Disappearing and reappearing dividends

https://doi.org/10.1016/j.jfineco.2021.06.029Get rights and content

Abstract

We decompose the decrease (1970s–2000) and subsequent recovery (2000–2018) in the fraction of dividend-paying firms. Changes in firm characteristics and proclivity to pay (probability of paying dividends conditional on characteristics) each drive half of the dividend disappearance. A higher proclivity drives 82% of the dividend reappearance. The remaining 18% is driven by a single characteristic: reduced earnings volatility. Changing characteristics are associated with low-profitability, high-earnings-volatility firms. Changing proclivity is associated with stable, profitable firms. Rather than dividend initiations or omissions, newly listed and delisted firms drive trends. Finally, the magnitude and duration of disappearing total payout is substantially smaller than that of dividends, indicating some substitution between dividends and repurchases.

Introduction

Payout policy is one of the most visible financial decisions a firm makes. Although these decisions are inherently important, they can also convey information to capital markets about a range of firms prospects, including growth opportunities (e.g.,  Miller and Rock, 1985), cash-flow volatility (Michaely et al., 2021), and corporate control (e.g.,  Jensen, 1986, Chetty, Saez, 2010). At the aggregate level, these changes have practical implications regarding factors such as equity value and discount rates (e.g.,  Shiller, 1981, Campbell, Shiller, 1988, Boudoukh, Michaely, Richardson, Roberts, 2007). Therefore, changes in payout policy have attracted significant attention from both market participants and researchers (for a comprehensive review of this literature, see  Farre-Mensa et al., 2014). For example, the discovery that a systemic shift in firms payout policies had begun in the late 1970s was met with justified interest and scrutiny.  Fama and French (2001) show a startling decline in the fraction of firms paying dividends (the phenomenon known as disappearing dividends) and  Grullon and Michaely (2002) show a systemic substitution of repurchases for dividends.

This paper extends and complements these earlier studies along several dimensions. We first show that the fraction of dividend-paying firms begins to increase around the turn of the century. We refer to this reversal of the disappearing dividends as “reappearing dividends.1 The central question in the paper is why the fraction of dividend-paying firms falls from 73% in 1978 to 23% in 2000 and then increases to 36% in 2018. Broadly speaking, we examine two not-mutually-exclusive possibilities. First, there could be changes in the relation between the fundamental characteristics and the decision to pay dividends; for example, firms dividend policy may have become less sensitive to changes in earnings. We define this as a decrease in the proclivity to pay dividends (probability of paying dividends conditional on firm characteristics). Second, there may have been changes in the distributions of fundamental characteristics, such as profitability and earnings volatility, known to be related to the decision to pay dividends. We then investigate the role of characteristics omitted from the empirical model. IPO year is an example of one such characteristic. If firms with IPOs later in the sample have a lower proclivity to pay dividends, the fraction of dividend payers will decrease as their presence increases. Finally, we examine changes in market conditions. For example, SEC rule 10-b made it easier for firms to repurchase shares in lieu of paying dividends and thus changed firms propensity to pay dividends.

We use a version of the logit model from Fama and French (2001), modified to include earnings volatility, to estimate the probability that a firm pays dividends. We find a substantial number of firms with characteristics similar to those of dividend payers in earlier years choosing not to pay dividends after 1977. This change in how firms set their dividend policy, or the changing proclivity, accounts for 47.38% of the disappearing dividends. The remaining 52.62% is driven by an influx of firms with characteristics, notably lower profitability and higher earnings volatility, that are different from those of previously listed dividend-paying firms. Naturally, the proclivity to pay dividends depends on the specific model being used. For example, we measure a greater change in the proclivity to pay dividends when we omit earnings volatility from the empirical model.

Reappearing dividends follow a different pattern from that of disappearing dividends. A substantial 81.72% of the reappearing dividends can be attributed to changes in the proclivity to pay dividends. Changing firm characteristics, especially earnings volatility, drive the remaining 18.28%. Except for earnings volatility, we find no reversal in the distribution of most firm characteristics during the 2000–2018 reappearing-dividends period. Although our sample of listed firms drops from a high of 4448 (1997) to 2187 (2018), the decline is not driven by unprofitable firms delisting at particularly high rates (consistent with  Grullon et al., 2019). Changes in most firm characteristics have less explanatory power in the reappearing-dividends phenomenon. The notable exception is changes in earnings volatility, which is an important factor during both the disappearing-dividend and reappearing-dividend eras.

A closer examination into changes in the proclivity to pay dividends reveals that firms listed after 1977 are responsible. Neither disappearing nor reappearing dividends occur in firms that had their IPO before 1978. Consistent with  Lintner (1956) dividend smoothing, firms rarely switch from payer to nonpayer status. However, a lower proclivity to pay dividends is visible particularly in high-profitability and low-earnings-volatility firms that were listed after the late 1970s. Disappearing dividends are then driven by the presence of new firms, with differing characteristics or proclivity to pay dividends. Reappearing dividends occur despite a consistently low number of dividend initiations. The appear to be driven by nonpaying firms delisting, especially firms with high earnings volatility and low proclivity to pay dividends. Firms cohort can thus be considered as a characteristic omitted from the empirical model.

Next, we find the substitution of repurchases for dividends accounts for much of the changing proclivity to pay dividends during the disappearing-dividends period. Compared with the gap between the actual and expected percentage of dividend payers, which reaches a trough of 21.7 percentage points (pp) and persists from the late 1970s until today, the gap between the actual and expected percentage of positive-payout firms is smaller in magnitude and shorter in duration. The gap exhibits a trough of 9.1 pp, and by 1998, the gap between actual and expected firms with positive payout closes. The smaller, shorter-lived gap between the actual and expected percentage of firms with positive payout suggests a smaller change in the total proclivity to pay out. This behavior is consistent with firms substituting share repurchases for dividends. However, firms do not substitute away from repurchases while dividends reappear. Instead, the percentage of firms repurchasing shares continues to increase throughout this time. In fact, repurchases are so popular that the actual number of firms with positive payout exceeds the expected number in almost every year since 2000. As a byproduct, the greater use of repurchases as a result of the regulatory change (rule 10b-18), as well as the impact of IPO cohort year, highlight the importance of changes in market structure and model specification (IPO cohort) on changes in proclivity over time.

Finally, we find that firms delisted from exchanges are the main reason for the reappearing dividends. The number of dividend-paying firms in the United States stays roughly constant from 2000 to 2018, while the total number of listed firms falls. We show that if these firms had not delisted, the percentage of firms that pay dividends would not have increased. Many of these non-dividend-paying firms delist due to a merger or acquisition. Indeed, when we construct a synthetic sample in which no mergers or acquisitions occur after 2000 (but delistings for other reasons still happen), we find that dividends reappear to only one-third of the extent that they actually did.

This paper extends our understanding of the dynamics of dividend policies through several new insights. We find that firm characteristics and the proclivity to pay dividends are each responsible for approximately half of the disappearing-dividends trend, and that both phenomena are attributable to firms listed after the late 1970s. Yet, since the turn of the 21st century, changing proclivity has accounted for most of the reappearance. We also find that most of the reappearing-dividends trend is driven by nonpaying firms delisting from exchanges. Combined with our decomposition of the reappearing-dividends trend, this implies that nonpaying firms that are expected to pay dividends delist at particularly high rates. The portion of the reappearing dividends that is explained by changing firm characteristics can be attributed to the high delisting rate of firms with high earnings volatility. Our findings also highlight the role of the repurchase-dividend substitution. When we consider dividends alone, modern-day firms ostensibly never fully recover the same proclivity to pay that we see in our sample in the 1960s and early 1970s. However, a different picture emerges once we account for repurchases: firms are more likely to pay out in the 2000s than they were before dividends disappeared.

We organize the rest of the paper as follows. In Section 2, we describe data and general trends. After estimating our baseline logit model (Section 3.1), we address how distributions of characteristics known to be related to the decision to pay dividends change over time (Section 3.2). Next, we ask how the relation between the fundamental characteristics in the logit model and the decision to pay dividends changes over time. Specifically, we quantify how much of dividend trends is due to changing characteristics vs. changing proclivity (Section 3.3) and what types of firms drive the changing proclivity to pay (Section 3.4). Finally, we discuss how other factors, such as potentially difficult-to-quantify firm characteristics or changes in the environment, affect the fraction of dividend-paying firms. We focus on the role of dividend persistence (Section 4.1), the impact of SEC rule 10b-18 (Section 4.2), and the high delisting rate after 2000 (Section 4.3).

Section snippets

Data, sample, and general trends

We begin with all firms in the CRSP/Compustat merged fundamentals annual database from 1963 to 2018 that traded on one of the major stock exchanges. We drop firm-year observations with either missing data, assets below $500,000, or book equity below $250,000. As per convention, we also exclude utilities (SIC 4900–4949) and financial firms (SIC 6000–6999). Finally, we analyze only common shares of U.S.-based publicly traded firms (CRSP share code 10 or 11). This filtering leaves 153,544 unique

Firm characteristics and the proclivity to pay dividends

Firms decision to pay dividends may change either because the firms have different characteristics or a different threshold at which they pay dividends. For example, dividend-paying firms can decide not to pay dividends because their profits are negative, a change in characteristics compared to firms listed earlier. Alternately, firms can choose to not pay despite being as profitable as previous dividend payers. That is, the threshold for paying dividends can change. We label this second reason

Model specification and changing market structure

In this section, we consider how other factors, such as firm characteristics omitted from the model or changes in the environment, affect the fraction of dividend-paying firms.

First, the empirical specifications in both  Fama and French (2001) and the main specification of this paper do not include past dividend status as one of the explanatory variables. The reason is that we want to quantify which characteristics are typical of firms that choose to pay dividends in the first place, and the

Discussion and conclusion

We examine the impact of changing firm characteristics and proclivity to pay dividends on dividend disappearance in the last quarter of the 20th century and dividend reappearance in the first two decades of the 21st century. Within our sample, changes in firm characteristics drive 53% of the dividend disappearance, and changes in the proclivity to pay dividends drive the remaining 47%. We find that, consistent with  Lintner (1956), few dividend-paying firms stop paying dividends. Rather, newer

References (19)

There are more references available in the full text version of this article.

Cited by (23)

  • Payout policy around the world

    2023, International Review of Financial Analysis
  • A bibliometric review of dividend policy literature

    2023, Research in International Business and Finance
  • Does stock market liquidity affect dividends?

    2022, Pacific Basin Finance Journal
    Citation Excerpt :

    Second, this study examines the dividend payout ratio in addition to the pay decision. This distinguishing measurement is important because the dividend literature observes that firms tend to smooth dividends over time, and the net payout yields have been increasing over time (Grullon et al., 2011), though the portion of firms paying dividends declined between the 1980s and the 2000s and then increased again (see, e.g., Farre-Mensa et al., 2014; Michaely and Moin, 2022). Therefore, not only might the results on the extensive margin not portray the full picture of the magnitude of firms' dividend policies, but the intensive margin (the amount of dividends that are paid) is also less vulnerable to time-series trends or latent variables than the extensive margin (the likelihood that dividends will be paid).

View all citing articles on Scopus

We appreciate helpful comments and suggestions from Ivo Welch, Rene Stulz, Kathleen Kahle, Nikolai Boboshko, Edward Mehrez, and seminar participants from Cornell University.

View full text