The Information Content of Dividends: Safer Profits, Not Higher Profits

Roni Michaely is Rudd Family Professor of Finance at Cornell University SC Johnson College of Business; Stefano Rossi is Professor of Finance and Gruppo Generali Chair in Insurance and Risk Management at Bocconi University; and Michael Weber is Assistant Professor of Finance and Cohen Keenoy Faculty Scholar at the University of Chicago Booth School of Business. This post is based on their recent paper.

Dividends represent one of the major financial decisions corporations make. Understanding both how capital markets evaluate dividends, and why firms pay dividends, are central to theories of asset pricing, portfolio allocation, capital structure, capital budgeting, cost of capital, and also to public economics, in particular regarding the effects of tax policy. Yet, despite extensive research, financial economists still do not fully understand why capital markets value dividends; why some firms pay dividends while others do not; or how a given firm’s payout policies are determined. Even firms with very similar observable characteristics such as age, earnings, and level of cash, display stark differences in terms of their dividend policies.

One credible and intuitive idea, dating back at least to Miller and Modigliani (1961), holds that dividend changes convey information about firms’ future prospects. This idea, later formalized by Miller and Rock (1985) and others, suggests that dividends conveys information, and might even signal future profits. Therefore, dividend changes should be followed by earnings or cash flow changes in the same direction. However, numerous empirical studies have failed to find evidence supporting this mechanism. In their review paper, DeAngelo, DeAngelo, and Skinner (2009) summarize the empirical evidence as follows: “we conclude that managerial signaling motives […] have at best minor influence on payout policy”. In this paper, we argue that prior theoretical and empirical work on dividend signaling might have barked at the wrong tree. Specifically, the literature has examined whether dividend changes signal changes in the level of future cash flows. We show that dividend signaling exist, but the signal is not about the expected level of cashflows (or earnings).

We show both theoretically and empirically that although dividends changes do signal, they signal future cash-flow volatility, and do not signal the level of future cash flows. If firms announce a dividend policy before current cash flows are realized, then firms are able to commit to a higher dividend, the lower their expected future cash flow volatility. Signaling is costly because, with imperfect access to capital markets, paying dividends comes with foregone investment opportunities, and higher-risk firms discover that imitating safer firms is too costly.

Our main prediction is that cash-flow volatility should decrease following a dividend increase, and should increase following a dividend decrease. Furthermore, larger dividend payments should carry more information, specifically, both larger decreases in cash-flow volatility and larger cumulative abnormal returns should be observed around the announcement of larger dividend increases.

Our model yields an additional and more nuanced cross-sectional prediction that speaks directly to the economic channel underlying our results. In our model, the cost of the signal is foregone investment opportunities. Consequently, following a dividend change, we expect a larger change in future cash-flow volatility for firms with smaller current earnings. The reason is that when the current earnings decrease, the foregone future investment opportunities at a given level of dividend increase. As a result, the same dollar of dividend should carry a larger information content for a lower earnings level.

As a byproduct, our model also helps understand the survey evidence that managers increase dividends when they believe the chance for future cuts are lower, e.g., Lintner (1956) and Brav et al. (2005). If dividend increases signal safer profits going forward, then firms can afford to keep future dividend payments stable after a dividend increase. Furthermore, if there is a cost to stopping dividend payouts, then the risk of incurring this cost is lower when earnings are more stable.

To test the model empirically, we need to estimate the volatility of cash flows, which presents two challenges. First, realized cash flows are non-stationary, implying that computing the standard deviation of realized earnings or cash flows will generate a biased estimate of cash-flow volatility. Second, we need a precise estimate of the volatility of cash flows, as opposed to, say, the volatility of assets or returns. To address these challenges, we borrow the methodology from asset pricing initially proposed by Campbell (1991) to study aggregate market return predictability. These studies argue that unexpectedly high returns follow positive news about higher future cash flows or news about lower future discount rates. Vuolteenaho (2002) extends this framework and applies it at the individual firm level. We follow Vuolteenaho (2002) to construct measures of cash flow and discount rate news and examine whether they vary around dividend events.

To implement this methodology, we begin by identifying four “dividend events” at the firm level: dividend increases and decreases (the intensive margin); and dividend initiations and omissions (the extensive margin). For each of these events, we estimate two firm-level vector auto regressions (VARs), one for the 60 months before the event and another for the 60 months after. These VARs first identify cash-flow and discount-rate news separately; we subsequently test whether cash-flow and discount-rate news following the dividend event differ from those before the event.

We find that the variance of cash-flow news is significantly lower after dividend increases and initiations, and the variance of cash-flow news is significantly higher after dividend decreases and omissions. Consistent with our theory, larger changes in dividends are associated with larger changes in cash-flow volatility in the expected direction, and announcements of larger changes in dividends are associated with larger cumulative abnormal returns in the same direction.

Importantly, we find that the same dollar of dividend paid is followed by a larger reduction in cash-flow volatility for firms with smaller current earnings. This result is consistent with our theory, in which dividends signal future cash-flow volatility, and the cost of the signal is foregone investment opportunities. The result is, however, inconsistent with an agency theory of dividends in which dividends themselves constitute “good news” because they come with lower private benefits of control.

What about the first moment of expected cash flows? Using our method, we revisit the earlier evidence on changes in the first moment of earnings following changes in dividends, and we confirm the earlier findings: corporate earnings generally do not change in the same direction as dividend changes, which is inconsistent with the traditional dividend signaling models. Furthermore, we find that discount-rate news does not change following dividend changes. Hence, any change in firm-level riskiness following dividend events relates exclusively to cash-flow volatility. This result reinforces the superiority of our approach relative to traditional measures of risk including beta (e.g., Grullon et al. (2002)). Our method can delineate the precise theoretical channel at play, while traditional measures fail to do so. Our results indicate that the evidence in Grullon et al. (2002) of a decrease in systematic risk–i.e., beta–following dividend increases is exclusively driven by changes in cash-flow volatility and not by changes in the riskiness of discount rates.

Finally, we examine share repurchases. Together with dividends, share repurchases constitute the firm’s overall payout policy. Prior empirical literature has documented several differences between dividend and share repurchase policies, most notably, the existence of substitution between them (e.g., Grullon and Michaely (2002)). In our framework, however, share repurchases are just yet another way to return cash to shareholders. As a result, we expect a pattern of changes in cash-flow volatility following share repurchases announcements similar to that following announcements of dividend increases and initiations. Consistent with our hypothesis, we find a strong decline in cash-flow volatility following share repurchase announcements and no changes in either the first moment of cash-flow news or discount-rate news. Also consistent with our hypothesis we find larger share repurchase programs associated with both larger reductions in cash-flow volatility and larger announcement returns. We conclude that announcements of changes to firms’ payout policy, whether through dividends or share repurchases, convey information about future changes in firms’ cash-flow volatility.

Our empirical approach presents four advantages relative to prior literature. First, by using a stock-return decomposition we directly measure variables that investors care about and that capital markets price. By contrast, approaches based exclusively on accounting information may measure variables that are not value-relevant and may thus be prone to criticism. Second, by using stock returns rather than realized earnings or cash flows, we are employing a framework that is not subject to the non-stationarity bias that arises when estimating cash flows from accounting information. Third, our approach also delivers a measure of the first moment of expected future cash flows, thereby allowing us to revisit prior empirical literature on the first moment of cash flows with a method whose parameters are market-based estimates. Fourth, a further benefit of our approach is the delivery of a measure of expected discount rates, thereby allowing us to test whether dividend changes convey information about changes in the firm’s discount rate.

Our results carry two implications. First, unlike prior literature we are able to support the claim that dividends signal firms’ future prospects: crucially, the signal is about expected cash-flow volatility, that is, the second moment of future cash flows and not the first moment. Signaling models of dividends were popular in the 1980s but they fell out of favor as the data did not support the models’ central prediction that the first moment of earnings should change in the same direction of dividend changes. Our paper shows that payout policy does convey information about cash-flow volatility in a manner consistent with signaling theory and inconsistent with a variety of alternative explanations. Second, while prior literature has documented that dividends and share repurchases have different features, we document a key shared attribute: both signal future changes in expected cash-flow volatility. One criticism of dividend signaling theories argues that many signaling models have been unable to account for the different features of dividend and share repurchases. We demonstrate that, with respect to future changes in cash-flow volatility, dividends and share repurchases convey similar information to the market.

The methodology we employ to measure the moments of the distribution of expected cash flows and discount rates, combined with our findings regarding firms’ conveying information about the second moment of future cash flows, suggest opportunities for future research exploring the motives of other corporate financial decisions using our approach. Our methodology may also be able to shed light on questions beyond finance. For example, a recent strand of economics literature has stressed ways in which aggregate uncertainty can affect firm investment dynamics. Researchers may now expand this line of reasoning to investigate the precise relevant source of firm-level uncertainty driving firms’ investment policies.

The complete paper is available for download here.

References

Brav, A., J. R. Graham, C. R. Harvey, and R. Michaely (2005). Payout policy in the 21st century. Journal of Financial Economics 77 (3), 483-527.

Campbell, J. Y. (1991). A variance decomposition for stock returns. The Economic Journal 101 (405), 157-179.

DeAngelo, H., L. DeAngelo, and D. J. Skinner (2009). Corporate payout policy. Foundations and Trends in Finance 3 (2-3), 95-287.

Grullon, G., R. Michaely, and B. Swaminathan (2002). Are dividend changes a sign of firm maturity? The Journal of Business 75 (3), 387-424.

Grullon, G. and R. Michaely (2004). The information content of share repurchase programs. The Journal of Finance 59 (2), 651-680.

Lintner, J. (1956). Distribution of incomes of corporations among dividends, retained earnings, and taxes. The American Economic Review 46 (2), 97-113.

Miller, M. H. and F. Modigliani (1961). Dividend policy, growth, and the valuation of shares. The Journal of Business 34 (4), 411-433.

Miller, M. H. and K. Rock (1985). Dividend policy under asymmetric information. The Journal of Finance 40 (4), 1031-1051.

Vuolteenaho, T. (2002). What drives firm-level stock returns? The Journal of Finance 57 (1), 233-264.

Both comments and trackbacks are currently closed.