Jofi 13196

THE JOURNAL OF FINANCE • VOL. LXXVIII, NO.
1 • FEBRUARY 2023
How Risky Are U.S. Corporate Assets?

TETIANA DAVYDIUK, SCOTT RICHARD, IVAN SHALIASTOVICH,
and AMIR YARON*
ABSTRACT
We use market data on corporate bonds and equities to measure the value of U.S.
corporate assets and their payouts to investors. In contrast to equity dividends, total
corporate payouts are highly volatile, turn negative when corporations raise capital,
and are acyclical. At the same time, corporate asset returns are similar to returns
on equity, and both are exposed to fluctuations in economic growth. To reconcile this
evidence, we argue that acyclical but volatile net repurchases mask the exposure of
total payouts’ cash components to economic growth risks. We develop an asset pricing
framework to quantitatively illustrate this economic channel.
THE KEY QUESTION IN FINANCIAL ECONOMICS concerns the trade-off between

risk and return in financial markets. An economic analysis of risk and return
has been a subject of the large body of macrofinance literature ranging from the
models of business-cycle risk of Mehra and Prescott (1985) and habits of Camp-
bell and Cochrane (1999) to the long-run risks of Bansal and Yaron (2004) and
rare disasters of Rietz (1988) and Barro (2006). Such analyses focus nearly
exclusively on per-share investments in the equity of a firm, sector, or the mar-
ket index. The price of such equity share is determined by the present value
* Tetiana Davydiuk is at Carnegie Mellon University. Ivan Shaliastovich is at the University
of Wisconsin Madison. Scott Richard is at the University of Pennsylvania. Amir Yaron is at the
University of Pennsylvania, NBER, and the Bank of Israel. We are thankful to Editor Stefan
Nagel; the Associate Editor; two anonymous referees; Sebastien Betermier; Harjoat Bhamra; Nuno
Clara; Winston Dou; Gustavo Grullon; Lars Lochstoer; Asaf Manela; Sebastien Plante; Abraham
Ravid; Michael Roberts; Juliana Salomao; Robert Stambaugh; Paul Whelan; Fernando Zapatero;
and seminar participants at Baylor University, BI Norwegian Business School, Federal Reserve
Bank, Goethe University, Hong Kong University of Science and Technology, John Hopkins Uni-
versity, Wharton Business School, University of Hong Kong; and conference participants at UBC
Winter Finance Conference, Utah Winter Finance Conference, Annual Young Scholars Finance
Consortium, HEC - McGill Winter Finance Workshop, Minnesota Macro Asset Pricing Conference,
Annual Conference in Financial Economics Research by Eagle Labs, SFS Cavalcade, European Fi-
nance Association Meeting, European Winter Finance Conference, Northern Finance Association
Meetings, European Summer Symposium in Financial Market, Early Career Women in Finance
Conference, CMU/OSU/PITT/PSU Conference, Econometric Summer Society Meeting, and Society
of Economic Dynamics Meeting. We have read The Journal of Finance disclosure policy and have
no conflicts of interest to disclose.
Correspondence: Tetiana Davydiuk, Department of Finance, Carnegie Mellon University, 5000
Forbes Avenue, Pittsburgh, PA 15213; e-mail: tetianad@andrew.cmu.edu
DOI: 10.1111/jofi.13196
© 2022 the American Finance Association.
141
142 The Journal of Finance®
of its payout, which is dividend per share. However, financial markets provide
a variety of other means through which corporations channel resources to and
from investors. Firms borrow money by issuing corporate bonds, and thus dis-
tribute funds through interest and principal payments. Corporations further
issue and repurchase equity and debt, which represents negative and positive
payouts to investors, respectively. The unifying feature of these distributions is
that they are readily accessible to investors through regular financial markets
by investing or disinvesting in the tradable corporate assets of firms, similar
to standard per-share equity investments.
A strategy that is particularly appealing economically is an investment in
the entire market supply of corporate assets. Corporate assets consist of trad-
able equity and debt, and corporate payouts comprise cash distributions in the
form of dividend and interest payments as well as the proceeds from stock and
debt issuances and repurchases. Holding the entire supply of corporate assets
enables investors to own the whole corporate sector and receive total corpo-
rate distributions through regular dividend and coupon payments and as well
as changes in entity and capital structure due to initial or seasoned share of-
ferings, stock and bond repurchases, mergers, acquisitions, bankruptcies, and
so on. Economically, the aggregate, as opposed to per-share equity strategy, is
the closest match to a typical macrofinance paradigm that features ownership
and transfer of aggregate resources between the representative firm and the
investor.1
Conceptually, accounting for total corporate payouts and valuations faces
multiple challenges in the data. Unlike market equity data, which are directly
available at a firm and index level, market data for debt are not as easily
accessible. Prior studies typically rely on book values for debt, which, as we
show, provide imperfect proxies for market valuations and payouts. In addi-
tion, many popular accounting measures do not accurately capture publicly
traded debt and the payouts on total corporate assets. Firms also own cor-
porate assets themselves, and such cross-holdings further affect the residual
payouts and valuations to investors.
To address these empirical challenges, we rely on a variety of sources for
equity and debt data at the firm and index level, such as the Bloomberg
Barclays corporate bond index2 database for market data on bonds, CRSP for
market data on equity, and Compustat, Flow of Funds, and Mergent for book
values of debt. We combine these data to characterize market prices, returns,
and payouts to the aggregate strategy of holding the entire supply of U.S.
corporate assets.
1 Corporations hold other claims to and against customers, suppliers, banks, governments, and
other economic entities: for example, accounts receivables, trade credits, bank loans, bank deposits,
policy, and tax liabilities. Generally, these claims cannot be directly accessed through regular fi-
nancial markets. As such, they are not part of our aggregate strategy.
2 The Barclays Capital Aggregate Bond Index was acquired by Bloomberg in 2016, and renamed
the Bloomberg Barclays U.S. Aggregate Bond Index. In August 2021, the index was further re-
named the Bloomberg U.S. Aggregate Bond Index.
How Risky Are U.S. Corporate Assets? 143
Our empirical analysis highlights several features of total corporate asset

payouts that are distinct from standard per-share equity dividends. First, total
asset payouts, which consist of equity dividends, debt coupon payments, and
net repurchases of equity and debt, are considerably more volatile than the eq-
uity dividend or interest payment components. Naturally, the extra variation
is due to net repurchases, which in the data are about 10 times more volatile
than dividend and coupon payments and, as such, are the central determinant
of the variation in total payouts. In fact, unlike equity dividends, total payouts
often turn negative, meaning that there are periods when the aggregate corpo-
rate sector receives funds from investors rather than paying them out. In our
sample from 1975 to 2017, total payouts go below zero about the third of the
time for equity, two-thirds of the time for bonds, and nearly half of the time for
corporate assets.
From the asset pricing perspective, the key question is whether the fluctua-
tions in asset payoffs are exposed to systematic economic risks. In particular,
a large body of macrofinance research hinges on payout comovements with
aggregate economic growth to characterize the asset’s risk compensation. Con-
sistent with this literature, we find that cash payouts on equity, debt, and as-
sets comove positively with real consumption and output growth at horizons
ranging from one quarter to five years. The correlations of changes in cash pay-
outs with economic growth rates, or, alternatively, the betas of cash payouts to
growth risks, are always positive. Quantitatively, the correlations of consump-
tion (output) growth with changes in asset cash payouts, which comprise equity
dividends and bond coupon payments, increase from 20% at a one-quarter hori-
zon to over 30% (40%) at a five-year horizon. Naturally, the long-run inference
in small samples is subject to a considerable amount of statistical uncertainty.
We find that the estimates of the cash payout cyclicality are significant at a 5%
level using Newey-West corrections or Hodrick (1992)-type inference based on
VAR(1). A more conservative long-term covariability inference approach, such
as Whitcher, Guttorp, and Percival (2000) or Müller and Watson (2018), pro-
duces very similar point estimates but substantially larger standard errors:
the correlation estimates are significant or borderline significant using 67%
confidence intervals (CIs).
While cash payouts appear procyclical, the empirical evidence indicates that
total payouts are acyclical both at short and long horizons. The correlations
and betas of total asset payouts with respect to consumption and output
growth are near zero or even negative, and are statistically insignificant from
zero at all considered frequencies and for any statistical inference method.
The difference in properties of total versus cash payouts can again be at-
tributed to the dynamics of net repurchases, whose volatile fluctuations are
uncorrelated with economic growth. Intuitively, while aggregate issuances
and repurchases tend to separately increase during economic expansions, on
net basis the two offset each other, resulting in acyclical total payouts. Much
of these adjustments takes place along the debt side of the corporations.
The risk in total asset payouts appears largely unrelated to fluctuations in
economic growth. At the same time, the returns on corporate assets are very
similar to the returns on equity: asset returns average 6.9%, comparable to

7.9% for equity, and the correlation between the two is in excess of 99%. Fur-
thermore, both asset and equity returns are positively exposed to movements
in consumption and output. Similar to cash payouts, most of the return corre-
lations with economic growth are statistically significant under the 95% CIs
based on Newey-West or VAR(1) inference, and fall inside the 67% CI based on
the long-term covariability inference of Whitcher, Guttorp, and Percival (2000)
and Müller and Watson (2018).
The joint evidence on total asset payouts and returns is thus reminiscent of
the Mehra and Prescott (1985) equity premium puzzle, restated and amplified
in the context of corporate assets. If fluctuations in total asset payouts are un-
related to movements in economic growth, why do corporate assets command
a large risk premium, and why are asset returns exposed to growth risks, es-
pecially at long horizons? To provide a potential explanation, we argue that
nonsystematic fluctuations in net repurchases mask the exposure of cash pay-
outs to economic growth risks. The total payouts appear acyclical in spite of
a considerable amount of systematic growth risk that affects corporate asset
prices and the risk premia.
The idea of volatile and nonsystematic aggregate net repurchases can be ap-
plied in various models and for different notions of aggregate risk. We embed it
in the long-run risk model of Bansal and Yaron (2004), which assigns a promi-
nent role to low-frequency growth risks. We show that the model can quantita-
tively account for the payout and return dynamics in the data, and reproduce a
large amount of statistical uncertainty around the small-sample estimates. We
provide a methodological contribution to the literature by proposing an alter-
native log-linearization of returns applicable to cases with negative payouts.
Related literature
Our focus on broader notions of payouts is related to several strands of the
literature. Closest to our work are Bansal and Yaron (2007), Larrain and Yogo
(2008), and Choi and Richardson (2016). Larrain and Yogo (2008) use standard
return decomposition to analyze the connection between total payouts and as-
set price fluctuations. Importantly, they measure payouts using book rather
than market values of debt as in our work. Bansal and Yaron (2007) focus on
total payouts in the equity market and provide evidence for equity return and
equity payout growth predictability. Relative to these studies, our main focus is
on the cyclicality and exposure of payouts to economic growth risks. This mar-
gin also differentiates us from Choi and Richardson (2016), who use market
data on bonds and equity to characterize and compare the dynamics of asset
and equity volatility.
Our analysis of total payout dynamics is also related to the question of the
cyclical behavior of debt and equity financing as in Covas and Haan (2011),
Jermann and Quadrini (2012), and Begenau and Salomao (2019). These stud-
ies typically rely on book values of debt from Compusat or the Flow of Funds
and highlight the importance of firm heterogeneity, such as across size.
Fama and French (2001) and Grullon and Michaely (2002) are among the
early papers that highlight the changing nature of firms’ payouts and an in-
creasing substitution from dividends to share repurchases. Guay and Har-
ford (2000), Jagannathan, Stephens, and Weisbach (2000), and Dittmar and
Dittmar (2004) discuss the role of repurchases as a preferred form of dis-
tributing the transitory component of earnings when dividend policy requires
financial commitment. Bansal, Dittmar, and Lundblad (2005) incorporate re-
purchases in their alternative measure of dividends to measure cash flow risk.
Boudoukh et al. (2007) find that total equity payouts, which include repur-
chases and issuances, provide stronger evidence for return predictability than
cash dividends alone. Dichev (2007) introduces dollar weighting for the evalu-
ation of average stock returns when investors’ payouts include share issuances
and repurchases. On the corporate finance side, Butler et al. (2011) consider
market-timing theories of net financing and the relation between the composi-
tion of net financing and future stock returns.
Our key findings are consistent with broad evidence in Bansal, Dittmar,
and Lundblad (2005), Julliard and Parker (2005), and Hansen, Heaton, and
Li (2008) and the basic premise of the long-run risks model of Bansal and
Yaron (2004), who identify low-frequency movements in economic growth as
a key source of risk in equity markets. For related work on corporate bond
returns, Bhamra, Kuehn, and Strebulaev (2010) and Chen (2010) show the
importance of low-frequency economic growth risks for the choice of capital
structure, leverage, and the riskiness of corporate bonds. Ferson, Nallareddy,
and Xie (2013) show the role of growth risks in the cross section of equity and
corporate bond returns.
Our empirical findings are also important for interpreting the expanding lit-
erature on production-based asset pricing (see Jermann (1998), Kaltenbrunner
and Lochstoer (2010), Croce (2014), and Kung and Schmid (2014), among many
others). In that literature, dividend dynamics are often countercyclical because
productivity improvements are associated with the desire to invest rather than
pay dividends. The encompassing notion of dividends in these models is closely
related to our total payouts. Our evidence for total payouts accords well with
the implications of these models, and can be viewed as a new and relevant
empirical benchmark for their evaluation.
The remainder of the paper is organized as follows. Section I provides the
main empirical analysis, and Section II contains robustness and extensions. In
Section III, we consider an economic model to interpret the empirical evidence.
Section IV concludes. All supportive evidence is relegated to an Appendix.
I. Empirical Analysis
A. Payouts and Valuations
In this section, we describe the key relationships between the valuations
and payouts that underlie our empirical measurements. Unlike the majority
of the literature, which considers per-share investments in equity, we focus on
the aggregate strategy defined as the claim to the entire supply of investable
corporate capital. As shown in Bansal and Yaron (2007) and Larrain and Yogo
(2008), the payoff on this aggregate strategy can be decomposed into standard
cash distributions in the form of dividends and interest payments, as well as
distributions associated with the issuance and repurchase of equity and debt.
For simplicity, we lay out our discussion for an individual asset. Corporate
sector quantities appropriately aggregate the valuations and payouts across
all firms’ corporate assets.
Borrowing from the analysis in section 5.1 of Larrain and Yogo (2008), we
start with a standard return-payoff relationship for holding one share of an
asset between period t and t + 1 (i.e., per-share strategy),
Pt Rt+1 = Pt+1 + CFt+1 . (1)
One share of an asset is bought at price Pt and next period earns gross return
Rt+1 . Its total dollar payoff, Pt Rt+1 , is split between a capital gain Pt+1 and the
cash payout CFt+1 . The cash payout corresponds to the cash dividend or coupon
payment on equity or bond, respectively.
Now consider an investment strategy that holds Nt shares of the asset. When
Nt corresponds to all publicly available shares, such investment tracks the
entire supply of a firm’s equity or debt capital, and after aggregation across all
firms and assets it defines a claim to the entire corporate sector (i.e., aggregate
strategy). At time t, the value of the strategy is Vt = Pt × Nt . Because it is
invested entirely in the underlying asset, its per-dollar return is equal to the
asset return Rt+1 . Hence, the total dollar payoff is Vt × Rt+1 . The total dollar
payoff represents the next-period capital gain Vt+1 and the total payout to the
investor Da,t+1 ,
Vt Rt+1 = Vt+1 + Da,t+1 . (2)
The capital gain is the market value of the available shares, Vt+1 = Pt+1 × Nt+1 ,
and the aggregate payout is made up of the aggregate cash payouts and share
issuances and repurchases,
Da,t+1 = Vt Rt+1 − Vt+1 = (Pt+1 + CFt+1 ) × Nt − Pt+1 × Nt+1

= Nt × CFt+1 + (Nt − Nt+1 ) × Pt+1 (3)
≡ Dt+1 + NREPt+1 ,
where Dt+1 ≡ Nt × CFt+1 is the aggregate cash payout and NREPt+1 = (Nt −
Nt+1 ) × Pt+1 is net repurchases. The latter corresponds to the difference be-
tween repurchases and issuances, NREPt+1 ≡ REPt+1 − ISSt+1 , and captures
the net transfer of resources out of the firm due to the change in the number
of shares. The outflow at date t + 1 is given by
REPt+1 ≡ −{Nt+1 − Nt }− Pt+1 ≥ 0. (4)

It is positive when there is a repurchase of existing shares, that is, when

Nt+1 − Nt ≤ 0. Similarly, issuances are given by
ISSt+1 ≡ {Nt+1 − Nt }+ Pt+1 ≥ 0. (5)
This represents the inflow of resources following a new issuance of shares

when Nt+1 is greater than Nt .
There are several important differences in the economic interpretation and
empirical measurement of the per-share and aggregate strategies in (1) and
(2). First, the two strategies earn the same return but have different dollar
payouts and valuations to investors. Indeed, because money is invested in
the same asset, the return is the same per every dollar. The way the return
is split between capital gains and the payout, however, varies across the two
strategies. By definition, the per-share payout on a stock is just a per-share
dividend. For an aggregate strategy, repurchases (selling shares back to the
firm) act like a positive dividend, while issuances (buying shares from the
firm) correspond to a negative dividend. Accounting for share issuances and
repurchases can thus affect the interpretation of positive and negative pay-
outs and their connection to market valuations. For example, a decrease in
dividends is often viewed as bad news for the firm, and may lead to a decrease
in the per-share price of equity. However, a decrease in total payout caused by
large share issuance can actually represent good times for a firm seeking more
resources for profitable investment opportunities. Investors may be willing to
provide new capital and bid up valuations at times of low or even negative
aggregate payouts.
This discussion also suggests that the distributions from the aggregate strat-
egy do not always have to be positive. For the per-share strategy, its payout is
given by the dividend or interest payments, which can never fall below zero.
For the aggregate strategy, however, resources can flow from investors to the
corporate sector at times of sizeable security issuances, and the need for new
corporate capital can be large enough to leave the aggregate investor with
a negative total payout. As long as investors receive other income to ensure
positive consumption, they may be willing to provide capital for the corporate
sector and accept a negative payout. We revisit the economic and empirical
implications of the negative payouts in further detail in Section II.D.
Finally, the aggregate strategy constitutes a better empirical benchmark for
many macrofinance models than the per-share strategy. The literature rou-
tinely uses per-share dividends and equity prices as data counterparts for
model payouts and valuations. This is justified under the assumption that eq-
uity is the only means of raising capital and dividends are the only means
of distributing payouts—but as we argue below, debt payments and asset re-
purchases are dominant components of total payouts and significantly affect
their properties. To take a broader notion of payouts into consideration, the
models should target total payouts, which encompass all aspects of resource
distributions in the forms of dividends, coupons, and stock and bond issuance
and repurchase. Our paper thus contributes to this literature by providing and
characterizing empirical measurements relevant for such analysis.
B. Data and Empirical Measurements

We use market data on prices, shares, and distributions to characterize the
returns and payouts on corporate assets accessible to investors through finan-
cial markets. For the benchmark analysis we focus on the aggregate market,
which consists of the entire cross section of public corporations, including the
financial sector. Financials such as banks, holding companies, and insurance
companies generate important value-added for investors and the economy and
hence are part of our aggregate strategy. In contrast, we exclude investment
vehicles such as mutual funds, closed-end funds, and exchange-traded funds,
which provide the means for investors to access the markets. Our analysis
effectively treats these establishments as part of the representative investor.
Separating their market value-added would require detailed data on holdings
and is beyond the scope of this paper. Notably, our analysis excludes other
forms of publicly traded bond instruments, such as residential asset-backed
securities (ABS) and mortgage-backed securities (MBS) and government debt.
Securitization enables institutions to repackage and sell claims to illiquid
assets to investors, so including noncommercial ABS and MBS would result in
double-counting of the valuations and payouts from these assets. Government
is not part of the corporate sector, so we exclude its debt from the analysis.
We also exclude government-sponsored enterprises (GSEs), which have the
implicit backing of the U.S. government.3 We present a variety of checks on
the components of corporate assets, the data sources, and our measures in
Section II.
Our measurement exercise faces several empirical challenges. Market data
for bond prices and distributions are not as readily available as for equities,
and thus the majority of studies in the literature have resorted to book rather
than market valuations.4 However, market values of debt, especially at long
maturities, can significantly drift away from their book values over time; we
examine this issue in Section II.C. Second, various types of accounting liabil-
ities (e.g., accounts payable and bank loans for firms, deposits for banks, and
policy liabilities for insurance companies) are not publicly traded, and hence
their valuations and payouts have to be excluded from our corporate asset mea-
sures. Finally, aggregating across all corporations raises a double-counting con-
cern because firms own corporate assets themselves (see Duchin et al. (2017)
and Darmouni and Mota (2020)). In our empirical implementation, we seek to
address these issues subject to limitations and availability of the data.
3 Before the Financial Crisis, the GSE equity value was less than 5% of the entire market, so
including GSEs would not materially affect our results.
4 It has been common to use book values to capture debt valuations or to approximate market
values by imputing the maturity distribution of long-term debt as pioneered by Brainard and
Shoven (1980). See also Hall et al. (1988), Richardson and Sloan (2003), and Larrain and Yogo
(2008).
Equity-related data. To measure equity-related variables, we use the Center

for Research in Security Prices (CRSP) Monthly Stock File. This data set
provides equity price per share (prc) and share data (shrout) at the individual
security level, as well as holding-period returns including and excluding
dividends, ret and retx, respectively. We include only common stocks listed
on the NYSE, Amex, NASDAQ, and NYSE Arca stock exchanges. Similar to
Boudoukh et al. (2007) and Larrain and Yogo (2008), we measure individual
stock i’s net repurchases in month t as the change in shares outstanding
valued at the month-end share price,5
nrepit = prc∗it−1 × shroutit−1

∗
× (1 + retxit ) − prc∗it × shroutit∗ , (6)
where the stock price and number of shares are appropriately adjusted
by the cumulative adjustment price and share factors that account for
splits and other corporate events, specifically, prc∗it = prcit /c f acprit and
shroutit∗ = shroutit × c f acshrit . Different from Boudoukh et al. (2007) but
similar to Larrain and Yogo (2008), we also account for changes in entity
structure due to initial public offerings (IPOs), mergers, acquisitions, and
exchanges. Conceptually, this is aligned with our aggregate strategy, which
calls for buying or selling asset shares as they become available or disappear
from financial markets, respectively. We use a firm’s market capitalization
in the first trading month to measure net issuances during the IPO. We use
CRSP delisting data to identify securities with delisting codes of 2xx and
3xx, and measure their delisting price (dlprc) and delisting return (dlretx) to
account for repurchases during mergers and acquisitions. In Section II.B, we
show that our main results are not sensitive to changes in entity structure.
We aggregate the firm-level data and compute market valuations, dividends,
returns, and net repurchases at the aggregate level.6
Corporations buy and hold equity shares of other firms. Equity cross-
holdings lead to double-counting in our measurements: payouts and valua-
tions of the same equity are counted once at the issuer level and a second time
through the valuations and payouts of the corporate buyer. We use table L.223
from the Flow of Funds to compute the market value of corporate equity held
by nonfinancials, banks, insurance companies, and funding corporations. In
5 This is equivalent to measuring net repurchases as the value of the change in the number of
shares over the period,

∗
nrepit = shroutit−1 − shroutit∗ × prc∗it .
Valuing net repurchases as the average of beginning- and end-of-month prices, as in Boudoukh
et al. (2007), instead of the month-end prices as in Larrain and Yogo (2008), does not impact
our results.
6 Bansal and Yaron (2007) and Welch and Goyal (2008) measure aggregate net repurchases
directly in the market index data as MCAPt−1 × (1 + VW RETXt ) − MCAPt , where MCAP is the
market capitalization and VWRETX is the value-weighted return excluding distributions. The
index and firm-based approaches treat differently firm exits for reasons other than mergers, ac-
quisitions, and exchanges; for example, defaults and bankruptcies would show up as positive net
repurchases using index data but not with our approach using firm-level data. Empirically, how-
ever, the two measures are quite similar.
our sample, average equity cross-holdings account for about 14% of aggregate
equity value, with this value dropping to under 10% over the last 15 years. We
use these data to adjust the market value of the aggregate equity available
to investors. Unfortunately, we do not have access to detailed holdings data
to determine the returns and payouts to the equity held by corporations. As
a first approximation, we assume that the returns to corporate-owned equity
are the same as those to the aggregate stock market index. This allows us to
infer the payouts to corporate equity holdings, and adjust the payouts on the
aggregate strategy.
Debt-related data. U.S. corporations issue a wide variety of debt instruments.
Unlike equity, however, debt obligations are typically traded at the over-the-
counter (OTC) dealer’s market. As such, there is no convenient centralized
platform to obtain market valuations and distributions for the total debt is-
sued by the firm and identify its publicly traded components. As we discuss
in greater detail in Section II.C, data sources like Flow of Funds or Mergent
report book values of debt, which could be inaccurate measures of its market
value, while data sources for market bond prices, such as the Trace database,
have limited coverage before the mid-2000s.7 Furthermore, accounting state-
ment data available through Compustat routinely lump public and bank debt
together. However, bank debt is not part of our aggregate strategy. It is gen-
erally not publicly traded, and in any case, adding bank debt would lead to
double-counting: banks are part of our aggregate investment strategy, and the
proceeds from loans they receive are already reflected in the valuations and
payouts on their corporate assets.
To tackle these issues, we use the bond market data from Bloomberg Bar-
clays Indices. The Indices are widely used in the financial industry because
of their accuracy and market coverage. Reported market capitalizations and
month-to-date index returns are updated on a daily basis, and our data are
taken on the last trading day of the month when bond prices are hand-marked
by traders. Unlike for equities, we do not have access to individual bond data,
so the payout and valuation computations are conducted at the index level.
Figure A.1 compares bond coverages by the Bloomberg Barclays Indices,
the Flow of Funds, Mergent, and Compustat. Based on the underlying book
values, Bloomberg Barclays Indices track Mergent quantities fairly well: the
two exhibit similar trends and growth over the sample period. Mergent book
values are below Bloomberg and Barclays prior to 1985. Mergent coverage
improves over time, and post-1985 the book value of bonds from Mergent
exceeds Bloomberg and Barclays by just under 20% on average, and by less
than 10% since the mid-2000s. This discrepancy is due in part to the fact that
Bloomberg Barclays Indices omit debt at shorter maturities. As we discuss
below, we supplement the Bloomberg Barclays index with other data sources
to help account for short-term debt. For the Flow of Funds, the data comprise
both public and private firms, and the total debt value is over 1.5 times that of
7 Ongoing work to construct reliable and comprehensive data sets for market bond prices in-
cludes Choi and Richardson (2016) and Gomes, Kilic, and Plante (2020).
Bloomberg Barclays or Mergent. Finally, Compustat data include bank debt, so

the level of total debt is several times larger than that based on other sources.
Bloomberg Barclays Indices represent many types of debt instruments, vary-
ing from debentures and asset-backed bonds to commercial paper issues, and
our goal is to measure all of the outstanding corporate debt. To capture long-
duration debt, we include the following subindices of the Bloomberg Bar-
clays U.S. Universal Index: Corporate Investment Grade (IG), Corporate High
Yield (HY), 144A Ex Aggregate, corporate issues of Mortgage-Backed Securi-
ties (CMBS), and Fixed Rate ABS. All of the bonds in the above subindices
have fixed-rate coupon, are fully taxable, include both senior and subordinate
debt, and have at least one year to maturity.8 Additional details for bond char-
acteristics are given in Table A.I. We further augment our debt measure with
corporate issues of taxable municipal bonds, in particular, Industrial Develop-
ment Revenue Bonds (IDR), Pollution Control Revenue Bonds (PCR), and U.S.
Convertibles Composite Index, since they are outside of the Universal Index.
To measure debt of short duration, we include the following Bloomberg
Barclays subindices: corporate issues of Asset-Backed Securities Floating
Rate (ABS FRN), Floating-Rate Notes (FRN), and Floating-Rate Notes High
Yield (FRN HY). The floating-rate securities in the above subindices may have
longer maturity, but their interest rate durations are typically less than one
year. We further augment our measure with short-term debt valuations from
Compustat and Flow of Funds. For many of these instruments, we have to
rely on reported book values. However, the value of short-duration debt is less
sensitive to movements in interest rates, and book values provide reasonable
assessments of the market valuations.
Specifically, we first include the short-term borrowing through commercial
paper. While the commercial paper amounts are sparsely populated in Compu-
stat, Flow of Funds track issuing and lending in the commercial paper markets
by investor class (table L.209). To compute the payouts accessible to the repre-
sentative investor, we consider financial and nonfinancial issuers and we use
a three-month commercial paper rate pre-1997 and three-month financial and
nonfinancial commercial paper rates post-1997 to construct coupon cash flows.
Second, we include long-term debt due in less than one year (Compustat
item dd1), which represents the total amount of long-term interest-bearing
obligations due in the current year. In addition to corporate debt, it includes
nontradable obligations, such as bank loans, mortgages, leases, and so on. The
database does not separate these items, so using the total value of debt due in
one year overestimates the value of tradable short-term debt, while excluding
it completely would underestimate it. We choose to incorporate the Compustat
measures in the benchmark analysis. In Section II.C, we provide robustness
checks using short-term debt quantities from the Mergent database over
8 The Universal Index excludes bonds that have less than one year to maturity as they become
money market eligible. Corporate issues of ABS and CMBS must have a remaining average life of
at least one year, while bonds that convert from fixed to floating rate will exit the subindices one
year prior to conversion.
a more recent sample. Coupon cash flows are constructed by multiplying

valuations by the commercial paper rate.
We combine the data across different sources to come up with our proxy for
the aggregate value of corporate bonds and decompose its total payouts into
cash (interest payments) and net repurchase components. Similar to equity, we
face a double-counting issue due to cross-holdings of corporate debt. One of the
largest buyers of corporate bonds is insurance companies, which are included
in our benchmark measures. To adjust for their bond holdings, we rely on the
aggregate corporate bond positions of the insurance companies from the Flow
of Funds. Specifically, we rely on tables L.116 and L.115 for life and property-
casualty insurers, respectively.9 We scale the reported market values by the
ratio of total assets of public insurance companies from Compustat to total
assets of public and private insurance companies from the Flow of Funds to
adjust for the fact that the Flow of Funds does not separate publicly traded
and closely held companies.10
Corporate bonds are also held by other financial and nonfinancial companies.
Flow of Funds table L.213 delineates bond holdings across categories of finan-
cial institutions besides insurance companies, which allows us to quantify and
adjust for the market value of their bond holdings. To adjust for cross-holdings
of corporate bonds held by nonfinancials, we use measures from Darmouni
and Mota (2020), who hand collect the data for 200 nonfinancial firms between
2000 and 2019 from the 10-K filings.11 The bond holdings of nonfinancials are
concentrated in a few large companies included in Darmouni and Mota (2020)
sample, which helps address potential concerns regarding the scope and cov-
erage of the measures. We also adjust the commercial paper borrowings by
netting out the amounts bought by nonfinancial and financial corporations.
We use our measures of bond holdings by corporate sector to adjust the
aggregate supply of corporate bonds and quantify the residual amounts held
by investors. To adjust the payouts, one requires information on the returns
on the bonds held by corporations. For life insurance companies, we collect
information on the average weights to IG and HY bonds in their bond portfo-
lios, as reported in the American Council of Life Insurers (ACLI) Fact Books
from 2001 onward. We apply these weights to the aggregate IG and HY bond
index returns to proxy for the returns on the bond holdings of insurance
companies.12 It is possible to refine these measures further using security- and
9 Life insurance companies divide their holdings into “general” and “separate” accounts depend-
ing on the nature of the policy instrument these assets support. Our goal is to capture the residual
corporate payouts to investors accessible through regular financial markets and without establish-
ing a client relationship with firms, so we treat asset holdings in the two accounts the same way.
We provide further details and robustness checks for insurance company holdings in Section II.A.
10 Flow of Funds data do not separate foreign bonds either, so foreign bond holdings further
contribute to the mismeasurement of payouts and valuations.

11 We thank the authors for sharing their data with us.
12 Since data on the HY bond index are available only starting in 1987, we assume that the
weight to IG bonds is equal to one prior to 1987. For the period between 1987 and 2001, we linearly
extrapolate the weight to IG bonds between the 1987 and 2001 values.
firm-level data as in Ellul, Jotikasthira, and Lundblad (2011) and Chodorow-

Reich, Ghent, and Haddad (2021). We leave these important extensions and
improvements for future research. We assess the sensitivity of our findings to
different assumptions on insurance company investments in Section II.A.
To proxy for the returns on bonds held by other financial institutions and
nonfinancial companies, we assume that corporate bonds held by these firms
are not significantly different from the aggregate market. Combining these
measures with the proxy for bond returns earned by insurance companies al-
lows us to compute the payouts on bonds held by corporations and adjust the
cash, net repurchases, and total payouts on the cross-holdings.
Macroeconomic data. We collect data on GDP and consumption, defined as
the sum of expenditures on nondurable goods and services, from the Bureau
of Economic Analysis (BEA) tables. Data on Consumer Price Index (CPI) come
from the Bureau of Labor Statistics. The price level is normalized to one in
December 2009. All nominal quantities are deflated by the CPI to obtain real
measures.
Our benchmark sample covers the period from 1975 to 2017, due to the avail-
ability of the bond data from the Bloomberg Barclays indices. In supplemental
analysis, we also use equity data that go back to 1949.
C. Empirical Evidence
C.1. Market Prices and Returns
We start our empirical analysis by describing the key properties of the mar-
ket values and returns to equity, bonds, and corporate assets of the U.S. corpo-
rate sector.
Figure 1 shows the evolution of the components of corporate debt, adjusted
for bond cross-holdings. For the purpose of the graph, corporate bond cross-
holdings are prorated according to the market values of each debt component.
As can be seen in Panel A, IG bonds make up the entirety of our measure
of long-term debt in the beginning of the sample. The role of other debt
instruments, especially HY bonds and 144A issues, has increased significantly
over time and helped fuel growth in the corporate debt market. By the end
of the sample, the real market value of long-term corporate debt has reached
5.5 trillion December 2009 dollars, with nearly a half of it consisting of non-IG
bonds.
Panel B of Figure 1 shows that short-duration corporate debt is made up
nearly entirely of debt due in one year and commercial paper, with floating-
rate notes and corporate issues of ABS entering in the early 2000s. Unlike
long-term corporate debt, whose value has been growing over time, the market
value of short-term debt increased from 0.1 trillion in 1975 to its maximum of
over 5.2 trillion in 2007, then fell precipitously during the Financial Crisis. It
remains at its late-1990s value of about 2.4 trillion by the end of 2017.13
13 This is consistent with the evidence in Kacperczyk and Schnabl (2010), who document a
significant decline in commercial paper during the Financial Crisis. Substitution to other sources
(A) (B)
Figure 1. Corporate debt. This figure shows the market values of the components of long-term
(Panel A) and short-term (Panel B) corporate debt, adjusted for cross-holdings. The data are real
annual observations from 1975 to 2017, expressed in trillions of December 2009 dollars. (Color
figure can be viewed at wileyonlinelibrary.com)
Figure 2 shows the dynamics of the market capitalization of bonds, equity,

and corporate assets. We find that aggregate equity is on average 3.7 times
larger than corporate debt. Recall that our measure of debt nets out cross-
holdings and excludes nontradable forms of debt such as bank loans, which
lowers the quantity of debt. Asset and equity values are more volatile than
debt, and also experience larger growth over time. The real market value of
U.S. corporate assets grew from 3.0 trillion in mid-1975 to 33.4 trillion at the
end of 2017, which consists of a 2.5 to 25.5 trillion increase in equity and a 0.5
to 7.9 trillion increase in debt.
Table I provides summary statistics for the returns on aggregate equity,
debt, and corporate assets. The mean real equity return is 7.91% and its stan-
dard deviation is 15.93%. The return on debt is smaller on average and is much
less volatile: its mean is 2.73% and its standard deviation is 5.77%. The asset
return is the weighted average of the two, with the weight tilted more to equi-
ties, which represent a larger fraction of asset value. As a result, the risk and
return on corporate assets is comparable to those on equities. The average as-
set return is 6.87%, its volatility is 12.67%, and it is nearly perfectly correlated
with equity returns.
C.2. Payouts
We next show the empirical evidence for payouts to debt, equity, and cor-
porate assets. Following the discussion in Section I.A, we separate the total
of financing, adverse selection, the inability of issuers to issue commercial paper, and institutional
constraints are potential reasons for the collapse.
Figure 2. Corporate assets. This figure shows the market values of corporate equity, debt, and
assets. Gray bars indicate the NBER recessions. The data are real annual observations from 1975
to 2017, expressed in trillions of December 2009 dollars. (Color figure can be viewed at wileyon-
linelibrary.com)
Figure 3. Corporate payouts. This figure shows cash payouts, net repurchases, and total pay-
outs on corporate equity, debt, and assets. Gray bars indicate NBER recessions. The data are real
annual observations from 1975 to 2017, expressed in trillions of December 2009 dollars. (Color
payouts on each of these instruments into cash distributions (dividends or

interest payments) and net repurchases.
Figure 3 shows the time series of total payouts and their cash and net repur-
chase components. Naturally, cash payouts are always positive. In contrast,
net repurchases can switch sign and turn negative about 85% of the time. At
times when firms choose a net distribution of resources, net repurchases are
positive, while they are negative when firms raise capital through new stock
and bond issuances. In the data, net repurchases are much more volatile than
Table I
Corporate Payouts and Returns in the Data
This table reports summary statistics for scaled changes in cash payouts, net repurchases, and
total payouts (Panels A, B, and C) and for returns (Panel D) on corporate equity, debt, and as-
sets. Payout changes are scaled by consumption level. The mean and standard deviation are in
percentage terms. The data are real annual observations from 1975 to 2017.
Panel A: Cash Payout
Equity Debt Asset
Mean 0.11 0.08 0.19

Std 0.24 0.42 0.56
AC(1) 0.09 0.35 0.33
Cross-Correlations
Debt 0.35
Asset 0.70 0.91
Panel B: Net Repurchase
Equity Debt Asset
Mean −0.09 −0.18 −0.27

Std 3.44 3.76 4.84
AC(1) −0.13 −0.19 −0.28
Cross-Correlations
Debt −0.10
Asset 0.63 0.71
Panel C: Total Payout
Equity Debt Asset
Mean 0.02 −0.09 −0.08

Std 3.53 3.57 4.78
AC(1) −0.11 −0.23 −0.29
Cross-Correlations
Debt −0.09
Asset 0.67 0.68
Panel D: Returns
Equity Debt Asset
Mean 7.91 2.73 6.87

Std 15.93 5.77 12.67
AC(1) −0.08 0.24 −0.06
Cross-Correlations
Debt 0.45
Asset 0.99 0.51
cash distributions and are a dominant component of total payouts. Net repur-
chases amplify the volatility of total payouts and make them occasionally turn
negative. In our sample, total payouts decrease below zero 33% of the time for
equity, 63% of the time for bonds, and 44% of the time for corporate assets.
Payout levels are nonstationary, so we need to consider changes to formally

analyze their statistical properties. However, standard measures of percent or
log growth rates do not work in our setting because payouts can be zero or neg-
ative. Instead, we compute a first difference in the payout scaled by the current
D −D
level of the aggregate economy to remove a trend, for example, D Ct
a,t
= a,t Ct a,t−1
D −D
for consumption Ct or D Yt
a,t
= a,t Yt a,t−1 for output Yt . Scaled payout changes do
not have easily interpretable units and can in principle depend on the choice
of scaling variable. For positive payouts, such as dividend and interest pay-
ments, the scaled change is approximately proportional to the standard log
growth rate, for example, D Ct
t
≈D
C
log Dt , so the log growth rates and scaled
changes are nearly perfectly correlated (see Figure B.9). For payouts that can
take negative values, empirically their variation far exceeds that of aggregate
consumption or output, so a scaled change mainly captures fluctuations in the
underlying payouts rather than the aggregate scaling factor.14 Another advan-
tage of the scaled changes is that they are additive, unlike the percent or log
growth rates. For example, because the level of corporate asset payouts is the
sum of equity and debt, the scaled changes in corporate asset payouts are equal
to a simple sum of scaled changes in equity and debt payouts. Similarly, scaled
changes in cash payouts and in net repurchases add up to scaled changes in
total payouts.
Figure 4 shows the time series of scaled changes in cash payouts D C
, net re-
Da
purchases NREPC
, and total payouts C
for equity, debt, and corporate assets.
Table I reports the key summary statistics for these variables. Below we high-
light several salient features of the total asset payouts (bottom right panel of
Figure 4) and how they compare to traditional equity dividend payments (top
left panel of Figure 4).
Moving from top to bottom in Figure 4 reveals that debt payouts represent a
significant component of total asset payouts. In fact, their contribution to the
level and variation in asset payouts exceeds that of equity, even though the
market value of debt is several times smaller than equity. For the cash compo-
nent, changes in debt payouts are almost twice as volatile as changes in equity
payouts, and the correlation between changes in debt and asset cash payouts
is over 90%. Net repurchases and total payouts on debt are also somewhat
more volatile than their equity counterparts, and have a higher correlation
with their respective asset counterparts of around 70%.
Moving from left to right in Figure 4 shows that fluctuations in net repur-
chases of equity and debt are an order of magnitude larger than dividend
and interest payments, respectively. This implies that total payouts, which are
equal to the sum of cash and net repurchases, are driven predominantly by
net repurchases. The volatilities of changes in total payouts are very similar to
14 We have also examined normalizations by previous-period consumption level, by the average
consumption level across the current and previous periods, by the constant nonlinear consumption
trend, and by excluding scaling all together. The results are nearly identical to the benchmark.
Figure 4. Changes in corporate payouts. This figure shows scaled changes in cash payouts,
net repurchases, and total payouts on corporate equity, debt, and assets. Payout changes are scaled
by consumption level. Gray bars indicate NBER recessions. The data are real annual observations
from 1975 to 2017. (Color figure can be viewed at wileyonlinelibrary.com)
those of net repurchases, and are nearly 10 times larger than the volatilities of
the cash payouts, as documented in Table I.
Overall, incorporating debt payouts and net repurchases significantly affects
the dynamics of total asset payouts relative to traditional equity dividend pay-
ments. The former is 20 times more volatile than the latter and can turn neg-
ative, and the two exhibit only a weak correlation of 25%.
C.3. Economic Growth Risk Exposure

We next assess the economic exposure of asset returns and payouts to fluc-
tuations in macroeconomic growth. The systematic risk stemming from vari-
ation in aggregate economic growth is one of the main tenets of macrofinance
research, from the business-cycle risk models of Mehra and Prescott (1985)
and Campbell and Cochrane (1999) to the long-run risk model of Bansal and
Yaron (2004). It is a natural starting point for analysis of economic risk in fi-
nancial markets.
Table II
Corporate Payout and Return Cyclicality
This table reports correlations of growth rates in consumption or output with scaled changes in
cash payouts, net repurchases, total payouts, and excess returns on corporate equity, debt, and as-
sets. Payout changes are scaled by consumption level. Changes and returns are quarter-to-quarter
(Panel A) or year-to-year (Panel B). The data are real quarterly observations from Q1.1975 to
Q4.2017.
Panel A: Quarterly Changes
Consumption Output
Cash Net Total Excess Cash Net Total Excess

Payout Repurchase Payout Returns Payout Repurchase Payout Returns
Equity 0.15 0.02 0.03 0.12 0.14 0.15 0.16 0.06

Debt 0.15 −0.00 0.00 0.11 0.22 0.02 0.02 −0.04
Asset 0.20 0.01 0.02 0.13 0.20 0.10 0.11 0.06
Panel B: Year-to-Year Changes
Consumption Output

Payout Repurchase Payout Returns Payout Repurchase Payout Returns
Equity 0.21 0.03 0.04 0.25 0.27 0.09 0.11 0.28

Debt 0.25 −0.06 −0.05 0.40 0.28 −0.13 −0.11 0.25
Asset 0.29 −0.02 −0.00 0.27 0.35 −0.02 0.00 0.28
Payouts. First, we examine the business-cycle behavior of payouts. Table II

reports the point estimates of the contemporaneous correlations of scaled
changes in payouts with log consumption or output growth rates. The corre-
lations are computed at a business-cycle quarterly frequency using quarter-to-
quarter or year-to-year changes in the variables of interest over the sample
period 1975 to 2017. Tables B.I, B.II, and B.III present additional evidence
across different adjustments and samples.
As shown in Table II, cash payouts are procyclical: the correlations of
quarter-to-quarter scaled changes in cash payouts on equity, debt, or corporate
assets with growth rates in consumption and output are always positive and
range between 14% and 22%. The point estimates increase to 21% to 35% when
payout changes are smoothed to reduce noise, such as when using year-to-year
changes. In contrast, changes in net repurchases are essentially acyclical. Most
of the correlations are nearly zero or even negative, and almost all of the esti-
mates are smaller than those for the corresponding cash payouts. Because net
repurchases drive the majority of variation in total payouts, it immediately
follows that total payouts are much less procyclical than cash payouts. Indeed,
our estimates suggest that total payouts on corporate assets are essentially
acyclical. The correlation of changes in total asset payouts with consumption
(output) growth is 0.02 (0.11) for quarter-to-quarter changes and zero for year-
to-year changes.
To expand the evidence beyond the short run, we consider the term structure
of cyclicality and compute multihorizon correlations of changes in payouts with
consumption or output growth,

Da,t Da,t+h
ρhc ≡ Corr + ··· + , ct + · · · + ct+h , (7)
Ct Ct+h

y Da,t Da,t+h
ρh ≡ Corr + ··· + , yt + · · · + yt+h (8)
Yt Yt+h
for h equal to 0,1,...,20 quarters. For h = 0, we recover the payout exposures

to business-cycle risk at a quarterly frequency as in Table II, while setting
h > 0 allows us to capture lower-frequency sensitivities of payouts to economic
growth risk. To assess statistical significance, we use Newey-West standard
errors with the number of lags equal to h + 1. In Section II.F, we consider al-
ternative statistical inference methods to evaluate significance of the estimates
at long horizons.
The term structures of the correlations, together with the 95% confidence
bands, are plotted in Figures 5 and 6. The evidence shows that the differ-
ences in the cyclicalities of cash, net repurchases, and total payouts persist
and increase at horizons beyond one quarter. Cash payouts on equity, debt,
and corporate assets remain positively exposed to economic growth risk at all
horizons considered. The correlations of equity cash payouts with consump-
tion (output) growth increase to over 20% (nearly 40%) within two years, while
the correlations of debt cash payouts with consumption and output increase
monotonically with the horizon and reach over 35% at a five-year maturity.
Procyclicality of cash payouts on equity and debt naturally gives rise to pro-
cyclicality of corporate asset cash payouts, whose correlations with consump-
tion (output) growth increase from 20% in the short run to over 30% (40%) at
five years. Interestingly, while the estimates of cash payout correlations are in-
significant for equity at long horizons and for debt in the short run, the statis-
tical support for asset cash payout cyclicality is more uniform across horizons.
Aggregating across various forms of capital distribution helps strengthen the
statistical evidence for the exposure of cash payouts to economic growth risks,
especially at long horizons.
Unlike cash distributions, net repurchases tend to be acyclical at all of the
horizons considered. The equity and debt components of net repurchases are
acyclical in the short run, consistent with the evidence in Table II. At medium
and long horizons, net repurchases of equity become mildly procyclical and
those of debt countercyclical. However, most of the correlation estimates are
statistically insignificant. In addition, the opposite cyclicality of equity and
debt net repurchases essentially cancels out at the level of corporate assets,
so that changes in net repurchases of corporate assets are uncorrelated with
consumption and output growth rates at both short and long horizons.
Figure 5. Term structure of corporate payout and return cyclicality. This figure shows
multihorizon correlations between real consumption growth and scaled changes in cash payouts,
net repurchases, total payouts, and excess returns on corporate equity, debt, and assets. Payout
changes are scaled by the consumption level. Solid lines show correlations at one-quarter to five-
year horizons, along with 95% confidence intervals (dashed lines). The data are real quarterly
observations from Q1.1975 to Q4.2017. Standard errors are Newey-West adjusted. (Color figure
can be viewed at wileyonlinelibrary.com)
Total payouts are equal to the sum of cash payouts and net repurchases.
Because net repurchases drive most of the fluctuations in total payouts, the
cyclicality patterns of total payouts are nearly identical to those of net repur-
chases. For corporate assets, the correlations of the changes in total payouts
with consumption or output are economically and statistically indistinguish-
able from zero at all horizons, and are smaller than those for cash payouts on
equity, debt, and corporate assets.
In the analysis above, we use correlations to gauge the link between pay-
outs and economic growth risk. Correlations are easy to interpret and compare
across measures and samples, and are particularly convenient in our context
because scaled changes do not have intuitive units—they depend on the level of
payouts relative to the scaling trend. However, from the asset pricing perspec-
tive it is covariances and betas, and not correlations, that characterize the risk
exposures of payoffs and returns and ultimately determine the risk premium.
We repeat our analysis by replacing correlations with the payout betas, com-
puted from the ordinary least squares (OLS) regressions of the payout changes
on the measures of economic growth; see Figures 7 and B.1 for evidence with
respect to consumption and output growth, respectively. By construction, betas
Figure 6. Term structure of corporate payout and return cyclicality with respect to out-
put. This figure shows multihorizon correlations between real output growth and scaled changes
in cash payouts, net repurchases, total payouts, and excess returns on corporate equity, debt, and
assets. Payout changes are scaled by the consumption level. Solid lines show correlations at one-
quarter to five-year horizons, along with 95% confidence intervals (dashed lines). The data are
real quarterly observations from Q1.1975 to Q4.2017. Standard errors are Newey-West adjusted.
(Color figure can be viewed at wileyonlinelibrary.com)
and correlations are of the same sign, so the beta- and correlation-based esti-
mates agree on the direction of cyclicality of the payouts. The term structure
patterns of the betas and their statistical significance are quite comparable
to the correlations as well. Cash payouts on equity, debt, and corporate as-
sets have positive betas to economic growth risks. The consumption betas of
debt and asset cash payouts are significant at long horizons, and those of eq-
uity cash payouts at short horizons. The output betas of equity and asset cash
payouts are significant at all horizons, and those of debt cash payouts at long
horizons. Nearly all of the betas of net repurchases are statistically insignifi-
cant from zero. Finally, the betas of total payouts are equal to the sum of the
betas of cash payouts and net repurchases, and they inherit the imprecision
of the latter. In particular, the consumption and output betas of asset total
payouts are estimated with a wide confidence band that always includes zero,
even though the betas of their cash components are significant in the majority
of cases. These findings support our benchmark cyclicality evidence based on
the correlations.
Returns. We use a similar approach to assess the exposure of returns to high-
and low-frequency movements in aggregate growth. Table II shows that the
Figure 7. Corporate payout and return betas to consumption. This figure shows multi-
horizon consumption betas of scaled changes in cash payouts, net repurchases, total payouts, and
excess returns on corporate equity, debt, and assets. Payout changes are scaled by the consumption
level. Betas (solid lines) and 95% confidence intervals (dashed lines) are based on OLS regressions
of the variables of interest on consumption growth rates at one-quarter to five-year horizons. The
data are real quarterly observations from Q1.1975 to Q4.2017. Standard errors are Newey-West
adjusted. (Color figure can be viewed at wileyonlinelibrary.com)
correlations of returns with measures of economic growth are nearly always

positive at a quarterly horizon, and strengthen to 25% to 40% at a year-to-year
frequency. We further consider the term structure of return cyclicality, and
compute multistep correlations of excess returns on equity, debt, and corporate
assets with consumption and output growth, by replacing payouts with returns
in equations (7) and (8).
We plot these correlations as a function of the horizon in Figures 5 and 6.
Virtually all of the correlations are positive. The estimates tend to be smaller
and insignificant in the short run, while they are larger in the medium and
long run and most become significant at the 5% level using Newey-West cor-
rections, especially with respect to output growth (see Section II.F for a discus-
sion of more conservative long-term statistical inference approaches). For ex-
ample, corporate asset return correlations with consumption growth increase
from 13% at a one-quarter horizon to 30% at annual and longer horizons. The
asset return correlations with output growth are statistically and economically
indistinguishable from zero at an annual frequency and nearly reach 50% at
five years. We find similar evidence when we consider the term structure of re-
turn betas to consumption and output growth risks (see Figures 7 and B.1). The
betas of equity, debt, and asset returns are positive and in most cases statisti-
cally significant. In sum, both the correlation and the beta evidence indicates
that asset market valuations are exposed to economic growth risk, especially
at low frequencies.
D. Economic Interpretation
Our empirical evidence shows that accounting for debt and net repurchases
significantly affects the dynamics of corporate payouts. Compared to equity
dividends, total asset payouts are considerably more volatile, can turn nega-
tive, and exhibit much smaller correlations with aggregate economic growth at
high and low frequencies. At the same time, returns on corporate assets are
quite similar to returns on equity, and both are exposed to economic growth
risk, especially at long maturities. The joint evidence on payouts and returns
extends and magnifies the equity premium puzzle of Mehra and Prescott
(1985) at the level of corporate assets. If fluctuations in corporate asset pay-
outs appear unrelated to aggregate economic growth, why do corporate assets
command a large risk premium and why are asset returns exposed to risks in
economic growth?
To provide a potential answer to this question, we first note that cash pay-
outs are quite exposed to economic growth fluctuations in the data, unlike ag-
gregate net repurchases. Net repurchases introduce substantial risk in total
payouts that dwarfs the variation in cash payments. However, if fluctuations
in net repurchases do not receive significant risk compensation (e.g., this risk
is not perceived as systematic by investors), they do not contribute to asset
valuations. At the same time, risk in cash payouts due to their exposure to
systematic fluctuations in economic growth becomes a dominant component
of corporate asset returns. This channel can reconcile the sizeable exposure
of corporate assets to economic growth risks with an apparent acyclicality of
their payoffs in the data.
To quantitatively validate the proposed mechanism, we embed it into the
long-run risks model in the spirit of Bansal and Yaron (2004) (see Section III).
The low-frequency movements in economic growth play a front role in this
framework, making it a natural modeling choice for a quantitative assessment
of joint cyclicality of payouts and returns. Our main argument, however, is
quite general and can be applied to other consumption-based asset pricing
models, such as the habit model of Campbell and Cochrane (1999) or the rare-
disaster models of Rietz (1988) and Barro (2006), and for different notions of
systematic risk. The consumption-based approaches above take the payout
dynamics as given, and exogenously specify their risk profile to match the
data. A deeper question is whether such properties of cash, net repurchases,
and total asset payouts can be economically rationalized through optimal
decisions of the corporations. We leave assessments of these channels for future
research.
II. Discussion, Extensions, and Robustness

In this section, we present a variety of checks to corroborate, expand, and
sharpen our benchmark empirical results.
A. Asset Cross-Holdings
In our benchmark analysis, we adjust the measurements to alleviate double-
counting of corporate payouts and valuations due to asset cross-holdings.
These concerns are particularly relevant in the context of insurance companies
and the financial sector in general. We perform multiple sensitivity checks to
ensure that the cross-holdings adjustments, while appropriate conceptually, do
not impact our central findings.
To proxy for the returns and payouts on the corporate bond holdings of the
insurance companies, we use the insurance industry weights from the ACLI
Fact Books and compute the portfolio average of the IG and HY bond index re-
turns. While these measurements can be refined using the individual security-
and firm-level data, our main conclusions do not appear to be sensitive to al-
ternative assumptions on the portfolio composition of the insurance company
holdings. For robustness, we recompute the payouts by assuming that the in-
surance companies invest in the aggregate bond market index or alternatively
only in IG bonds. The correlations of the alternative payout measures with the
benchmark exceed 90% both in levels and in scaled changes, and these alter-
native adjustments do not affect the cyclicality of total corporate payouts, as
shown in Figure B.2.
The life insurance companies further divide their holdings into general and
separate accounts. The former underlies guaranteed fixed-income liabilities,
such as standard life insurance policies, while the latter supports risk pass-
through investments such as variable annuities.15 Both fixed and variable
products require a client relationship with the firm—neither are sold directly
to investors through regular financial markets, and neither are included in the
underlying Bloomberg Barclays Indices—and thus we treat the two accounts
in the same way in our benchmark analysis. However, because the risk pass-
through products and lines of business supported by the separate accounts
have features of tradable corporate assets, for robustness we run our measure-
ments not excluding the assets held in separate accounts. As Figure B.2 shows,
this adjustment has only a marginal impact on the cyclicality of total payouts.
Alternatively, we drop the insurance companies from the analysis alto-
gether. Bloomberg Barclays Indices disaggregate corporate bond data by sec-
tors, which allows us to remove corporate bonds issued by insurance compa-
nies. We further remove their debt due in less than one year, commercial paper,
and equity. Because we no longer face double-counting regarding their holdings
of corporate bonds, commercial paper, and equity, we do not need to net them
out from our aggregate quantities. On average, excluding the insurance com-
panies increases our measure of the market value of corporate assets by 6%: a
15 See p. 11 of the Fact Books by American Council of Life Insurance (2020).
reduction in the bonds and equity issued by insurance companies is more than
offset by the inclusion of the corporate bonds they own. Removing insurance
companies does not impact the cyclicality of payouts, as can be seen in the top
panel of Figure B.3.
We next consider dropping the financial sector. In particular, we remove cor-
porate bonds issued by all financials from the Bloomberg Barclays Indices, and
we remove their debt due in the current year, commercial paper, and equity.
We no longer net out corporate assets held by the financial sector. Excluding
financials reduces the market value of aggregate corporate assets by 10%, on
average. As shown in the middle row of Figure B.3, limiting the sample to
nonfinancials does not materially affect our key findings. Total asset payouts
and net repurchases remain acyclical, whereas cash payouts are procyclical,
though quantitatively weaker than in the benchmark.16
Finally, we keep all of the benchmark sectors but drop cross-holding ad-
justments. While this strategy no longer captures the residual distributions
from the corporate sector to the aggregate investor, it can be rationalized as a
market-value-weighted portfolio of all the bonds and equities of U.S. corpora-
tions, that is, the aggregate corporate asset index. As the last row of Figure B.3
shows, the cyclicality of payouts on such an index is nearly identical to that
in the benchmark.
B. Corporate Structure and Equity Ownership

Our benchmark analysis relies on total number of outstanding equity shares
to measure valuations and payouts on corporate equity. Our analysis treats
an increase in the number of shares, initial or seasoned, as a new issuance,
whereas a decrease in shares is considered a repurchase. However, some of the
variation in equity shares is caused by cross-agent changes in firm ownership
rather than investor payouts. For instance, a significant amount raised in IPOs
involves the ownership claims of existing owners becoming publicly traded.
Similarly, equity-based compensation (warrants, restricted stocks) redistribute
the claims on a firm’s equity from nonemployee shareholders to employees. The
question is whether these considerations conceptually and empirically affect
the measurements, interpretations, and validity of our aggregate strategy.
Conceptually, our aggregate strategy takes the perspective of a regular
investor who maintains an arm’s length relationship with firms and invests
in their public corporate assets through regular financial markets. Such an
investor is not a customer of any firm (e.g., does not hold deposits with or
borrow from the intermediary, or receives goods and services from firms), nor
is it a supplier (does not sell goods or extend trade credit) or employee (does
not receive compensation and pension). From an asset pricing perspective, this
investor is only concerned with the financial payouts on corporate investable
assets, in the form of cash payouts and net repurchases, and not with how they
16 Cash payouts in the nonfinancial sector are strongly procyclical before the Crisis, similar to
the benchmark.
are generated and used. For example, an increase in the number of shares
represents a flow of capital from the investor into the firm, regardless of
whether the firm is investing in a new project, acquiring a new firm, or using
the shares to compensate a CEO or venture capitalist. Of course, one can en-
tertain a broader notion of payouts beyond investable corporate assets, which
could include claims to both public and private companies and claims to wages
and employee compensation, among others. For these payouts, the nature of
cross-agent transfers matters, and certain flows have to be netted out to avoid
double-counting, much in the same way as we adjust for the cross-holdings of
firms’ corporate assets. While these notions of payouts are beyond the scope
of our paper, we can examine whether aspects of the ownership structure are
empirically important for our results.
First, we consider the impact of IPOs. IPOs represent a transfer of owner-
ship from private to public investors and are counted in the equity net repur-
chase component of our corporate asset payouts. This is consistent with our
aggregate strategy, which calls for buying asset shares as soon as they become
accessible through the regular financial markets. To examine whether the IPO
events are critical for our main findings, we drop observations when a security
appears in the sample for the first time. On average, this reduces measured
equity issuances by 33%, but it has no impact on the cyclicality of corporate
asset payouts and its components, as shown in the top row of Figure B.4.17
Second, we consider changes in entity structure due to mergers and acqui-
sitions. Our benchmark analysis accounts for mergers and acquisitions, such
that a reduction in the market value of the target is offset by the share is-
suance or the price appreciation of the acquirer. Merger waves are known to be
cyclical, so it is natural to ask whether this cyclicality is related to our findings.
Excluding mergers reduces measured equity repurchases by 53% on average,
which, if anything, makes net repurchases and total asset payouts more coun-
tercyclical, as shown in the middle row of Figure B.4.
Next, we examine the role of executive compensation in the form of re-
stricted shares. Unlike shares available to the general public (float), restricted
shares are not available for public trading. However, both count toward the
number of outstanding shares that we use in our benchmark analysis. We
use the Factset database to obtain data on public float and merge it with
our CRSP/Compustat sample. In the post-1990 period, we match about three-
quarters of CRSP/Compustat (nearly 85% post-2000) in terms of the number of
firms, and over 80% in terms of market capitalization. As shown in Figure B.5,
equity net repurchases based on public float are very similar to those based on
total outstanding shares, with the correlation between them over 95% in levels
and nearly 90% in scaled changes.18
17 Excluding first observations likely overestimates the importance and magnitude of IPOs. Gao,
Ritter, and Zhu (2013) find that the number of IPOs has decreased precipitously since 2000. They
estimate the average proceeds raised in the IPOs to be $28.1 billion dollars annually between 1980
and 2012, which would amount to just under 20% of our measured equity issuances.
18 The Factset measures of equity net repurchases have a correlation of 68% with our bench-
mark measure of net repurchases, and the latter is more volatile as can be seen in Figure B.5. Due
Finally, similar to the majority of the literature, our analysis is conducted

from the U.S. perspective and relates payouts on corporate assets publicly
traded in the United States to economic risks in U.S. consumption and output.
However, a substantial amount of the investable assets of U.S. corporations
are held by non-U.S. households. Conversely, the portfolios of U.S. investors
may be diversified in international markets, so the systematic risk may not be
driven entirely by U.S.-specific factors. Tracing the geography of payout flows
is a challenging question that goes beyond the scope of this paper and most of
the macrofinance literature that we benchmarking against. For suggestive evi-
dence, we examine the robustness of our main results to using alternative mea-
sures of economic risks in international macroeconomic data. We collect quar-
terly GDP data for major industrialized countries and measure global output
as value-weighted GDP across countries. In addition, we remove the United
States from the GDP sample and use equal weights to reduce the impact of
the United States. Figure B.6 shows that the term structures of asset payout
cyclicality with respect to global GDP are nearly identical to those based on
U.S. data.
C. Debt Measurements
Our empirical measure of corporate debt relies on Bloomberg Barclays In-
dices, supplemented by short-term debt data from the Flow of Funds and
Compustat. Our data sources provide broad coverage of the aggregate cor-
porate bond market (Figure A.1) and deliver accurate and relevant market-
based measures of payouts and valuations of tradable corporate debt, net of
cross-holdings. Below we argue that other popular sources of debt data are
less suited for this purpose, with important implications for the measurement
and interpretation of valuations and payouts.
A common approach in the literature is to use book values of debt, which
are available in the Flow of Funds or Compustat. However, book values ig-
nore interest rate risk and are at best stale historical proxies for its market
values. Figure 8 shows the log ratio of book value to market value of debt
in our sample.19 On average, book value of debt exceeds its market value by
about 2%. However, the wedge between the two fluctuates significantly over
time, rising to nearly 25% in the early 1980s and dropping to –7% in the mid-
2000s. The discrepancy is related primarily to the level of interest rate risk: its
correlation with the BAA corporate rate is 88%. The differences between the
book and market values get more pronounced at longer maturities due to
higher sensitivity to interest rate risk. For Bloomberg Barclays Indices, which
represent bonds with one year to maturity or above, the gap between their book
to data availability, the Factset measure is constructed for a subset of firms in CRSP/Compustat,
uses end-of-year rather than end-of-month price data, and excludes IPOs and mergers and acqui-
sitions.
19 For simplicity, in this section we ignore netting out debt cross-holdings and use aggregate
unadjusted debt data.

Figure 8. Book-to-market value of corporate debt. This figure shows the logarithm of book
to market value of corporate debt (left axis) and the BAA corporate yield (right axis). The data
are quarterly observations from Q1.1975 to Q4.2017. (Color figure can be viewed at wileyonlineli-
brary.com)
and market values can be as high as 48%. Book values are also less volatile
than the market values: the standard deviation of log growth rate of the mar-
ket value of debt is 2.6% versus 1.9% for book values (4.4% vs. 2.3% for bonds
with longer duration). Finally, using book values significantly distorts returns
on corporate debt. The correlation between debt returns computed using book
versus market data (keeping the payout component the same) is 66%, and the
former is less volatile at 3.9% versus 4.5%, annualized.
Because book values do not provide accurate measurements of debt, changes
in book values are noisy proxies for bond net repurchase. Bond issuances gener-
ally happen at par, so the book value of debt would increase nearly one-to-one
with new bond issuance. Repurchases, however, do not need to occur at par.
Quantitatively, quarterly changes in book values have a 78% correlation with
net repurchases. Due to noise, changes in book value are more volatile than
net repurchases, and tend to exhibit more spikes.
Another challenge for empirical analysis is to correctly identify the cash
component of corporate debt payouts. Reported interest expenses in Compu-
stat (item xint) combine corporate bond coupon payments together with non-
tradable bank and other forms of debt, which should be excluded from our
measurements. Interest expenses from Compustat thus overestimate public
debt cash payouts—in our sample, it is two to three times larger than our
benchmark payout series (Figure 9). As an alternative, National Income and
Product Accounts (NIPA) tables report net interest expense, which is interest
paid less interest received by private enterprises (item A453RC1 for all firms
and B471RC1 for nonfinancials). While this measure helps address issues with
Figure 9. Corporate debt cash payout. This figure shows our benchmark measure of corporate
debt cash payout, aggregate interest expenses for all firms and nonfinancials from the NIPA, and
aggregate interest expenses from Compustat. The data are real annual observations from 1975
to 2017 and are expressed in trillions of December 2009 dollars. (Color figure can be viewed at
wileyonlinelibrary.com)
cross-holdings and bank debt, it also includes other forms of payments such as
those for implicit deposit and borrower services by financial intermediaries,
mortgage and home improvement loans, royalties, interest to and from the
rest of the world, and so on. As a result, interest expenses from NIPA are only
weakly related to our aggregate cash payments on debt. As shown in Figure 9,
the level of interest expenses from NIPA, for all firms and nonfinancials, are on
average similar to our benchmark cash payment on debt. However, they exhibit
fairly different dynamics. The scaled changes in our benchmark cash payouts
on debt have a correlation of 37% (23%) with the scaled changes in NIPA in-
terest expenses of all firms (nonfinancials). Moreover, unlike benchmark cash
payments on debt, which are procylical at all frequencies, the NIPA interest
expenses are acyclical and even countercylical in the data (see Figure B.7).20
While acylicality of interest expenses strengthens our argument for a lack of
cyclicality of total payouts, the measurement issues of interest expenses in the
NIPA data make it less suitable for our main analysis.
We further find that our aggregate asset payout measure is quite distinct
from other, earnings-based measures popular in the literature. Earnings cap-
ture profits generated by the firm during the period, and aggregate earnings
are often used to measure the performance of the corporate sector. Aggregate
earnings, however, are conceptually different from corporate asset payouts.
They represent an accounting, rather than a cash flow measure of distribu-
20 The bottom panel of Figure B.7 also shows lack of cyclicality for the interest expense measure
of Larrain and Yogo (2008) in the annual sample from 1927 to 2004.
tions. First, they contain retained earnings, which represent the capital not
paid out to investors. Second and most importantly, earnings do not incorpo-
rate asset repurchases and issuances, which we argue are the dominant part
of corporate asset payouts. Accordingly, aggregate earnings are more aligned
with equity cash distributions than with asset total payouts. In our sample,
the annual growth rate in aggregate earnings has a 33% correlation with
changes in equity cash payouts, while its correlation with changes in total
asset payouts is negative at –30%. Consistent with the evidence in Longstaff
and Piazzesi (2004), corporate earnings are also quite procyclical in the data,
unlike aggregate asset payouts.
Recall that our benchmark measure of long-term debt due in the current
year from Compustat can include other forms of nontradable debt. To address
this concern, we can alternatively use the book value of bonds maturing in less
than one year from Mergent. Mergent does not have reliable coverage of short-
term debt prior to 2000. As such, in the years prior to 2001 we proxy for the
debt due in the current year with the Bloomberg Barclays book value of long-
term bonds scaled by the average Mergent share of short-term component in
its overall debt over the period 2001 to 2007.21 As shown in the bottom row of
Figure B.4, using this alternative measure does not affect our key results.
D. Negativity of Payouts
In the data, total payouts from the aggregate strategy can turn negative,
unlike per-share dividends and interest payments. The negativity of payouts
does not present a conceptual economic problem. At the individual firm level,
total payouts are naturally negative at IPOs and at other equity and bond
issuances. Intuitively, investors are willing to provide capital to firms with
profitable growth opportunities, and they price in future positive payoffs. The
evidence in our paper suggests that this logic extends to the entire corporate
sector: at certain economic times, firms in aggregate raise more capital than
they distribute and therefore leave investors with negative payouts. From an
economic modeling perspective, consumption is always positive of course, and
investors use wages, savings, or other forms of background income to finance
the public sector.
Standard present-value arguments can be applied to connect valuations
to payouts. Using the Euler equation for the per-share return and iterating
forward, we obtain
∞

Pt = Et Mt,t+ jCFt+ j , (9)
j=1
where Mt,t+ j denotes the stochastic discount factor (SDF) between time t
and t + j. A transversality condition, lim j→∞ Et (Mt,t+ j Pt+ j ) = 0, ensures the
21 The results are nearly identical if we instead use Compustat long-term debt due in the cur-
rent year scaled by the average short-term debt in Mergent over short-term debt in Compustat
over the period 2001 to 2007.
existence of the valuations and rules out bubbles. Similarly, we can apply the
Euler condition to the aggregate strategy,
∞

Vt = Et Mt,t+ j Da,t+ j
j=1
∞
∞

= Et Mt,t+ j Dt+ j + Et Mt,t+ j NREPt+ j , (10)
j=1 j=1
under the transversality condition lim j→∞ Et (Mt,t+ jVt+ j ) = 0.

The valuation equations above can include both positive and negative pay-
outs. The value of a per-share investment is guaranteed to be positive because
per-share payouts are never negative. For an aggregate strategy, economic in-
tuition suggests that the capital inflows to the corporate sector are generally
associated with profitable growth opportunities, which would ensure that the
present value of total payouts is positive.
While the negativity of payouts may not be an issue in economic models, it is
less common in the context of self-financing investment strategies. To provide a
nonnegative payout counterpart to our aggregate strategy, we consider a man-
aged portfolio that tracks returns and nonnegative payouts to the total supply
of equity, debt, or corporate assets of the corporate sector. A managed portfolio
receives the total payout generated by the underlying asset class. Similar to
the benchmark, it distributes positive payouts to investors. Different from the
benchmark, it finances negative payouts by selling off portions of the original
investment to ensure nonnegative distributions to investors.22 Note that the
managed portfolio delivers exactly the same return as the underlying asset it
is fully invested in, but the total value of the managed portfolio goes down over
time as it draws down its capital to finance negative payouts. In that sense,
it is not an exact equivalent to our aggregate strategy, but rather serves as a
robustness check on the quantitative importance of negative distributions.
We run managed strategies separately for equity, debt, and corporate as-
sets.23 By construction, managed payouts are zero when the underlying
payouts go negative. At other times the payouts on the managed portfolio track
but are below the levels of the underlying distributions, since the portfolio is
selling off parts of its initial investment (for further details, see Appendix C
and Figure C.1). The top panel of Figure 10 shows the term structure of cycli-
cality for the payouts on the managed portfolios. Naturally, the managed cash
payouts are acyclical for each asset class—they correspond to the total payouts
on the underlying asset class scaled by the number of shares. Net repurchases
22 For simplicity, we abstract from buying securities back.

23 An alternative approach to manage corporate asset investment is to consider a portfolio of
stocks and bonds, and sell off only those assets that generate a negative payout. This alternative
managed portfolio effectively corresponds to the sum of the managed portfolios of equity and debt.
Empirically, it heavily draws down debt, so its payouts and returns are close to the managed
investment in equity alone.
Figure 10. Managed portfolio payout cyclicality. This figure shows multihorizon correlations
between real consumption growth and scaled changes in cash payouts, net repurchases, and total
payouts on the managed portfolio. The managed portfolio maintains nonnegative distributions to
investors by selling off the underlying principal. Payout changes are scaled by the consumption
level. Solid lines show correlations at one-quarter to five-year horizons, along with 95% confidence
intervals (dashed lines). The data are real quarterly observations from Q1.1975 to Q4.2017. Stan-
dard errors are Newey-West adjusted. (Color figure can be viewed at wileyonlinelibrary.com)
on managed portfolios represent selling off securities at times of negative pay-

outs and are acyclical. Finally, the mild procyclicality of the managed pay-
outs on equity combines with acyclicality of debt, so that the managed payouts
on corporate assets are essentially acyclical. Hence, the negativity of payouts,
while economically significant, does not play a critical role for the cyclicality
of payouts.
The discussion above focuses on evaluating the dynamics of the payouts and
the issues pertaining to negative distributions. Dichev (2007) further argues
that positive and negative payouts from share repurchases can impact compu-
tations of historical stock returns themselves. Rather than standard equal av-
eraging of buy-and-hold returns over the sample, he suggests dollar-weighing
the returns as is commonly done in the evaluation of investment projects. The
dollar-weighted return is just the internal rate of return on buying an asset at

the beginning of the period, collecting positive and negative distributions over
time, and selling it at the end of the period. In our sample, the dollar-weighted
returns on equity, debt, and assets are 7.6%, 4.2%, and 6.8%, respectively, and
are quite similar to the benchmark numbers reported in Table I.
E. Acyclicality of Net Repurchases

Our evidence suggests that net repurchases are not exposed to economic
growth risks at short and long horizons, which affects the cyclicality of total
payouts. To help interpret the acyclicalilty of net repurchases themselves, it is
helpful to consider its issuance and repurchase components separately. How-
ever, this requires individual firm data, which we can only obtain for equities
from the CRSP Monthly Stock File. With this caveat in mind, we focus on the
term structure of cyclicality of equity payouts and split equity net repurchases
into issuances and repurchases.24 Panel A of Figure B.8 shows the results
for the benchmark sample from 1975 to 2017, while in Panel B we consider a
longer sample that starts in 1949. The results from both samples are consis-
tent with our main findings: the scaled changes in cash payouts on equity are
procyclical, especially in the long sample, while the changes in net repurchases
are acyclical and the changes in total equity payouts appear acyclical as well.
Interestingly, while changes in net repurchases are acyclical, both of its com-
ponents are quite procyclical in the data: the correlations of scaled changes in
issuances and repurchases with consumption growth are all positive. A po-
tential explanation for these findings is that our measures aggregate across
firms that have different needs for capital. In good times, some firms in the
cross section face good investment opportunities and thus raise capital through
issuances. Other firms may opt to distribute profits to investors, which can
be done through cash dividends or repurchases. Repurchases may be the
preferred form of distributing the transitory component of earnings, as divi-
dend policy requires a financial commitment (e.g., Lintner (1956)), consistent
with the evidence in Guay and Harford (2000), Jagannathan, Stephens, and
Weisbach (2000), and Dittmar and Dittmar (2004). In both cases, aggregate
issuances and repurchases increase. This makes them procyclical separately,
while on a net basis the two effects offset each other, which leads to acyclical
net repurchases at the aggregate level.
The analysis of debt and equity financing at the aggregate level can fur-
ther mask heterogeneity in firms’ policies across their size, level of financial
constraint, and other characteristics. This firm heterogeneity can play an im-
portant role to help account for inconclusive evidence for the cyclical behavior
of debt and equity financing at the aggregate level; see, for example, Covas and
Haan (2011), Jermann and Quadrini (2012), and Begenau and Salomao (2019).
24 Following the literature, we attribute a firm’s net repurchases to repurchases (issuances) if
the number of shares outstanding decreases (increases) over the month, and we aggregate the
firm-level measures to aggregate index repurchases and issuances.
Figure 11. VAR-implied term structure of corporate payout and return cyclicality. This
figure shows VAR-implied multihorizon correlations between real consumption growth and scaled
changes in cash payouts, net repurchases, total payouts, and excess returns on corporate equity,
debt, and assets. Payout changes are scaled by the consumption level. The VAR(1) is run sepa-
rately for equity, debt, and assets, and is fitted to two out of three payout measures together with
consumption growth (excess returns and consumption growth for the excess return correlations).
Solid lines show correlations at one-quarter to five-year horizons, along with 95% confidence in-
tervals (dashed lines). The data are real quarterly observations from Q1.1975 to Q4.2017. (Color
F. Measurements and Inference

We perform several robustness checks to assess the validity of our results
to alternative measurements, statistical inference methods, samples, and
adjustments.
We first examine the statistical inference on the multiyear correlations. In
the main analysis, we compute Newey-West standard errors with the num-
ber of lags exceeding the horizon of the correlation. However, the Newey-West
corrections may not be reliable in small samples and with a large number of
overlapping observations; see, for example, Andrews and Monahan (1992) and
Den Haan and Levin (1997), among many others. We conduct two alternative
approaches to assess and verify the statistical evidence at long horizons.
First, motivated by the Hodrick (1992) corrections commonly used in the re-
turn predictability literature, we stack the variables of interest into a VAR(1)
model and compute the VAR-implied estimates and standard errors of the cor-
relations at any horizon in the future. This avoids statistical issues of small-
sample overlapping data at the cost of putting additional VAR structure on the
underlying variables. Figure 11 shows the cyclicality evidence from VARs run
separately for equity, debt, and corporate asset payouts.25 Similar to the bench-
mark, cash payouts are quite procyclical at short and long frequencies, and the
estimates for asset cash payouts are statistically significant at all horizons.
Net repurchases are largely acyclical, so that the total asset payouts exhibit
insignificant correlations with consumption growth at any horizon. The VAR
approach further improves on the statistical inference for the cyclicality of re-
turns. As shown in Figure 11, most of the return correlations with consumption
growth are now statistically significant.
Under the second approach, we implement the long-run covariability infer-
ence of Müller and Watson (2018), which is based on low-frequency transforma-
tions of the data. Table III presents the evidence for the cyclicality of payouts
and returns at 16- and 20-quarter frequencies. Similar to Müller and Watson
(2018), we report 67% and 90% CIs for each estimate. The magnitudes of the
correlations are very similar to the benchmark estimates in Figures 5 and 6,
while the standard errors increase substantially relative to the Newey-West or
VAR-based standard errors. Cash payouts are procyclical, and nearly all of the
estimates are significant or borderline significant under 67% CIs. Most of the
correlations of net repurchases and total payouts with economic growth rates
are near or below zero, and are estimated with substantial noise. We also con-
sider related wavelet-based inference on the correlations, following Whitcher,
Guttorp, and Percival (2000). The results, presented in Appendix D and Table
D.I, are very similar to Müller and Watson (2018) evidence.
Finally, we examine the evidence over various samples and seasonality ad-
justments. Specifically, we consider the correlations of equity payouts in a
longer sample going back to 1949; and seasonally adjust quarterly payout
changes using a band-pass filter or x12 ARIMA model. The results reported
in Tables B.I, B.II, and B.III are very similar to our benchmark findings, espe-
cially at the level of asset payouts.
III. Model
Our empirical evidence suggests that total asset payouts are acyclical at
short and low frequencies. At the same time, corporate assets command a
risk premium and are exposed to economic growth risk, especially in the long
run. To reconcile this evidence, we argue that asset payouts are dominated
by acyclical net repurchases that mask growth risk exposure of cash payouts.
We develop a long-run risks valuation framework to quantitatively assess the
plausibility of our economic explanation. Independently, we make a method-
ological contribution to the literature by providing an alternative extension of
the log-linearization framework of Campbell and Shiller (1988) to cases with
negative payouts.
25 For each payout variable (e.g., changes in equity cash payments), we include one of the two
other payouts (e.g., changes in equity net repurchases) together with the measure of economic
growth. Including all three payout measures leads to singularity issues as the total payout is by
definition equal to the sum of cash and net repurchases.
Table III
Long-Term Corporate Payout and Return Cyclicality
This table reports the 16-quarter (Panel A) and 20-quarter (Panel B) frequency correlations of growth rates in consumption or output with scaled
changes in cash payouts, net repurchases, total payouts, and excess returns on corporate equity, debt, and assets. Correlations and confidence intervals
are computed following Müller and Watson (2018). The data are real quarterly observations from Q1.1975 to Q4.2017.
Panel A: 16 Quarter Frequency
Consumption Output

Payout Repurchase Payout Return Payout Repurchase Payout Return
Equity Est. 0.21 0.08 0.08 0.21 0.38 0.10 0.13 0.34
67% CI −0.01, 0.41 −0.09, 0.33 −0.08, 0.33 −0.01, 0.41 0.12, 0.56 −0.04, 0.36 −0.01, 0.36 0.08, 0.52
90% CI −0.21, 0.53 −0.22, 0.41 −0.21, 0.42 −0.10, 0.53 −0.01, 0.65 −0.16, 0.45 −0.13, 0.46 −0.02, 0.65
Debt Est. 0.21 −0.13 −0.12 0.49 0.28 −0.30 −0.30 0.47
67% CI −0.00, 0.41 −0.36, 0.06 −0.35, 0.08 0.28, 0.64 0.00, 0.42 −0.43, −0.01 −0.39, −0.01 0.30, 0.64
90% CI −0.10, 0.53 −0.44, 0.16 −0.43, 0.20 0.05, 0.72 −0.08, 0.56 −0.56, 0.08 −0.56, 0.09 0.08, 0.72
Asset Est. 0.25 −0.04 −0.01 0.23 0.34 −0.08 −0.01 0.36
67% CI 0.00, 0.42 −0.30, 0.10 −0.21, 0.13 0.00, 0.42 0.08, 0.52 −0.33, 0.08 −0.21, 0.13 0.13, 0.55
90% CI −0.10, 0.56 −0.38, 0.30 −0.36, 0.34 −0.09, 0.55 −0.03, 0.64 −0.38, 0.21 −0.36, 0.30 −0.01, 0.65
(Continued)
177
178
Table III—Continued
Panel B: 20 Quarter Frequency
Consumption Output

Equity Est. 0.12 0.01 0.02 0.21 0.37 0.08 0.12 0.36
67% CI −0.04, 0.39 −0.13, 0.27 −0.10, 0.30 −0.01, 0.42 0.08, 0.58 −0.06, 0.36 −0.04, 0.38 0.08, 0.63
90% CI −0.21, 0.53 −0.33, 0.41 −0.32, 0.42 −0.13, 0.59 −0.04, 0.68 −0.21, 0.47 −0.20, 0.54 −0.02, 0.65
Debt Est. 0.20 −0.08 −0.08 0.47 0.20 −0.16 −0.13 0.49
67% CI −0.01, 0.41 −0.35, 0.08 −0.35, 0.08 0.26, 0.64 −0.01, 0.42 −0.38, 0.01 −0.38, 0.04 0.30, 0.65
90% CI −0.13, 0.56 −0.46, 0.21 −0.44, 0.30 0.02, 0.72 −0.13, 0.63 −0.56, 0.13 −0.56, 0.16 0.05, 0.75
The Journal of Finance®
Asset Est. 0.21 −0.04 −0.01 0.30 0.33 −0.08 −0.01 0.38
67% CI −0.01, 0.42 −0.33, 0.08 −0.30, 0.12 0.00, 0.45 0.02, 0.55 −0.33, 0.08 −0.30, 0.12 0.13, 0.64
90% CI −0.13, 0.59 −0.43, 0.30 −0.39, 0.33 −0.09, 0.63 −0.08, 0.64 −0.44, 0.21 −0.38, 0.31 −0.01, 0.65
A. Economic Setup
Preferences. We consider a discrete-time endowment economy, in the spirit
of Bansal and Yaron (2004) and a subsequent long-run risks literature. The
preferences of the representative agent are characterized by the Kreps and
Porteus (1978) recursive utility of Epstein and Zin (1989) and Weil (1989),
1−γ
1−γ
θ
1−γ 1θ
Ut = (1 − β )Ct θ
+ β(Et Ut+1 ) , (11)
where Ct is consumption, β is the subjective discount factor, γ is the risk-

aversion coefficient, and ψ is the intertemporal elasticity of substitution (IES).
1−γ
For ease of notation, the parameter θ is defined as θ ≡ 1− 1 . Note that when θ =
ψ
1, that is, γ = 1/ψ, the recursive preferences collapse to expected power utility,
in which case the agent is indifferent to the timing of the resolution of uncer-
tainty of the consumption path. When the risk aversion is above (below) the
reciprocal of the IES, the agent prefers early (late) resolution of uncertainty.
Epstein and Zin (1989) show that the asset pricing restriction for any asset
return rt+1 satisfies a standard Euler condition:
Et [exp{mt+1 + rt+1 }] = 1. (12)
The log of the intertemporal marginal rate of substitution mt+1 is defined as
θ
mt+1 = θ log δ − ct+1 + (θ − 1 )rc,t+1 , (13)
ψ
where ct+1 = log(Ct+1 /Ct ) is the log growth rate of aggregate consumption,
and rc,t is the log return on the asset, which delivers aggregate consumption as
dividends (the wealth portfolio).
Consumption dynamics. As in Bansal and Yaron (2004), the consumption
growth rate contains a small predictable component xt that determines the
conditional expectation of consumption growth. The volatility of fundamental
shocks is time-varying and is captured by the state variable σt2 :
ct+1 = μc + xt + σt ηt+1 ,
xt+1 = ρx xt + ϕx σt et+1 ,

σt+1
2
= σ02 + ν σt2 − σ02 + σω ωt+1 . (14)
The parameters ρx and ν capture the persistence of the expected growth and
volatility news, and σ0 , ϕx , and σw govern the unconditional scales of shocks to
realized and expected consumption and consumption volatility, respectively.
Corporate sector payouts. We focus on the total assets of the corporate sector
and provide a parsimonious exogenous specification for the cash and net re-
purchase components of total asset payouts. For simplicity, we do not consider
capital structure decisions regarding issuances and repurchases of equity and

debt; we leave these model extensions for future research.
Following Bansal, Dittmar, and Lundblad (2005), Bansal and Yaron (2007),
and Hansen, Heaton, and Li (2008), cash payouts are cointegrated in logs with
the consumption level:

Dt
log ≡ μs + st . (15)
Ct
The cointegrating residual st is stationary, persistent, and exposed to the low-

frequency growth risk:
st+1 = ρs st + φs xt + ϕs σt ut+1 . (16)
Parameters μs , ρs , and ϕs determine the unconditional level, persistence, and

volatility of the cash payout dynamics, and φx governs its exposure to expected
growth risks.
To accommodate net repurchases, we first define consumption adjusted for
net repurchases Ht :
Ht ≡ Ct − NREPt = Ct − REPt + ISSt . (17)
Economically, we expect consumption net of repurchases to be positive—

repurchases are a capital distribution from firms to investors, supplemental to
cash dividends, coupons, and other sources of income, all of which are used to
finance consumption expenditures. We therefore expect Ct > REPt and hence
Ht > 0.26 We assume that the log of Ht is driven by independent and identically
distributed (i.i.d.) shocks t ,
Ht
log = μh + ϕh σt εt , (18)
Ct
where μh and ϕh capture the unconditional level and volatility of the process.
Our specification ensures that net repurchases are cointegrated with aggre-
gate consumption in logs. The same is true for cash and total payouts. At the
same time, the level of net repurchases can be negative, unlike the cash pay-
outs. Finally, unlike cash payouts, we assume that net repurchases are not
exposed to low-frequency fluctuations in economic growth.
The four shocks ηt+1 , et+1 , ωt+1 , and εt+1 are i.i.d. standard Normal. We allow
the correlation between the transitory shocks to consumption growth and cash
payout growth to help match the data:
cov(ηt+1 , ut+1 ) = α. (19)
26 In our sample, equity repurchases are on average 7% and never exceed 25% of total consump-
tion. The net repurchases never exceed 7% of total consumption at equity, debt, or asset levels.
Economically, adding consumption ensures cointegration of asset payouts with consumption, and
is different from Boudoukh et al. (2007), who add a constant to the net yield to make it positive.
B. Model Solution
Valuation of consumption claim. For tractability, we consider an approxi-
mate solution to the model based on the log-linearization of consumption and
asset returns. The log-linearization of the consumption return is standard and
follows Campbell and Shiller (1988). Specifically,

Vc,t+1 + Ct+1
rc,t+1 = log ≈ κ0,c + κ1,c vcc,t+1 + ct+1 − vcc,t , (20)
Vc,t
where vcc,t = log( VCc,tt ) is the valuation of the consumption claim, and κ0,c and
κ1,c are the linearization coefficients, which are determined in equilibrium
by the unconditional level of the consumption asset’s valuation. Under log-
linearization, the consumption return, the value of the consumption claim, and
hence the SDF are linear in the underlying states of the economy and can be
solved in closed form. As shown in Appendix E, the value of the consumption
claim is given by
vcc,t = A0,c + A1,c xt + A2,c σt2 , (21)
and the equilibrium log SDF satisfies
mt+1 = m0 + mx xt + mσ σt2 − λη σt ηt+1 − λe ϕe σt et+1 − λw σw ωt+1 . (22)
The exposures of the consumption asset and the market prices of risk are
pinned down by the model and preference parameters (see Appendix E). The
economic interpretation of the long-run risks model is that when agents have
a preference for timing of uncertainty resolution, the short-run, long-run, and
volatility risks (η, e, and ω, respectively) are priced and determine the risk com-
pensation in asset markets. Specifically, for γ > 1 and ψ > 1, the consumption
claim requires a positive risk premium because the consumption asset return
is low in bad times of low realized or expected consumption growth (λη , λe > 0,
and A1,c > 0) or high consumption volatility (λw < 0 and A2,c < 0). Quantita-
tively, the risk premia for expected consumption and volatility risks are mag-
nified by the persistence of these shocks.
Valuation of corporate assets. The payouts from the corporate sector are de-
termined by the cash and net repurchases components (see equations (15) and
(18), respectively). We extend the Campbell and Shiller (1988) log-linearization
approach and rewrite the return on the corporate assets in (2) as
Vd,t+1 + Dt+1 + NREPt+1 Vd,t+1 + Dt+1 + Ct+1 − Ht+1

Rd,t+1 = =
Vd,t Vd,t
Vd,t+1 Dt+1 Ht+1
Ct+1 1 + Ct+1
+ Ct+1
− Ct+1
= · Vd,t
, (23)
Ct
Ct
where Vd,t is the value of the asset. Notably, all of the ratios in the last equa-
tion are positive: consumption, prices, cash payouts, and consumption adjusted
for net repurchases are all above zero. We then log-linearize the expression
above around the unconditional log values of vcd , dc, and hc to derive the log-
linear approximation for the asset return,
rd,t+1 ≈ κ0,d + κ1,d vcd,t+1 + ct+1 + κ2,d dct+1 + κ3,d hct+1 − vcd,t , (24)
V
where vcd,t = log( Cd,tt ) is the log asset value to consumption ratio, dct = log( Dt
Ct
)
is the log ratio of the asset cash payouts to consumption, and hct = log( Ct ). The
Ht
expressions for the log-linearization coefficients are provided in Appendix E.

Our log-linear approximation in (24) nests a standard case for the consump-
tion asset in (20). Indeed, when there are no net repurchases and cash payouts
are equal to consumption, we have hct = dct = 0. When net repurchases are
part of the investor’s payoff, total payout can be negative and can no longer
be used to scale valuations and define growth rates. This is why we have to
switch to consumption to scale all of the quantities, and rewrite the payouts in
terms of positive cash and the adjusted net repurchase components. Further-
more, our approach is different from the linearizations in Bansal and Yaron
(2007) and Larrain and Yogo (2008). These papers log-linearize returns around
positive gross issuance and repurchase components. Due to data limitations,
we cannot separate issuances and repurchases at the asset level and instead
model net repurchases directly.
We can use the log-linearization solution to asset returns in (24), the corpo-
rate payout dynamics in (15) to (18), and the equilibrium SDF in (22) to solve
for the equilibrium asset valuations. The corporate valuations are linear in the
economic states,
vcd,t = A0,d + A1,d xt + A2,d σt2 + A3,d st . (25)
Similar to the consumption asset, for typical model parameters corporate as-
sets are risky—they fall in bad times of low economic growth (A1,d > 0) or high
consumption volatility (A2,d < 0). Asset prices also increase at times of a pos-
itive gap between cash payouts and consumption (A3,d > 0)—because the gap
st is persistent, it signifies higher future cash payments to investors.
C. Implications for Payouts and Valuations

We calibrate the model and assess whether it can quantitatively account for
our empirical evidence. As is common in this literature, we calibrate the model
at a monthly frequency, and use simulations to target the data at an annual
horizon. We report the median and percentiles for the model statistics based on
10,000 Monte Carlo simulations with 43 × 12 monthly observations each that
match the length of the historical data. We also show the population values
that correspond to a long sample of 10,000 annualized observations.
Consumption and corporate payouts. Table IV reports the parameter values
for the model. In the spirit of the long-run risks literature, the consumption cal-
ibration features persistent low-frequency movements in the expected growth
Table IV
Configuration of Model Parameters
This table reports the model parameters. The model is calibrated at a monthly frequency.
Preferences δ γ ψ
0.9992 10 1.5
Consumption μc ρx ϕx σ0 ν σω
0.0024 0.985 0.038 0.005 0.999 0.000001
Cash Payout μs ρs φs ϕs α
−2.65 0.96 6 5 −0.15
Net Repurchase μh ϕh
0.045 22
Table V
Model Implications: Consumption
This table reports the data and model properties of real consumption growth. The summary statis-
tics in the data are computed in annual samples from 1929 to 2017 and from 1975 to 2017. The
median and 2.5%, 5%, 95%, and 97.5% values capture the model moment distributions across the
small samples whose size equals the data. Population values correspond to a long simulation of
the model. Means and volatilities are expressed in percentage terms.
Data Model
1929 to 2017 1975 to 2017 Med 2.5% 5% 95% 97.5% Pop
E(·) 2.85 2.66 2.88 1.38 1.61 4.22 4.43 2.90

σ (·) 2.80 1.59 1.95 1.18 1.27 3.03 3.29 2.21
AC(1) 0.39 0.25 0.43 0.10 0.16 0.67 0.70 0.52
AC(2) 0.07 0.18 0.18 −0.18 −0.14 0.49 0.55 0.32
AC(5) −0.07 −0.19 0.04 −0.29 −0.23 0.31 0.36 0.18
and consumption volatility. The persistence of expected growth component is

set at 0.985, and that of volatility shocks at 0.999. The scales of the expected
growth and volatility shocks are rather small to account for the empirical prop-
erties of the macroeconomic fundamentals in the data.
Table V shows that our model can match salient properties of the consump-
tion data. The table reports the mean, standard deviation, and persistence of
consumption growth at one, two, and five lags in the data and in the model.
The data moments are computed for the benchmark sample from 1975 to 2017
as well as for a long sample going back to 1929. The median model values are
close to the data, and in all cases the data values are within the model CIs.
We next calibrate the dynamics of the cash payouts and net repurchases.
The cointegrating residual between cash payouts and consumption is
moderately persistent (ρs = 0.96) and is exposed to the expected growth
fluctuations (φs = 6). The ratio of net repurchases to consumption is unpre-
dictable and driven by its own i.i.d. shock, and we set its overall level and
scale to target the unconditional moments of net repurchases in the data. As
shown in Table VI, the model can successfully capture the key properties of
Table VI
Model Implications: Corporate Asset Payouts
This table reports the data and model properties of scaled changes in cash payouts, net repur-
chases, and total payouts on corporate assets. Payout changes are scaled by the consumption level.
Summary statistics in the data are computed in the annual sample from 1975 to 2017. The me-
dian and 2.5%, 5%, 95%, and 97.5% values capture the model moment distributions across the
small samples whose size equals the data. Population values correspond to a long simulation of
the model. Means and volatilities are expressed in percentage terms.
Panel A: Cash Payout
Model
Data Med 2.5% 5% 95% 97.5% Pop
E(·) 0.19 0.21 0.06 0.08 0.37 0.41 0.22

σ (·) 0.56 0.59 0.36 0.39 0.95 1.05 0.65
AC(1) 0.33 0.24 −0.08 −0.02 0.46 0.50 0.26
AC(2) −0.28 −0.03 −0.34 −0.29 0.23 0.27 −0.01
AC(5) −0.21 −0.05 −0.34 −0.30 0.21 0.24 −0.05
% o f Neg Payouts 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Panel B: Net Repurchase
Model
Data Med 2.5% 5% 95% 97.5% Pop
E(·) −0.27 −0.16 −0.44 −0.38 0.07 0.13 −0.16

σ (·) 4.84 5.25 3.22 3.43 7.96 8.39 5.52
AC(1) −0.28 −0.47 −0.67 −0.64 −0.28 −0.24 −0.51
AC(2) −0.29 −0.02 −0.37 −0.30 0.29 0.33 0.02
AC(5) 0.13 0.00 −0.34 −0.29 0.27 0.31 −0.00
% o f Neg Payouts 86.05 92.86 80.95 83.33 100.00 100.00 92.86
Panel C: Total Payout
Model
Data Med 2.5% 5% 95% 97.5% Pop
E(·) −0.08 0.06 −0.24 −0.18 0.31 0.37 0.06

σ (·) 4.78 5.26 3.25 3.46 8.00 8.45 5.55
AC(1) −0.29 −0.47 −0.66 −0.64 −0.27 −0.23 −0.50
AC(2) −0.30 −0.02 −0.37 −0.30 0.29 0.33 0.02
AC(5) 0.13 0.00 −0.35 −0.28 0.25 0.32 −0.00
% o f Neg Payouts 44.19 30.95 11.90 14.29 50.00 53.57 32.24
payouts in the data. Changes in net repurchases are nearly 10 times more
volatile than changes in cash payouts. This leads to highly volatile total asset
payouts dominated by shocks to net repurchases, as in the data. In the model,
net repurchases and total payouts can go negative, whereas cash dividends are
always positive. Similar to the data, asset net repurchases in the model are
negative 93% of the time, while total payouts go below zero 30% of the time.
Figure 12. Model implications for corporate asset payout and return cyclicality. This
figure shows multihorizon correlations between real consumption growth and scaled changes in
cash payouts, net repurchases, total payouts, and excess returns on corporate assets in the data
and in the model. The payouts include cash payouts, net repurchases, and total payouts. The data
(solid line) are real quarterly observations from Q1.1975 to Q4.2017. Model median (circles) and
95% confidence interval (dashed line) are based on a long simulation of the model. Panel A shows
the results for the benchmark model, and Panel B for the restriction of the model to power utility.
Changes in annual cash payouts are mildly persistent both in the model and
in the data, while changes in net repurchases and in total payouts actually
have a negative autocorrelation. The model structure captures this evidence
because i.i.d. shocks to the levels of net repurchases lead to negative autocor-
relation for the changes. Total payouts are dominated by net repurchases, and
inherit their negative persistence.
Finally, the model can successfully reproduce the short- and long-run cycli-
cality of payouts in the data, as shown in the top panel of Figure 12. Scaled
changes in cash payouts are positively correlated with consumption growth
rates, while changes in net repurchases and total payouts are essentially
acyclical at all horizons. The data estimates are comparable to the model and
are within the model CI. Notably, CIs implied by the model are quite wide,
consistent with the data and our argument that net repurchases introduce
substantial noise in measuring the exposures of aggregate payouts to economic
growth risks in small samples.
Asset prices and economic risk. We calibrate the preference parameters to
the standard values in the literature. In the benchmark specification, risk
aversion is set at 10, and the IES parameter is 1.5. This configuration implies
Table VII
Model Implications: Corporate Asset Prices
This table reports the data and model properties of the real risk-free rate and corporate asset
returns. The summary statistics in the data are computed in the annual sample from 1975 to 2017.
The median and 2.5%, 5%, 95%, and 97.5% values capture the model moment distributions across
the small samples whose size equals the data. Population values correspond to a long simulation
of the model. Means and volatilities are expressed in percentage terms.
Model
Moment Data Median 2.5% 5% 95% 97.5% Pop
Risk-Free Return
E(r f ) 0.79 1.64 0.46 0.63 2.41 2.52 1.62
σ (r f ) 1.75 0.80 0.40 0.46 1.38 1.57 1.04
Asset Return
E(rd ) 6.87 5.24 −0.19 0.73 8.56 9.35 5.04
σ (rd ) 12.67 10.85 7.12 7.57 18.85 22.13 12.22
a preference for early resolution of uncertainty and a strong substitution

effect, which enable the model to generate sizeable risk compensation and
realistic asset price dynamics (see Bansal and Yaron (2004)). We also restrict
the model to the power utility case by setting the IES to the reciprocal of the
risk-aversion coefficient.
Table VII shows the model implications for the key asset pricing moments,
such as the mean and standard deviation of the risk-free rate and the asset
return. The model replicates quite well a relatively low level and volatility of
the risk-free rate in the data. The level is 0.79% in the data relative to 1.64% in
the model, and its volatility is 1.75% in the data and 0.81% in the model. The
asset returns are risky: the asset risk premium is 6.87% in the data relative
to 5.23% in the model, and the asset return volatility is about 12.67% and
10.99%, respectively.
Table VIII decomposes the conditional variation in the SDF, asset returns,
and asset payouts into components related to each primitive economic shock.
As discussed earlier, net repurchases are the key drivers of the short-run
variation in total payouts. In fact, shocks to net repurchases εt+1 explain
nearly all of the variation in next-quarter total payouts, log(Dt+1 a
/Ct+1 ). The
long-run distribution of payouts, however, is governed by their exposure to
persistent fluctuations in expected consumption risk, which originates in
the cash component of payouts (ϕs > 0.) Preference for early resolution of
uncertainty magnifies the contribution of persistent expected consumption
risk to asset prices. Fluctuations in expected growth explain 30% of the
conditional variation in the SDF and close to 40% in asset returns. The rest
comes from the volatility shocks and, to a lesser extent, shocks to realized
consumption growth. In contrast, net repurchases risks are unpriced and
contribute about 30% to the return variation, down from over 99.9% for asset
payouts. Thus, zeroing out fluctuations in net repurchases by setting ϕh = 0
Table VIII
Model Decomposition of Risk
This table reports the percentage contributions of each economic risk in the model to the condi-
tional variation in next-period corporate asset total payouts, stochastic discount factor, and asset
returns. The left columns show the results for the full model, and the right columns show the
evidence under power utility.
Innovations
Benchmark Power Utility
Total Asset Total Asset

Shocks Payout SDF Return Payout SDF Return
Consumption 0.00 5.85 2.59 0.00 100.00 0.04

Exp. Consumption 0.00 29.75 38.38 0.00 0.00 6.27
Volatility 0.00 64.40 26.50 0.00 0.00 1.47
Cash Payout 0.02 0.00 3.62 0.02 0.00 6.19
Net Repurchase 99.98 0.00 28.92 99.98 0.00 86.04
would considerably change the dynamics of total asset payouts, but would not
materially change the model risk premium.27
The risk exposures of the asset payouts, together with the risk preferences
of the agent, have direct implications for the equilibrium properties of asset
returns. As shown in Figure 12, the equilibrium returns inherit and magnify
the exposure of cash payouts to the expected growth risks and are correlated
with economic growth rates at high and low frequencies, as in the data.
For comparison, we consider a restricted version of the model in which we
do not change the dynamics of the payouts but assume power utility. With
no preference for timing of uncertainty resolution, the expected growth and
volatility risks are unpriced, and most of the variation in payouts and returns
now comes from fluctuations in net repurchases (see bottom panel of Ta-
ble VIII). Furthermore, because the IES implied by the power utility is below
one, the exposure of returns to expected growth risks is actually negative,
which results in countercyclical rather than procyclical returns (bottom panel
of Figure 12) and the absence of significant risk compensation. Thus, the asset
pricing evidence helps assess and validate the plausibility of the economic
environment and, in particular, the dynamics of payouts and investors’ prefer-
ence for risk. In our framework, preference for early resolution of uncertainty
and the exposure of cash payouts to low-frequency growth risk, together with
a novel ingredient of volatile and acyclical net repurchases, go a long way to
quantitatively account for the joint evidence of cash, net repurchases, total
payouts, and returns in the data.
27 The volatility of net repurchases is exposed to consumption volatility, so that net repurchases
affect the asset exposure to volatility risks. Net repurchases also impact the values of the steady
states for the log-linearization of returns. These effects are quite small.
IV. Conclusions
We measure the market value of U.S. corporate assets and their payouts to
investors. Our measure of total payouts includes not only equity dividends and
bond coupon payments (cash payouts), but also net transfers in the form of re-
purchases and new issuances of equity and debt. We show that incorporating
debt and net repurchases affects the key properties of the aggregate payouts
in the data. First, total asset payouts often turn negative, meaning that there
are periods in which investors finance the aggregate corporate sector. Second,
net repurchases are highly volatile and are a dominant component of total pay-
outs. Third, while cash payouts are procyclical, total payouts appear acyclical
at short and long frequencies. At the same time, corporate asset returns are
about as risky as equity returns, and both are exposed to fluctuations in eco-
nomic growth.
We develop an asset pricing framework to interpret the empirical evidence.
In the model, net repurchases are acyclical and highly volatile, and as such
mask the exposure of the cash component of total payouts to low-frequency
economic risks. The model can quantitatively account for the joint dynamics of
consumption growth, cash, net repurchases, and total asset payouts, together
with asset returns.
Several extensions of our paper would be fruitful to pursue in future
work. On the empirical side, it is important to consider cross-sectional differ-
ences in valuations and payouts across firms. On the theory, future research
should provide an economic environment to endogenize the corporate financ-
ing and payout decisions. Finally, it would be interesting to extend the analysis
beyond the public sector to capture private firms. We leave these extensions for
future research.
Initial submission: February 9, 2018; Accepted: August 6, 2021

Editors: Stefan Nagel, Philip Bond, Amit Seru, and Wei Xiong
Appendix A: Bond Data
Table A.I
Bloomberg Barclays Index Data
This table reports characteristics of bond subindices from Bloomberg Barclays that are used in
the construction of corporate debt. Corporate issues of ABS and CMBS (indicated with *) have a
restriction on the remaining average life rather than minimum maturity.
Minimum
Maturity/
Index Start Quality Minimum Average
Date Issue Size Life
Long-Term Debt Components

U.S. Corporate IG Jan-73 IG $250m 1 year
U.S. Corporate HY Jan-83 HY $150m 1 year
U.S. 144a Ex-Aggregate Feb-98 IG $250m 1 year
U.S. Commercial MBS Jun-99 IG, HY $250m 1 year∗
U.S. Fixed-Rate ABS Jan-92 IG $500m Deal Size 1 year∗
$25m Tranche Size
U.S. Tax-Exempt Municipals Jan-73 IG $250m 1 year
U.S. Convertibles Composite Jan-03 IG, HY $250m 1 month
Short-Term Debt Components
U.S. Floating-Rate ABS May-05 IG $500m Deal Size 1 year∗
$25m Tranche Size
U.S. Floating-Rate Notes Oct-03 IG $300m 1 month
(13 months
prior Apr-07)
U.S. Floating-Rate Notes HY Mar-06 HY $150m 1 year
Figure A.1. Bond data coverage. This figure shows book values of bonds from Bloomberg Bar-
clays, Flow of Funds, Mergent, and Compustat (long-term and short-term) databases. The data are
real quarterly observations from 1975 to 2015, in trillions of December 2009 dollars. (Color figure
Appendix B: Robustness Checks

We perform a variety of checks to assess the robustness of our payout cycli-
cality results. Specifically, in addition to the benchmark sample from 1975 to
2017, we consider a shorter sample that stops in 2006 before the Financial Cri-
sis, as well as the sample period excluding the Financial Crisis (i.e., excluding
the period from Q1.2008 to Q2.2009). For equity data, we also provide results
for the 1949 to 2017 and 1949 to 2006 samples. In addition, we look at changes
in quarterly payouts seasonally adjusted through the band-pass filter and the
X-12 ARIMA filter, and we look at year-to-year changes. Finally, we include
sensitivity checks for the components of payouts and for measures of equity
and debt.
Table B.I
Corporate Equity Payout Cyclicality
This table reports correlations of growth rates in consumption with scaled changes in cash payouts,
net repurchases, issuances, repurchases, and total payouts on corporate equity. Payout changes are
scaled by the consumption level. Payouts are sampled at a quarterly frequency, and are seasonally
adjusted through either a band-pass or X12-ARIMA filter or by computing year-to-year changes.
The data are real quarterly observations over various sample periods.
Cash Net Total

Payout Repurchase Issuance Repurchase Payout
Sample: 1949 to 2017

Band-passed 0.26 −0.16 0.23 0.13 −0.15
X12 ARIMA 0.18 −0.04 0.11 0.11 −0.04
Year-to-year 0.27 −0.09 0.17 0.16 −0.08
Band-passed 0.20 −0.22 0.28 0.14 −0.21
X12 ARIMA 0.19 −0.06 0.14 0.05 −0.09
Year-to-year 0.29 −0.14 0.19 0.11 −0.13
Band-passed 0.27 0.04 0.09 0.18 0.05
X12 ARIMA 0.17 0.02 0.05 0.17 0.01
Year-to-year 0.22 0.04 0.12 0.23 0.05
Band-passed 0.03 −0.08 0.18 0.19 −0.08
X12 ARIMA 0.18 −0.04 0.10 0.08 −0.08
Year-to-year 0.20 −0.03 0.14 0.16 −0.03
Sample: 1975 to 2017 (Excluding Financial Crisis)
Band-passed 0.23 0.00 0.12 0.18 0.01
X12 ARIMA 0.12 −0.06 0.08 0.09 −0.08
Year-to-year 0.15 0.02 0.10 0.18 0.03
Table B.II
Debt Payout Cyclicality
This table reports correlations of growth rates in consumption with scaled changes in cash payouts,
net repurchases, and total payouts on corporate debt. Payout changes are scaled by the consump-
tion level. Payouts are sampled at a quarterly frequency, and are seasonally adjusted through
either a band-pass or X12-ARIMA filter or by computing year-to-year changes. The data are real
quarterly observations over various sample periods.
Cash Net Total

Payout Repuchase Payout

Band-passed 0.32 −0.02 0.01
X12 ARIMA 0.16 0.03 0.05
Year-to-year 0.25 −0.06 −0.05
Band-passed 0.17 −0.06 −0.05
X12 ARIMA 0.04 0.03 0.03
Year-to-year 0.05 −0.00 0.00
Band-passed 0.32 −0.06 −0.03
X12 ARIMA 0.05 0.04 0.03
Year-to-year 0.12 0.06 0.07
Table B.III
Corporate Asset Payout Cyclicality
This table reports correlations of growth rates in consumption with scaled changes in cash pay-
outs, net repurchases, and total payouts on corporate assets. Payout changes are scaled by the
consumption level. Payouts are sampled at a quarterly frequency, and are seasonally adjusted
through either a band-pass or X12-ARIMA filter or by computing year-to-year changes. The data
are real quarterly observations over various sample periods.
Cash Net Total

Payout Repurchase Payout

Band-passed 0.37 0.02 0.05
X12 ARIMA 0.24 0.03 0.04
Year-to-year 0.29 −0.02 −0.00
Band-passed 0.15 −0.11 −0.10
X12 ARIMA 0.13 −0.05 −0.05
Year-to-year 0.15 −0.05 −0.04
Band-passed 0.36 −0.03 −0.01
X12 ARIMA 0.12 −0.05 −0.05
Year-to-year 0.17 0.04 0.05
Figure B.1. Corporate payout and return betas to output. This figure shows multihorizon
output betas of scaled changes in cash payouts, net repurchases, total payouts, and excess returns
on corporate equity, debt, and assets. Payout changes are scaled by the consumption level. Betas
(solid lines) and 95% confidence intervals (dashed lines) are based on OLS regressions of variables
of interest on output growth rates at one-quarter to five-year horizons. The data are real quarterly
Figure B.2. Corporate asset payout cyclicality: Insurance company cross-holdings ad-
justments. This figure shows multihorizon correlations between real consumption growth and
scaled changes in cash payouts, net repurchases, and total payouts on corporate assets under al-
ternative adjustments for cross-holdings of bonds. To compute returns on bonds held by insurance
companies, we alternatively assume that insurance companies invest in the aggregate bond index
(Panel A) or investment-grade bonds only (Panel B). In Panel C, we adjust for the bond cross-
holdings reported on general accounts only. Payout changes are scaled by the consumption level.
Solid lines show correlations at one-quarter to five-year horizons, along with 95% confidence inter-
vals (dashed lines). The data are real quarterly observations from Q1.1975 to Q4.2017. Standard
errors are Newey-West adjusted. (Color figure can be viewed at wileyonlinelibrary.com)
Figure B.3. Corporate asset payout cyclicality: Robustness I. This figure shows multihori-
zon correlations between real consumption growth and scaled changes in cash payouts, net repur-
chases, and total payouts on corporate assets under alternative payout measurements. We exclude
insurance companies in Panel A and the financial sector entirely in Panel B. In Panel C, we ig-
nore cross-holding adjustments for equity and debt. Solid lines show correlations at one-quarter to
five-year horizons, along with 95% confidence intervals (dashed lines). The data are real quarterly
Figure B.4. Corporate asset payout cyclicality: Robustness II. This figure shows multi-
horizon correlations between real consumption growth and scaled changes in cash payouts, net
repurchases, and total payouts on corporate assets under alternative payout measurements. We
exclude IPOs in Panel A and we exclude mergers and acquisitions in Panel B. In Panel C, we use
the Mergent database to compute the value of short-term bonds. Solid lines show correlations at
one-quarter to five-year horizons, along with 95% confidence intervals (dashed lines). The data are
real quarterly observations from Q1.1975 to Q4.2017. Standard errors are Newey-West adjusted.
Figure B.5. Corporate equity net repurchases. This figure shows benchmark measures of
net equity repurchases, together with net equity repurchases for matched firms in the Factset
database based on public float or total outstanding number of shares. The repurchases are in
levels (Panel A) or changes scaled by the consumption level (Panel B). The data are real annual
observations from 1990 to 2017, and are expressed in trillions of December 2009 dollars. (Color
Figure B.6. Corporate asset payout cyclicality: Global risks. This figure shows multihori-
zon correlations between real global output growth and scaled changes in cash payouts, net repur-
chases, and total payouts. Measures of global output include value-weighted global GDP (Panel A),
value-weighted global GDP excluding the United States (Panel B), and equally weighted global
GDP (Panel C). To compute global GDP, we use OECD quarterly output data for 17 major in-
dustrialized countries, such as the United States, Canada, France, Germany, Italy, Japan, the
United Kingdom, Australia, Belgium, Denmark, the Netherlands, New Zealand, Norway, Portu-
gal, Spain, Sweden, and Switzerland. Solid lines show correlations at one-quarter to five-year
horizons, along with 95% confidence intervals (dashed lines). The data are real quarterly obser-
vations from Q1.1975 to Q4.2017. Standard errors are Newey-West adjusted. (Color figure can be
viewed at wileyonlinelibrary.com)
Figure B.7. Debt cash payout cyclicality. This figure shows multihorizon correlations between
real consumption growth and scaled changes in cash payouts on debt. Panel A shows results for
NIPA interest expenses for all firms, Panel B for nonfinancials, and Panel C for interest expenses
computed in Larrain and Yogo (2008). Solid lines show correlations at one-quarter to five-year hori-
zons, along with 95% confidence intervals (dashed lines). The data are real quarterly observations
from Q1.1975 to Q4.2017 in Panels A and B, and annual observations from 1929 to 2004 in Panel C.
Standard errors are Newey-West adjusted. (Color figure can be viewed at wileyonlinelibrary.com)
Figure B.8. Corporate equity issuance and repurchase cyclicality. This figure shows mul-
tihorizon correlations between real consumption growth and scaled changes in net repurchases,
and gross repurchases and issuances of corporate equity. Solid lines show correlations at one-
quarter to five-year horizons, along with 95% confidence intervals (dashed lines). The data are
real quarterly observations from Q1.1975 to Q4.2017 (Panel A) and Q1.1949 to Q4.2017 (Panel B).
Standard errors are Newey-West adjusted. (Color figure can be viewed at wileyonlinelibrary.com)
Figure B.9. Scaled changes versus log growth rates. This figure shows changes in cash pay-
outs on corporate equity, debt, and assets scaled by the consumption level, together with the log
growth rates of cash payouts adjusted by the average ratio of cash payouts to the consumption
level (dashed line). The data are real annual observations from 1975 to 2017. (Color figure can be
viewed at wileyonlinelibrary.com)
Appendix C: Negativity of Payouts

Under our benchmark approach, the aggregate payout Da,t can be negative.
Here we provide an illustration of the managed strategy, which eliminates neg-
ative payouts by selling shares in the asset.
A managed portfolio holds st shares in the underlying asset with return Rt+1 .
It distributes a positive payout to investors. If the asset payout is negative, it fi-
nances it by selling off a portion of the shares, so st+1 goes down. For simplicity,
we do not consider buying back the asset.
Using the issuance/repurchase algebra of Section I.A, it follows that the mar-
ket value of the fund is st Vt , and its total gains next period can be reallocated
between the market value and the share-adjusted payouts:
st Vt Rt+1 = st (Vt+1 + Da,t+1 ) = st+1Vt+1 + D̃a,t+1 . (C1)
The total managed payout can be decomposed into the cash payout st Da,t+1 and
the net repurchase term (st − st+1 )Vt+1 :
D̃a,t+1 = st Da,t+1 + (st − st+1 )Vt+1 . (C2)
The share dynamics are dictated by the fund’s objective to keep payouts
nonnegative. For D̃a,t+1 to be bigger or equal to zero, st+1 cannot exceed
st (1 + Da,t+1 /Vt+1 ). Hence, the share dynamics follow
st+1 = min(st , st (1 + Da,t+1 /Vt+1 )). (C3)

Figure C.1. Managed portfolio. This figure shows the managed payouts on corporate equity,
debt, and assets (left panel), as well as the evolution of the number of shares in a managed port-
folio (right panel). The managed strategy sells off a portion of the underlying investment to fi-
nance a negative payout. The data are real quarterly observations from Q1.1975 to Q4.2017, and
are expressed in trillions of December 2009 dollars. (Color figure can be viewed at wileyonlineli-
brary.com)
Appendix D: Wavelet Analysis

The wavelet correlation between two stochastic processes x and y for scale
λ j = 2 j−1 equals
(x) (y)

cov W j,t , W j,t

ρxy λ j = (x)
(y)
12 , (D1)
var W j,t var W j,t
(x) (y)
where W j,t and W j,t are the scale λ j maximal overlap discrete wavelet trans-
form (MODWT) coefficients for x and y, respectively. Since this is just a cor-
relation coefficient between two random variables on a scale-by-scale basis,
−1 ≥ ρxy (λ j ) ≤ 1 for all j. The MODWT coefficient for a stochastic process u is
defined as
L j −1
(u)
W j,t = h̃ j,l ut−l , (D2)
l=0
where {h̃ j,0 , . . . , h̃ j,L j −1 } are the wavelet filter coefficients from a Daubechies
compactly supported wavelet family, with L j = (2 j − 1)(L − 1) + 1.
We estimate the sample wavelet correlation by simply using the estimators
of wavelet covariance and wavelet variance, respectively,
1
T−1
(x) (y)
γ̂xy λ j = W j,t W j,t (D3)
Tj
t=L j −1
and
1 (x)
2
T−1

ν̂x2 λ j = W j,t , (D4)
Tj
t=L j −1
with Tj = T − L j + 1.
Whitcher, Guttorp, and Percival (2000) establish a central limit theorem for
the estimator of wavelet correlation,

γ̂xy λ j
ρ̂xy λ j = , (D5)
ν̂x λ j ν̂y λ j
and construct an approximate confidence interval (CI).

An approximate 100(1 − 2p)% CI for ρxy (λ j ) is given by
⎧ ⎫
⎨ (1 − p ) ⎬
−1
tanh tanh−1 ρ̂xy λ j ± , (D6)
⎩ T − L
− 3 ⎭ j j
with L
j = (L − 2)(1 − 2− j ).
Table D.I
Long-Term Cyclicality Evidence, Wavelet Analysis
This table reports the 8-to-16-quarter (Panel A) and 16-to-32-quarter (Panel B) frequency correlations of growth rates in consumption or output
with scaled changes in cash payouts, net repurchases, total payouts, and excess returns on corporate equity, debt, and assets. The correlations and
confidence intervals are computed following the wavelet analysis of Whitcher, Guttorp, and Percival (2000). The data are real quarterly observations
from Q1.1975 to Q4.2017.
Panel A: 8-to-16-Quarter Frequency
Consumption Output

Equity Est. 0.41 −0.02 −0.01 0.39 0.42 −0.01 −0.00 0.40
67% CI 0.21, 0.58 −0.25, 0.21 −0.24, 0.21 0.19, 0.57 0.22, 0.59 −0.24, 0.21 −0.23, 0.22 0.19, 0.57
90% CI 0.05, 0.68 −0.39, 0.35 −0.38, 0.36 0.03, 0.67 0.06, 0.69 −0.38, 0.36 −0.37, 0.37 0.03, 0.67
Debt Est. 0.32 −0.11 −0.08 0.49 0.34 −0.12 −0.08 0.48
67% CI 0.11, 0.51 −0.33, 0.11 −0.30, 0.15 0.29, 0.64 0.12, 0.52 −0.34, 0.11 −0.30, 0.15 0.29, 0.64
90% CI −0.05, 0.62 −0.46, 0.27 −0.43, 0.30 0.14, 0.73 −0.03, 0.63 −0.47, 0.26 −0.44, 0.30 0.14, 0.72
Asset Est. 0.43 −0.05 −0.02 0.41 0.45 −0.05 −0.02 0.41
67% CI 0.23, 0.60 −0.27, 0.18 −0.25, 0.20 0.20, 0.58 0.24, 0.61 −0.27, 0.18 −0.25, 0.20 0.20, 0.58
90% CI 0.07, 0.69 −0.41, 0.33 −0.39, 0.35 0.05, 0.68 0.09, 0.70 −0.41, 0.32 −0.39, 0.35 0.05, 0.68
(Continued)
203
204
Table D.I—Continued
Panel B: 16-to-32-Quarter Frequency
Consumption Output

Equity Est. 0.62 0.14 0.16 0.48 0.62 0.14 0.17 0.48
67% CI 0.34, 0.80 −0.23, 0.47 −0.21, 0.48 0.15, 0.71 0.34, 0.80 −0.22, 0.47 −0.20, 0.49 0.16, 0.71
90% CI 0.10, 0.87 −0.45, 0.64 −0.43, 0.65 −0.10, 0.82 0.10, 0.87 −0.44, 0.65 −0.42, 0.66 −0.09, 0.82
Debt Est. 0.44 −0.34 −0.29 0.43 0.46 −0.34 −0.29 0.43
67% CI 0.11, 0.69 −0.62, 0.01 −0.58, 0.07 0.09, 0.68 0.13, 0.70 −0.62, 0.01 −0.58, 0.07 0.09, 0.68
90% CI −0.15, 0.80 −0.75, 0.26 −0.72, 0.32 −0.16, 0.79 −0.12, 0.81 −0.75, 0.26 −0.72, 0.32 −0.16, 0.80
The Journal of Finance®
Asset Est. 0.54 −0.17 −0.10 0.48 0.55 −0.16 −0.09 0.48
67% CI 0.23, 0.75 −0.49, 0.20 −0.44, 0.26 0.15, 0.71 0.24, 0.75 −0.49, 0.20 −0.43, 0.27 0.15, 0.71
90% CI −0.02, 0.84 −0.66, 0.42 −0.62, 0.48 −0.10, 0.81 −0.01, 0.84 −0.66, 0.43 −0.61, 0.48 −0.10, 0.82
Appendix E: Model Solution

The equilibrium consumption claim loadings are given by

1 1 θ 2
A0,c = log(δ ) + 1 − μc + κ0,c + κ1,c A2,c (1 − ν )σ0 + κ1,c A2,c σω
2
,
1 − κ1,c ψ 2
1− 1
ψ
A1,c = ,
1 − κ1,c ρx

1 − ψ1 (γ − 1 ) 2
κ1,c ϕx
A2,c =− 1+ .
2 1 − κ1,c ν 1 − κ1,c ρx
The market prices of risks are
λη = γ , λe = −(θ − 1 )κ1,c A1,c ϕx , λω = −(θ − 1 )κ1,c A2,c .
The log-linearization coefficients for the corporate asset satisfy

κ0,d = log 1 + exp{vcd } + exp dc − exp hc − κ1,d vcd − κ2,d dc − κ3,d hc,
exp{vcd }
κ1,d = ,
1 + exp{vcd } + exp dc − exp hc

exp dc
κ2,d = ,

exp hc
κ3,d =− .
The loadings for the corporate claim are given by

1
A0,d = m0 + μc + κ0,d + κ1,d A2,d (1 − ν )σ02 + κ2,d μs (1 − ρs ) + · · ·
1 − κ1,d
2
+ κ3,d μh + 0.5 κ1,d A2,d − λω σω2 ,

1 1 κ2,d φs
A1,d = 1− + ,
1 − κ1,d ρx ψ 1 − κ1,d ρs
1 2 2
A2,d = m2 + 0.5 1 − λη + κ1,d A1,d ϕx − λe + · · ·
1 − κ1,d ν
2
+ κ1,d A3,d + κ2,d ϕs2 + 2α 1 − λη κ1,d A3,d + κ2,d ϕs + κ3,d

2
ϕh2 ,
κ2,d ρs
A3,d = .
1 − κ1,d ρs
REFERENCES
American Council of Life Insurance, 2020, Life Insurance Fact Book (Institute of Life Insurance).
Andrews, Donald W. K., and J. Christopher Monahan, 1992, An improved heteroskedasticity and
autocorrelation consistent covariance matrix estimator, Econometrica 60, 953–966.
Bansal, Ravi, Robert F. Dittmar, and Christian T. Lundblad, 2005, Consumption, dividends, and
the cross section of equity returns, Journal of Finance 60, 1639–1672.
Bansal, Ravi, and Amir Yaron, 2004, Risks for the long run: A potential resolution of asset pricing
puzzles, Journal of Finance 59, 1481–1509.
Bansal, Ravi, and Amir Yaron, 2007, The asset pricing macro nexus and return cash-flow pre-
dictability, in 2007 Meeting Papers, Number 18, Society for Economic Dynamics.
Barro, Robert, 2006, Rare disasters and asset markets in the twentieth century, Quarterly Journal
of Economics 121, 823–866.
Begenau, Juliane, and Juliana Salomao, 2019, Firm financing over the business cycle, Review of
Financial Studies 32, 1235–1274.
Bhamra, Harjoat, Lars-Alexander Kuehn, and Ilya Strebulaev, 2010, The levered equity risk pre-
mium and credit spreads: A unified framework, Review of Financial Studies 23, 645–703.
Boudoukh, Jacob, Roni Michaely, Matthew Richardson, and Michael R. Roberts, 2007, On the
importance of measuring payout yield: Implications for empirical asset pricing, Journal of
Finance 62, 877–915.
Brainard, William C., and John B. Shoven, 1980, The financial valuation of the return to capital,
Brookings Papers on Economic Activity 2, 453–511.
Butler, Alexander W., Jess Cornaggia, Gustavo Grullon, and James Peter Weston, 2011, Corporate
financing decisions, managerial market timing, and real investment, Journal of Financial
Economics 101, 666–683.
Campbell, John, and John Cochrane, 1999, By force of habit: A consumption-based explanation of
aggregate stock market behavior, Journal of Political Economy 107, 205–251.
Campbell, John Y., and Robert J. Shiller, 1988, The dividend-price ratio and expectations of future
dividends and discount factors, Review of Financial Studies 1, 195–228.
Chen, Hui, 2010, Macroeconomic conditions and the puzzles of credit spreads and capital struc-
ture, Journal of Finance 65, 2171–2212.
Chodorow-Reich, Gabriel, Andra Ghent, and Valentin Haddad, 2021, Asset insulators, Review of
Financial Studies 34, 1509–1539.
Choi, Jaewon, and Matthew Richardson, 2016, The volatility of a firm’s assets and the leverage
effect, Journal of Financial Economics 121, 254–277.
Covas, Francisco, and Wouter J. Den Haan, 2011, The cyclical behavior of debt and equity finance,
American Economic Review 101, 877–899.
Croce, Mariano M., 2014, Long-run productivity risk. A new hope for production-based asset pric-
ing?, Journal of Monetary Economics 66, 13–31.
Darmouni, Olivier, and Lira Mota, 2020, The financial assets of non-financial firms, Working pa-
per, Columbia Business School.
Dichev, Ilia D., 2007, What are stock investors? Actual historical returns? Evidence from dollar-
weighted returns, American Economic Review 97, 386–401.
Dittmar, Amy, and Robert F. Dittmar, 2004, Stock repurchase waves: An explanation of the trends
in aggregate corporate payout policy, Working Paper, Ross School of Business.
Duchin, Ran, Thomas Gilbert, Jarrad Harford, and Christopher Hrdlicka, 2017, Precautionary
savings with risky assets: When cash is not cash, Journal of Finance 72, 793–852.
Ellul, Andrew, Chotibhak Jotikasthira, and Christian T. Lundblad, 2011, Regulatory pressure and
fire sales in the corporate bond market, Journal of Financial Economics 101, 596–620.
Epstein, Larry G., and Stanley E. Zin, 1989, Substitution, risk aversion, and the temporal behav-
ior of consumption and asset returns: A theoretical framework, Econometrica: Journal of the
Econometric Society 57, 937–969.
Fama, Eugene, and Kenneth French, 2001, Disappearing dividends: Changing firm characteristics
or lower propensity to pay?, Journal of Financial Economics 60, 3–43.
Ferson, Wayne, Suresh Nallareddy, and Biqin Xie, 2013, The “out of sample” performance of long-
run risk models, Journal of Financial Economics 107, 537–556.
Gao, Xiaohui, Jay R. Ritter, and Zhongyan Zhu, 2013, Where have all the IPOs gone?, Journal of
Financial and Quantitative Analysis 48, 1663–1692.
Gomes, João F., Mete Kilic, and Sebastien Plante, 2020, Leverage risk and investment: The case
of gold clauses in the 1930s, Working paper, Wharton School.
Grullon, Gustavo, and Roni Michaely, 2002, Dividends, share repurchases, and the substitution
hypothesis, Journal of Finance 57, 1649–1684.
Guay, Wayne, and Jarrad Harford, 2000, The cash-flow permanence and information content of
dividend increases versus repurchases, Journal of Financial Economics 57, 385–415.
Hall, Bronwyn H., Clint Cummins, Elizabeth Laderman, and Joy Mundy, 1988, The R&D master
file documentation, NBER Working Paper No. t0072.
Hansen, Lars Peter, John C. Heaton, and Nan Li, 2008, Consumption strikes back? Measuring
long-run risk, Journal of Political Economy 116, 260–302.
Hodrick, Robert, 1992, Dividend yields and expected stock returns: Alternative procedures for
inference and measurement, Review of Financial Studies 5, 357–386.
Jagannathan, Murali, Clifford P. Stephens, and Michael Weisbach, 2000, Financial flexibility and
the choice between dividends and stock repurchases, Journal of Financial Economics 57, 355–
384.
Jermann, Urban, 1998, Asset pricing in production economies, Journal of Monetary Economics 41,
257–275.
Jermann, Urban, and Vincenzo Quadrini, 2012, Macroeconomic effects of financial shocks,
American Economic Review 102, 238–271.
Julliard, Christian, and Jonathan Parker, 2005, Consumption risk and the cross-section of ex-
pected returns, Journal of Political Economy 113, 185–222.
Kacperczyk, Marcin, and Philipp Schnabl, 2010, When safe proved risky: Commercial paper during
the financial crisis of 2007–2009, Journal of Economic Perspective 24, 29–50.
Kaltenbrunner, Georg, and Lars A. Lochstoer, 2010, Long-run risk through consumption smooth-
ing, Review of Financial Studies 23, 3141–3189.
Kreps, David M., and Evan L. Porteus, 1978, Temporal resolution of uncertainty and dynamic
choice theory, Econometrica 46, 185–200.
Kung, Howard, and Lukas Schmid, 2014, Innovation, growth, and asset prices, Journal of Finance
70, 1001–1037.
Larrain, Borja, and Motohiro Yogo, 2008, Does firm value move too much to be justified by subse-
quent changes in cash flow?, Journal of Financial Economics 87, 200–226.
Levin, Haan, Wouter J. Den, and Andrew T., 1997, A practitioner’s guide to robust covariance
matrix estimation, Handbook of Statistics 15, 299–342.
Lintner, John, 1956, Distribution of incomes of corporations among dividends, retained earnings,
and taxes, American Economic Review 46, 97–113.
Longstaff, Francis A., and Monika Piazzesi, 2004, Corporate earnings and the equity premium,
Journal of Financial Economics 74, 401–421.
Mehra, Rajnish, and Edward C. Prescott, 1985, The equity premium: A puzzle, Journal of Mone-
tary Economics 15, 145–161.
Müller, Ulrich K., and Mark W. Watson, 2018, Long-run covariability, Econometrica 86,
775–804.
Richardson, Scott A., and Richard G. Sloan, 2003, External financing and future stock returns,
Rodney L. White Center for Financial Research Working Paper No. 03–03.
Rietz, Thomas, 1988, The equity risk premium: A solution, Journal of Monetary Economics 22,
117–131.
Weil, Philippe, 1989, The equity premium puzzle and the risk-free rate puzzle, Journal of Monetary
Economics 24, 401–421.
Welch, Ivo, and Amit Goyal, 2008, A comprehensive look at the empirical performance of equity
premium prediction, Review of Financial Studies 21, 1455–1508.
Whitcher, Brandon, Peter Guttorp, and Donald B. Percival, 2000, Wavelet analysis of covariance
with application to atmospheric time series, Journal of Geophysical Research 105, 14,941–
14,962.
Supporting Information
Additional Supporting Information may be found in the online version of this
article at the publisher’s website:
Replication Code.

Jofi 13196

Uploaded by

Copyright:

Available Formats

Jofi 13196

Uploaded by

Document Information

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

Jofi 13196

Uploaded by

Copyright:

Available Formats

THE JOURNAL OF FINANCE • VOL. LXXVIII, NO.

How Risky Are U.S. Corporate Assets?

THE KEY QUESTION IN FINANCIAL ECONOMICS concerns the trade-off between

* Tetiana Davydiuk is at Carnegie Mellon University. Ivan Shaliastovich is at the University

Our empirical analysis highlights several features of total corporate asset

similar to the returns on equity: asset returns average 6.9%, comparable to

Pt Rt+1 = Pt+1 + CFt+1 . (1)

Vt Rt+1 = Vt+1 + Da,t+1 . (2)

Da,t+1 = Vt Rt+1 − Vt+1 = (Pt+1 + CFt+1 ) × Nt − Pt+1 × Nt+1

REPt+1 ≡ −{Nt+1 − Nt }− Pt+1 ≥ 0. (4)

It is positive when there is a repurchase of existing shares, that is, when

ISSt+1 ≡ {Nt+1 − Nt }+ Pt+1 ≥ 0. (5)

This represents the inflow of resources following a new issuance of shares

B. Data and Empirical Measurements

Equity-related data. To measure equity-related variables, we use the Center

nrepit = prc∗it−1 × shroutit−1

shares over the period,

Bloomberg Barclays or Mergent. Finally, Compustat data include bank debt, so

a more recent sample. Coupon cash flows are constructed by multiplying

contribute to the mismeasurement of payouts and valuations.

firm-level data as in Ellul, Jotikasthira, and Lundblad (2011) and Chodorow-

Figure 2 shows the dynamics of the market capitalization of bonds, equity,

payouts on each of these instruments into cash distributions (dividends or

Panel A: Cash Payout

Equity Debt Asset

Mean 0.11 0.08 0.19

Panel B: Net Repurchase

Equity Debt Asset

Mean −0.09 −0.18 −0.27

Panel C: Total Payout

Equity Debt Asset

Mean 0.02 −0.09 −0.08

Equity Debt Asset

Mean 7.91 2.73 6.87

Payout levels are nonstationary, so we need to consider changes to formally

14 We have also examined normalizations by previous-period consumption level, by the average

C.3. Economic Growth Risk Exposure

Panel A: Quarterly Changes

Cash Net Total Excess Cash Net Total Excess

Equity 0.15 0.02 0.03 0.12 0.14 0.15 0.16 0.06

Panel B: Year-to-Year Changes

Cash Net Total Excess Cash Net Total Excess

Equity 0.21 0.03 0.04 0.25 0.27 0.09 0.11 0.28

Payouts. First, we examine the business-cycle behavior of payouts. Table II

for h equal to 0,1,...,20 quarters. For h = 0, we recover the payout exposures

correlations of returns with measures of economic growth are nearly always

II. Discussion, Extensions, and Robustness

B. Corporate Structure and Equity Ownership

Finally, similar to the majority of the literature, our analysis is conducted

unadjusted debt data.

under the transversality condition lim j→∞ Et (Mt,t+ jVt+ j ) = 0.

22 For simplicity, we abstract from buying securities back.

on managed portfolios represent selling off securities at times of negative pay-

dollar-weighted return is just the internal rate of return on buying an asset at

E. Acyclicality of Net Repurchases

24 Following the literature, we attribute a firm’s net repurchases to repurchases (issuances) if

F. Measurements and Inference

Panel A: 16 Quarter Frequency

Cash Net Total Excess Cash Net Total Excess