreases. The coefficient for large positive changes (−0.0080) is smaller than the coefficient on DIV for large negative changes (0.0164). In addition, the differences in the coefficients for large and small changes are statistically significantly different from one another, so that even across these changes, these differences are not symmetric. Therefore, the effect gets stronger the more the market declines, providing further evidence of increasing and asymmetric shareholder preference for dividend-paying stocks in declining markets. Thus, the answer to our central question – Do investors prefer dividend-paying firms to non-dividend-paying firms in declining markets? – is yes. We also find that the effects become stronger the larger the decline. 4.3. Quality differences? Allen et al. (2000) suggest that better firms pay dividends. Therefore, it is possible that, although the previous regressions controlled for beta, size, and the book value of equity, the results in Tables 2 and 3 are due to some quality difference in the firms not captured by those variables. To investigate this possibility, we first divide our sample into “better” and “worse” firms. First, similar to DeAngelo et al. (1992), we segment our data into positive and negative earnings. Generally speaking, negative net income should be correlated with “worse” firms. Therefore, if we were only capturing the fact that better firms pay dividends and worse firms do not, we should not find that our results hold once the firms are segmented by positive or negative net income. We therefore segment our firms based on whether the firm had positive or negative net income in the previous year. As shown in Panel A of Table 4, dividend-paying firms do better than non-dividend-paying firms by more in declining markets for both the negative net income and positive net income samples. As our second proxy of quality, we use prior year's return-on-assets (ROA) and evenly divide the sample into high and low ROA firms, and repeat our test. As shown in Panel B of Table 4, we again find that both the low ROA and high ROA groups indicate that dividend-paying firms outperform non-dividend-paying firms by more in declining markets. To verify these results further, we re-run our Fama–MacBeth regressions but include an additional variable to control for profitability or quality directly: rit −rFt = αiT + γiT βt + μ iT LnðMktcapÞt + ηiT LnðBVEquityÞt + φiT Profitabilityt + δiT DIVt + εit

ð2Þ

Panel A of Table 5 shows the results when we use return-on-assets (ROA), and Panel B has the results for return-on-equity (ROE), as measures of profitability. In each case, our results continue to hold. As a further quality test, we include Tobin's Q as a further measure of quality. We estimate Tobin's Q using the methodology suggested by Chung and Pruitt (1994). As shown in Table 5, our results continue to hold. In all three panels, dividend-paying stocks outperform non-dividend-paying stocks by more 21 We also perform a variety of robustness checks, not tabulated for brevity. For example, we re-run our Fama and MacBeth (1973) regressions in Eq. 1 for both size quintiles and volume quintiles, and our results continue to hold. We also find that our results hold for both NYSE-Amex and Nasdaq firms separately, so that our results are not a function of exchange listing. We have segmented by industry, and find results in the opposite direction for only two industries: agricultural and forestry and finance, insurance and real estate. In addition, as noted previously, our results are robust to using bull and bear markets as defined by Ned Davis Research instead of using advancing or declining months as defined by the S&P 500, or examining returns for economic expansions and contractions as defined by the NBER. Our results are also robust to adjusting for risk using either the CAPM or the Fama and French (1993) three factor model. In all cases, dividendpaying stocks outperform non-dividend-paying stocks.

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

465

Table 4 Fama–MacBeth risk adjusted returns segmented by profitability. This table contains the coefficients of ordinary least squares across equally weighted portfolios of dividend-paying and non-dividend-paying firms. The regressions take the form: rit −rFt = αit + γit βt + μ it LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividendpaying firm in month t. In Panel A profitability is defined by either positive or negative net income. In Panel B profitablity is defined as Return on Assets. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less.

Panel A: profitability = +/− net income Negative net income Declining markets Advancing markets Differences Positive net income Declining markets Advancing markets Differences Panel B: profitability = ROA Low ROA

High ROA

Declining markets Advancing markets Differences Declining markets Advancing markets Differences

Intercept

β

Ln(Mktcap)

Ln(BVEquity)

DIV

− 0.0520 ⁎⁎ − 0.0140 ⁎⁎ ⁎⁎ − 0.0070 ⁎⁎ 0.0264 ⁎⁎ ⁎⁎

− 0.0420 ⁎⁎ 0.0237 ⁎⁎ ⁎⁎ − 0.0300 ⁎⁎ 0.0277 ⁎⁎ ⁎⁎

0.0049 ⁎ 0.0038 ⁎ ⁎⁎ − 0.0006 ⁎ − 0.0030 ⁎⁎ ⁎⁎

− 0.0140 ⁎ − 0.0190 ⁎⁎ ⁎⁎ − 0.0140 ⁎⁎ − 0.0180 ⁎⁎ ⁎⁎

0.0107 ⁎⁎ − 0.0090 ⁎⁎ 0.0197 ⁎⁎ 0.0028 ⁎⁎ − 0.0060 ⁎⁎ 0.0088 ⁎

− 0.0360 ⁎⁎ − 0.0030 ⁎⁎

− 0.0390 ⁎⁎ 0.0220 ⁎⁎ ⁎⁎ − 0.0300 ⁎⁎ 0.0284 ⁎⁎ ⁎⁎

0.0035 ⁎⁎ 0.0022 ⁎⁎ ⁎⁎ − 0.0010 ⁎⁎ − 0.0040 ⁎⁎ ⁎⁎

− 0.0130 ⁎⁎ − 0.0170 ⁎⁎ ⁎⁎ − 0.0140 ⁎⁎ − 0.0190 ⁎⁎ ⁎⁎

0.0123 ⁎⁎ − 0.0013 ⁎⁎ 0.0136 ⁎⁎ 0.0016 ⁎⁎ − 0.0070 ⁎⁎ 0.0086 ⁎

− 0.0040 0.0314 ⁎⁎ ⁎⁎

⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level.

in down markets by about 1.1% to 1.2%. Therefore, our results do not seem to be driven simply by better firms being more able to pay dividends. Overall, the results in Tables 2 to 5 indicate that the first major empirical prediction of dividend-paying stocks having differential outperformance depending on the state of the market. Even after adjusting for time period, risk, or the quality/

Table 5 Fama–MacBeth returns controlling for profitability. This table contains the average coefficients of monthly ordinary least squares of dividend-paying and nondividend-paying firms. We run the regressions cross-sectionally each month for every firm, as in Fama and MacBeth (1973). The coefficients reported below are the average coefficients for each group. The regressions take the form: rit −rFt = αiT + γiT βt + μ iT LnðMktcapÞt + ηiT LnðBVEquityÞt + φiT Profitabilityt + δiT DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, Profitability is the firm's prior year's return on assets, return on equity or Tobin's Q, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividend-paying firm in month t. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less. Tobin's Q is calculated in accordance with Chung and Pruitt (1994). Intercept

β

Ln(Mktcap)

Ln(BVEquity)

Profitability

DIV

Panel A: ROA Declining markets Advancing markets Differences

− 0.0220 ⁎⁎ 0.0124 ⁎ ⁎⁎

− 0.0350 ⁎⁎ 0.0261 ⁎⁎ ⁎⁎

0.0013 0.0001 ⁎⁎

− 0.0140 ⁎⁎ − 0.0180 ⁎⁎ ⁎⁎

0.0341 ⁎⁎ 0.0273 ⁎⁎ ⁎⁎

0.0063 ⁎⁎ − 0.0050 ⁎⁎ 0.0113 ⁎⁎

Panel B: ROE Declining markets Advancing markets Differences

− 0.0230 ⁎⁎ 0.0125 ⁎ ⁎⁎

− 0.0360 ⁎⁎ 0.0261 ⁎⁎ ⁎⁎

0.0017 ⁎⁎ − 0.0008 ⁎⁎

− 0.0140 ⁎⁎ − 0.0180 ⁎⁎ ⁎⁎

0.0117 ⁎⁎ 0.0128 ⁎⁎

0.0069 ⁎⁎ − 0.0050 ⁎⁎ 0.0119 ⁎⁎

Panel B: Tobin's Q Declining markets Advancing markets Differences

− 0.0260 ⁎⁎ 0.0090 ⁎⁎

− 0.0370 ⁎⁎ 0.0254 ⁎⁎ ⁎⁎

0.0023 ⁎⁎ − 0.0002 ⁎⁎

− 0.0130 ⁎⁎ − 0.0160 ⁎⁎ ⁎⁎

⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level.

⁎

− 0.0002 0.0006 ⁎

0.0078 ⁎⁎ − 0.0040 ⁎⁎ 0.0118 ⁎⁎

466

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

profitability of the firm, dividend-paying stocks differentially outperform non-dividend-paying stocks across advancing and declining markets. This result is robust to a variety of measures of quality. 5. Signaling and free cash flow The results cannot be attributed to a difference in quality between dividend-paying and non-dividend-paying stocks. Instead, there seems to be something important about the payment of dividends. The reason why firms pay dividends have been hotly debated. One suggestion is that firms pay dividends to constrain free cash flow, as suggested by Easterbrook (1984) and Jensen (1986) and others. Another reason why firms might pay dividends is in order to signal future earnings, as noted by Lintner (1956), Bhattacharya (1979), John and Williams (1985), and Miller and Rock (1985), among others. We therefore check to see if our results could be explained by either the free cash flow or signaling hypotheses. 5.1. Fama and MacBeth (1973) Regressions The free cash flow literature has suggested that firms use dividends to constrain the use of cash flow and overinvestment. To investigate if our results are due to these effects, both of which are related to the free cash flow hypothesis, we divide our sample two different ways. First, as in Lang et al. (1991), we divide our sample into low and high cash flow firms using the technique in Lehn and Poulsen (1989) to define cash flows and then run the Fama and MacBeth (1973) regression in Eq. (1). As shown in Panel A of Table 6, the coefficient on the DIV dummy variable was positive and significant at the 1% level in the declining markets, and negative and significant at the 1% level in advancing markets, for both the low and high cash flow groups. More notably, the difference in the coefficients across the two coefficients was positive and significant at the 1% level for both for the low (0.0156) and high (0.0078) cash flow groups. A related issue raised by the free cash flow hypothesis is overinvestment. Lang and Litzenberger (1989) suggest that firms which overinvest will have a Tobin's Q that is less than one. Lang et al. (1991) also suggest that firms with poor investment opportunities will have a Tobin's Q less than one. Similar to Lang and Litzenberger (1989), Lang et al. (1991), Denis et al. (1994),

Table 6 Fama–MacBeth risk adjusted returns segmented by free cash flow/Tobin's Q. This table contains the coefficients of ordinary least squares across equally weighted portfolios of dividend-paying and non-dividend-paying firms by groups based on either Free Cash Flow or Tobin’s Q. The regressions take the form: rit −rFt = αit + γit βt + μ it LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividendpaying firm in month t. Free Cash Flow is defined as in Lehn and Poulsen (1989): FCF = INC–TAX–INTEXP–PFDDIV–COMDIV where FCF is free cash flow, INC is operating income before depreciation, TAX is total income taxes minus change in deferred taxes, INTEXP is the gross interest expense on short- and long-term debt, PFDDIV is the total amount of preferred dividend requirement on preferred stock, and COMDIV is the total dollar amount of dividends declared on common stock. We standardize the FCF by total assets of the firm. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less. Panel A: FCF

Low FCF

High FCF

Declining markets Advancing markets Differences Declining markets Advancing markets Differences

Intercept

β

− 0.0310 ⁎⁎ − 0.0001 ⁎⁎ − 0.0140 ⁎⁎ 0.0225 ⁎⁎ ⁎⁎

− 0.0380 ⁎⁎ 0.0248 ⁎⁎ ⁎⁎ − 0.0330 ⁎⁎ 0.0261 ⁎⁎ ⁎⁎

Intercept

β

− 0.0240 ⁎⁎ 0.0067 ⁎⁎ − 0.0250 ⁎⁎ 0.0163 ⁎⁎ ⁎⁎

− 0.0390 ⁎⁎ 0.0192 ⁎⁎ ⁎⁎ − 0.0350 ⁎⁎ 0.0299 ⁎⁎ ⁎⁎

Ln(Mktcap) 0.0032 ⁎⁎ 0.0032 ⁎⁎ 0.0004 − 0.0030 ⁎

Ln(BVEquity)

DIV

− 0.0140 ⁎ − 0.0170 ⁎⁎ ⁎⁎ − 0.0140 ⁎⁎ − 0.0181 ⁎⁎ ⁎⁎

0.0116 ⁎⁎ − 0.0040 ⁎⁎ 0.0156 ⁎⁎ 0.0038 ⁎⁎ − 0.0040 ⁎⁎ 0.0078 ⁎⁎

Ln(BVEquity)

DIV

− 0.0150 ⁎ − 0.0170 ⁎⁎ ⁎ − 0.0110 ⁎⁎ − 0.0140 ⁎⁎ ⁎

0.0122 ⁎⁎ 0.0013 0.0109 ⁎⁎ 0.0015 ⁎⁎ − 0.0100 ⁎⁎ 0.0115 ⁎⁎

Panel B: Tobin's Q

Low Tobin's Q

High Tobin's Q

Declining markets Advancing markets Differences Declining markets Advancing markets Differences

⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level.

Ln(Mktcap)

⁎

0.0018 ⁎ 0.0006

0.0027 ⁎⁎ − 0.0009 ⁎⁎

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

467

Yoon and Starks (1995), we segment firms into high (Q N 1) and low (Q b 1) Tobin's Q to separate out those firms with good investment opportunities and those which do not overinvest from those that do. We then rerun the regression in Eq. (1). Panel B of Table 6 indicates that for Q b 1 firms, the coefficient on the DIV dummy variable is positive and significant at the 1% level in declining markets, but insignificant in advancing markets. This result provides some slight support for the free cash flow hypothesis, as we would expect the dividend payment to be more binding in down markets for firms that overinvest. For the Q N 1 firms, the coefficient on the DIV dummy variable is positive in declining markets and negative in advancing markets and both are significant at the 1% level. The difference across advancing and declining markets for Q N 1 firms is also significant at the 1% level. However, the difference in the coefficients for Q N 1 firms across advancing and declining markets (0.0115) is not much different than the difference for Q b 1 firms (0.0109), suggesting that our main result, dividend-paying firms do better in declining markets, is invariant to Tobin's Q. The results in Table 6 provide marginal support at best for the idea that the main results are due to the free cash flow hypothesis. The result holds for both low and high cash flow firms, although the effect is twice as strong in low cash flow firms, where the dividend may be more binding, especially in declining markets. The result also holds across high and low Q firms equally, although there is very slight evidence that for firms with Q b 1, the primary effect comes from declining markets. Alternatively, the signaling literature has suggested that dividend-paying firms use dividends to signal future earnings. Therefore, to see if our results are due to this effect, we segment our data based on the future earnings of the firm. Panel A of Table 7 segments our sample based on the next year's earnings measured as earnings before interest, taxes, depreciation, and amortization scaled by total assets (EBITDA/TA), and then we run the Fama and MacBeth (1973) regression found in Eq. (1). The coefficient on the DIV dummy variable was positive and significant at the 1% level in the declining markets, and negative and significant at the 1% level in advancing markets, for both the low future EBITDA and high future EBITDA groups. More notably, the difference in the coefficients was positive and significant at the 1% level for both for the low future EBITDA group (0.0154) and for the high EBITDA group (0.0075). In Panel B of Table 7 we segment the sample based on next year's earnings per share (EPS) and get substantially the same results, although the coefficient on the DIV dummy variable for the high EPS group is not significantly different from zero and the difference across the coefficients in the high EPS group is only significant at the 5% level. The data in Table 7 does not seem to indicate that it is signaling that is causing dividend-paying firms to do better in declining markets. One might expect that if it were signaling, the effect would only occur in high future earning firms. However, the effect is still seen in both low future earning firms and high future earning firms even when future earnings is measured in two different ways. In addition, the difference in outperformance is almost double the size for the low future earners as it is for the high future earners. Therefore, it does not appear that signaling explains the effects seen previously that dividend-paying firms outperform non-dividend-paying firms by more in declining markets than advancing ones.

Table 7 Fama–MacBeth risk adjusted returns segmented by future earnings. This table contains the coefficients of ordinary least squares across equally weighted portfolios of dividend-paying and non-dividend-paying firms by earnings groups. The regressions take the form: rit −rFt = αit + γit βt + μit LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividendpaying firm in month t. In Panel A future earnings are defined as the next years EBITDA standardized by total assets and in Panel B future earnings are defined as the next year's EPS. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less. Intercept

β

− 0.0301 ⁎⁎ − 0.0020 ⁎⁎ − 0.0150 ⁎⁎ 0.0209 ⁎⁎ ⁎⁎

0.0390 ⁎⁎ 0.0242 ⁎⁎ ⁎⁎ − 0.0320 ⁎⁎ 0.0267 ⁎⁎ ⁎⁎

− 0.0310 ⁎⁎ 0.0012 ⁎⁎

− 0.0370 ⁎⁎ 0.0241 ⁎⁎

0.0022 ⁎⁎ 0.0017 ⁎⁎ ⁎⁎

− 0.0001 0.0318 ⁎⁎ ⁎⁎

− 0.0300 ⁎⁎ 0.0286 ⁎⁎ ⁎⁎

− 0.0010 − 0.0040 ⁎

Ln(Mktcap)

Ln(BVEquity)

DIV

− 0.0140 ⁎⁎ − 0.0180 ⁎⁎ ⁎⁎ − 0.0130 ⁎⁎ − 0.0160 ⁎⁎ ⁎⁎

0.0124 ⁎⁎ − 0.0030 ⁎⁎ 0.0154 ⁎⁎ 0.0035 ⁎⁎ − 0.0040 ⁎⁎ 0.0075 ⁎⁎

− 0.0150 ⁎⁎ − 0.0190 ⁎⁎ ⁎⁎ − 0.0130 ⁎⁎ − 0.0160 ⁎⁎ ⁎⁎

0.0077 ⁎⁎ − 0.0070 ⁎⁎ 0.0147 ⁎⁎ 0.0002 − 0.0071 ⁎⁎ 0.0073 ⁎

Panel A: future earnings = EBITDA/TA Low future earnings

High future earnings

Declining markets Advancing markets Differences Declining markets Advancing markets Differences

Panel B: future earnings = EPS Low future earnings Declining markets Advancing markets Differences High future earnings Declining markets Advancing markets Differences ⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level.

0.0036 ⁎⁎ 0.0037 ⁎⁎ 0.0006 ⁎ − 0.0002

468

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

5.2. Dividend changes analysis As the regression analysis did not indicate that either free cash flow or signaling explains why dividend-paying firms do better than non-dividend-paying firms in declining markets, we now restrict our attention to only dividend-paying firms so that we can examine dividend changes. Previous work on dividend changes did not examine the effect of advancing and declining markets. Dividend changes in declining markets could make dividends more or less binding. At the same time, in declining markets investors' perceptions of future profits tend to be lower, while investors tend to have positive outlooks on future earnings during advancing markets. Increasing dividends in declining markets therefore provides a much stronger indication about the future than a similar increase during an advancing market. Similarly, during a declining market, maintaining a dividend provides investors with reassuring information, while doing the same in advancing markets likely provides less additional information. Thus, we would expect in declining markets, dividend increases will have higher price reactions and dividend decreases will have less negative price reactions than comparable changes in advancing markets. Further, if maintaining a dividend provides information, then in declining markets firms that do not change their dividend payments should have higher abnormal returns than firms with no dividend change in advancing markets. We test this empirical prediction by examining the market's reaction to dividend changes and no changes during advancing and declining markets first, and later examine if there are different effects if we segment on either free cash flow or signaling variables. We obtain changes in quarterly-dividends from CRSP from 1970 to 2007. The only restrictions we place on the sample is that there must be five days of returns surrounding the announcement listed on CRSP, the dividend is paid on ordinary common shares of U.S.-incorporated companies, and the change is not a dividend initiation or omission. Our sample comprises 6344 firms with 38,032 increases, 9596 decreases, and 199,856 no changes. Following Brown and Warner (1980, 1985) we estimate the abnormal returns using a modified market model: ARi = ri −rm

ð3Þ

where ARi is the abnormal return for firm i, ri is the return on firm i and rm is the equally weighted market index return. To calculate the cumulative abnormal returns (CARs), we sum the ARs over the five-day period (−2.2) around the announcement date supplied by CRSP. We do not estimate market parameters based on a time period before each change because some firms have frequent dividend changes. Thus, there would be a high probability that previous changes would be included in the estimation period, which would make the beta estimates less meaningful. We first examine the overall effects of increasing, decreasing, or not changing dividends depending on the state of the market. Panel A of Table 8 shows that price reactions to dividend increases are less in advancing markets (0.678%) than in declining markets (1.148%). This 0.470% difference is significant at the 1% level for both the Student t-test and the two nonparametric tests. In addition, dividend decreases have less negative returns if announced during declining markets (− 0.494%) than during advancing markets (−0.517%), and this difference is also significant at the 1% level for all tests. Notably, firms that maintain their current dividend payments in declining markets experience significant, positive abnormal returns (0.195%), but for firms that maintained their current dividend levels in advancing markets experience a significant negative return (−0.084%). The 0.279% difference in abnormal returns between firms that maintained their dividend in advancing and advancing markets is significant at the 1% level for the parametric and nonparametric tests, indicating a much stronger and positive response for firms that just maintain their dividend in declining markets over the negative response for maintaining a dividend in advancing markets. Clearly, dividends convey some information, as dividend increases in declining markets (and to a lesser extent, dividend decreases in advancing markets) seem to be very important. More importantly, not changing an existing dividend matters differentially in declining and advancing markets. These results seem to support DeAngelo and DeAngelo (2006). The results in Panel A of Table 8 may be due to either the free cash flow or signaling hypotheses, although it might provide a bit more support for the signaling hypothesis. Increasing a dividend in a declining market may become a bit more binding, but clearly it is also a strong signal. More interestingly, simply maintaining a dividend in a declining market might signal that the situation is not that dire for the firm, but the binding should remain constant (assuming free cash flow issues do not increase when markets are doing poorly), so this result seems to be more in favor of a signaling explanation than a free cash flow one. To investigate this further, we segment dividend increases, decreases, and no change into high and low cash flow (Panel B) and high and low future earnings (Panel C) to investigate the free cash flow and signaling hypotheses, respectively. Panel B indicates that dividend increases for both high and low cash flow firms are better in declining markets. Dividend decreases matter more for low cash flow firms in declining markets. Interestingly, for high cash flow firms, dividend decreases do not have statistically significant effect during declining markets. However, the difference across advancing and declining markets is also significant, but only non-parametrically. High cash flow firms also do not have any significant results when they maintain their dividend in advancing markets, but do in declining markets. However, when firms maintain their dividends, the difference between advancing and declining markets is significant for both high and low cash flow firms, although it is bigger for low cash flow firms (0.471%) than high cash flow firms (0.208%), although in either case it should be just as binding, as there was no change in the dividend (assuming that cash flow generated by the firm does not drop or is not expected to drop in declining markets, which may not be the case). Panel C examines the results of dividend changes for low and high future earnings firms and finds generally similar results, with a few exceptions. Most notably, dividend decreases for low future earners are worse in declining markets than in advancing markets, consistent with signaling theory. This result is different than the result for cash flow in Panel B; while the result in Panel B

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

469

Table 8 Cumulative abnormal returns for dividend changes in advancing and declining markets. We calculate the cumulative abnormal returns (CARs) for the five days (− 2 and 2) around the announcement (day 0) of a dividend change for firms with high and low future earnings. We use a modified market model to estimate abnormal returns: ARi = ri − rm where ri is the return on firm i and rm is the equally weighted market index return. We eliminate the usual estimation period due to the high probability of previous dividend changes for firms during the estimation period. We report the CARs for 38,032 increases, 9596 decreases, and 199,856 no changes announced between 1970 and 2007 for 6344 firms. Our reports cover all markets: advancing markets, when the S&P 500 index return is greater than zero; and declining markets, when the S&P 500 index return is zero or less. Free cash flow is measured as in Lehn and Poulsen (1989). High (low) free cash flows are defined as the firm having free cash flow that is great than (less than or equal to) the median free cash flow for all firms in the sample for the year prior to the dividend payment. Future earnings are defined as the net income for the year after the dividend payment. High (low) future earnings are defined as the firm having earnings that are great than (less than or equal to) the median earnings for all firms in the sample for the year after the dividend payment.

Panel A: overall Dividend increase Dividend decrease No change Panel B. Free cash flow Dividend increases High cash flow Low cash flow Dividend Decreases High cash flow Low cash flow No change High cash flow Low cash flow Panel C: future earnings Dividend increases High future earnings Low future earnings Dividend decreases High future earnings Low future earnings No change High future earnings Low future earnings

Advancing markets (%)

Declining markets (%)

Difference (%)

0.678 ⁎⁎ − 0.517 ⁎⁎ − 0.084 ⁎

1.148 ⁎⁎ − 0.494 ⁎⁎ 0.195 ⁎⁎

− 0.470 ⁎⁎, a, b − 0.023 ⁎, a, b − 0.279 ⁎⁎, a,b

0.603 ⁎⁎ 0.899 ⁎⁎

1.019 ⁎⁎ 1.560 ⁎⁎

− 0.416 ⁎⁎, a, b − 0.661 ⁎⁎, a, b

− 0.263 ⁎ − 1.361 ⁎⁎

− 0.164 − 1.157 ⁎⁎

− 0.099 a, b − 0.204 ⁎⁎, a,b

− 0.054 − 0.164 ⁎⁎

0.154 ⁎⁎ 0.307 ⁎⁎

− 0.208 ⁎⁎, a, b − 0.471 ⁎⁎, a, b

0.989 ⁎⁎ 1.546 ⁎⁎

− 0.411 ⁎⁎, a, b − 0.628 ⁎⁎, a, b

− 0.154 ⁎⁎ − 1.257 ⁎⁎

− 0.047 ⁎⁎ − 1.394 ⁎⁎

− 0.107 ⁎, a, b 0.137 ⁎⁎, a, b

− 0.057 − 0.147 ⁎⁎

0.143 ⁎⁎ 0.318 ⁎⁎

− 0.200 ⁎, a, b − 0.465 ⁎⁎, a, b

0.578 ⁎⁎ 0.918 ⁎⁎

a

Indicates the Wilcoxon sign-rank test is significant at the 1% level. Indicates the Kruskal–Wallis test is significant at the 1% level. ⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level. b

for dividend decreases for low cash flow firms indicated that the CARs were larger (less negative) for declining markets than for advancing markets and therefore were similar to the results in the rest of the paper, the results in Panel C for dividend decreases for low future earnings suggest the opposite, namely that the CARs are even more negative for declining markets. Therefore, this provides some weak support for the signaling theory for poor future performers, but only based on market conditions. Interestingly, maintaining a dividend for high future earners in advancing markets is not significant, but is for declining markets. The difference across advancing and declining markets for both high and low future earnings is still significant. Again, the difference in return across advancing and declining markets is bigger for low future earners (0.465%) than for high future earners (0.200%). Overall, both the regression and dividend change results indicate that the results are not primarily driven by either the free cash flow or signaling hypotheses. While the regressions provided some weak support for the free cash flow hypothesis, the dividend changes provided some weak evidence for the signaling hypothesis. In both cases, however, the main results continued to hold overall and were found for both hypotheses. 6. Cash payments As a final check, we make sure that the results are not due to the actual receipt of cash or the size of the cash payment. 6.1. Cash payments Kalay and Michaely (2000) note that time series variation in returns may be related to the actual dividend payment itself. Therefore, one possibility is that dividend-paying firms outperform non-dividend-paying firms simply because of the return in the month the firm paid the dividend, and that in the remaining months when no dividend is paid, returns for dividend and nondividend-paying firms are similar. Dividends may matter more in declining markets simply due to the cash payment itself. If this is true, then there may be no information in the fact that a firm continues to pay a dividend, but only value in the dividend payment

470

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

Table 9 Average return for advancing and declining markets for dividend-paying stocks during months with no dividend payments. Panel A of this table reports the average monthly return to dividend- and non-dividend-paying stocks from 1970 to 2007. We include dividend-paying stocks only for months during which a quarterly dividend-paying stock does not pay a dividend. Panel B reports the coefficients of ordinary least squares across equally weighted portfolios of dividendpaying and non-dividend-paying firms. The regressions take the form: rit −rFt = αit + γit βt + μit LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividend-paying firm in month t. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less. Panel A — Returns Advancing markets Non-Dividend-paying All Stocks

3.68%

Declining markets Dividend-paying 3.67%

Difference 0.01%

Non-Dividend-paying − 3.79%

Dividend-paying

Difference

Diff. of Diff. a

− 2.21%

− 1.58% ⁎⁎, b, c

1.59% ⁎⁎

Panel B — Fama–MacBeth regressions

Declining markets Advancing markets Differences

Intercept

β

Ln(Mktcap)

Ln(BVEquity)

DIV

− 0.0280 ⁎⁎ 0.0076 ⁎⁎

− 0.0380 ⁎⁎ 0.0261 ⁎⁎ ⁎⁎

0.0026 ⁎ 0.0001 ⁎⁎

− 0.0140 ⁎⁎ − 0.0180 ⁎⁎ ⁎⁎

0.0058 ⁎⁎ − 0.0060 ⁎⁎ 0.0118 ⁎⁎

a

Significance was only tested using parametric tests for the Differences of Differences. Indicates the Wilcoxon sign-rank test is significant at the 1% level. Indicates the Kruskal-Wallis test is significant at the 1% level. ⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level. b c

itself when it is received. Similarly, any asset pricing strategy that involves dividend capture or tax clienteles should not have different results in non-dividend-paying months. Thus, the third main empirical prediction suggests that investors should still prefer dividend-paying stocks over non-dividend-paying stocks not only during those months in which the dividend is paid, but also in those months between dividend payments. To verify that our results are not driven by the cash payment, we study the eight months of the year when dividend-paying stocks do not pay dividends. We eliminate the returns for dividend-paying firms in the month the dividend is paid and compare the returns of non-dividend-paying firms to the returns of dividend-paying firms in months with no dividend payments. In Panel A of Table 9, we find that for advancing markets, dividend-paying and non-dividend-paying firms have about the same average monthly return (3.67%), but in declining markets, dividend-paying firms (− 2.21%) still significantly outperform nondividend-paying firms (−3.79%), even when the dividend-paying firms' returns exclude the dividend return. The 1.59% difference of differences is significant at the 1% level. We also re-estimate the Fama and MacBeth (1973) model in Eq. (1) separately for advancing and declining markets and find in Panel B that again the coefficient on the dividend dummy is significantly larger in declining markets (0.0058) than in advancing markets (− 0.0060) even during months when these firms are not paying a dividend. Dividend-paying firms outperform nondividend-paying firms by over 1.18% more in declining markets than advancing markets, even when the months with the dividend payment is excluded. Thus, it is not the receipt of the cash itself that causes the asymmetric response by investors in advancing and declining markets. Overall, these results support the main finding that investors differentially prefer dividend-paying stocks in declining markets.

6.2. Dividend yield Previous tests focus only on whether or not firms pay dividends, not the magnitude of the dividend payments. We investigate if our results are sensitive to the size of the dividend yield. We start by examining only dividend-paying firms, and divide them into quintiles based on their dividend yield. The results in Table 10 indicate that each quintile has different intercepts for the Fama and MacBeth (1973) regressions. Panel A indicates that in advancing markets, the alphas monotonically decrease as the dividend yield increases, indicating that higher yields actually produce lower returns in advancing markets. However, Panel B indicates that for declining markets, the relation between dividend yield and return is not monotonic, as the coefficients on the intercept for quintiles 2 and 3 are not significantly different from zero. In fact, while there is a large difference in alphas between the lowest dividend yield group and second quintile dividend yield group, there is little difference in alphas between the second, third, fourth, and highest dividend yield groups. While the F-test of all five quintiles indicates that they are different from one another, the F-test of all but the lowest dividend group indicates that there is no difference based on dividend-yield. These results suggest that the asymmetric response of non-dividend-paying and dividend-paying stocks to declining markets is not related to dividend yield, but

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

471

Table 10 Fama–MacBeth adjusted returns segmented by dividend yield. This table contains the coefficients of ordinary least squares across equally weighted portfolios of dividend-paying and non-dividend-paying firms. The regressions take the form: rit −rFt = αit + γit βt + μit LnðMktcapÞt + ηit LnðBVEquityÞt + δit DIVt + εit where rit − rFt is the return on a stock in month t minus the three-month Treasury-bill return for month t, β is the firm's beta measured for the prior year for month t, Ln(Mktcap) is the natural log of the firm's market capitalization for month t, Ln(BVEquity) is the natural log of the firm's book value of equity for month t, and DIV is an indicator variable that equals one if the firm is a classified as a dividend-paying firm in month t and zero if the firms is classified as a non-dividend-paying firm in month t. The data are from CRSP and Compustat and run from January 1970 to December 2007. Advancing markets are when the S&P 500 index return is greater than zero and declining markets are when the S&P 500 index return is zero or less. We divide dividend-paying firms into quintiles based on their dividend yield. β

Ln(Mktcap)

Ln(BVEquity)

Panel A: dividend yield comparisons in advancing markets Lowest dividend yield 0.0218 ⁎⁎ 2 0.0215 ⁎⁎ 3 0.0171 ⁎⁎ 4 0.0154 ⁎⁎ Highest dividend yield 0.0100 ⁎⁎ ⁎⁎ F-test of all five ⁎⁎ F-test of highest 4

0.0252 ⁎⁎ 0.0255 ⁎⁎ 0.0231 ⁎ 0.0219 ⁎⁎ 0.0222 ⁎⁎ ⁎⁎

− 0.0020 ⁎ − 0.0020 ⁎ − 0.0010 ⁎ − 0.0009 − 0.0005 ⁎

− 0.0160 ⁎⁎ − 0.0140 ⁎⁎ − 0.0100 ⁎⁎ − 0.0120 ⁎⁎ − 0.0060 ⁎ ⁎⁎

Panel B: dividend yield comparisons Lowest dividend yield 2 3 4 Highest dividend yield F-test of all five F-test of highest 4

− 0.0280 ⁎⁎ − 0.0310 ⁎⁎ − 0.0300 ⁎⁎ − 0.0320 ⁎⁎ − 0.0340 ⁎⁎ ⁎

− 0.0010 − 0.0006 ⁎ − 0.0004 − 0.0002 − 0.0006

− 0.0160 ⁎⁎ − 0.0140 ⁎⁎ − 0.0120 ⁎⁎ − 0.0110 ⁎⁎ − 0.0070 ⁎ ⁎⁎

0.0038 ⁎⁎ 0.0014 ⁎⁎

− 0.0150 ⁎⁎ − 0.0200 ⁎⁎ ⁎⁎

0.0069 ⁎ − 0.0030 ⁎ 0.0099 ⁎⁎

0.0039 ⁎⁎ 0.0013 ⁎⁎

− 0.0150 ⁎⁎ − 0.0200 ⁎⁎ ⁎⁎

0.0037 ⁎ − 0.0070 ⁎⁎ 0.0107 ⁎⁎

Intercept

in declining markets − 0.0120 ⁎⁎ − 0.0070 − 0.0060 − 0.0040 ⁎ − 0.0040 ⁎ ⁎⁎ not significant

Panel C: comparison of non-dividend-paying and each dividend yielding group Lowest dividend yield Declining markets − 0.0330 ⁎⁎ − 0.0390 ⁎⁎ Advancing markets 0.0029 0.0247 ⁎⁎ ⁎⁎ ⁎⁎ Differences Dividend yield group 2 Declining markets − 0.0330 ⁎⁎ − 0.0390 ⁎⁎ Advancing markets 0.0031 0.0246 ⁎⁎ ⁎⁎ ⁎⁎ Differences Dividend yield group 3 Declining markets Advancing markets Differences Dividend yield group 4 Declining markets Advancing markets Differences Dividend yield group 5 Declining markets Advancing markets Differences

DIV

− 0.0330 ⁎⁎ 0.0030 ⁎⁎

− 0.0390 ⁎⁎ 0.0245 ⁎⁎ ⁎⁎

0.0039 ⁎⁎ 0.0013 ⁎⁎

− 0.0140 ⁎⁎ − 0.0190 ⁎⁎ ⁎⁎

0.0043 ⁎ − 0.0080 ⁎⁎ 0.0123 ⁎⁎

− 0.0330 ⁎⁎ 0.0028 ⁎⁎

− 0.0400 ⁎⁎ 0.0243 ⁎⁎ ⁎⁎

0.0040 ⁎⁎ 0.0014 ⁎⁎

− 0.0150 ⁎⁎ − 0.0190 ⁎⁎ ⁎⁎

0.0046 ⁎ − 0.0080 ⁎⁎ 0.0126 ⁎⁎

− 0.0330 ⁎⁎ 0.0019 ⁎⁎

− 0.0400 ⁎⁎ 0.0247 ⁎⁎ ⁎⁎

0.0042 ⁎⁎ 0.0017 ⁎ ⁎⁎

− 0.0140 ⁎⁎ − 0.0190 ⁎⁎ ⁎⁎

0.0083 ⁎ − 0.0110 ⁎⁎ 0.0193 ⁎⁎

⁎⁎ Indicates t-test is significant at the 1% level. ⁎ Indicates t-test is significant at the 5% level.

rather to the existence of dividend payments themselves, as increasing dividend yield reduces performance in advancing markets and, beyond the lowest dividend yield group, is seemingly unrelated to performance in declining markets. As a further test, Panel C examines the difference between the non-dividend-paying stocks and stocks by dividend yield quintile. For each quintile, the coefficients on the DIV dummy variable are significant and significantly different for declining and advancing markets, indicating that the differences between dividend-paying and non-dividend paying stocks in advancing and declining markets exist for all dividend yields. While the magnitude of the differences gets larger as the dividend yield quintiles increase, the increases from one quintile to the next are not large, except for the jump from second highest and highest dividend yield. More importantly, the overall result occurs for all quintiles. Collectively, these results suggest that it is the payment itself, and not the size of the payment, which drives dividend-paying firms to outperform non-dividend-paying firms in declining markets. Along with the results in Table 9, these findings imply that it is unlikely that the tax clientele/dividend capture hypotheses are driving the outperformance in declining markets.

472

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

7. Conclusion We find evidence that investors are concerned with firms' dividend policies. Our results indicate that dividend-paying stocks outperform non-dividend-paying stocks by approximately 1% to 2% more in declining markets than in advancing markets. Further, these results hold when we control for risk, different definitions of advancing and declining markets, size, liquidity, industry groups, and for different sub-periods. We also find that these differences increase the more the market decreases. These results seem not to be a function of the quality of the firm, based on past profitability, future profitability, cash flow, or Tobin's Q. Our tests indicate that the main results hold for both high and low cash flow and Tobin's Q firms, as well as high and low future profitability firms. Different tests provided only weak support for either the free cash flow hypothesis in one case and signaling in the other, suggesting that the main result is not primarily caused either by issues related to free cash flow, overinvestment, or signaling. We also show that investors respond asymmetrically to dividend increases, decreases, and no changes, based on the state of the market, and that dividend-paying firms outperform non-dividend-paying firms even in the months with no dividend payments. The differential results between advancing and declining markets also do not seem to be driven by either the free cash flow or signaling hypotheses, so, similar to many other papers, we cannot declare a firm winner. Results indicate a larger difference between the non-dividend-paying and low-dividend-yield portfolios than among the dividend-paying portfolios themselves. This finding suggests that our results are not due to reasons related to tax clienteles or dividend capture strategies. Similar to Baker and Wurgler (2004), we conclude shareholders are not indifferent to dividend policy. Instead, they value dividends most highly in during declining markets. In this way, our results provide support for the point in DeAngelo and DeAngelo (2006) that dividends are not irrelevant, in that we have found conditions where they matter that is robust to a large variety of controls and checks. Overall, shareholders in dividend-paying stocks do better than investors in non-dividend-paying stocks, particularly in market downturns. References Aharony, Joseph, Swary, Itzhak, 1980. Quarterly dividends and earnings announcements and stockholders' returns: an empirical analysis. J. Finance 35, 1–12. Allen, Franklin, Bernardo, Antonio E., Welch, Ivo, 2000. A theory of dividends based on tax clienteles. J. Finance 55, 2499–2536. Allen, Franklin, Michaely, Roni, 2003. Payout policy. In: Constantinides, G., Harris, M., Stulz, R. (Eds.), North-Holland Handbooks of Economics. Elsevier. Ang, Andrew, Chen, Joseph, 2002. Asymmetric correlations in equity portfolios. J. Financ. Econ. 63, 443–494. Ang, Andrew, Chen, Joseph, Xing, Yuhang, 2006. Downside risk. Rev. Financ. Stud. 19, 1191–1239. Baker, Malcolm, Wurgler, Jeff, 2004. A catering theory of dividends. J. Finance 59, 1125–1165. Bhattacharya, Sudipto, 1979. Imperfect information, dividend policy, and ‘The bird in the hand’ fallacy. Bell J. Econ. 10, 259–270. Black, Fischer, 1976. The dividend puzzle. J. Portfolio Manage. 2, 5–8. Black, Fischer, Scholes, Myron, 1974. The effects of dividend yield and dividend policy on common stock prices and returns. J. Financ. Econ. 1, 1–22. Blau, Benjamin M., Fuller, Kathleen P., 2008. Flexibility and dividends. J. Corp. Finance 14, 133–152. Blau, Benjamin M., and Kathleen P. Fuller, 2009, “Signaling, Free Cash Flow, and “Nonmonotonic” Dividends”, working paper, University of Mississippi. Blume, Marshall E., 1980. Stock returns and dividend yields: some more evidence. Rev. Econ. Stat. 62, 567–577. Brennan, Michael, 1970. Taxes, market valuation, and financial policy. Nat. Tax J. 23 (4), 417–429. Brennan, Michael J., Thakor, Anjan, 1990. Shareholder preferences and dividend policy. J. Finance 45, 993–1019. Brown, Stephen, Warner, Jerold, 1980. Measuring security price performance. J. Financ. Econ. 8, 205–258. Brown, Stephen, Warner, Jerold, 1985. Using daily stock returns: the case of event studies. J. Financ. Econ. 14, 3–32. Christie, William, 1990. Dividend yield and expected returns: the zero-dividend puzzle. J. Financ. Econ. 28, 95–125. Chung, Kee H., Pruitt, Stephen W., 1994. A simple approximation of Tobin's Q. Financ. Manage. 23 (3), 70–74. Connolly, Robert, Stivers, Chris, Sun, Licheng, 2005. Stock market uncertainty and the stock-bond return relation. J. Financ. Quant. Anal. 40, 161–194. Cook, Douglas O., Krigman, Laurie, Chris Leach, J., 2004. On the timing and execution of open market repurchases. Rev. Financ. Stud. 17 (2), 463–498. DeAngelo, Harry, DeAngelo, Linda, 2006. The irrelevance of the MM dividend irrelevance theorem. J. Financ. Econ. 79, 293–315. DeAngelo, Harry, DeAngelo, Linda, Skinner, Douglas J., 1992. Dividends and losses. J. Finance 47 (5), 1837–1863. DeAngelo, Harry, DeAngelo, Linda, Skinner, Douglas J., 2004. Are dividends disappearing? Dividend concentration and the consolidation of earnings. J. Financ. Econ. 72 (3), 425–456. DeBondt, W., Thaler, R., 1987. Further evidence on investor overreaction and stock market seasonality. J. Finance 42, 557–581. Denis, David J., Osobov, Igor, 2008. Why do firms pay dividends? International evidence on the determinants of dividend policy. J. Financ. Econ. 89, 62–82. Denis, David J., Denis, Diane K., Sarin, Atulya, 1994. The information content of dividend changes: cash flow signaling, overinvestment, and dividend clienteles. J. Financ. Quant. Anal. 29, 567–587. Easterbrook, Frank H., 1984. Two agency-cost explanations of dividends. Am. Econ. Rev. 74, 650–659. Elton, Edwin J., Gruber, Martin J., 1970. Marginal stockholder tax rates and the clientele effect. Rev. Econ. Stat. 52, 68–74. Elton, Edwin, Gruber, Martin, Rentzler, Joel, 1983. A simple examination of the empirical relationship between dividend yields and deviations from the CAPM. J. Bank. Finance 7, 135–146. Fama, Eugene F., French, Kenneth R., 1992. The cross-section of expected returns. J. Finance 47, 427–465. Fama, Eugene F., French, Kenneth R., 1993. Common risk factors in the returns on stocks and bonds. J. Financ. Econ. 33, 3–56. Fama, Eugene F., French, Kenneth R., 2001. Disappearing dividends: changing firm characteristics or lower propensity to pay? J. Financ. Econ. 60, 3–43. Fama, Eugene F., MacBeth, James D., 1973. Risk, return, and equilibrium: empirical tests. J. Polit. Econ. 81, 607–636. Goldstein, Michael A., Nelling, Edward F., 1999. REIT Return behavior in advancing and declining stock markets. Real Estate Finance 15, 68–77. Grinblatt, Marc, Han, Bing, 2005. Prospect theory, mental accounting, and momentum. J. Financ. Econ. 78, 311–339. Grullon, Gustavo, Michaely, Roni, 2002. Dividends, share repurchases and the substitution hypothesis. J. Finance 57, 1649–1684. Grullon, Gustavo, Michaely, Roni, Swaminathan, Bhaskaran, 2002. Are dividend changes a sign of firm maturity? J. Bus. 75, 387–424. Hong, Yongmiao, Jun Tu, and Guofu Zhou, 2003, “Asymmetric Correlation of Stock Returns: Statistical Tests and Economic Evaluation”, working paper, Washington University. Howe, K., He, J., Kao, G., 1992. One-time cash flow announcements and free cash-flow theory: share repurchases and special dividends. J. Finance 47, 1963–1975. Jagannathan, Ravi, Wang, Yong, 2007. Lazy investors, discretionary consumption, and the cross-section of stock returns. J. Finance 62, 1623–1661. Jensen, Michael C., 1986. Agency costs of free cash flow, corporate finance, and takeovers. Am. Econ. Rev. 76, 323–329. Jensen, Michael C., Meckling, William H., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. J. Financ. Econ. 3, 305–360. John, Kose, Williams, J., 1985. Dividends, dilution, and taxes: a signaling equilibrium. J. Finance 40, 1053–1070. Julio, Brandon, Ikenberry, David L., 2004. Reappearing dividends. J. Applied Corporate Finance 16, 89–100.

K.P. Fuller, M.A. Goldstein / Journal of Corporate Finance 17 (2011) 457–473

473

Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk. Econometrica 263–291. Kalay, Avner, Michaely, Roni, 2000. Dividends and taxes: a re-examination. Financ. Manage. 29 (2), 55–75. Keim, Donald B., 1985. Dividend yields and stock returns: implications of abnormal January returns. J. Financ. Econ. 14, 473–489. Lang, Larry H.P., Litzenberger, Robert H., 1989. Dividend announcements: cash flow signaling vs. free cash flow hypothesis? J. Financ. Econ. 24, 181–191. Lang, Larry H.P., Stulz, Rene M., Walkling, Ralph A., 1991. A test of the free cash flow hypothesis: the case of bidder returns. J. Financ. Econ. 29, 315–335. Lehn, Kenneth, Poulsen, Annette, 1989. Free cash flow and stockholder gains in going private transactions. J. Finance 44, 771–787. Lintner, John, 1956. Distribution of incomes of corporations among dividends, retained earnings, and taxes. Am. Econ. Rev. 46, 97–113. Lintner, John, 1965. The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets. Rev. Econ. Stat. 47, 13–37. Litzenberger, R., Ramaswamy, K., 1979. The effects of personal taxes and dividends on capital asset prices: theory and empirical evidence. J. Financ. Econ. 7, 163–195. Litzenberger, R., Ramaswamy, K., 1980. Dividends, short selling restrictions, tax induced investor clientele and market equilibrium. J. Finance 35, 469–482. Litzenberger, R., Ramaswamy, K., 1982. The effects of dividends on common stock prices: tax effects or information effects. J. Finance 37, 429–443. Markowitz, Harold, 1952. The utility of wealth. J. Polit. Econ. 60, 151–158. Markowitz, Harold, 1959. Portfolio Selection. Yale University Press, New Haven. Michaely, Roni, Thaler, Richard H., Womack, Kent L., 1995. Price reactions to dividend initiations and omissions: overreaction or drift? J. Finance 50, 573–608. Miller, Merton, Modigliani, Franco, 1961. Dividend policy, growth and the valuation of shares. J. Bus. 34, 411–433. Miller, Merton, Rock, Kevin, 1985. Dividend policy under asymmetric information. J. Finance 40, 1031–1051. Ng, Serena, 2007. Wall St. J. B4 Saturday/Sunday, April 14-15. Pettit, Richardson R., 1972. Dividend announcements, security performance, and capital market efficiency. J. Finance 27, 993–1007. Sharpe, William F., 1964. Capital asset prices: a theory of market equilibrium under conditions of risk. J. Finance 19, 425–442. Shefrin, Hersh, Statman, Meir, 1984. Explaining investor preference for cash dividends. J. Financ. Econ. 13, 53–282. Shefrin, Hersh, Statman, Meir, 1985. The disposition to sell winners too early and ride losers too long: theory and evidence. J. Finance 40, 777–790. Smith, Adam, 1776. The Wealth of Nations. http://www.bibliomania.com/2/1/65/112/frameset.html1776. Thaler, Richard, 1980. Toward a positive theory of consumer choice. J. Econ. Behav. Organ. 1, 39–60. Yoon, Pyung Sig, Starks, Larua, 1995. Signaling, investment opportunities, and dividend announcements. Review of Financ. Studies 8, 995–1018.