Dividend smoothing and credit rating changes

Abstract

This paper examines the impact of credit rating changes on firms' dividend smoothing behavior, considering for the first time the ‘big three’ credit rating agencies (Standard and Poor's, Fitch and Moody's). Using a hand-collected sample of credit rating changes for firms listed at the S&P500 that are involved in dividend payments, we implement the traditional Lintner's [1956. “Distribution of Incomes of Corporations Among dividends, Retained Earnings and Taxes.” American Economic Review 46: 97–113] model and we initially verify the fact that firms smooth their dividend payments. Then we consider the effect of credit rating changes on smoothing behavior and we show the presence of an asymmetric impact on credit rating changes to dividend smoothing behavior. In particular, on average, a credit rating downgrade among any of the three credit rating agencies forces firms to engage in less smoothing, whereas a credit rating upgrade has only a marginal positive effect on dividend smoothing. Finally, our key results remain valid for firms with high level of financial pressure and under various robustness checks.

Keywords:

• Dividend smoothing
• credit ratings
• financial pressure

JEL Classifications:

• C23
• G24
• G3
• G35

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For example, credit ratings determine whether institutional investors such as banks or pension funds are allowed to invest in the company's securities.

2 Brav et al. (2005) argue that managers are willing either to abandon profitable investment opportunities with positive net present value (NPV) or to raise new capital from external market instead of reducing dividend payments. In addition, Larkin, Leary, and Michaely (2016) reports that market is willing to pay a premium to the shares of firms that pay a dividend consistently.

3 See Lintner (1956), Baker, Farrelly, and Edelman (1985), and Brav et al. (2005).

4 See Brav et al. (2005), Skinner (2008), Leary and Michaely (2011), Lambrecht and Myers (2012) and Larkin, Leary, and Michaely (2016).

5 In short, the ‘information hypothesis’ postulates that multiple ratings add value-relevant information. If CRAs do not perfectly agree or there is uncertainty about credit quality, while issuers have more accurately information about their own credit quality, according to the ‘shopping hypothesis’, the issuers will solicit for extra rating only if they are sure about the improvement to their average ratings. The ‘regulatory certification’ hypothesis points out issuers will solicit for extra rating due to precautionary motives, only in case there is fear to experience a future downgrade to high yield status. In this case the extra CRA will play the role of tiebreaker (see Bongaerts, Cremers, and Goetzmann 2012 for more details).

6 Asimakopoulos et al. (2017) show that a different information hypothesis (regarding the time dimension) matters for the investors and policy makers via a more detailed monitoring of the market, albeit in a different setup.

7 For further details see the National Association of Insurance Commissioners (NAIC) guidelines or the Basel II Accord.

8 Opler et al. (1999), Gao, Harford, and Li (2013) and Asimakopoulos, Asimakopoulos, and Fernandes (2019) have shown that firms' financial policies have a significant role in corporate cash management policies particularly, in maintaining the optimal level of cash.

9 Despite Lintner's (1956) work is so old, his findings still hold for many firms and more recent time periods (e.g. Fama and Babiak 1968; Brav et al. 2005; Leary and Michaely 2011).

10 Brav et al. (2005) argue that managers are prepared to raise external capital or even forego profitable investments in order to avoid cutting dividends.

11 Similarly, according to Kisgen (2006) because financial frictions are associated with credit ratings, they are a major concern to a firm manager when making decisions. As a result, the inclusion of credit ratings at the capital structure framework can improve the understanding and accuracy of capital structure behavior. He points out that companies that take into account credit rating upgrades or downgrades have fewer debts than companies that do not pay attention to credit rating changes.

12 Denis (2011) claim that firms with costly external financing can undertake valuable investments opportunities only by keeping larger cash reserves and as a consequence they alter their dividend smoothing behavior.

13 Alessi, Wolverson, and Sergie (2013) and Duff and Einig (2007) argue that Moody's and S&P jointly control 80% of the global market and Fitch further controls 15%.

14 A firm is assumed to pay dividends consistently when it pays a positive dividend every consecutive year for the period that it is included in our sample. Therefore, if a firm in our sample does not pay a dividend every year then it is excluded from our empirical analysis. However, we also checked our sample, before excluding firms that do not pay dividends every year consistently, and we could not find any firms that stopped paying dividends in a given year and then resumed their dividend payments in a later year.

15 We also show in the appendix a figure with the evolution of unique firms over time in our sample.

16 We would like to note that in our sample we the mean value of institutional ownership is at about 69.6% with 13.5% standard deviation. This indicates clearly that our firms experience high levels of institutional ownership and as a result we do not expect that ownership structure affects our dividend smoothing results.

17 This line of thought does not provide any expectation about the direction of dividend changes. From one side, according to pecking order the relationship between dividends and investment opportunities could be negative due to cash flow being used to finance investments (La Porta et al. 2000; Fama and French 2001; DeAngelo, DeAngelo, and Stulz 2006). On the other side, the relationship between investment opportunities and dividends could be positive considering that managers are reluctant to cut dividend in the future since dividend cuts are perceived as a bad signal to capital providers. As a result, according to the substitute theory firms' managers with growth opportunities are willing to distribute a higher dividend to attract capital providers as they are sure that they can retain the higher dividend in the future. Thus, the effect of growth to dividends is ambiguous.

18 In the appendix we also provide a table (Table 11) that summarizes our variable definitions.

19 In the appendix we also provide a table (Table 10) that shows the association between credit rating changes and dividend changes as numbers of firm-year observations.

20 We do not include any special dividends in our analysis.

21 We get even more pronounced differences between the two samples if we consider firms that changed their dividends more than a certain threshold, e.g. 30%, towards any direction. In this case we would avoid firms that change their dividends only slightly and consistently with a dividend smoothing behavior.

22 In this section, we consider any change in dividend as an indication of lower dividend smoothing. We acknowledge the fact that minor changes of dividends over time do not necessarily break the dividend smoothing behavior. We do consider this in our analysis later on when we assess specifically increases and decreases of dividends larger than a certain threshold resulted from our smoothing analysis in the next section. In this section, we just analyze a general impact of credit rating changes on dividend behavior.

23 Our reference to dividend changes is about dividend per share changes, $\mathrm{\Delta }DP{S}_{i,t}$.

24 Note that here we also drop Market-to-Book ratio due to collinearity with Tobins'Q.

25 Note that we take as given the established payout policy of each firm, focusing on dividend smoothing behavior instead.

26 See Lintner (1956), Fama and Babiak (1968), Brav et al. (2005), Farre-Mensa, Michaely, and Schmalz (2014) among others.

27 Following Becker and Milbourn (2011) arguments we have reduced our sample for Fitch on a sub-sample from 2000–2017 and we re-estimated equation (5(5) $\begin{array}{rl}{D}_{i,t}& =a_0+a_1{D}_{i,t-1}+a_2{E}_{i,t}+{\beta }_{1}Fitc{h}_{i,t}^{\mathrm{up}}{D}_{i,t-1}+{\beta }_{2}Fitc{h}_{i,t}^{\mathrm{down}}{D}_{i,t-1}\\ & +{\gamma }_{1}Fitc{h}_{i,t}^{\mathrm{up}}{E}_{i,t}+{\gamma }_{2}Fitc{h}_{i,t}^{\mathrm{down}}{E}_{i,t}+{\epsilon }_i\end{array}$(5) ). Our key results, as reported in Table 12 in the Appendix, remain robust.

28 Firms may hold cash instead to distribute them in the form of cash dividends, to protect themselves against the adverse cash flow shocks that might force them to miss investment opportunities due to costly external finance (Bates, Kahle, and Stulz 2009; Gao, Harford, and Li 2013).

29 Note that these results remain consistent even if we use the modified Lintner model as in Chen, Da, and Priestley (2012).

30 In addition, Benito and Young (2007) and Guariglia and Yang (2016) show that financial pressure in the form of debt-servicing costs has a negative effect on firms' employment and investment decisions.

31 Denis (2011) claims that firms with costly external financing can undertake valuable investments opportunities only by keeping larger cash reserves.

32 We have also experimented with different control variables and the results remain unaffected.

33 We need to use the nearest neighbor matching approach for total assets and market-to-book ratio because it is impossible to find two firms with the exact same level of assets and market-to-book ratio.

34 We also experiment with two lags on credit rating changes at the interaction terms and the results are similar with Table 7.

35 The lagged dividend interaction term with the lagged credit rating change is not statistically significant in any Model of Table 7.

36 We acknowledge the fact that some credit rating changes in our sample might happen towards the end of the year and after the announced dividends. However, our result do not seem to be affected by the existence of this possible time inconsistency showing that on average past credit rating changes do not affect current dividend smoothing. In addition, our main results are also confirmed via appropriate causality tests.

37 We also tested the impact of individual CRAs changes on dividend smoothing under different time periods and the results remain valid. However, we do not show the results here to save space.

38 We have also experimented with different threshold values (i.e. 25% or 35%) and our key results remain consistent.