Declining Propensity To Pay A Re-Examination of The Lifecycle Theory

Journal of Corporate Finance 27 (2014) 345–366
Contents lists available at ScienceDirect
Journal of Corporate Finance

journal homepage: www.elsevier.com/locate/jcorpfin
Declining propensity to pay? A re-examination of the

lifecycle theory
Monica L. Banyi a,1, Kathleen M. Kahle b,⁎
a
McIntire School of Commerce, University of Virginia, United States
b
Eller College of Management, University of Arizona, United States
a r t i c l e i n f o a b s t r a c t
Article history: Our results indicate that the declining propensity to pay is a function of the changing composition
Received 10 September 2013 of firms over time and not a declining propensity in individual firms themselves. In particular, the
Received in revised form 10 June 2014 propensity to pay is greater than expected following the 2003 dividend tax cut. The decade a firm
Accepted 14 June 2014
went public is also a major determinant of its initial payout policy. Finally, while the strength of
Available online 22 June 2014
the relation between earned/contributed capital and payout propensity declines across IPO de-
cades, there is still a lifecycle effect — within a given IPO cohort, the likelihood of payout increases
JEL classification: as firms age.
G35
© 2014 Elsevier B.V. All rights reserved.
Keywords:
Dividends
Repurchases
Payout policy
Earned equity
Lifecycle theory
1. Introduction
Since Fama and French (2001), numerous papers have attempted to explain the declining propensity to pay dividends. DeAngelo
et al. (2006) propose a lifecycle theory of dividends, in which the probability that a firm pays dividends is positively related to its mix
of earned and contributed capital. They suggest that firms with low retained earnings as a proportion of total equity (RE/TE) or total
assets (RE/TA) tend to be in the capital infusion stage of their lifecycle and are less likely to pay dividends, whereas firms with high
RE/TE (or RE/TA) are more mature and profitable, and thus are more likely to pay dividends. Their results are consistent with this
hypothesis, but also indicate that firms whose earned equity makes them reasonable candidates to pay dividends have a reduction
in the propensity to pay that is twice that found in Fama and French (2001). But has the propensity to pay declined?
We begin by examining whether the DeAngelo et al. (2006) results hold not only for dividends, but also for repurchases. Julio and
Ikenberry (2004) document a pronounced shift away from dividends and towards repurchases beginning in 1992, and show that by
1997, the percentage of earnings paid out as repurchases exceeds the percentage paid out as dividends. Consistent with this shift, we
find a positive and monotonic relation between earned/contributed capital and both the fraction of firms that pay dividends and the
fraction that repurchase.
We next examine the influence of earned/contributed capital on payouts according to the decade in which the firm went public.
Even though DeAngelo et al. (2006) find that the earned/contributed capital mix is positively related to the payment of dividends, this
⁎ Corresponding author at: University of Arizona, Eller College of Management, 1130 E. Helen St., Tucson, AZ 85721-0109, United States. Tel.: + 1 520 621 7489;
fax: + 1 520 621 1261.
E-mail addresses: mlb6s@virginia.edu (M.L. Banyi), kkahle@eller.arizona.edu (K.M. Kahle).
1
Tel.: +1 434 243 1582.
http://dx.doi.org/10.1016/j.jcorpfin.2014.06.001
0929-1199/© 2014 Elsevier B.V. All rights reserved.
346 M.L. Banyi, K.M. Kahle / Journal of Corporate Finance 27 (2014) 345–366
relation might not hold equally for all firms. Brown and Kapadia (2007) find that the increase in idiosyncratic risk in the U.S.
stock market is due to new listings by riskier firms. Other papers (Fama and French, 2001; Grullon and Michaely, 2002; Julio
and Ikenberry, 2004; Skinner, 2008) provide evidence that these newly listed firms are less likely to pay dividends and more
likely to repurchase. Consequently, the relation between earned/contributed capital and dividends might decline in the 1980s
and 1990s due to the influx of recent IPOs, which are not only less profitable and riskier but whose managers are also less in-
clined to initiate dividends (Brav et al., 2005). Consistent with this hypothesis, we find that a firm's “starting point” matters.
The relation between earned/contributed capital and the likelihood of dividend payment remains constant over time for
those firms that went public in the 1970s or earlier. However, the more recently a firm has gone public, the less likely it is to
pay dividends in its initial years and the weaker the relation between RE/TA and dividend payment. Our results on repurchases
and total payout show similar trends.
We also examine the joint effects of IPO cohort and time on the relation between RE/TA and the fraction of firms that make share-
holder payouts. Our results indicate that while repurchases replace dividends as the preferred form of payout for firms that went pub-
lic in the 1980s or later, repurchases supplement the dividend policy of older firms. Further, we find that in spite of the fact that RE/TA
is not a good measure of lifecycle in recent IPO cohorts, a lifecycle effect does exist: within a given IPO cohort, the likelihood of making
shareholder payouts increases as firms age.
In multivariate logistic specifications, we continue to find that the impact of RE/TA on the likelihood of payout declines across IPO
decades. Our results do not simply reflect firm age. When we jointly examine the effect of firm age vs. the decade in which the firm
went public, we find that while the effect of RE/TA on payouts increases as firms age, indicating that older firms are more likely to
return cash to shareholders than younger firms, this effect is mitigated in firms that went public in the 1990s or later. Thus even
after controlling for age, the impact of RE/TA on payouts has declined in recent IPO cohorts. We also explore other factors that
might drive the payout propensity differences among IPO cohorts, including different methods of accounting for repurchases, omitted
risk variables, and executive compensation. However, none of these explanations fully resolves the issue.
Given the decrease in the impact of RE/TA on payouts over time and IPO decade, we next examine the effect of these issues on the
predictive accuracy of payout models. We find that the predictive ability of our payout model also declines across IPO decades. While
the model does a fairly good job of predicting the total number of dividend payers, it underpredicts the number of payers in the pre-
1970s cohort but overpredicts payers in the 1990s cohort. Further, the model is increasingly less able to correctly identify which firms
pay dividends. For example, whereas 82.1% of actual payers are predicted to pay in the pre-1970s cohort, only 49.9% of actual payers
are predicted to pay in the 1990s cohort. Using an estimation period that occurs after the SEC enacted Rule 10b-18 improves the
prediction accuracy, but within each RE/TA category, the accuracy of the model still declines across IPO decades.
Finally, given our finding that prediction accuracy improves when using an estimation period after the enactment of Rule 10b-18,
we reexamine whether the conditional propensity to pay has decreased over time. Other researchers (Skinner, 2008; Grullon and
Michaely, 2002; Grullon et al., 2011) find a declining propensity to pay dividends. However, these papers either examine an uncon-
ditional propensity to pay that does not allow for the changing composition of firms over time or model the conditional propensity to
pay using an estimation period prior to the enactment of Rule 10b-18, when open market repurchases were less common and divi-
dends the norm. After controlling for changing firm characteristics and using an estimation period after 1982, we find little evidence
of a wide-spread deficit in the propensity to pay. These results hold for both dividends and total payouts and also hold not only over
time but also by IPO decade, firm age, and tax regime. In fact, using our model, we find that the actual number of payers is greater than
predicted after the 2003 dividend tax cut. Overall our results indicate that the declining propensity to pay found in earlier papers is
largely a function of the change in the composition and characteristics of firms in the population, combined with regulatory and
tax regimes that have altered payout preferences, and not a declining propensity in the individual firms themselves across time.
Our finding that payout preferences have changed over time in response to regulatory and tax regimes is not inconsistent with
the catering theory of Kulchania (2013) and Jiang et al. (2013). However, we still find strong evidence of a lifecycle effect in that
the likelihood of payout increases as firms mature.
Our paper proceeds as follows. Section 2 provides a review of the relevant literature. Section 3 contains our sample selection and
examines the impact of repurchase accounting method on RE/TE and RE/TA. Section 4 examines the effect of time and IPO decade on
the ability of RE/TA to predict payout policy. Section 5 examines the prediction accuracy of our model and evaluates the propensity to
pay across time, IPO decade, age quintiles, tax regimes, and option intensity. Section 6 concludes.
2. Literature review and hypotheses
2.1. The lifecycle theory
There is no formal model that relates a firm's lifecycle to its dividend policy. However, Grullon et al. (2002) suggest the following:
as firms mature, their investment opportunity sets become smaller, resulting in a declining rate of reinvestment and a decrease in
growth and risk. This declining reinvestment gives rise to excess cash that can be paid out to shareholders. Consistent with this
hypothesis, they find that firms that increase (decrease) dividends experience significant decreases (increases) in systematic risk.
DeAngelo et al. (2006) use the lifecycle theory to examine the disappearing dividend phenomena of Fama and French (2001). They
find that the fraction of firms that pays dividends is high when retained earnings comprise a large portion of total equity (or total
assets) and falls to near zero when equity includes mostly contributed rather than earned capital. DeAngelo et al. conclude that
their findings are consistent with a lifecycle theory in which the probability that a firm pays dividends is positively related to its
mix of earned and contributed capital.
M.L. Banyi, K.M. Kahle / Journal of Corporate Finance 27 (2014) 345–366 347
Multiple corporate finance theories suggest that repurchases are close substitutes for dividends.2 If repurchases substitute for div-
idends over time, then the motives for repurchases should be similar to the motives for dividends. Consequently, we examine whether
the lifecycle theory of dividends extends to repurchases and total payouts. If the motives behind repurchases are similar to those of
dividends, we expect a positive relation between repurchases (or total payouts) and earned/contributed capital. Alternatively, if
repurchases are used to distribute transient shocks (Guay and Harford, 2000; Jagannathan et al., 2000), they may be used by younger,
less profitable firms and the relation may be muted.
2.2. Repurchases, lifecycle, and the propensity to pay
Recent literature documents a decline in the propensity to pay dividends.3 While the lifecycle theory appears to explain the rela-
tion between the dividend payout decision and the earned/contributed capital mix, it does not explain the declining propensity to pay.
Julio and Ikenberry (2004) suggest that one contributing factor for the decreasing percentage of dividend payers in the 1990s is the
increased proportion of newly public firms during this time period. Since IPO firms tend to be young, risky firms with abundant in-
vestment opportunities but limited resources, they are unlikely to commit to paying dividends. If these firms have excess cash,
they prefer to distribute this cash using share repurchases. In fact, Brav et al. (2005) find evidence that while dividend paying firms
view repurchases as a substitute, repurchasing firms do not see dividends as a close substitute, and that most firms would prefer to
repurchase if free of their dividend history. Three-quarters of the non-dividend paying firms in their survey indicate that they
might never initiate dividend payments. Likewise, Skinner (2008) suggests that firms that only pay dividends are largely extinct,
and repurchases are the dominant form of payout.
We hypothesize the following: the relation between dividends and earned/contributed capital decreases over time; this decrease
is driven largely by the increasing proportion of recent IPO firms exhibiting characteristics associated with lower propensities to pay
dividends and with stronger preferences for share reacquisitions.
3. Data and methodology
3.1. Sample selection
Our sample selection procedure mirrors DeAngelo et al. (2006). We select industrial firms (specifically, firms outside of SIC codes
4900–4999 and 6000–6999) which trade on the NYSE, Nasdaq, and Amex exchanges (i.e. CRSP exchange codes 1, 2, or 3) and are in-
corporated in the U.S. according to Compustat. As in DeAngelo et al. (2006), we begin our sample in 1973. Our sample ends in 2011,
whereas DeAngelo et al. (2006) end in 2002. For our initial analyses, we exclude any observation with missing data on income before
extraordinary items (IB), dividends (DVC), retained earnings (RE), or assets (AT) on Compustat. Purchases of common stock (PRSTKC)
is set equal to zero if it is missing or has a combined data code, although our results are robust to deleting these observations. This
procedure results in 15,291 unique firms and 151,356 firm-year observations over our sample period. We then form three categories
of payers based on cash distributions in year t: Dividend Payers are firms with non-zero common dividends (DVC) and comprise 34.8%
of our sample; Repurchase Firms are firms with non-zero purchases of common stock (PRSTKC) and comprise 32.1% of our sample;
and Payout Firms are either Dividend Payers, Repurchase Firms, or both and comprise 51.5% of our sample.4
We place further constraints on the sample in our IPO decade and logit analyses. For the IPO decade analysis, we collect IPO dates
and initial listing dates from CRSP. We use the earliest of these as an estimate of the year in which the firm went public. When
examining firm age, we use the founding date to calculate a firm's age each year.5 In the logit analysis, we require that all control
variables are available, which results in the loss of an additional 21,436 firm-year observations. We winsorize all variables at the
1st and 99th percentiles. See Appendix A for a list of the control variables and their definitions.
3.2. Time periods and IPO cohorts
We divide firms into time periods by decade and into IPO cohorts based on the decade in which the firm went public. We believe
that decades delineate our sample for several reasons. First, Nasdaq firms were added to CRSP and Compustat in the early 1970s,
which resulted in a substantial change in the composition of firms available from these data sources. Multiple events occurred around
1980 that lead us to believe the 1980s are different than the 1970s. First, in 1979, revisions to the Employee Retirement Income
2
John and Williams (1985) suggest that firms may switch from dividends to repurchases since repurchases are more tax effective. In agency models, such as Jensen
(1986), and signaling models, such as Miller and Rock (1985), repurchases and dividends play a similar role. Empirically, both Grullon and Michaely (2002) and Skinner
(2008) find evidence of a substitution effect and suggest that share repurchases are now the dominant form of payout.
3
For example, see Fama and French (2001) and DeAngelo et al. (2004).
4
Grullon et al. (2011) examine net payouts, defined as dividends plus repurchases minus equity issuance. However, Fama and French (2005) document that shares
issued to employees comprise a large portion of the “equity issuance” captured by Compustat and that seasoned equity offerings themselves are rare. Since we are in-
terested in payout decisions that are under a firm's control, we examine payout rather than net payout. We also examine an alternate definition of payout. Specifically,
we follow Grinstein and Michaely (2005) and use a two-year window to define repurchasing firms; thus repurchasing firms are firms that repurchase in either the cur-
rent or previous year. Using this alternate definition increases the percent of firms that repurchase (payout) but does not change our results.
5
IPO and founding dates are from SDC, Jay Ritter's website (http://bear.cba.ufl.edu/ritter/ipodata.htm), and the Jovanovic and Rousseau (2001) data. We thank Amy
Dittmar for providing us with the Jovanovic and Rousseau data. CRSP begins coverage of Nasdaq stocks on December 14, 1972. For Nasdaq stocks with this CRSP list date,
we examine Compustat data for the earliest year of coverage. If Compustat has data on the firm prior to 1970, we consider it to have gone public in the 1960s; otherwise
it is classified as a 1970s IPO firm.
Security Act of 1974 (ERISA) allowed pension funds to provide venture capital, erasing any doubt that non-dividend paying firms were
acceptable investments. Second, the promulgation of Rule 10b-18 in 1982 provided safe harbor provisions for firms against stock ma-
nipulation charges if certain volume, timing, and price conditions were observed in the share repurchase transaction; this regulation
resulted in a gradual increase in repurchases throughout this decade. Finally, Fama and French (2001) show that the universe of firms
tilts towards small, unprofitable, high market-to-book firms after 1980. The combination of these changes led to a dichotomy: firms
going public before 1980 operated in a “dividend is expected” world and felt pressure to initiate sticky dividend payments which they
could not alter without facing market consequences. Firms that went public in the 1990s, however, had more flexibility to return
shareholder capital by other, more efficient means.
The 1990s differ from the 1980s in that the number of IPOs, particularly VC-backed and technology IPOs, increased substantially.
The environment changed again in March of 2000 when the tech bubble burst. Numerous small, unprofitable dot-coms ran out of cap-
ital and were acquired or liquidated, and the number of IPOs fell drastically. Gao et al. (2012) show that this drop in IPO volume was
concentrated in small firms: between 1980 and 2000, an average of 165 small firm IPOs occurred each year; this average fell to 30 per
year between 2001 and 2009. Meanwhile, the average age of firms on their IPO dates increased substantially relative to the 1990s (Jay
Ritter, “IPO statistics for 2011 and earlier years”), which suggests that the 2000s cohort of firms may have greater means to initiate
shareholder payouts than their predecessors in the 1990s. Finally, the 2003 JOBS Act provided tax incentives to pay dividends.
Appendix B provides further justification for our splits. This table replicates Table 1 of Baker and Wurgler (2004) but uses our re-
gression sample screens and updates the data through the 2000s. We first divide the firms into dividend paying and non-paying firms.
We then examine the number of old firms and the number of new lists in each category. New lists are firms that enter our sample in
that year. The number of new lists among non-dividend paying firms increases substantially across decades: It averages 72 per year in
the 1970s, increases to 322 in the 1980s and 385 in the 1990s, and drops to 155 in the 2000s. In comparison, new lists by dividend
payers fall over time. We also examine the percentage of new lists that pay dividends (ListPay). From 1975 to 1979, 42.8% of new
lists pay dividends, similar to the 41.7% reported by Baker and Wurgler (2004) over the 1970s. This is a substantial drop from the
61.1% reported in Baker and Wurgler (2004) for the 1960s. The percentage of new lists in our sample that pay dividends falls to
13.7% in the 1980s, 8.1% in the 1990s, and 7.6% in the 2000s. Baker and Wurgler (2004) show a similar decline in their dividend ini-
tiating percentages across time. Overall, these numbers are consistent with the fact that the substantial economic and institutional
changes have impacted firms' payout policies across the past three decades.
3.3. Accounting for repurchases and retained earnings
In the Introduction, we discuss how the advent of open market repurchases and the changing composition of firms may contribute
to the decline over time in the ability of earned/contributed capital to predict dividends. A third potential reason for this decline is
related to the two different accounting methods used to record share repurchases and their varying impacts on reported retained
earnings. Generally accepted accounting principles allow firms to use one of two methods to account for repurchases: the retirement
method or the treasury stock method.6 Though the two methods have the same impact on total book value of equity and shares out-
standing, each method has a very different mechanical impact on the relations among the equity accounts and specifically the RE/TE
and RE/TA ratios. These mechanical impacts bias RE/TE to a greater extent than RE/TA, especially for firms with low common equity.7
The end result is that while many firms report negative retained earnings due to the accumulation of accounting losses over the firm's
life, even mature, profitable firms can report negative retained earnings. For example, Microsoft reported negative retained earnings
between 2005 and 2012, largely as a result of its continued repurchase policy. Microsoft is incorporated in Washington State which
does not allow treasury stock; thus, Microsoft must retire repurchased shares and reduce paid-in capital and retained earnings for
the repurchase cost, thereby reducing RE, TE, and TA. Dell is also known for making large repurchases but reports positive retained
earnings. Unlike Microsoft, Dell uses the treasury stock method and records the purchase price of the reacquired shares in a treasury
stock account, which reduces Dell's total equity (TE) and total assets (TA) but does not affect reported RE. Unfortunately, the choice
between the two accounting treatments is more than likely not random, as it is influenced by the firm's chosen state of incorporation.
6
Treasury stock accounts were not separately captured in Compustat until after 1982, around the time of the enactment of the Rule 10b-18 Safe Harbor. However,
many firms, particularly those incorporated in states which follow the Model Business Corporation Act (MBCA), are legally precluded from using a treasury stock ac-
count. The MBCA was created by the American Bar Association in 1950. The Act operates as a template for other states to use when deciding corporate law. This Act
has been adopted in some form by 24 states, although California, New York, and Delaware do not follow the MBCA. California corporate law, however, still precludes
the use of a treasury stock account. The major “competitor” to the MBCA is the Delaware General Corporation law. We find that 52.7% of our repurchasing firm-year
observations report a non-zero treasury stock account.
7
If a firm retires repurchased shares, it should reduce the capital accounts by the initial issue price of the stock and retained earnings for any difference between the
original issue price and the requisition price. Any reduction in retained earnings is most likely less than the corresponding decrease in total equity, as the latter absorbs
the reduction in both the capital account and retained earnings account. For a repurchasing firm that reports an initial RE/TE relation between 0 and 1, the repurchase
transaction reduces RE/TE, but this reduction is generally less than an equivalent cash dividend (which reduces retained earnings and total equity by the same amount).
If a firm uses the treasury stock method, the firm records the shares repurchased as treasury stock at the market price paid for the reacquired shares. This action reduces
total book value of equity but has no mechanical impact upon retained earnings, so firms using a treasury stock account report higher retained earnings (RE) relative to a
similar firm using the retirement method. Thus, for a treasury stock firm reporting an initial RE/TE relation between 0 and 1, a repurchase transaction mechanically in-
creases RE/TE, since the accounting transaction reduces the total book value of equity with no corresponding change to the retained earnings account (i.e. RE is un-
changed as TE declines). In such situations, RE/TE is biased upwards, even though economically, the firm is reducing this relation by returning invested capital and
accumulated earnings to the selling shareholders. RE/TA also increases since RE is unchanged while assets decline when cash is used to repurchase the shares. However,
since assets generally are larger than total equity, the mechanical increase is comparatively smaller. Consistent with these biases, we find that 26.9% of all repurchasing
observations with non-zero treasury stock accounts have RE/TE N0.90, while only 13.0% of firms without treasury stock accounts have RE/TA N0.90.
In conclusion, a firm's use of the retirement method is problematic for researchers who expect that the firm's retained earnings
account reflects its earned equity less dividends paid to common shareholders and that the firm's treasury stock account captures
repurchased shares. Retirement firms will report lower contributed capital, lower retained earnings, and zero treasury stock. Although
DeAngelo et al. (2006) focus on the ratio of earned equity to total equity (RE/TE) to proxy for lifecycle, we purport that the use of RE/TA
is less impacted by mathematical distortions than RE/TE for firms with low total common equity; RE/TA also allows the inclusion of
negative common equity firms. Thus, we use RE/TA as our primary measure of lifecycle in the remainder of our tests and analyses.
In untabulated results, when we use RE/TE, our results are not qualitatively altered. Following DeAngelo et al. (2006), we refer
to RE/TA as earned versus contributed capital although we recognize that this does not completely depict the relation in the two
components of equity.
4. Univariate results of RE/TA and lifecycle
4.1. Payout as a function of RE/TA
Table 1 examines the proportion of firms that distribute cash in the form of dividends, repurchases, or either, based on the ratio of
retained earnings to total assets. In each year from 1973 to 2011, we calculate the proportion of payers for firms grouped by ratio level,
starting with firms that have negative RE/TA ratios and moving up in increments of 0.10. The numbers reported are the medians of the
annual proportions over 1973–2011 and the median number of firms in each category over the 39-year period. The first two rows
show a strong monotonic and positive relation between the proportion of firms that pay dividends and their earned/contributed cap-
ital mix. Only 4.1% of firms with negative RE/TA pay dividends. The percent of firms paying dividends rises to 21.8% for RE/TA between
0.00 and 0.10, and then rises steadily for each subsequent 0.10 ratio group through RE/TA between 0.80 and 0.90. As in DeAngelo et al.
(2006), the percent of dividend payers declines slightly when RE/TA is 0.90 or greater.
The next two rows of Table 1 examine the proportion of firms that repurchase. Again, there is a strong monotonic relation between
the proportion of firms that repurchase and their earned/contributed capital mix. Compared to the dividend results, firms with
negative RE/TA are more likely to repurchase. Over 17% of firms with negative RE/TA repurchase shares. This number rises steadily,
though less quickly when compared to the dividend results, for each subsequent 0.10 increase in RE/TA, reaching 76.0% when
RE/TA is 0.90 or greater. The high proportion of repurchasing firms in low RE/TA categories is consistent with our discussion in
Section 3.3, which describes how the use of the retirement method by repurchasing firms biases RE downward, and in the extreme,
causes otherwise profitable firms to report negative RE.
The final two rows of Table 1 examine the proportion of firms that pay dividends and/or repurchase. Over 19% of firms with
negative retained earnings either pay dividends, repurchase, or both. This number increases to 44.0% for RE/TA between 0.00 and
0.10 and increases steadily for each subsequent category. Over 90% of firms with RE/TA greater than or equal to 0.70 make some
sort of cash payout.
4.2. Payout as a function of RE/TA and IPO decade
We next examine the impact of a firm's IPO date on the proportion of firms that pay out. Table 2 divides our sample firms into
groups based on the decade in which they went public. As shown in Panel A, firms that went public before 1970 are more likely to
pay dividends than the overall sample in Table 1, regardless of the ratio of earned to contributed capital. Over 48% of firms with
RE/TA between 0.00 and 0.10 pay dividends. This proportion increases to 67.6% for firms with RE/TA between 0.10 and 0.20 and
increases steadily to 100% when RE/TA is between 0.80 and 0.90, before falling slightly in the highest RE/TA group. Firms that went
public in the 1970s are also more likely to pay dividends than the overall sample in Table 1, although the difference is not as large
as for the pre-1970s IPO firms. Firms that went public in the 1980s are less likely to pay dividends at each level of RE/TA than in
the overall sample. However, the proportion still increases almost monotonically as RE/TA increases, reaching a peak of 70.0%
Table 1
Proportion of firms making payouts based on the ratio of retained earnings to assets.The sample includes firm-year observations from 1973 to 2011 for U.S. incorporat-
ed, industrial firms with CRSP exchange codes 1, 2, or 3 and non-missing data on Compustat for dividends (DVC), assets (AT), retained earnings (RE), common equity
(CEQ), and earnings before taxes (IB). Firms are allocated annually to categories of RE/TA based upon the level of earned equity divided by total assets. We report the
results for three samples: Dividend Payers are firms with non-zero common dividends; Repurchasers are firms with non-zero purchases of common stock (PRSTKC);
and Payout Firms are members of Dividend Payers and/or Repurchasers. For each year, we calculate the number of firms making the respective payout and divide this by
the total number of firms. The median values for the resulting 39 observations for each RE/TA category in each payout sample are reported.
Ratio of retained earnings to total equity (RE/TA)
b0.00 0.00–0.10 0.10–0.20 0.20–0.30 0.30–0.40 0.40–0.50 0.50–0.60 0.60–0.70 0.70–0.80 0.80–0.90 0.90+
Proportion Dividend Payers 0.0413 0.2177 0.3343 0.4312 0.5129 0.5726 0.6546 0.7419 0.7922 0.8235 0.7179
Total number of firms 1496 350 491 500 388 282 184 126 63 23 22
Proportion Repurchasers 0.1734 0.2756 0.3358 0.3703 0.4102 0.4693 0.5000 0.5191 0.5526 0.6216 0.7600
Proportion Payout Firms 0.1938 0.4400 0.5500 0.6327 0.7114 0.7829 0.8269 0.8696 0.9091 0.9412 0.9487
when RE/TA is between 0.80 and 0.90. Compared to the 1970s IPO cohort, the proportion paying dividends is substantially lower at
each level of RE/TA. This difference is likely due to both the ability to repurchase shares after 1982 and the stickiness of dividends
once such payments commence. The proportion of firms paying dividends at each level of RE/TA continues to fall for firms that
went public in the 1990s; for these firms, the proportion paying dividends is greater than 50% only when RE/TA is between 0.70
and 0.90. Very few firms that went public in the 2000s have RE/TA ratios greater than 0.30, so it is difficult to draw conclusions
from this group. Overall, Panel A of Table 2 shows that the relation between dividends and earned/contributed capital is weaker for
firms that went public more recently. Specifically, young firms are unlikely to pay dividends even if their RE/TA is high, while older
firms are likely to pay dividends even if RE/TA is low.
Table 2
Proportion of firms making payouts based on IPO decade and RE/TA.The sample includes firm-year observations from 1973 to 2011 for U.S. incorporated, industrial
firms with CRSP exchange codes 1, 2, or 3 and non-missing data on Compustat for dividends (DVC), assets (AT), retained earnings (RE), common equity (CEQ), and
earnings before taxes (IB). Firms are allocated annually to categories of RE/TA based upon the level of earned equity/assets. We report the results for three samples:
Dividend Payers are firms with non-zero common dividends; Repurchasers are firms with non-zero purchases of common stock (PRSTKC); and Payout Firms are mem-
bers of Dividend Payers and/or Repurchasers. Within each IPO decade, we annually calculate the number of firms making the respective payout and divide this by the
total number of firms. The median values for the resulting annual observations in each RE/TA category are reported.
Ratio of earned equity to total assets (RE/TA)
b0.00 0.00–0.10 0.10–0.20 0.20–0.30 0.30–0.40 0.40–0.50 0.50–0.60 0.60–0.70 0.70–0.80 0.80–0.90 0.90+
Panel A: Dividend Payers

IPO pre-1970s (39 years)
Proportion 0.1508 0.4805 0.6761 0.7787 0.8511 0.9065 0.9032 0.9403 0.9429 1.0000 0.9167
Number of firms 106 72 97 121 113 86 69 50 24 10 11
IPO 1970s (39 years)
Proportion 0.0593 0.2609 0.3926 0.4930 0.5758 0.6364 0.6842 0.7619 0.7879 0.8333 0.8333
Number of firms 140 55 78 76 82 64 46 34 19 6 5
Proportion 0.0403 0.1666 0.2723 0.3060 0.3384 0.3835 0.4530 0.5264 0.5896 0.7000 0.5000
Number of firms 417 130 126 109 90 64 40 27 17 7 9
Proportion 0.0316 0.1673 0.1696 0.1403 0.2152 0.2409 0.3646 0.4000 0.5736 0.5357 0.3934
Number of firms 809 150 172 145 91 66 45 30 13 7 8
Proportion 0.0481 0.1850 0.1576 0.1889 0.2209 0.2154 0.2000 0.2154 0.7000 0.3500 0.5833
Number of firms 525 98 68 45 19 13 7 5 5 2 2
Panel B: Repurchasers
Proportion 0.2051 0.3333 0.4124 0.4265 0.4915 0.5943 0.6111 0.6078 0.5926 0.5714 0.7386
Number of firms 106 72 97 121 113 86 69 50 24 10 11
Proportion 0.1860 0.3125 0.3246 0.3409 0.3671 0.3913 0.4889 0.5143 0.5385 0.7500 0.8000
Number of firms 140 55 78 76 82 64 46 34 19 6 5
Proportion 0.1718 0.3116 0.3392 0.3825 0.3972 0.4525 0.4741 0.5125 0.5240 0.6000 0.7639
Number of firms 417 130 126 109 90 64 40 27 17 7 9
Proportion 0.2031 0.3445 0.3845 0.4380 0.4690 0.4875 0.5921 0.6578 0.8153 1.0000 1.0000
Number of firms 809 150 172 145 91 66 45 30 13 7 8
Proportion 0.2264 0.3144 0.3876 0.3246 0.4022 0.4083 0.3636 0.6905 0.6000 0.3000 0.7500
Number of firms 525 98 68 45 19 13 7 5 5 2 2
Panel C: Payout Firms

Proportion 0.3036 0.6250 0.7813 0.8523 0.8989 0.9398 0.9518 0.9643 0.9677 1.0000 1.0000
Number of firms 106 72 97 121 113 86 69 50 24 10 11
Proportion 0.2284 0.4153 0.5769 0.6667 0.7230 0.7826 0.8462 0.8571 0.9048 1.0000 1.0000
Number of firms 140 55 78 76 82 64 46 34 19 6 5
Proportion 0.2002 0.3960 0.5250 0.5181 0.5947 0.6270 0.6882 0.7462 0.8417 1.0000 0.8869
Number of firms 417 130 126 109 90 64 40 27 17 7 9
Proportion 0.2238 0.4216 0.4623 0.5051 0.5674 0.5793 0.7230 0.7883 0.9024 1.0000 1.0000
Number of firms 809 150 172 145 91 66 45 30 13 7 8
Proportion 0.2480 0.4130 0.4637 0.4919 0.5188 0.4664 0.4444 0.6905 1.0000 0.8000 1.0000
Number of firms 525 98 68 45 19 13 7 5 5 2 2
Table 3
Frequency and form of payouts by IPO decade and time period.The sample includes U.S. incorporated, industrial firms from 1973 to 2011 with CRSP exchange codes 1, 2,
or 3 and nonmissing data on Compustat for dividends (DVC), assets (AT), retained earnings (RE), common equity (CEQ), and earnings before taxes (IB). Dividend Payers
are firms with non-zero common dividends; Repurchasers are firms with non-zero purchases of common stock (PRSTKC); and Payout Firms are members of Dividend
Payers and/or Repurchase Firms. Percentages represent the percentage of firms making the respective form of corporate payout, sorted by IPO decade and time.
N Percentage of firms
Dividend Payer Repurchaser Payout Firm
IPO pre-1970
1973–1979 12,896 74.8 31.5 80.8
1980–1989 11,892 75.4 39.2 82.0
1990–1999 7,305 68.5 49.0 77.4
2000–2011 5,006 70.3 55.9 80.2
All years 37,099 73.1 40.7 80.4
IPO during the 1970s
1973–1979 9,801 48.2 27.2 60.8
1980–1989 11,485 42.5 29.2 56.5
1990–1999 5,493 42.9 36.0 58.6
2000–2011 3,073 47.1 48.5 66.7
All years 29,852 44.9 31.8 59.3
1980–1989 15,245 12.2 21.6 29.7
1990–1999 15,640 19.1 27.7 38.3
2000–2011 7,119 28.5 42.2 54.0
All years 38,004 18.1 27.9 37.8
1990–1999 19,293 9.6 21.9 28.2
2000–2011 18,213 13.9 37.2 42.5
All years 37,506 11.7 29.3 35.1
2000–2011 8,895 9.9 26.8 32.2
Panel B of Table 2 examines the relation between IPO decade, earned/contributed capital, and the proportion of repurchasing firms.
As with dividend payers, the proportion of repurchasing firms generally increases monotonically within each IPO cohort. However,
unlike dividend payers, the proportion does not decrease across the IPO cohorts. For example, when RE/TA is between 0.20 and
0.30, the proportion of repurchasing firms ranges from 32.5% to 43.8%. The proportion of firms that repurchased in the pre-1970s co-
hort (42.6%) is almost the same as the fraction that paid in the 1990s cohort (43.8%). Within most RE/TA levels, firms that went public
in the 1990s or 2000s are more likely to repurchase than firms that went public in the 1970s and 1980s. A comparison of Panel A and
Panel B also shows that firms that went public in the 1980s or later are more likely to repurchase than to pay dividends, consistent
with the enactment of Rule 10b-18 in 1982.
Panel C of Table 2 examines firms that pay dividends, repurchase, or both. Again, the proportion of firms making payouts generally
increases monotonically within each IPO cohort as RE/TA increases. Firms that went public in the 1970s or earlier are more likely to
make cash payouts than firms that went public later, likely due to the stickiness of dividends. For more recent IPO cohorts, however,
the relation between RE/TA and payouts is driven by the increasing prevalence of repurchases.
4.3. Payout as a function of RE/TA, IPO decade, and time
The results in Table 2 could be driven by the fact that firms that went public in recent decades are younger, less profitable, and thus
less likely to return cash to shareholders. Consequently, we next examine the impact of both IPO decade and time period on the pro-
portion of firms that pay out. Table 3 reports the results. As discussed in Section 3.2, we divide our sample into four different time pe-
riods: 1973–1979, 1980–1989, 1990–1999, and 2000–2011. For the pre-1970 IPO cohort, the likelihood of paying dividends is fairly
constant over time, while the likelihood of repurchasing increases from 31.5% in the 1970s to 55.9% in the 2000s. This is consistent
with the well-documented increase in share repurchases over time. The 1970s IPO cohort firms are less likely to pay dividends or re-
purchase than the pre-1970 firms, but again the proportion paying dividends does not show a clear time trend while the proportion
repurchasing increases over time. Firms in the 1980s IPO cohort are substantially less likely to pay dividends than firms that went
public earlier, and the proportion of dividend payers is lower than the proportion of repurchasers. However, both the likelihood of
repurchasing and the likelihood of paying dividends increase over time in this group.8 The likelihood of paying dividends declines
further in the 1990s IPO cohort; however, the likelihood of paying a dividend in the first decade of public life (9.6%) is only slightly
less than a 1980s firm in its first decade (12.2%).
8
We examine whether the increase in payouts over time is due to non-paying firms exiting the sample or sample firms initiating payouts by examining the subset of
firms that survive for an extended period of time. In the 1980s IPO cohort, firms with 15 or more annual observations from their IPO year through 2011 still show a
monotonic increase in both the proportion of dividend payers and the proportion making any form of shareholder payout. Similar results obtain if we require 20 or more
annual observations. These findings indicate that our results are not driven by non-payers exiting the sample over time.
Overall, several things are clear from Table 3. First, the percentage of firms making payouts decreases across IPO cohorts, although
this trend is most pronounced for dividend paying firms. Second, for the earlier IPO decades, the proportion of dividend payers re-
mains fairly constant across time. Third, in later IPO cohorts the likelihood of making a payout, whether in the form of dividends,
repurchases or both, increases over time. While not surprising based on the lifecycle hypothesis, this result is contrary to most studies
which show that payouts in general have decreased over time.
For brevity, in future analyses we tabulate the results only for dividends and total payouts. Our rationale is that while dividends are
the historical form of payment, Boudoukh et al. (2007) show that total payouts to shareholders are more informative of asset prices
than the particular form of payout. Boudoukh et al. (2007) also find that the dividend process exhibits a structural shift around the
promulgation of Rule 10b-18, but the total payout yield does not. We discuss the repurchase results when appropriate.
Table 4 adds the impact of RE/TA to the Table 3 analysis to examine whether the explanatory power of RE/TA on payout propensity
has decreased over time or in more recent IPO cohorts. Panel A reports the dividend results. In the pre-1970s IPO cohort, the propor-
tion of firms paying dividends is fairly constant across time periods within most RE/TA levels. We note an exception when RE/TA is less
than 0.20, as the proportion of firms paying dividends actually increases across time. Although Panel A of Table 2 shows that firms in
recent IPO cohorts are less likely to pay dividends, dividend propensity within an RE/TA group increases over time in the 1980s or later
IPO cohorts, e.g. within a given RE/TA level, firms in the 1980s IPO cohort are more likely to pay dividends in the 1990s and 2000s than
they are in the 1980s. Overall, the results in Panel A indicate that (1) the decade in which a firm went public is a stronger determinant
of the relation between earned/contributed capital and the likelihood of paying dividends than is the time period itself, and (2) for
firms in the 1980s or later IPO cohorts, the likelihood of paying dividends increases over time as the firms age.
Panel B examines firms' total payout propensity.9 The results are consistent with Panel A. Firms in more recent IPO cohorts are less
likely to make payouts. Within a cohort, however, the likelihood of paying out increases as RE/TA increases and tends to increase
across time, especially for firms in the 1980s or later cohorts. Interestingly, firms that went public in the 2000s are more likely to
make payouts in the first decade of their life compared to firms that went public in the 1980s or 1990s. This finding could result
from the favorable dividend and capital gains tax rates in effect after 2003, in addition to the reduced number but increased age of
the IPOs made after 2000 (Gao et al., 2012).
4.4. Multivariate analysis of RE/TA and lifecycle
4.4.1. Payout as a function of RE/TA

We next apply Fama and French's (2001, Section 3.5) Fama and MacBeth (1973)-based methodology to test whether the proba-
bility of paying out depends systematically on RE/TA. Specifically, we examine logistic regressions of the decision to make shareholder
payouts. Explanatory variables include RE/TA and measures of profitability, growth, size, cash balances, and risk. We run separate re-
gressions for each of our 39 sample years to obtain a time series of coefficients and report t-statistics adjusted for serial correlation
using the Newey and West (1987) procedure (out to lag 10). Table 5 reports the coefficients and their marginal effects.
Panel A examines the dividend payout decision. Model 1 relates the decision to pay dividends to profitability, growth, size, and RE/
TA. Consistent with Fama and French (2001) and DeAngelo et al. (2006), the probability that a firm pays dividends is positively related
to its profitability, size, and RE/TA and negatively related to growth. Models 2 and 3 replicate Model 1 but incorporate cash holdings
and cash flow volatility, respectively, into the regressions. The impact of cash on dividend payment is negative and significant, which
is consistent with high cash balances proxying for firm risk (Bates et al. (2009)). High cash flow volatility lowers the likelihood of div-
idend payments. The inclusion of these variables has little impact on the other control variables. Model 4 adds an indicator variable
equal to one if the firm paid dividends in the previous year. The coefficient on the dividend indicator is positive and significant,
consistent with the stickiness of dividends (Lintner (1956)). Inclusion of this dummy substantially reduces the marginal impact of
RE/TA on dividend payments, although it remains significant. Overall, our results support DeAngelo et al.'s (2006) conclusion that a
firm's decision to pay dividends depends on its mix of earned vs. contributed capital.
Panel B of Table 5 examines the payout decision.10 As in the dividend regressions, the decision to pay out is positively related to
profitability and size, and negatively related to growth. The effect of cash holdings on shareholder payouts is negative except in the
specification controlling for past dividend status, where we note a positive and significant impact; the change in sign is likely due
to the fact that once we control for past dividend behavior, the repurchase portion of total payout is positively related to a firm's
cash holdings. Firms with high cash flow volatility are less likely to make payouts. As in the dividend model, inclusion of the lagged
dividend indicator reduces the impact of the other variables, including RE/TA, but the marginal effect of each variable upon the pro-
pensity to pay out remains significant. Overall, our results indicate that a firm's earned/contributed capital is an important factor in
determining not only whether it pays dividends but also whether it makes payouts in general.
9
An examination of repurchases shows similar patterns to those found in Table 2 and Panel A of Table 4. Further, a comparison of repurchases to dividends shows
evidence of a substitution effect: Firms that went public in the 1980s are more likely to pay dividends than to repurchase in the 1980s if RE/TA is greater than 0.3; how-
ever, in the 1990s, dividends are more likely only for firms with RE/TA greater than 0.6. By the 2000s, repurchases are as likely as or more likely than dividends in almost
all RE/TA categories.
10
We also examine logistic regressions of the decision to repurchase shares. In contrast to Panel A, the decision to repurchase is positively related to a firm's cash bal-
ances. Firms with greater cash flow volatility are less likely to repurchase, and firms that paid a dividend in the previous year are more likely to repurchase; this is con-
sistent with the idea that dividends and repurchases can be complementary and are both used by firms with free cash flow. The decision to repurchase is positively
related to RE/TA. The coefficient on RE/TA is much lower than in Panel A; however, in Model 4, the marginal effect of RE/TA on a repurchase is slightly higher (0.09)
than its effect on dividends (0.08). We also note that the pseudo R-squares are one-tenth of those reported in Panel A.
Table 4
Proportion of firms making payouts by IPO decade, time period, and RE/TA.The sample includes firm-year observations from 1973 to 2011 for U.S. incorporated, indus-
trial firms with CRSP exchange codes 1, 2, or 3 and non-missing data on Compustat for dividends (DVC), assets (AT), retained earnings (RE), common equity (CEQ) and
earnings before taxes (IB). Firms are allocated annually to categories of RE/TA based upon the level of earned equity divided by total assets. We report the results for two
samples: Dividend Payers are firms with non-zero common dividends; and Payout Firms are members of Dividend Payers and/or Repurchasers, which are identified as
firms with non-zero purchases of common stock (PRSTKC). Within each IPO decade and time period, we annually calculate the number of firms making the respective
payout and divide this by the total number of firms. The median values for the resulting observations in each category are reported. Panel A examines dividend payers
and Panel B payout firms.
Panel A Ratio of earned equity to total assets (RE/TA)
IPO pre-1970s b0.00 0.00–0.10 0.10–0.20 0.20–0.30 0.30–0.40 0.40–0.50 0.50–0.60 0.60–0.70 0.70–0.80 0.80–0.90 0.90+
1973–1979
Proportion 0.0580 0.2817 0.5903 0.7904 0.8694 0.9224 0.9301 0.9200 0.9565 1.0000 0.0000
Number of firms 119 115 199 347 405 303 166 95 34 6 0
1980–1989
Proportion 0.1216 0.4223 0.6347 0.7848 0.8873 0.9228 0.9294 0.9425 0.9490 1.0000 0.8452
Number of firms 120 79 142 207 218 199 125 64 34 12 6
1990–1999
Proportion 0.1496 0.5097 0.7051 0.7836 0.8417 0.8564 0.8899 0.9450 0.9599 1.0000 0.8258
Number of firms 129 75 96 114 98 77 60 41 24 11 11
2000–2011
Proportion 0.2056 0.5907 0.7232 0.7482 0.7968 0.8762 0.8679 0.9375 0.9199 1.0000 0.9333
Number of firms 66 27 52 68 60 54 35 21 13 10 15
IPO in the 1970s
1973–1979
Proportion 0.0396 0.1967 0.3728 0.5365 0.6489 0.7273 0.7714 0.8333 0.8750 0.8330 1.0000
Number of firms 202 122 234 255 225 165 105 64 21 7 1
1980–1989
Proportion 0.0421 0.2275 0.3908 0.4842 0.6044 0.6666 0.7234 0.7889 0.7698 0.7321 0.5000
Number of firms 239 115 163 167 175 124 71 49 26 5 4
1990–1999
Proportion 0.0586 0.2290 0.4629 0.5212 0.5841 0.6014 0.6470 0.7248 0.7500 0.8542 0.9167
Number of firms 134 48 73 73 70 60 45 33 20 4 5
2000–2011
Proportion 0.1024 0.2660 0.3160 0.4507 0.5173 0.5834 0.6092 0.7454 0.8293 0.8571 0.8661
Number of firms 41 19 22 37 42 28 20 20 13 7 7
IPO in the 1980s
1980–1989
Proportion 0.0184 0.1090 0.1600 0.2198 0.3154 0.3388 0.4343 0.5258 0.4500 1.0000 0.2500
Number of firms 752 225 246 203 129 57 30 17 7 1 2
1990–1999
Proportion 0.0386 0.1685 0.2877 0.3060 0.3264 0.3485 0.4327 0.5000 0.5691 0.6364 0.6500
Number of firms 628 184 192 162 130 90 58 40 22 11 8
2000–2011
Proportion 0.0765 0.2590 0.3990 0.3446 0.3847 0.4750 0.4794 0.5998 0.6667 0.6905 0.5000
Number of firms 199 55 56 61 63 49 34 25 16 7 10
IPO in the 1990s
1990–1999
Proportion 0.0286 0.1676 0.1636 0.1315 0.1937 0.2027 0.3215 0.3905 0.6250 0.7500 0.0000
Number of firms 1085 373 278 156 100 45 35 11 2 1 1
2000–2011
Proportion 0.0428 0.1585 0.2279 0.2554 0.2945 0.3190 0.4161 0.4452 0.5736 0.5357 0.4118
Number of firms 714 123 135 134 91 79 52 36 18 8 8
IPO in the 2000s
2000–2011
Proportion 0.0481 0.1850 0.1576 0.1889 0.2209 0.2154 0.2000 0.2154 0.7000 0.3500 0.5833
Number of firms 525 98 68 45 19 13 7 5 5 2 2
Panel B Ratio of earned equity to total assets (RE/TA)
IPO pre-1970s b0.00 0.00–0.10 0.10–0.20 0.20–0.30 0.30–0.40 0.40–0.50 0.50–0.60 0.60–0.70 0.70–0.80 0.80–0.90 0.90+
1973–1979
Proportion 0.1961 0.4437 0.7014 0.8357 0.9002 0.9493 0.9518 0.9579 0.9714 1.0000 0.0000
Number of firms 119 115 199 347 405 303 166 95 34 6 0
1980–1989
Proportion 0.3122 0.5561 0.7348 0.8475 0.9197 0.9489 0.9710 0.9704 0.9646 1.0000 1.0000
Number of firms 120 79 142 207 218 199 125 64 34 12 6
(continued on next page)

Table 4 (continued)
Panel B Ratio of earned equity to total assets (RE/TA)
IPO pre-1970s b0.00 0.00–0.10 0.10–0.20 0.20–0.30 0.30–0.40 0.40–0.50 0.50–0.60 0.60–0.70 0.70–0.80 0.80–0.90 0.90+
1990–1999
Proportion 0.2968 0.6372 0.7944 0.8645 0.8840 0.9243 0.9536 0.9720 0.9808 1.0000 0.9129
Number of firms 129 75 96 114 98 77 60 41 24 11 11
2000–2011
Proportion 0.3912 0.7509 0.8316 0.8590 0.8807 0.9256 0.9344 0.9506 0.9643 1.0000 1.0000
Number of firms 66 27 52 68 60 54 35 21 13 10 15
IPO in the 1970s
1973–1979
Proportion 0.2045 0.3934 0.5614 0.6250 0.7156 0.8000 0.8556 0.9063 0.9167 0.9167 1.0000
Number of firms 202 122 234 255 225 165 105 64 21 7 1
1980–1989
Proportion 0.1923 0.4819 0.5671 0.6253 0.7140 0.7645 0.8274 0.8529 0.8610 0.8831 0.8750
Number of firms 239 115 163 167 175 124 71 49 26 5 4
1990–1999
Proportion 0.2148 0.4037 0.6198 0.6826 0.7468 0.7694 0.7836 0.8528 0.9230 1.0000 1.0000
Number of firms 134 48 73 73 70 60 45 33 20 4 5
2000–2011
Proportion 0.3175 0.4422 0.5695 0.6830 0.7663 0.8000 0.8536 0.8441 0.9375 0.9545 1.0000
Number of firms 41 19 22 37 42 28 20 20 13 7 7
IPO in the 1980s
1980–1989
Proportion 0.1766 0.3061 0.3774 0.3879 0.4851 0.5291 0.6158 0.7183 0.5476 1.0000 1.0000
Number of firms 752 225 246 203 129 57 30 17 7 1 2
1990–1999
Proportion 0.1860 0.3640 0.4583 0.5104 0.5947 0.6092 0.6859 0.7321 0.8235 0.9615 1.0000
Number of firms 628 184 192 162 130 90 58 40 22 11 8
2000–2011
Proportion 0.2786 0.5414 0.6579 0.6417 0.6571 0.7269 0.7822 0.8338 0.8995 0.9500 0.8452
Number of firms 199 55 56 61 63 49 34 25 16 7 10
IPO in the 1990s
1990–1999
Proportion 0.1625 0.3286 0.3220 0.3670 0.4578 0.4662 0.5767 0.6250 1.0000 1.0000 1.0000
Number of firms 1085 373 278 156 100 45 35 11 2 1 1
2000–2011
Proportion 0.2632 0.4755 0.5702 0.5939 0.6251 0.7068 0.7691 0.8310 0.8760 0.9500 0.9412
Number of firms 714 123 135 134 91 79 52 36 18 8 8
IPO in the 2000s
2000–2011
Proportion 0.2480 0.4130 0.4637 0.4919 0.5188 0.4664 0.4444 0.6905 1.0000 0.8000 1.0000
Number of firms 525 98 68 45 19 13 7 5 5 2 2
While Table 5 reports the mean of the annual marginal effects at the means (MEMs), MEMs for continuous variables measure the
instantaneous rate of change, which may not be close to the effect on P(Y = 1) of a one unit increase in X. Consequently, we also
examine probabilities at fixed levels of RE/TA while holding the other variables at their annual means. Using Model 3, the mean pre-
dicted probability of paying dividends is 7.7% when RE/TA is −0.2. It increases to 17.2% when RE/TA is 0, to 34.5% when RE/TA is 0.2,
and to 54.1% when RE/TA is 0.4. The increases in probabilities across RE/TA categories are economically significant and coincide with
the proportions reported in Table 1. Using Model 3 to predict payouts in Panel B, the probability of paying out is 43.1% when RE/TA is
−0.2. It increases to 51.3% when RE/TA is 0, to 59.2% when RE/TA is 0.2, and to 65.4% when RE/TA is 0.4. Using Model 4, the lagged
dividend indicator subsumes much of the impact of RE/TA, and the changes in probabilities resulting from increased RE/TA levels
grow smaller for both payout types.
4.4.2. Payout as a function of RE/TA, IPO decade, and time

Table 5 shows that the earned/contributed capital mix is an important determinant of the payout decision, even after controlling
for other variables. However, our univariate results in Tables 2 and 3 indicate that the impact of RE/TA on dividend payment declines
over time due to the influx of recently public firms with lower propensities to pay dividends. Consequently, we next control for both
IPO decade and time period. For each IPO cohort, a logit analysis of payout decision is run as a function of RE/TA, the control variables,
and a fiscal year indicator. Panel A (B) uses the control variables in Model 3 (4) of Table 5. For brevity, Table 6 reports only the
coefficients on RE/TA for each IPO decade.
Table 5
Logistic regression of the decision to payout.The sample includes all observations with non-missing data for total equity to total assets (TE/TA); profitability (ROA); sales growth (SGR); size (NYE); cash holdings (CASH); cash flow
volatility (Σ5); and an indicator variable for payment of dividends in the prior year (Divt − 1). Panel A examines the dividend decision and Panel B examines the payout decision. The coefficients reported are the mean value of the
fitted coefficients for 39 annual Fama–MacBeth logistic regressions run for each year 1973–2011. The t-statistics (in parentheses) use Newey–West (1987) adjusted standard errors (lag 10). Marginal effects (in italics) are the
M.L. Banyi, K.M. Kahle / Journal of Corporate Finance 27 (2014) 345–366

means of the annual marginal effects at the means. *, **, and *** indicate significance of the marginal effects at the 0.10, 0.05 and 0.01 levels, respectively.
Panel A: dividend decision Panel B: payout decision
(1) (2) (3) (4) (1) (2) (3) (4)
Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect
RE/TA 4.94 0.64 4.85 0.62 4.74 0.60 1.95 0.08 1.82 0.28 1.79 0.28 1.71 0.26 0.60 0.06
(6.17) *** (5.94) *** (6.06) *** (3.74) *** (2.41) *** (2.35) *** (2.34) *** (2.59) ***
TE/TA −2.54 −0.32 −2.02 −0.26 −1.84 −0.23 0.11 0.00 −0.25 −0.03 −0.55 −0.01 0.01 0.01 0.29 0.04
(−13.64) *** (−10.91) *** (−12.52) *** (0.62) (−1.06) ** (−7.94) (0.03) (2.38) ***
ROA 2.33 0.31 2.48 0.33 2.38 0.32 5.40 0.22 1.52 0.24 3.00 0.24 1.43 0.23 1.89 0.20
(3.89) *** (4.00) *** (3.81) *** (4.62) *** (−2.62) *** (12.91) *** (2.47) *** (2.46) ***
SGR −0.86 −0.11 −0.88 −0.11 −0.85 −0.11 −0.34 −0.01 −0.55 −0.10 −0.55 −0.09 −0.53 −0.09 −0.30 −0.04
(−7.86) *** (−7.37) *** (−6.87) *** (−1.96) *** (−8.11) *** (−7.94) *** (−8.08) *** (−5.21) ***
NYE 3.19 0.41 3.22 0.41 3.23 0.41 2.04 0.08 2.98 0.52 3.00 0.52 3.00 0.52 1.48 0.20
(7.84) *** (8.28) *** (8.48) *** (15.53) *** (12.42) *** (12.91) *** (13.82) *** (9.77) ***
CASH −1.75 −0.22 −1.68 −0.21 −0.55 −0.02 −0.36 −0.07 −0.29 −0.06 0.42 0.04
(−5.99) *** (−6.59) *** (−1.76) *** (−2.25) *** (−2.26) *** (2.18) ***
Σ5 −9.03 −1.17 −5.34 −0.23 −7.77 −1.17 −3.65 −0.37
(−2.06) *** (−2.05) *** (−2.10) *** (−1.99) ***
Div t − 1 5.61 0.21 3.42 0.43
(34.49) *** (28.19) ***
Int −1.14 −1.20 −0.77 −4.05 −0.57 −0.59 −0.19 −1.14
(−14.80) (−14.22) (−4.65) (−22.53) (−6.45) (−6.37) (−1.02) (−17.01)
Pseudo R2 0.3527 0.3583 0.3622 0.7587 0.2186 0.2194 0.2211 0.3993
355
Table 6
Logistic regression of the decision to payout by IPO decade and time period.The sample includes U.S. incorporated, industrial firms with CRSP exchange codes 1, 2, or 3
and non-missing data on Compustat for dividends (DVC), assets (AT), retained earnings (RE), common equity (CEQ), and earnings before taxes (IB). We eliminate any
observation without all control variables required for the logistic model, including profitability (ROA), sales growth (SGR), size (SIZE), cash holdings (CASH), and cash
flow volatility (Σ5). For each payout sample, a logit analysis of the propensity to make a payout is run as a function of RE/TA, the control variables, and a fiscal year in-
dicator. Panel A uses Model 3 of Table 5, while Panel B uses Model 4 of Table 5. The analysis is run separately for each IPO decade within the respective payout sample.
We report the RE/TA coefficients for each IPO decade, as well as the marginal effects at the means for each IPO decade calculated over each time period. The time periods
are defined as follows: time 1 is from 1973 through 1979; time 2 runs from 1980 through 1989; time 3 runs from 1990 through 1999; and time 4 includes 2000 through
2011.
Dividend firms Payout firms
Pre-70s 1970s 1980s 1990s 2000s Pre-70s 1970s 1980s 1990s 2000s
Panel A
Coefficients (t-stats)
RE/TA 5.635 4.911 2.778 1.291 0.912 3.435 2.106 0.562 0.186 0.131
(51.20) (47.22) (34.37) (17.98) (6.77) (38.52) (33.52) (21.50) (13.17) (5.10)
N 35,995 26,409 32,422 29,104 5,990 35,995 26,409 32,422 29,104 5,990
Pseudo R2 0.379 0.273 0.268 0.201 0.255 0.309 0.169 0.1472 0.121 0.106
Marginal effects by time period

1 RE/TA 0.696 0.941 0.400 0.426
N 12,268 7,122 12,268 7,122
2 RE/TA 0.612 0.826 0.256 0.343 0.411 0.107
N 11,668 11,090 11,459 11,668 11,090 11,459
3 RE/TA 0.628 0.751 0.335 0.094 0.354 0.376 0.112 0.034
N 7,189 5,289 14,524 13,323 7,189 5,289 14,524 13,323
4 RE/TA 0.611 0.734 0.387 0.130 0.069 0.329 0.352 0.110 0.039 0.026
N 4,870 2,908 6,439 15,781 5,990 4,870 2,908 6,439 15,781 5,990
Panel B
Coefficients (t-stats)
RE/TA 2.652 1.859 1.139 0.168 0.203 1.102 0.662 0.143 0.164 0.101
(16.29) (12.48) (10.48) (3.40) (1.69) (12.63) (12.50) (10.84) (10.48) (4.08)
N 35,995 26,409 32,422 29,104 5,990 35,995 26,409 32,422 26,958 5,990
Pseudo R2 0.749 0.731 0.735 0.643 0.620 0.531 0.415 0.1964 0.190 0.172
Marginal effects by time period

1 RE/TA 0.098 0.107 0.081 0.084
N 12,268 7,122 12,268 7,122
2 RE/TA 0.096 0.084 0.030 0.069 0.083 0.050
N 11,668 11,090 11,459 11,668 11,090 11,459
3 RE/TA 0.100 0.080 0.036 0.005 0.079 0.083 0.050 0.025
N 7,189 5,289 14,524 13,323 7,189 5,289 14,524 13,323
4 RE/TA 0.110 0.087 0.047 0.006 0.007 0.079 0.085 0.052 0.027 0.018
N 4,870 2,908 6,439 15,781 5,990 4,870 2,908 6,439 15,781 5,990
Panel A shows that the coefficient on RE/TA declines monotonically across IPO decades for both dividend payers and payout firms.
As in Table 5, we also report the instantaneous MEMs of RE/TA by IPO decade and time period. The results are similar to our univariate
results: the impact of RE/TA on the probability of paying out declines more across IPO cohort than over time. The marginal effects are
quite small for firms that went public in the 1990s or later, relative to firms that went public in the 1970s or earlier. These results are
consistent with RE/TA becoming a less meaningful proxy measure of lifecycle for firms that have more recently gone public.
Panel B of Table 6 uses Model 4 of Table 5. Inclusion of the lagged dividend indicator decreases both the coefficients and the
marginal effects of RE/TA, indicating that lagged dividend status usurps much of the power of RE/TA. The marginal effects decrease
by a factor of about ten for the dividend firms. However, the overall conclusions are similar to the model without the lagged dividend
indicator: the impact of RE/TA on the likelihood of shareholder payout declines across IPO decades for firms that went public in the
1970s or later.
To aid in interpretation, we again use Model 3 of Table 5 to examine the probability of paying out at fixed levels of RE/TA while
holding all other variables at their mean. For 1970s cohort firms, the probability of paying dividends during 1973–1979 is 25.8%
when RE/TA is 0 and increases to 46.3% (68.2) when RE/TA is 0.2 (0.4). We find similar probabilities at each level of RE/TA during
the 1980–1989 and 1990–1999 time periods. The probability decreases slightly in the last time period, to 23.3% when RE/TA is zero
and 57.5% when RE/TA is 0.4. For 1980s cohort firms, the probability of paying dividends during 1980–1989 is 18.1% when RE/TA is
0 and increases to 26.4% (36.5) when RE/TA is 0.2 (0.4). For this cohort, the likelihood of paying dividends increases for each RE/TA
level over the subsequent time periods, such that the probability of paying dividends during 2000–2011 is 25.2% when RE/TA is
zero and increases to 44.8% when RE/TA is 0.4. By the 2000s cohort, the probability of paying dividends is 13.0% when RE/TA is
zero and increases to 16.6% when RE/TA is 0.4. Overall, at any given RE/TA level, the probability of paying dividends declines across
IPO cohorts. For the 1980s and later cohorts, the probabilities increase slightly over time periods. By the 2000s cohort, however,
the increase in the probability of paying dividends as RE/TA increases is small.
Table 7
Logistic regression of the decision to payout — age quintile regressions.Panel logistic regressions modeling firm payout decisions, run on quintiles formed annually on a
firm's age, where age is defined as the founding date. The sample period is 1982–2011 (N = 67,226) for observations with all available information. IPOpost90 is an
indicator variable for firms with an IPO year during or after 1990. All models include the control variables in Model 3 of Table 5 and year fixed effects (results not tab-
ulated). Robust t-statistics are reported in parenthesis. Mean marginal effects are reported in italics. *, **, and *** indicate significance of the marginal effects to the 0.10,
0.05 and 0.01 levels respectively.
Age quintile
Youngest Oldest
(1) (2) (3) (4) (5)
Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect Coeff Meffect
Panel A: propensity to make dividend payments

RE/TA 1.91 0.03 2.46 0.06 3.24 0.16 3.06 0.50 4.49 0.83
(4.20) *** (7.92) *** (16.53) *** (19.87) *** (23.58) ***
RE/TA ∗ IPOpost90 −1.75 −0.03 −1.84 −0.04 −1.98 −0.10 −1.55 −0.25 −1.67 −0.31
(−3.85) *** (−5.05) *** (−5.60) *** (−6.64) *** (−7.16) ***
IPOpost90 −1.16 −0.02 −0.89 −0.02 −0.31 −0.02 −0.45 −0.07 −0.43 −0.08
(−6.07) *** (−7.03) *** (−3.40) *** (−5.51) *** (−5.43) ***
Intercept −1.73 −0.75 −0.19 0.11 −0.08
(−8.33) (−4.65) (−1.25) (0.67) (−0.30)
Control variables Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
N observations 12,300 13,528 13,934 13,891 13,573
Pseudo R2 0.234 0.2503 0.2416 0.2547 0.3834
Panel B: propensity to make shareholder payouts

RE/TA 0.37 0.07 0.55 0.12 0.63 0.15 1.45 0.36 3.02 0.35
(4.78) *** (5.69) *** (6.43) *** (11.11) *** (10.85) ***
RE/TA ∗ IPOpost90 −0.30 −0.06 −0.34 −0.07 −0.39 −0.09 −0.96 −0.24 −1.63 −0.19
(−3.98) *** (−3.57) *** (−3.67) *** (−5.66) *** (−5.79) ***
IPOpost90 −0.69 −0.14 −0.62 −0.13 −0.43 −0.10 −0.38 −0.09 −0.38 −0.04
(−5.82) *** (−8.69) *** (−9.13) *** (−7.07) *** (−4.74) ***
Intercept −0.70 −0.22 0.13 0.57 0.28
(−4.83) (−1.68) (0.94) (3.16) (0.99)
Control variables Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
N observations 12,300 13,528 13,934 13,891 13,573
Pseudo R2 0.0785 0.116 0.1228 0.1582 0.3069
We perform a similar analysis of firm payouts. While the probability of making payouts is greater than the probability of paying
dividends in any given IPO cohort and time period combination, we still find that the increase in the probability of payout as RE/TA
increases is small in the 1990s and 2000s IPO cohort firms. By the 2000s cohort, the probability of payout is 39.5% when RE/TA is
zero and increases to 40.7% when RE/TA is 0.4.
4.4.3. IPO decade vs. age?

Table 6 indicates that RE/TA is less important in predicting dividend and payout policy in recent IPO cohorts. It is possible that we
are simply picking up an age effect, in which firms are less likely to pay out when they are young and less profitable, but increase
payouts as they mature. Consequently, Table 7 examines the effect of both age and IPO cohort. Specifically, in each year, we split
firms into quintiles based on their founding date, as determined from Jay Ritter's website (http://bear.cba.ufl.edu/ritter/ipodata.
htm) and the Jovanovic and Rousseau (2001) data. Requiring founding dates decreases the overall sample to 67,226 observations,
but there are still over 12,000 observations in each age quintile. We run logistic regressions on each age quintile separately using
Model 3 of Table 5. Since earlier results show that the payout propensity of firms that went public in the 1990s or later is substantially
different than firms that went public earlier, we include a dummy variable equal to one if the firm went public in 1990 or later
(IPOpost90) and an interaction between this dummy and RE/TA.
Panel A of Table 7 shows the results for dividend policy. After controlling for other determinants of the propensity to pay dividends,
the marginal effect of RE/TA increases as firms age. The marginal effect of the interaction between RE/TA and the IPOpost90 on
dividend propensity is negative and significant in each age quintile, indicating that regardless of age, the impact of RE/TA on the pro-
pensity to pay is weaker in firms that went public in the 1990s or later.11, 12 After controlling for these effects, however, the average
firm in the youngest (oldest) quintile is 2% (8%) less likely to pay dividends if it went public in the 1990s or later.
11
For robustness, we use dummies equal to one if the firm went public in the mid-1980s (after the enactment of Rule 10b-18) and find similar results.
12
The coefficient on cash flow volatility switches from negative and significant in the four younger age quintiles to positive and significant in the oldest quintile. This
result is driven by the years surrounding the recent financial crisis, when volatility was high but older firms continued to pay dividends. If we exclude the 2008–2011
observations from the regression, the coefficient is negative but not significant.
Panel B models the propensity to make payouts. As with dividends, the marginal effect of RE/TA on a firm's payout propensity
increases as a firm ages and the interaction variable is negative and significant in all age quintiles. The impact of RE/TA on the
propensity to pay is quite low in the younger quintiles. In each age quintile, however, firms in the 1990s or later IPO cohorts
are less likely to make payouts. Overall, Table 7 indicates that our earlier results do not simply capture an age effect: while an
increase in RE/TA has a greater impact on the likelihood of payout as firms age, this impact is attenuated in firms from more re-
cent IPO vintages.
4.4.4. What drives the difference across IPO cohorts?

The results above beg the question: what drives the payout propensity differences among IPO cohorts? On average, firms in the
later cohorts are younger and less profitable, but based on fundamentals, those that are profitable and have high RE/TA could still
pay out at the same level as prior cohorts. We examine three potential explanations: differences in methods of accounting for
repurchases, omitted risk variables, and executive compensation.
We allude to the first explanation in Section 3.2: Due to the varying impacts of the two methods of accounting for repurchases,
RE/TA does not consistently reflect a firm's earned-contributed capital relation across time or firms. To explore this, we examine
how accounting methods change over time for three groups of Delaware firms13: a) firms making no recent repurchases; b) firms
making repurchases within the past four years with a treasury stock account and c) firms making repurchases within the past four
years with no treasury stock account and assumed to use the retirement method. As discussed previously, a distribution paid as a div-
idend, a share repurchase recorded with the treasury method, or a share repurchase recorded with the retirement method will impact
RE/TA differently. If the choices between distribution forms and accounting methods are random, no patterns in the RE/TA categories
across IPO cohorts and the three categories of Delaware firms should emerge.
In untabulated results, we find differently. First, although the treasury method is the preferred method for frequently
repurchasing firms, the use of the retirement method increases across IPO cohorts, growing from 22% for the pre-1970 IPO co-
hort to nearly half of all repurchasing firms in the 2000s cohort. Second, within each IPO cohort, a higher proportion of retire-
ment firms report negative RE. This proportion grows over time, such that in the 1990s IPO cohort, 54.3% of retirement method
firms report negative retained earnings compared to only 36.6% of the treasury firms. Since repurchases have become the pre-
ferred form of payout over time, the growing use of the retirement method combined with this increasing preference for
repurchases indicates that RE's ability to capture a firm's cumulative earnings less returns to shareholders changes over time
and is systematically related to other characteristics of the firm, including its IPO cohort. These biases impact the usefulness
of RE/TA (or RE/TE) as a proxy for a firm's stage in its lifecycle.
We also attempt to control for accounting method by creating an indicator variable equal to one for firms incorporated in
MBCA states, where treasury stock accounts are prohibited. We collect data on state of incorporation from the 10-K, 10-Q,
and 8-K filings in the WRDS SEC Analytics Suite for each firm-year from 1993 to 2011. The results are reported in regressions
(1) and (5) of Table 8. MBCA status itself does not impact the dividend decision. However, the marginal effect on the MBCA
indicator is positive and marginally significant in the payout regression, indicating that firms incorporated in states that do
not allow treasury stock are more likely to payout. The interaction between MBCA and RE/TA is significantly positive for
both payout samples. Since firms required to use the retirement method have downward-biased RE/TA, the model appears
to adjust accordingly and attribute a more positive effect to RE/TA in MBCA states. However, firms going public in the 1990s
or later are still less likely to pay out.
The second explanation to explain payout differences among IPO decades is that IPO cohort proxies for an omitted variable, such as
risk, that is not captured in RE/TA itself. While our Table 5 regressions include cash flow volatility, it is possible that cash flow risk does
not fully capture firm risk. Consequently, we examine two other risk measures: accounting risk and idiosyncratic risk. In terms of
accounting risk, for a firm to pay out, it may not only need high RE/TA but also consistent profitability, which is not yet experienced
by the younger IPO cohorts. We identify high earnings volatility firms as those with a five-year standard deviation of earnings before
interest, taxes, and dividends (deflated by average assets over the five year period) above the annual median and include both an
indicator variable for these firms and an interaction between this indicator and RE/TA. Idiosyncratic risk is the standard deviation
of residuals from a regression of daily excess stock returns on the market factor (Hoberg and Prabhala, 2009). Systematic risk is the
standard deviation of the predicted values from this regression. We define high idiosyncratic (systematic) risk firms as firms
whose idiosyncratic (systematic) risk is above the annual median.
Regression (2) of Table 8 examines the effect of idiosyncratic and systematic risks on dividend payment. The likelihood of paying
dividends is negatively related to both measures of risk. The marginal effect on the interaction of idiosyncratic risk with RE/TA is also
negative and significant, indicating that the influence of RE/TA on dividend payment is not as strong in firms with high idiosyncratic
risk. The marginal effects on the IPOpost90s indicator and its interaction with RE/TA are negative and significant, indicating that idi-
osyncratic and systematic risks do not explain the reduction in payout propensity in the later IPO cohorts. Regression (6) examines the
effect of these risks on the payout decision; the results are similar to those in regression (2).
In regression (3), the high earnings volatility indicator is negative and significant, implying that firms with volatile earnings
are less likely to pay dividends. The marginal effect on the interaction with RE/TA is negative and significant also, as the influence
13
Since the choice of accounting method is impacted by state of incorporation, we focus on Delaware firms which a) have a choice between accounting
methods, b) comprise over 50% of firms on Compustat, and c) face the least stringent legal requirements to make payouts. By doing so, we control for differences in
the choice of state of incorporation but we recognize that the inferences drawn may not necessarily generalize to our full sample of firms.
Table 8
Logistic regression of the decision to payout — alternative specifications.The sample period is 1993–2011 for models using MBCA incorporation information, 1982–2011 for models using idiosyncratic risk or earnings variance, and
1992–2011 for models using stock option data. IPOpost90s is an indicator variable for firms with an IPO after 1990; high idiosyncratic and systematic risk firms report these respective measures above the annual median, where the
risk measures are calculated as per Hoberg and Prabhala (2009). High earnings variance firms report a five year EBIT variance above the annual median. High option firms report stock options to common shares outstanding above
the annual median. All models include control variables as per Model 3 of Table 5 as well as year fixed effects (results not tabulated below). Robust t-statistics are reported in parenthesis. Mean marginal effects (Mf) are reported in
italics. *, **, and *** indicate significance of the marginal effects to the 0.10, 0.05 and 0.01 levels respectively.
Dividend payer Payout firm
(1) (2) (3) (4) (5) (6) (7) (8)
M.L. Banyi, K.M. Kahle / Journal of Corporate Finance 27 (2014) 345–366

Coeff Mf Coeff Mf Coeff Mf Coeff Mf Coeff Mf Coeff Mf Coeff Mf Coeff Mf
RE/TA 3.46 0.19 3.86 0.34 4.19 0.52 3.62 0.86 0.97 0.24 1.59 0.40 1.87 0.47 0.21 0.41
(34.84) *** (18.91) *** (19.93) *** (17.25) *** (13.94) *** (12.31) *** (14.37) *** (12.50) ***
RE/TA ∗ IPOpost90s −2.01 −0.11 −2.62 −0.23 −2.61 −0.32 −1.86 −0.44 0.86 −0.21 −0.63 −0.16 −0.78 −0.19 −1.60 −0.31
(−12.25) *** (−11.35) *** (−10.46) *** (−7.21) *** (−12.44) *** (−9.00) *** (−10.25) *** (−10.94) ***
IPOpost90s −0.73 −0.04 −0.60 −0.05 −0.55 −0.06 −0.73 −0.17 −0.73 −0.18 −0.59 −0.16 −0.50 −0.13 −0.43 −0.08
(−20.30) *** (−7.74) *** (−6.47) *** (−11.24) *** (32.56) *** (−14.89) *** (−11.03) *** (−9.17) ***
High idiosyncratic risk −1.00 −0.09 −0.78 −0.21
(−18.94) *** (−23.68) ***
High systematic risk −0.31 −0.03 −0.31 −0.08
(−6.71) *** (−13.87) ***
Idiosyncratic risk ∗ RE/TA −0.43 −0.04 −0.84 −0.21
(−2.29) *** (−5.97) ***
Systematic risk ∗ RE/TA −0.05 0.00 −0.12 −0.03
(−0.31) (−5.67)
High earnings volatility −0.51 −0.06 −0.48 −0.12
(−8.98) *** (−13.76) ***
Earnings volatility ∗ RE/TA −0.75 −0.09 −1.03 −0.26
(−4.08) *** (−8.22)
High options −0.61 −0.14 −0.06 −0.01
(−9.33) *** (−1.24)
High options ∗ RE/TA −0.23 −0.05 −0.42 −0.08
(−1.02) (−2.40) **
High ownership −0.15 −0.04 −0.05 −0.01
(−2.27) ** (−1.31)
High ownership ∗ RE/TA 0.26 0.06 0.13 0.02
(1.17) (1.29)
MBCA firm 0.00 0.00 0.04 0.01
(−0.01) (1.66) *
MBCA ∗ RE/TA 0.74 0.04 0.30 0.07
(4.32) *** (4.35) ***
N observations 63,046 103,744 84,528 24,292 63,046 103,744 84,528 24,292
Pseudo R2 0.3700 0.4144 0.3915 0.3256 0.2010 0.2441 0.2390 0.2055
359
of RE/TA on dividend payment is not as strong in firms with high earnings volatility. After controlling for earnings volatility how-
ever, the marginal effects on the IPOpost90s indicator and its interaction with RE/TA remain negative and significant. Similar re-
sults obtain in regression (7) for the payback firms. Thus earnings volatility cannot explain the payout propensity reduction in
younger IPO cohorts.
The third possibility is that executive compensation, specifically growth in stock options, has made firms less likely to pay
dividends over time.14 To test this hypothesis, we collect Execucomp data on both the number of shares and the number of un-
exercised options held by top managers in the S&P 1500 from 1992 to 2011. We measure ownership and option intensity as the
ratio of each to shares outstanding. Option intensity increases over time for these firms. Median option intensity increases from
0.73% in 1992 to a high of 2.89% in 2002, before declining in the mid-2000s. To determine whether increased option use has
decreased the likelihood of payouts, we include a high ownership (option) dummy set equal to one if a firm's managerial
ownership (option intensity) is above the annual median. Regression (4) of Table 8 shows that dividend propensity is negatively
related to both high ownership and high options. However, firms that went public in the 1990s or later are still less likely to pay
dividends than other firms, and the marginal effect of RE/TA on dividend payment is still lower for the later IPO cohorts. Regres-
sion (8) shows similar results for the influence of options on payout. Thus, option usage does not explain the reduction in payout
propensity in later IPO cohorts.
5. Prediction accuracy and the propensity to pay
The previous sections show that the ability of RE/TA to predict payouts decreases for firms that have gone public in recent decades.
In this section, we begin by examining the effect of this decreasing power of RE/TA on the prediction accuracy of models of firm
payout. We finish by re-examining the propensity of firms to pay dividends or make payouts over time.
5.1. Prediction accuracy
We examine prediction accuracy across IPO decades and RE/TA groups by comparing the actual number of dividend payers over
the predicted period (1987 to 2011) to the expected number of payers, using the parameters of a prediction model from an earlier time
period (the estimation period). To calculate the expected number of payers, we use the parameters from a pooled logit model of our
estimation period to generate the probability of payout for each firm in our predicted period. We choose two different estimation
periods. First, we use an estimation period from 1973 through 1978, which is similar to the period used by DeAngelo et al. (2006).
Second, we select an estimation period from 1982 to 1986, which occurs after the enactment of Rule 10b-18 and models a different
shareholder payout regulatory regime.15 For brevity, we do not tabulate results for all combinations of estimation periods and IPO
cohorts, but show a representative sample that demonstrates our findings.
Panels A and B of Table 9 report the results using the 1973–1978 estimation period to predict dividend propensity in 1987–
2011 for the pre-1970 and 1990 IPO cohorts, respectively. For the pre-1970s cohort firms, the model predicts that 9703 obser-
vations will pay dividends, compared to the 10,314 observations that actually pay dividends, a − 4.1% dividend prediction error
rate. The majority of the errors occur at low (b0.2) or high (N0.9) RE/TA levels; the prediction accuracy is good for RE/TA be-
tween 0.2 and 0.5. We also examine the percentage of actual payers who are predicted to pay and refer to these observations
as correctly predicted payers. The correctly predicted payers is above 83% when RE/TA is greater than 0.2. Overall, in the pre-
1970s cohort, more firms pay dividends than expected and the percentage of correctly predicted payers is high, indicating
that the model is accurate.
Panel B uses the 1973–1978 estimation period to predict dividend propensity for the 1990 IPO cohort. Compared to the pre-1970
cohort, the model predicts that more firms should pay (4660) than actually do (3442), consistent with a declining propensity to pay in
recent IPO cohorts. Further, the model does a poor job of correctly predicting which firms will pay: the percent correctly predicted is
only 49.9%, compared to 82.1% in the pre-1970 IPO cohort. These results are consistent with the 1973–1978 estimation period poorly
identifying dividend paying firms in IPO cohorts occurring after the enactment of Rule 10b-18.
Panel C uses the 1982–1986 estimation period to predict dividend propensity in the 1990s IPO cohort. This period falls after the
enactment of Rule 10b-18 and should better reflect dividend propensity for later IPO cohorts. Consistent with this notion, the dividend
error rate is only 2.4%, compared to 4.2% when using the 1973–1978 estimation period. The ability of this model to correctly identify
dividend payers is still low, although it performs better in the lowest RE/TA groups. Since the majority of firms fall into these lower RE/
TA categories, the 1982–1986 period does a better job of predicting for the largest proportion of firms. In untabulated results, we also
find that the 1982–1986 estimation period more accurately identifies repurchasing firms (and payout firms) than the 1973–1978
period.
Panel D uses the 1982–1986 estimation period to predict the total payout propensity of the 1990s IPO cohort. When predicting
total payouts rather than dividends, the model predicts that fewer firms will pay (9045) than actually do (10,432). Compared to
14
Frydman and Saks (2010) and Murphy (2013) show that the use of options as a form of compensation was modest in the 1960s but had become the largest com-
ponent of compensation by the mid-1990s. Since the value of stock options is negatively related to future dividend payments (unless the options are dividend
protected), managers maximize the value of their options by substituting repurchases for dividend growth. Consistent with this motivation, Kahle (2002) finds that
firms are more likely to repurchase, as opposed to pay dividends, if managerial wealth would be negatively impacted by the payment of dividends.
15
In robustness, we also use an estimation period from 1989 to 1992. Results are similar to 1982–1986.
Table 9
Model estimation accuracy.A logistic model is run for the specified estimation period and includes the control variables in Model 3 of Table 5. The coefficients from the
pooled model are used to determine the predicted payout propensity for observations between 1987 and 2011. The prediction results are then sorted by RE/TA category
and IPO decade. The dividend (payout) error rate of the category is the deficit of dividend payers (payout firms) to the total number of firms in the RE/TA category,
where the deficit is defined as the number of expected payers minus the number of actual payers.
Ratio of earned equity to total equity (RE/TA) All

firms
b0.00 0.00–0.10 0.10–0.20 0.20–0.30 0.30–0.40 0.40–0.50 0.50–0.60 0.60–0.70 0.70–0.80 0.80–0.90 0.90+
Panel A: 16,553 observations between 1973 and 1978 are used to predict dividend propensity for pre-1970s IPO firms
Total firms 2,374 1,322 1,808 2,293 2,127 1,832 1,197 857 436 260 285 14,791
Actual payers 407 692 1,265 1,769 1,781 1,605 1,078 801 404 252 260 10,314
Expected payers 149 432 986 1,718 1,792 1,662 1,148 840 433 258 285 9,703
Correctly predicted 73 245 844 1,473 1,598 1,490 1,044 789 403 250 260 8,469
payers
% correctly predicted 0.179 0.354 0.667 0.833 0.897 0.928 0.968 0.985 0.998 0.992 1.000 0.821
Deficit (258) (260) (279) (51) 11 57 70 39 29 6 25 (611)
Dividend error rate −10.9% −19.7% −15.4% −2.2% 0.5% 3.1% 5.8% 4.6% 6.7% 2.3% 8.8% −4.1%
Panel B: 16,553 observations between 1973 and 1978 are used to predict dividend propensity for the 1990s IPO firms
Total firms 15,099 3,732 3,361 2,521 1,702 1,111 717 467 208 82 106 29,106
Actual payers 526 478 581 452 421 303 267 211 115 42 46 3,442
Expected payers 80 244 498 802 888 745 585 431 200 81 106 4,660
Correctly predicted 31 97 208 249 277 229 231 197 112 41 46 1,718
payers
Deficit (446) (234) (83) 350 467 442 318 220 85 39 60 1,218
Dividend error rate −3.0% −6.3% −2.5% 13.9% 27.4% 39.8% 44.4% 47.1% 40.9% 47.6% 56.6% 4.2%
Panel C: 17,176 observations between 1982 and 1986 are used to predict dividend propensity for 1990s IPO firms
Total firms 15,099 3,732 3,361 2,521 1,702 1,111 717 467 208 82 106 29,106
Actual payers 526 478 581 452 421 303 267 211 115 42 46 3,442
Expected payers 114 302 501 698 707 593 485 374 182 79 106 4,141
payers
Deficit (412) (176) (80) 246 286 290 218 163 67 37 60 699
Dividend error rate −2.7% −4.7% −2.4% 9.8% 16.8% 26.1% 30.4% 34.9% 32.2% 45.1% 56.6% 2.4%
Panel D: 17,176 observations between 1982 and 1986 are used to predict payout propensity for 1990 IPO firms
Total firms 15,099 3,732 3,361 2,521 1,702 1,111 717 467 208 82 106 29,106
Actual payers 3,432 1,384 1,463 1,235 982 692 519 378 181 71 95 10,432
Expected payers 1,214 1,309 1,603 1,562 1,181 839 566 397 191 78 105 9,045
payers
Deficit (2,218) (75) 140 327 199 147 47 19 10 7 10 (1,387)
Payout error rate −14.7% −2.0% 4.2% 13.0% 11.7% 13.2% 6.6% 4.1% 4.8% 8.5% 9.4% −4.8%
Panel C, however, the model does a better job of correctly predicting which firms will payout. Over 75% of firms with RE/TA N0.3 and
over 93% of firms with RE/TA N0.7 are correctly predicted to pay.
Overall, Table 9 shows several things. First, an estimation period after the enactment of Rule 10b-18 better predicts payout
propensity in recent IPO cohorts, compared to an earlier estimation period. Both estimation periods underestimate the number
of low RE/TA firms making dividend payments, however. Second, regardless of which estimation period is used, the model does
a poor job correctly identifying the firms that pay dividends, particularly at low RE/TA levels. This result is consistent with dif-
ferences in accounting methods for share repurchases muddling the relationship between accounting RE and a firm's earned
capital. Third, while fewer firms pay dividends than expected in the later IPO cohorts, more firms pay out than expected. Com-
pared to dividends, the error rate for total payouts is much lower at high RE/TA levels, consistent with repurchases substituting
for dividends over time.
5.2. The propensity to pay
Both DeAngelo et al. (2006) and Grullon et al. (2011) find evidence of a declining propensity to pay dividends over time.
Grullon et al. also find an overall payout deficit, i.e. the expected number of payers predicted by their model is greater than
the actual number of payers. We re-examine these results in light of our findings on the importance of model and estimation
period in predicting payouts. As in Grullon et al. (2011), our first model includes the Fama and French (2001) controls, including
Table 10
Deficits in propensity to pay.The deficit in the propensity to pay is defined as the expected proportion of payers less the actual proportion of payers. The deficit is mea-
sured annually and the mean and median deficits for each time period are reported. Four separate prediction models are used to predict payout propensities: the Fama–
French model is used by Grullon et al. (2011) and includes SIZE, M/B, ROA_FF, and SGR (N = 16,041). Model 1 is as per Table 5 and includes RE/TA, TE/TA, ROA, SGR, and
SIZE (N = 16,553). Model 3 is as Table 5, and adds CASH and Σ5 (N = 16,553) to Model 1. These three prediction models pool observations from 1973 to 1978 to make
out of sample predictions for observations in years 1987 to 2011. The final model, Model 3B, includes the same control variables as Model 3 in Table 5 but uses a later
prediction period, 1982 to 1986 (N = 17,176), to make out of sample predictions for observations in years 1987 to 2011. Panel A examines the deficit in the propensity
to pay dividends while Panel B examines the deficit in the propensity to make shareholder payouts, including dividends and repurchases.
Fama–French Model 1 Model 3 Model 3B
Mean Median Mean Median Mean Median Mean Median
Panel A
1987–1989 0.065 0.067 0.110 0.109 0.021 0.022 −0.013 −0.013
1990–1999 0.114 0.120 0.135 0.133 0.034 0.034 0.010 0.007
2000–2011 0.058 0.093 0.163 0.168 0.005 0.003 −0.014 −0.021
Panel B
1987–1989 0.060 0.060 0.070 0.067 −0.034 −0.041 −0.020 −0.026
1990–1999 0.095 0.097 0.090 0.104 −0.033 −0.012 0.019 0.032
2000–2011 −0.071 0.014 0.024 0.033 −0.160 −0.160 −0.101 −0.094
a slightly revised definition for ROA than in our previous analyses (ROA_FF). Using the 1973–1978 period, we estimate the logit
parameters using three separate models: the Fama–French controls described above, Model 1 in Table 5, and Model 3 in Table 5.
Our final model also uses Model 3 of Table 5, but uses the 1982–1986 estimation period; we refer to this as Model 3B. We gen-
erate the probability that each firm pays out in 1987 through 2011 by applying the estimated parameters from our logit models
to the value of each firm's determinants of the decision to pay (RE/TA, profitability, growth, etc.).16 We then calculate the deficit
(expected percentage of payers − actual percentage of payers) for each year of our predicted period and calculate the mean and
median of these annual deficit measures. Negative deficits indicate that more firms paid than expected, conditional on firm
characteristics.
Panel A of Table 10 shows the prediction accuracy for dividend payers by time period. Using the Grullon et al. (2011) methodology,
the numbers in Table 10 are more summarized than the individual deficits reported in Table 9. The first two columns replicate Fig. 3 of
Grullon et al. (2011) but with our sample firms. Using this Fama–French model, the mean deficit in the propensity to pay dividends is
positive, i.e. the expected number of payers is greater than the actual number of payers, which is consistent with Grullon et al. (2011).
The deficits increase from the 1980s to the 1990s, but decrease in the 2000s.17 The next two columns perform the same analysis but
using Model 1 of Table 5. Using this model, the deficit in the propensity to pay is positive and increases over time. Columns 5 and 6
include cash and cash flow volatility as control variables, as in Model 3 of Table 5. Inclusion of these variables reduces the deficit in
the propensity of pay dividends fourfold in the 1980s and 1990s, and in the 2000s, the median deficit falls to 0.005! The final two col-
umns of the table also use Model 3 of Table 5, but with the 1982–1986 estimation period. We refer to this as Model 3B. Using this later
estimation period, there is no evidence of a deficit in the propensity to pay dividends in the 1980s or the 2000s, and the deficit in the
1990s is small.
Panel B examines the models for total payouts. With the Fama–French model, we find no evidence of a time trend in the propensity
to payout. The deficits are positive in the 1980s and 1990s, indicating that fewer firms than expected return cash to shareholders, be-
fore switching to negative in the 2000s. Using Model 1 of Table 5, the results in the 1980s and 1990s are similar to the Fama–French
model. In the 2000s, the deficit declines but remains positive. When we include cash and cash flow volatility in the model, more firms
return cash than expected in each period. Finally, using the 1982–1986 estimation period, the prediction accuracy improves (deficits
are smaller) compared to the 1973–1978 estimation period, and in the late 1980s and 2000s more firms return cash to shareholders
than expected. Since both the estimation and prediction periods in Model 3B occur after the enactment of Rule 10b-18, the improved
predictive accuracy for repurchases seems consistent with the Safe Harbor regulatory regime.
Since Model 3B estimated over 1982–1986 performs the best in Tables 9 and 10, we use this model and estimation period to ex-
amine the deficit in the propensity to pay for subsamples of our firms. Panel A of Table 11 examines the propensity to pay dividends.
The first three columns focus on the decade in which the firm went public. In the pre-1970s cohort, the mean deficit is −0.056, indi-
cating that actual payers are greater than predicted, consistent with our earlier results that dividend propensity is stronger for firms
going public in earlier decades. For the 1980s cohort, both the median and mean deficits are close to 0. The deficit becomes positive for
the 1990s cohort before turning negative again in the 2000s. Overall the results suggest that only firms that went public in the 1990s
have a deficit in the propensity to pay dividends.
16
We use 1987–2011 prediction period so that regardless of which estimation period is used, our out-of-sample period is the same.
17
We differ from Grullon et al. (2011) on key variables in our respective regression models, such as market to book, price volatility and cash volatility. Thus our sample
populations are slightly different. In untabulated results, we use our models and Grullon et al.'s sample exclusions and find similar results. Further, if we exclude the
2008–2011 observations from our sample, we find that the deficits increase in the 2000s relative to the 1990s.
Table 11
Deficits in propensity to pay by subsamples.The deficit in the propensity to pay is defined as the expected proportion of payers less the actual proportion of payers. The
deficit is measured annually and the mean and median deficits for each category are reported. Panel A (B) examines the propensity to pay dividends (make shareholder
distributions) using Model 3 in Table 5. The prediction model pools observations from 1982 to 1986 and makes out of sample predictions for observations in years 1987
to 2011. The IPO decade splits the deficits based on the year of the firm's listing date on CRSP. The age quintile split is based on the earliest of the firms founding date or
the SDC or CRSP listing date. The tax regime splits are based upon changes in capital gains and dividend tax rates. High option firms report executive stock options as a
percentage of shares outstanding above the median level of options for the time period (1992–2011). Likewise, low option firms report the percentage of stock options
outstanding below the median level for the pooled sample of firms.
IPO decade Age quintile Tax regime Options
Mean Median Mean Median Mean Median Mean Median
Panel A High
Pre-1970s −0.056 −0.050 5 (oldest) −0.012 0.003 1987 −0.009 −0.009 1992–1999 0.093 0.096
1970s −0.020 −0.015 4 0.014 0.012 1988–1992 −0.002 −0.006 2000–2011 0.056 0.062
1980s −0.008 0.003 3 0.021 0.024 1993–1995 0.008 0.004 Low
1990s 0.022 0.021 2 0.007 0.005 1996–2002 0.012 0.012 1992–1999 0.073 0.075
2000s −0.020 −0.018 1 −0.040 −0.036 2003–2011 −0.012 −0.026 2000–2011 0.023 0.022
Panel B High
Pre-1970s −0.003 −0.011 5 (oldest) 0.026 0.015 1987 −0.019 −0.019 1992–1999 0.214 0.22
1970s −0.036 −0.043 4 −0.025 −0.037 1988–1992 0.013 0.004 2000–2011 −0.032 −0.04
1980s −0.069 −0.100 3 −0.063 −0.099 1993–1995 0.054 0.055 Low
1990s −0.039 −0.068 2 −0.078 −0.102 1996–2002 −0.044 −0.064 1992–1999 0.143 0.149
2000s −0.138 −0.115 1 −0.069 −0.090 2003–2008 −0.103 −0.096 2000–2011 0.051 0.058
The next columns examine the deficit by firm age. The deficits are small in all age quintiles. The oldest and youngest firms have
negative deficits, indicating that more firms pay than expected. The positive deficits in the middle quintiles are consistent with the
positive deficit for 1990s IPO cohort firms, which Table 3 indicates has the most post-1982 observations and likely comprises the
largest portion of young to middle-aged firms.
We next examine the deficit by tax regime. The tax regimes are based upon changes in the capital gains and dividend tax rates.18 In
1987, more firms pay dividends than predicted, which is consistent with the reduced dividend tax rate in 1987 compared to the es-
timation period. However, we measure our time periods by fiscal reporting years, which may differ from the timing of the tax regimes.
Further, since dividend and capital gains tax rates changed several times between 1986 and 1988, we cannot draw strong conclusions
from the 1987 tax regime. The deficit is small but positive from 1993 to 2002, but there are more dividend payers than expected from
2003 through 2011, again consistent with the reduction in dividend taxes during this last time period. These tax regime results are
also consistent with Kulchania (2013), who finds that the relation between the dividend and repurchase premia is related to tax
effects.
In the last columns of Panel A, we examine the results for the high vs. low option intensity sample discussed in Section 4.4.4. Since
this sample is limited to Execucomp firms, it is not completely comparable to the broader sample. However, comparison of firms with
high vs. low option intensity shows that firms with high option intensity have a higher deficit, and thus are less likely to pay dividends
than firms with low option intensity. The deficit is lower in the 2000s than in the 1990s, however, consistent with the tax regime
results.
Panel B of Table 11 examines the deficits for payout firms rather than dividend paying firms and shows little evidence of deficits in
the propensity to pay across our subsamples. The deficit is the greatest, indicating that fewer firms pay than expected, in firms with
high options intensity during the 1990s. The deficits are most negative in the youngest firms, in firms that went public in the 2000s,
and during the most recent tax regime. Our overall results are inconsistent with the notion that firms have become less likely to make
shareholder payouts over time. Instead we find that propensities to pay change over IPO decade, tax regime and overall regulatory
regime. If these factors influence the repurchase premia, then our results are not inconsistent with Jiang et al. (2013) and
Kulchania (2013), who propose a catering motive for repurchases in which the choice of payout depends on investors' relative pref-
erences for dividends vs. repurchases.
6. Conclusion
Consistent with the DeAngelo et al. (2006) lifecycle theory, we find a positive relation between earned/contributed capital
and not only the fraction of firms that pay dividends, but also the fraction that repurchase. However, we find that the strength
of the relation between RE/TA and the payment of dividends has declined over time. This trend is due less to a chronological time
trend and more to the influx of new IPO firms starting in the 1980s — firms not only less profitable and riskier, but also less
18
Specifically, our first regime is 1987, when the top tax rate on dividends (capital gains) was 38.5% (28%). From 1988 to 1992 the rates on the two were approximate-
ly equal. From 1993 to 1995, the top rate on dividends (39.6%) was again higher than the rate on capital gains (28%). From 1996 to 2002, the top rate on dividends
remained at 39.6% while the rate on capital gains fell to 20%. Finally, from 2003 to 2011, the top rate on qualified dividends again equaled the top rate on capital gains,
and was the lowest dividend tax rate in history (15%).
inclined to start paying dividends (Brav et al., 2005) and more inclined to repurchase shares. While the influence of earned/
contributed capital on dividend payment remains constant over time for firms that went public in the 1970s or earlier, the
more recent a firm's public offering, the less RE/TA influences a firm's propensity to pay dividends. In spite of the fact that RE/
TA is not a good measure of lifecycle in recent IPO cohorts, a lifecycle effect does exist: within a given IPO cohort, the likelihood
of making shareholder payouts increases as firms mature.
This declining relation between RE/TA and the likelihood of payout across IPO decades remains even after controlling for size, prof-
itability, growth, total equity, cash balances, dividend history, and firm age. We also find that the influence of RE/TA on shareholder
payouts is weaker for firms experiencing high earnings volatility or high idiosyncratic risk. Even after controlling for these factors,
firms that went public in the 1990s or later are less likely to pay dividends (or payout) than firms that went public earlier. While
our results still support a lifecycle theory, they indicate that RE/TA alone is not a sufficient measure of lifecycle — a firm's “starting
point,” as measured by its public offering cohort, is also an important determinant.
Given the declining marginal impact of RE/TA on payout propensity across IPO decade, we examine the predictive accuracy of the
model. We find that using an estimation period after the enactment of Rule 10b-18 better predicts payout propensity in recent IPO
cohorts, compared to an earlier estimation period. Regardless of which estimation period is used, the model does a poor job of correct-
ly predicting which firms pay dividends, particularly at low RE/TA levels. However, the error rate for total payouts is lower at higher
RE/TA levels, consistent with repurchases substituting for dividends over time. We conclude that the ability of RE/TA to proxy for a
firm's lifecycle stage has declined over time due to economic and institutional changes which have influenced the propensity of
each IPO cohort to make shareholder payments. Additionally, the increasing use of the retirement method to record repurchases af-
fects the ability of RE to capture the earned capital equally for all firms.
Finally, we reexamine the propensity to make shareholder payouts. After controlling for the changing characteristics of firms and
using an estimation period after the enactment of the repurchase Safe Harbor, we find little evidence that the propensity to pay div-
idends or make any shareholder payout declines over time. In fact, we find that the number of firms that pay dividends or make any
shareholder payout in the 2000s is greater than expected. An examination of tax regimes indicates that this is at least partially attrib-
utable to the 2003–2011 period, when taxes on payouts are reduced. These results contrast sharply with other papers that show that
the propensity to pay dividends has declined substantially over time.
Acknowledgments
We thank Dan Dhaliwal, Amy Dittmar, Lillian Fickle, Stuart Gillan (the editor), Jarrad Harford, Jean Helwege, Vince Intintoli, Sandy
Klasa, Dave Mauer, Phil Shane, an anonymous referee, and seminar participants at Texas A&M University and the 2012 Virginia Ac-
counting Research Conference for helpful comments.
Appendix A. Variable definitions
Variable Definition
Age The number of years since the firms' founding, CRSP listing date, or SDC IPO date, whichever is earliest.
Cash holdings (CASH) The ratio of cash and marketable securities to the book value of total assets.
Dividend dummy (DIVt − 1) Dummy variable set to one if the firm paid a common dividend in that year, and zero if it did not.
Cash flow volatility (Σ5) The mean of the standard deviation of cash flow/assets over 5 years for firms in the same industry, as defined
by two-digit SIC code, where cash flow is measured as EBITDA − interest − taxes − common dividends paid.
Idiosyncratic risk The standard deviation of residuals from a regression of its daily excess stock returns (raw returns less the riskless
rate) on the market factor (i.e., the value-weighted market return less the riskless rate).
Market to book (M/B) The book value of total assets − book value of equity + market value of equity, divided by the book value of total assets.
MBCA MBCA is an indicator of 1 for any firm incorporated in a state that does not recognize treasury stock. These states exclude
DE, TX, MO, NJ, NY, DC, KS, OK, PA, LA, OH, AL, WI, IL, NV, IN, RI, VA, AK, ID, GA (which allows treasury shares if permitted
by the firm's articles of incorporation), FL before June 1990, MD before 1996, and MA before 2004.
NYE The percentile in which the firm falls on the distribution of equity market values for NYSE firms in year t.
Payout Firm Any firm which makes a dividend payment and/or a share repurchase in the current year.
RE/TA The ratio of retained earnings to total assets.
Repurchase Firm Any firm which makes a share repurchase in the current year.
ROA Earnings before extraordinary items plus interest expense plus deferred taxes from the income statement (if available),
all divided by total assets.
ROA_FF The operating income before depreciation scaled by the book value of assets.
SGR The annual percentage change in total sales.
Systematic risk The standard deviation of the predicted value from the regression used to define idiosyncratic risk.
TE/TA The ratio of total common equity to total assets.
Appendix B. Propensity of dividend payments over time 1975–2011
Payers include firms reporting non-zero common dividends (DVC) in the fiscal year ending in the respective calendar year. Old
firms met the sample screens in a prior sample year. New List firms are firms which meet the sample screens for the first time,
where the sample screens require the following available information: total equity to total assets (TE/TA); profitability (ROA); sales
growth (SGR); size (NYE); cash holdings (CASH); cash flow volatility (CF Volatility); and an indicator variable for payment of
dividends in the prior year (DIVt − 1). ListPay is the proportion of New List firms which pay dividends in the year they are included
as New Lists.
Year Payer Nonpayer % ListPay Total firms Total New List
Total Old New List Total Old New List
1975 1871 1802 69 1039 945 94 42.3 2910 163

1976 1955 1925 30 897 859 38 44.1 2852 68
1977 1998 1964 34 799 747 52 39.5 2797 86
1978 1976 1948 28 750 722 28 50.0 2726 56
1979 1953 1863 90 882 734 148 37.8 2835 238
1970s avg 1951 1900 50 873 801 72 42.8 2824 122
1980 1933 1836 97 1103 808 295 24.7 3036 392
1981 1815 1773 42 1270 1082 188 18.3 3085 230
1982 1693 1649 44 1600 1246 354 11.1 3293 398
1983 1599 1556 43 1728 1487 241 15.1 3327 284
1984 1540 1496 44 1996 1644 352 11.1 3536 396
1985 1454 1400 54 2093 1691 402 11.8 3547 456
1986 1318 1291 27 2155 1890 265 9.2 3473 292
1987 1240 1201 39 2350 1961 389 9.1 3590 428
1988 1226 1161 65 2498 2034 464 12.3 3724 529
1989 1202 1159 43 2402 2132 270 13.7 3604 313
1980s avg 1502 1452 50 1920 1598 322 13.7 3422 372
1990 1168 1133 35 2399 2181 218 13.8 3567 253
1991 1129 1107 22 2452 2208 244 8.3 3581 266
1992 1134 1114 20 2472 2191 281 6.6 3606 301
1993 1145 1106 39 2693 2298 395 9.0 3838 434
1994 1148 1108 40 2953 2466 487 7.6 4101 527
1995 1177 1129 48 3067 2614 453 9.6 4244 501
1996 1148 1117 31 3247 2793 454 6.4 4395 485
1997 1116 1074 42 3465 2878 587 6.7 4581 629
1998 1071 1037 34 3301 2871 430 7.3 4372 464
1999 956 936 20 3086 2782 304 6.2 4042 324
1990s avg 1119 1086 33 2914 2528 385 8.1 4033 418
2000 857 843 14 2989 2662 327 4.1 3846 341
2001 766 760 6 2845 2520 325 1.8 3611 331
2002 725 716 9 2650 2510 140 6.0 3375 149
2003 792 785 7 2379 2280 99 6.6 3171 106
2004 851 845 6 2233 2136 97 5.8 3084 103
2005 886 870 16 2108 1970 138 10.4 2994 154
2006 866 845 21 2052 1908 144 12.7 2918 165
2007 845 825 20 2000 1850 150 11.8 2845 170
2008 801 795 6 1985 1801 184 3.2 2786 190
2009 742 728 14 1928 1847 81 14.7 2670 95
2010 765 761 4 1781 1692 89 4.3 2546 93
2011 676 667 9 1468 1386 82 9.9 2144 91
2000s avg 798 787 11 2202 2047 155 7.6 2999 166
References
Baker, M., Wurgler, J., 2004. A catering theory of dividends. J. Financ. 60, 1125–1166.
Bates, T., Kahle, K., Stulz, R., 2009. Why do U.S. firms hold so much more cash than they used to? J. Financ. 64, 1985–2021.
Boudoukh, J., Michaely, R., Richardson, M., Roberts, M.R., 2007. On the importance of measuring payout yield: implications for empirical asset pricing. J. Financ. 62,
877–915.
Brav, A., Graham, J., Harvey, C., Michaely, R., 2005. Payout policy in the 21st century. J. Financ. Econ. 77, 483–527.
Brown, G., Kapadia, N., 2007. Firm-specific risk and equity market development. J. Financ. Econ. 84, 358–388.
DeAngelo, H., DeAngelo, L., Skinner, D.J., 2004. Are dividends disappearing? Dividend concentration and the consolidation of earnings. J. Financ. Econ. 72, 425–456.
DeAngelo, H., DeAngelo, L., Stulz, R., 2006. Dividend policy and the earned/contributed capital mix: a test of the life-cycle theory. J. Financ. Econ. 81, 227–254.
Fama, E., French, K., 2001. Disappearing dividends: changing firm characteristics or lower propensity to pay? J. Financ. Econ. 60, 3–43.
Fama, E., French, K., 2005. Financing decisions: who issues stock? J. Financ. Econ. 76 (3), 549–582.
Fama, E.F., MacBeth, J.D., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political Economy 81, 607–636.
Frydman, C., Saks, R., 2010. Executive compensation: a new view from a long-term perspective. 1936–2005. Rev. Financ. Stud. 23, 2099–2138.
Gao, X., Ritter, J., Zhu, Z., 2013. Where Have All the IPOs Gone? J. Financ. Quant. Anal. 48 (6), 1663–1692.
Grinstein, Y., Michaely, R., 2005. Institutional holdings and payout policy. J. Financ. 60 (3), 1389–1426.
Grullon, G., Michaely, R., 2002. Dividends, share repurchases and the substitution hypothesis. J. Financ. 57 (4), 1649–1684.
Grullon, G., Michaely, R., Swaminathan, B., 2002. Are dividend changes a sign of firm maturity. J. Bus. 35 (3), 387–424.
Grullon, G., Paye, B., Underwood, S., Weston, J., 2011. Has the propensity to pay out declined? J. Financ. Quant. Anal. 46 (1), 1–24.
Guay, W., Harford, J., 2000. The cash flow permanence and information content of dividend increases vs. repurchases. J. Financ. Econ. 57, 385–415.
Hoberg, G., Prabhala, N., 2009. Disappearing dividends, catering, and risk. Rev. Financ. Stud. 22 (1), 79–116.
Jagannathan, M., Stephens, C.P., Weisbach, M.S., 2000. Financial flexibility and the choice between dividends and stock repurchases. J. Financ. Econ. 57, 355–384.
Jensen, M.C., 1986. Agency costs of free cash flow, corporate finance, and takeovers. Am. Econ. Rev. 76, 323–329.
Jiang, Z., Kim, K., Lie, E., Yang, S., 2013. Share repurchases, catering, and dividend substitution. J. Corp. Financ. 21, 36–50.
John, K., Williams, J., 1985. Dividends, dilution, and taxes: a signaling equilibrium. J. Financ. 40, 1053–1070.
Jovanovic, B., Rousseau, P., 2001. Why wait? A century of life before IPO. Am. Econ. Rev. 101, 336–341.
Julio, B., Ikenberry, D., 2004. Reappearing dividends. J. Appl. Corp. Financ. 16 (4), 89–100.
Kahle, K., 2002. When a buyback isn't a buyback: open market repurchases and employee options. J. Financ. Econ. 63, 235–261.
Kulchania, M., 2013. Catering driven substitution in corporate payouts. J. Corp. Financ. 21, 180–195.
Lintner, J., 1956. Distribution of incomes of corporations among dividends, retained earnings, and taxes. Am. Econ. Rev. 46, 97–113.
Miller, M., Rock, K., 1985. Dividend policy under asymmetric information. J. Financ. 40, 1031–1051.
Murphy, K., 2013. Executive compensation: where we are, and how we got there. In: Constantinides, G., Harris, M., Stulz, R. (Eds.), Handbook of the Economics of
Finance, vol. II. Elsevier, Amsterdam, pp. 211–356.
Newey, W., West, K., 1987. A simple, positive definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55, 703–708.
Skinner, D.J., 2008. The evolving relation between earnings, dividends, and stock repurchases. J. Financ. Econ. 87 (3), 582–609.

Declining Propensity To Pay A Re-Examination of The Lifecycle Theory

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Declining Propensity To Pay A Re-Examination of The Lifecycle Theory

Uploaded by

Copyright:

Available Formats

Journal of Corporate Finance 27 (2014) 345–366

Contents lists available at ScienceDirect

Journal of Corporate Finance

Declining propensity to pay? A re-examination of the

2. Literature review and hypotheses

2.1. The lifecycle theory

2.2. Repurchases, lifecycle, and the propensity to pay

3. Data and methodology

3.1. Sample selection

3.2. Time periods and IPO cohorts

3.3. Accounting for repurchases and retained earnings

4. Univariate results of RE/TA and lifecycle

4.1. Payout as a function of RE/TA

4.2. Payout as a function of RE/TA and IPO decade

Ratio of retained earnings to total equity (RE/TA)

Ratio of earned equity to total assets (RE/TA)

Panel A: Dividend Payers

Panel C: Payout Firms

Dividend Payer Repurchaser Payout Firm

4.3. Payout as a function of RE/TA, IPO decade, and time

4.4. Multivariate analysis of RE/TA and lifecycle

4.4.1. Payout as a function of RE/TA

Panel A Ratio of earned equity to total assets (RE/TA)

IPO in the 1970s

IPO in the 1980s

IPO in the 1990s

IPO in the 2000s

Panel B Ratio of earned equity to total assets (RE/TA)

(continued on next page)

Panel B Ratio of earned equity to total assets (RE/TA)

IPO in the 1970s

IPO in the 1980s

IPO in the 1990s

IPO in the 2000s

4.4.2. Payout as a function of RE/TA, IPO decade, and time

M.L. Banyi, K.M. Kahle / Journal of Corporate Finance 27 (2014) 345–366

Panel A: dividend decision Panel B: payout decision

(1) (2) (3) (4) (1) (2) (3) (4)

Dividend ﬁrms Payout ﬁrms

Marginal effects by time period

Marginal effects by time period

(1) (2) (3) (4) (5)

Panel A: propensity to make dividend payments

Panel B: propensity to make shareholder payouts

4.4.3. IPO decade vs. age?

4.4.4. What drives the difference across IPO cohorts?

Dividend payer Payout ﬁrm

(1) (2) (3) (4) (5) (6) (7) (8)

M.L. Banyi, K.M. Kahle / Journal of Corporate Finance 27 (2014) 345–366

5. Prediction accuracy and the propensity to pay

5.1. Prediction accuracy

Ratio of earned equity to total equity (RE/TA) All

5.2. The propensity to pay

Fama–French Model 1 Model 3 Model 3B

Mean Median Mean Median Mean Median Mean Median

IPO decade Age quintile Tax regime Options

Mean Median Mean Median Mean Median Mean Median

Appendix A. Variable deﬁnitions

Appendix B. Propensity of dividend payments over time 1975–2011

Year Payer Nonpayer % ListPay Total ﬁrms Total New List

Total Old New List Total Old New List

1975 1871 1802 69 1039 945 94 42.3 2910 163