Upload
will-hendricksen
View
9
Download
0
Tags:
Embed Size (px)
DESCRIPTION
This paper contends that cash flows are better predictors of returns than earnings...
Citation preview
Electronic copy available at: http://ssrn.com/abstract=2472571
Are Cash Flows Better Stock Return Predictors Than Profits?
Stephen Foerster, Ivey Business School, Western University*John Tsagarelis, Primes Corp. Grant Wang,
Highstreet Asset Management Inc.
August 6, 2014
Abstract
Novy-Marx (2013) shows that profitability, measured by gross profits-to-assets, predictsthe cross-section of average returns just as well as book-to-market ratios do. We findthat, in our 1994-2013 sample of S&P 1500 stocks, cash flow measures are even betterpredictors of stock returns than various income statement-based measures. We presenta procedure for transforming indirect cash flow method statements into disaggregatedand more direct estimates of cash flows from operations and other sources. We thenderive ‘direct method’ cash flow measures and form portfolio deciles based on thesemeasures. Stocks in the highest cash flow decile outperform those in the lowest cashflow decile by over 10% annually after controlling for well-known risk factors. Ourresults are robust to the investment horizon, and controlling for sector differences. Wealso show that, in addition to operating cash flow information, cash taxes and capitalexpenditures provide incremental predictive power.
JEL Codes: G10, G14Keywords: direct cash flow method; cash flows; stock return predictability
————————————————————————–* Corresponding author. Ivey Business School, Western University, 1255 WesternRoad, London, ON N6G 0N1, (519) 661-3726, [email protected]. We wish to thankMatthias Hanauer, Darren Henderson, Rick Robertson and seminar participants atWestern University for helpful comments and suggestions. All errors remain our ownresponsibility.
Electronic copy available at: http://ssrn.com/abstract=2472571
1 Introduction
Investors rely on financial information, such as profitability and cash-related measures, to
assess a firm’s intrinsic equity value and predict the cross-section of average returns. Fama
and French (2006) find that more profitable firms have higher expected returns. Novy-
Marx (2013) shows that profitability, measured by gross profits-to-assets, predicts the cross-
section of average returns just as well as book-to-market ratios do. Fama and French (2013)
incorporate the profitability measure as a new factor, extending their well-known three-factor
model. Ball et al. (2014) show that an operating profitability measure that better matches
current expenses and revenues is an even better predictor of returns and also show that
results depend on whether the denominator is total assets or the market value of equity.
Thus the search continues for financial information that better predicts stock returns.
While the income statement has long been at the center of financial statement analysis,
well-documented shortcomings1 and the recent studies such as Ball et al. (2014) show that
not all components of the income statement have the same predictive power. Notorious
bankruptcies including Enron and WorldCom, graphically illustrate that profitable GAAP
income statements can simultaneously co-exist with negative operating or free cash flow for
the same company (see Appendix A for an example). Novy-Marx’s (2013) intuition that
the farther down the income statement one goes, the more “polluted” profitability measures
become and less related to “true” or economic profitability certainly rings true. Yet it is not
obvious that any accounting profit measures, regardless of where on the income statement
it appears, should be superior to cash-based measures of profitability. This friction between
accrual earnings and cash flows has spawned considerable academic interest in the field of
“quality of earnings”2 and has contributed to investors renewed interest in cash flow related
1See Sloan (1996) and Markham (2006). See also more generally, American Institute of Certified PublicAccountants (1973). The committee, chaired by Robert Trueblood, urged the accounting profession tobe more responsive to investors and creditors in providing information about the companies they audited.This was in reaction to the Securities and Exchange Committee’s (SEC) growing pressure on accountantsto forsake management and serve new constituencies of investors, creditors, rating agencies, analysts andanyone else with an interest in a company’s financial statements.
2See O’Glove (1987), Kellogg and Kellogg (1991) and Siegel (1991).
1
measures. However, it is unclear how the common indirect cash flow statement3 should be
used: should investors focus on operating cash flows? Free cash flows? Pre-tax or post-
interest costs? Investors’ frustration with both the income statement and the indirect cash
flow statement has inspired periodic reviews by the International Accounting Standards
Board (IASB) and the U.S. Financial Accounting Standard Boards (FASB).4 While the
FASB (2008b) study provides a series of far reaching insights and highlighted the potential
benefits of disaggregating financial information using a direct cash flow methodology, it fails
to outline how these techniques can be used today and provides no evidence of their efficacy.
Our study attempts to fill that void.
Our main finding is that ‘direct cash flow’ measures are generally better stock return
predictors than indirect cash flow measures, which in turn tend to be better than various
income statement profitability measures that focus on either gross profits, operating profits,
or net income. We first outline a systematic process for combining existing income and
cash flow statements to generate a ‘direct cash flow’ approximation. That is, we create a
new statement that disaggregates operating, financing, tax, and non-operating cash flows
to isolate recurring value creation activities. We then create a series of cash-based financial
ratios and compare them to Ball et al.’s (2014) operating profitability measure, Novy-Marx’s
(2013) gross profitability measure, as well as the traditional return on assets (ROA)5 measure,
all of which have total assets in the denominator. Consistent with Ball et al. (2014), we
repeat our analysis using market value of equity as the denominator. We show that our
new measures generate greater risk-adjusted returns than those based on standard income
statement information. Our results are robust across investment horizons and risk factors
3See Goyal (2004), who estimates that 97% of firms use the indirect rather than direct method. Anecdo-tally, the current number is probably close to 100%.
4The accounting standard for presenting the statement of cash flows is documented in the FinancialAccounting Standards Board (FASB) (1987), known simply as FASB 95 or FAS 95. In particular, see page6 that states “The information provided in a statement of cash flows...should help investors...to assess theenterprise’s ability to generate positive future net cash flows.” See also FASB (2008b) for a discussion paperon proposed changes in the standard. For an overview of the proposed changes see Reilly (2007).
5In unreported results we repeat our analysis with return on capital measures as well as the traditionalreturn on equity measure. The results are qualitatively the same.
2
including controlling for sector differences.
While the intrinsic value of a firm is the present value of future cash flows, there are many
proxies for future cash flows.6 Variation and inconsistency in financial statement presentation
formats creates difficulties for investors who wish to understand what to extrapolate and to
analyze how sustainable the value creating activities are. While FASB specifically endorses
the accrual accounting system, we believe accrual accounting7 also creates many potential
forecasting problems for investors. A key aspect of the accrual system is its ability to smooth
out ‘temporary’ fluctuations in cash flows (e.g., see Dechow (1994) and Dechow, Kothari and
Watts (1998)). The contrast between cash basis reporting and earnings reported under the
accrual system is highlighted in FASB (2008b, p. 9): “[Accrual accounting recognizes that]
the buying, producing, selling, and other operations of an entity during a period, as well as
changes in fair value and other events that affect its economic resources and claims to them,
often do not coincide with the cash receipts and payments of that period.” The accrual
system, therefore, implicitly assumes that by recognizing economic events independent of
the timing of the underlying cash flows, users can rely on the income statement as an
unbiased decision predictor. For example, an increase in accounts receivable stemming from a
customer’s delayed payment (which simultaneously reduces cash flows and increases accruals
by the same amount) leaves both revenues and net income unchanged as if full payment
was received. While FASB sees accruals as improving the ability of earnings to measure firm
performance,8 we can also view the lack of cash inflows from the tardy customer and the
resulting differences in accrual and cash statements as creating the potential for securities
mispricings. More generally, research suggests earnings targets influence accounting decisions
6As FASB (2008b, p. 3) outlines in their discussion paper, “Transactions or events recognized in financialstatements today are not described or classified in the same way in each of the statements...That makes itdifficult...for users who want to assess the quality of an entity’s earnings by comparing operating incomewith operating cash flows.”
7The term “accrual” is used in a general sense and includes both accruals and deferrals.8See FASB (2008a) at paragraph 44: “Information about enterprise earnings and its components measured
by accrual accounting generally provides a better indication of enterprise performance that does informationabout current cash receipts and payments.”
3
thereby creating strong incentives to bias accruals either upward or downward.9 This pattern
of persistent “timing” differences between the two statements that fails to converge is not
uncommon.10
In an effort to improve the usefulness of information both the IASB and FASB undertook
a detailed study to explore how the presentation of financial information can be modified
to meet investors needs. While FASB “encourages” firms to report cash flows using the
direct cash flow method (DCFM), very few firms actually do. In the FASB (2008b) exposure
draft, the IASB and FASB examined using the DCFM to measure a company’s operating
performance rather than the commonly used indirect method that starts with net income
and makes adjustments. According to FASB (2008b, p. 45) the indirect approach includes a
major deficiency: “it derives the net cash flow from operating activities without separately
presenting any of the operating cash receipts and payments.” Additionally, it merely recon-
ciles information and is not a valid substitute for operating cash inflows and outflows. Under
the DCFM, cash flows are segregated between cash receipts from customers less cash pay-
ments to suppliers, financing activities, tax impacts and other non-recurring, non-operational
activities (see Appendix A for an example of the indirect and direct cash flow methods). The
motivation for this compartmentalization is to isolate recurring cash flows from business ac-
tivities that are additive to net present value, segregate financing activities that are not and
flag on-off gains or losses.11 As Lee (2014) notes, information that is useful in forecasting
future “residual income” (i.e., above a normal rate of return) is the type of information
that directs investor attention to the sustainability of future cash flows, which requires more
normalized measures of performance.
It is our belief that segregating cash flows into their common underlying economic factors
9See Graham, Harvey and Rajgopal (2005) and regarding abnormal accruals see Healy and Wahlen (1999).10For example, we investigated the U.S. railway industry. Between 1993 and 2012, reported trailing four-
quarter net income consistently and persistently exceeded trailing four-quarter free cash flows for each ofCSX Corp., Norfolk Southern Corp. and Union Pacific Corp.
11See Esplin et al. (2014). See also, Nissim and Penman (2001), Fairfield and Yohn (2001), Soliman (2008),and Fairfield, Kitching and Tang (2009). Also, many financial statement analysis textbooks recommendthe operating/financial disaggregation for profitability forecasting and firm valuation: see Penman (2013),Lundholm and Sloan (2006).
4
permits investors to better understand and model the magnitude, frequency, timing and
volatility of recurring cash flows. Unfortunately, the cost of preparing cash flow statements
under the DCFM and the hesitancy by firms to reveal the information provided by them has
resulted in very few companies adopting this approach, despite the FASB’s encouragement.
Accordingly, there is very little research linking the DCFM and future stock returns.12 Our
study attempts to fill this gap by creating a proxy for DCFM statements from existing
financial statements and testing their efficacy as it relates to predicting future stock returns.
Our study lends support to the IASB’s and FASB’s initiatives to require a direct method
financial reporting regime.
Our research is most closely related to and extends the work of Novy-Marx (2013) and
Ball et al. (2014). While Novy-Marx (2013) argues that gross profitability is the cleanest
accounting measure of true economic profitability, and Ball et al. (2014) derive an improved
operating profitability measure, we argue that even “cleaner” measures can be derived by
focusing on the cash flow statement and cash-related items: we focus on cash-related mea-
sures that better reflect cash that is available to equity holders.13 We empirically test the
relationship between DCFM-based cash flow estimates and future stock returns in the U.S.
equity market in order to test our conjecture that disaggregated direct cash flows permit
investors to isolate recurring from non-recurring value adding activities, leading to superior
return predictions. First, we provide a template to show how DCFM cash flows can be esti-
mated from income statements and indirect cash flow statements by segregating cash flows
from operating activities, financing activities, taxes, and other non-operating activities. Sec-
12Krishnan and Largay (2000) examine the usefulness of direct method cash flow information and theextent to which this information is more beneficial in predicting future operating cash flows compared withthe indirect method. Orpurt and Zang (2009) report evidence that suggests market participants utilize directmethod disclosures to better forecast future operating performance. Farshadfar and Monem (2013) examinedata from Australia from 1992 to 2004 and find that the cash flow components as reported in the DCFMhave a greater ability to predict future cash flows as compared with an aggregate cash flow measure.
13Our work also complements previous studies that investigate cash flows and investments (e.g., Lakon-ishok, Shleifer, and Vishny (1994), Hackel, Livnat, and Ra (1994), Hackel and Livnat (1991)); accruals andinvestments (e.g., Sloan (1996), Xie (2001), DeFond and Park (2001), Thomas and Zhang (2002), Richard-son, Sloan, Soliman, and Tuna (2005), and Livnat and Santicchia (2006)); and cash flows, accruals, andinvestments (e.g., Houge and Loughran (2000), and Livnat and Lopez-Espinosa (2008)).
5
ond, we construct new profitability measures based on various estimates of cash flows from
operations less capital expenditures, adjusted for financing and taxes. Consistent with Ball
at al. (2014), we use both total assets and market value of equity in the denominator of these
ratios. These ratios are compared with Novy-Marx’s (2013) gross profitability measure, Ball
et al.’s (2014) operating profitability measure, and traditional return on assets as well as
earnings-to-price (EP) and related market value of equity-based measures. We find a signif-
icant positive relationship between the derived cash-based ratios and future stock returns as
measured by high-low portfolio returns, information ratios, and risk-adjusted alphas. Stocks
in the highest cash flow decile outperform those in the lowest cash flow decile by over 10%
annually after controlling for well-known risk factors. In contrast, other profitability and
earnings-based ratios generally have relatively weaker relationships with future returns in
our sample of S&P 1500 stocks in the 1994 to 2013 period. We also show that, in addition
to operating cash flow information, cash taxes and capital expenditures information provide
incremental predictive power.
The remainder of the paper is organized as follows. Section 2 presents our direct method
cash flow template that segregates cash from operations, financing, taxes, and other. Section
3 describes our data and methodology. Sections 4 and 5 present our main results while section
6 presents robustness checks. Section 7 investigates cash flow components and section 8
concludes.
2 The Direct Cash Flow Template
As an alternative to the traditional income and cash flow statements we propose the following
direct cash flow template and an alternative set of cash-based capital efficiency and valuation
ratios. Unlike the indirect approach that begins with net income and reconciles to operating
cash flows by reversing non-cash activities, the direct cash flow template adheres to the
culinary principle of mise en place. That is, we organize and sort financial information
6
into clusters of homogeneous business activities thereby permitting the reader to understand
how value is being created and what can be extrapolated. In this way, operating activities
associated with a competitive advantage are value enhancing and are likely to repeat, while
financing activities are not. Tax structures and other non-operating activities may be value
enhancing attributes but are likely unsustainable. This overview is consistent with earlier
studies that demonstrated that cash-based component of earnings are more persistent and
therefore are of “higher quality” than accrual based components (Sloan (1996)).
While we recognize that this template does not generate the ‘actual’ direct cash flows
as envisioned by FASB (2008b) (since corporations are not required to publish segregated
operating from non-operating cash flows) we nevertheless feel it provides an improved method
over non-transformed accounting data for estimating intrinsic values. While all the individual
components of the cash flows are publicly available, we conjecture that their availability
alone does not ensure it is integrated in a timely manner and used by investors to estimate
intrinsic values. Patterns must be isolated, revealed and applied before stock prices reflect
this information. Our analysis also highlights what types of information are most informative.
The direct cash flow template we use is presented in Table 1 and is summarized as follows.
We first directly estimate net cash flows from operating activities. We begin with the main
cash inflow of sales, adjusted for changes in accounts receivable, deferred revenues, and other
cash inflows from operations. We subtract cost of goods sold as well as selling, general, and
administrative expenses, then adjust for changes in accounts payable from operations and
changes in inventories. The result is our estimate of Net Cash Flows from Operations (A).
Next we estimate Net Cash Flows from Operations After Financing Activities (B) by
subtracting interest expenses and adjusting for other financing income or expenses. We
then estimate Net Cash Flows from Operations After Financing and Tax Activities (C) by
subtracting cash flows related to tax activities—including taxes on the income statement,
adjusted for changes in account payable taxes and deferred taxes. Finally, we estimate
Net Cash Flows from Operations After Financing, Tax, and Extraordinary Activities (D)
7
by accounting for other non-operating activities, including discontinued operations, foreign
exchange, and pension-related items. For completeness, in order to reconcile this cash flow
estimate with the traditional Free Cash Flows to Equity Holders, we account for cash flows
from investing activities. Here we simply subtract the cash outflows associated with capital
expenditures.
3 Data and Methodology
The fundamental and pricing data come from Standard & Poor’s Xpressfeed North American
database (Xpressfeed). The investment universe is the S&P 1500 which consists of the largest
500 stocks by market capitalization, the mid-cap 400, and small-cap 600. The choice of S&P
1500 ensures that these stocks are truly investable and the tested investment strategies
are more relevant to practitioners. The historical S&P 1500 constituents also comes from
Xpressfeed. Each month we obtain a list of stocks included in S&P 1500 at that time, thus
preventing any survivorship bias. As is common in studies that rely on accounting metrics,
and given some major structural differences between financial and non-financial firms, for
most of our reported analysis we exclude financial firms from our sample (including banks,
insurance firms, and REITs), on average, about 15% of the sample. We use data from
October 1994 to December 2013 as cash flow statements only became mandatory in the
U.S. since 1987, and actual index constituents for the S&P 1500 only become available since
October 1994.
The major fundamental variables used in this study are: operating activities-net cash flow
(Xpressfeed mnemonic OANCF), net sales (SALE), accounts receivable decrease/increase
(RECCH), cost of goods sold (COGS), selling, general and administrative expense (XSGA),
depreciation and amortization income statement (DP), depreciation and amortization cash
flow statement (DPC), funds from operations – other (FOPO), accounts payable and accrued
liabilities - increase/decrease (APALCH), inventory – decrease/increase (INVCH), assets
8
and liabilities – other – net change (AOLOCH), sale of property, plant and equipment and
investments – gain/loss (SPPIV), interest and related expense – total (XINT), income taxes
— total (TXT), income taxes – accrued – increase/decrease (TXACH), deferred taxes – cash
flow (TXDC), special items (SPI), discontinued operations (DO), extraordinary items (XI),
capital expenditure (CAPX) and income before extraordinary items – available for common
(IBCOM). Following industry practice, we create trailing twelve month (TTM) values for the
fundamental variables. These TTM values are updated with each quarter’s new information
and are combined with monthly returns data to conduct the tests. Despite having quarterly
updates and to avoid any look-ahead biases, we only use accounting data that has been
lagged by four months.14
Table 2 presents summary statistics for the fundamental variables for companies that are
in the S&P 1500 at any time during our sample period. There are 164,385 company-quarter
total observations from 1992 to 2013. Mean (median) net sales are $4,352 million ($914m);
net cash flows from operating activities are $513m ($81m); capital expenditures are $287m
($40m); and income before extraordinary items available for common shareholders are $237m
($36m).
Based on our template definitions in Table 1, we create various direct and indirect free
cash flow measures. FCF AFAT represents net cash flows from operations after financing
and tax activities (C) - capex; FCF AF represents net cash flows from operations after
financing activities (B) - capex; and FCF O represents net cash flows from operations (A) -
capex. These direct cash flow measures (DCFM) are compared with the indirect cash flow
measure FCF IM defined as operating activities net cash flow (OANCF) - capex. Note that
this indirect measure, OANCF, results in the same number as net cash flows from operations
after financing, tax, and extraordinary activities, (D) in Table 1.
These free cash flow definitions are then used to construct various cash returns on assets
and free cash flow yield metrics. The variable CFAFAT/TA is the cash return on assets
14The four month lag is conservative. In our un-tabulated analysis of S&P 1500 firms between 1993 and2013, 99% of these firms filed 10-Q statements within three months after the quarter end.
9
measured as the free cash flow direct method measure FCF AFAT divided by total assets.
The variables CFAT/TA, CFO/TA, and CFIM/TA are alternative return on assets measures
that have as the numerator FCF AT, FCF O, and FCF IM, respectively. We compare these
measures to a number of accounting-based measures. The operating profit measure, OP/TA,
has as the numerator operating profits (as estimated by Ball et al., 2014, subtracting from
sales the cost of goods sold and selling, general & administrative costs excluding research &
development) and total assets as the denominator. Novy-Marx’s (2013) gross profit to total
assets, GP/TA, has as the numerator gross profit as measured as sales less cost of goods sold.
The traditional return on assets ratio is measured as IBCOM (income before extraordinary
items available for common shareholders) divided by total assets. The price yield measures
are similar to the measures just described expect total assets are replaced by the market
value of equity (MVE) in the denominator. Note that the traditional earnings yield, or the
inverse of the price-to-earnings ratio, EP, is measured as IBCOM or net income divided by
the market value of equity, and is noted simply as NI/MVE.
Each month, we rank the S&P 1500 firms by a particular return on asset or yield measure
and divide the universe into deciles from low (P1) to high (P10). One-month-ahead portfolio
returns are calculated on a value-weighted basis and equal-weighted basis (we also consider
other horizons). In standard fashion, a long-short portfolio return is estimated as the spread
return difference between the highest (P10) and lowest (P1) portfolio returns. We then test
for the significance of the return differences.
4 Cash Flow and Profitability Return and Yield Mea-
sures and One-Month-Ahead Returns
Table 3 presents results comparing the one-month-ahead returns for the various portfolios
(P1 through P10), as well as the P10-P1 high-low portfolio return spread, sorted on the
basis of the previous month’s return on assets or yield measure, for the entire October
1994 to December 2013 period. We also show the standard deviation of the P10-P1 return
10
measure, the t-statistic of the significance of the P10-P1 return, the minimum and maximum
monthly P10-P1 observations, and the information ratio (IR) measured as the annualized
P10-P1 return divided by the annualized standard deviation of the return difference. Panel A
presents results measured by value-weighted returns.15 Not surprisingly, across all measures
there is a general (but not perfect) monotonic relationship, with higher measures tending to
exhibit higher one-month-ahead returns than lower measures.
Panel A presents results based on value-weighted returns. The upper part of the panel
presents measures with total assets in the denominator. Returns generally increase somewhat
monotonically across the low to high portfolios for the various measures. We see that the
Novey-Marx (2013) GP/TA variable has the largest P10 return among the return on asset
variables but not the biggest P10-P1 spread since many of the cash flow measures have
more cross-sectional variation. For the portfolios sorted on the basis of our three DCFM
efficiency measures (CFAFAT/TA, CFAF/TA and CFO/TA) the high-low (P10-P1) return
differences are significantly positive (t-statistics ranging from 2.14 to 2.73), with monthly
return differences ranging from 0.64% to 0.82% (or equivalently 8.0% to 10.3% annually).
Information ratios range from 0.489 to 0.623. For the indirect method cash flow return
measure (CFIM/TA), the high-low monthly return difference drops to 0.49% (6.0% annually)
but not significant, and the information ratio drops to 0.356. The return difference for the
portfolios sorted on the basis of operating profits-to-total assets (OP/TA) is very similar to
that based on the CFIM/TA variable at 0.49% (6.0% annually) and not significant and with
a similar information ratio of 0.356. The high-low portfolio returns based on the gross profit-
to-total assets (GP/TA) measure is also similar at 0.49% (6.1% annually), but given the lower
standard deviation of returns is significant (t-statistic of 2.10); additionally, the information
ratio is higher at 0.480. Finally, for the return on assets (NI/TA) measure, the monthly high-
low return difference is only 0.06% (0.7% annually) and is not significantly different from
zero and the information ratio drops to 0.040. Among these cash flow return measures, the
15We also repeat our analysis including financial firms and results are substantively similar.
11
best performing high-low portfolio is based on the sorting using the FCF DM AF measured
(net cash flows from operations after financing activities (B in Table 1) less capex). This
same measure also has the highest information ratio.
The lower part of the panel presents measures with market value of equity in the de-
nominator. We also see that all of the return differences for the portfolios sorted on the
basis of our three direct method cash flow yield measures (CFAFAT/MVE, CFAT/MVE
and CFO/MVE) as well as the indirect method cash flow return measure (CFIM/MVE)
are significantly positive (t-statistics ranging from 1.91 to 3.03), with monthly return differ-
ences ranging from 0.55% to 0.77% (or 6.9% to 9.6% annually). Information ratios range
from 0.436 to 0.692. The return difference for the portfolios sorted on operating profit yield
(OP/MVE) is 0.70% (8.7% annually) and marginally significant, while the information ratio
is lower at 0.380. For the gross profit yield (GP/MVE), the high-low return difference is
0.58% (7.2% annually) but not significant and the information ratio drops to 0.313. Finally,
for the earnings yield measure (NI/MVE), the monthly high-low return difference is 0.43%
(5.2% annually) and is not significant, and the information ratio is only 0.271. Among these
cash flow return measures, the best performing long-short portfolio is again based on the
sorting using CFAF/MVE (net cash flows from operations after financing (B in Table 1)
less capex) and generates the highest information ratio. The AFAT and AF DCFM measure
sortings result in higher information ratios than for the indirect cash flow measure sorting
(CFIM/MVE).
Table 3, Panel B presents results based on equal-weighted returns. Recall that since
our sample includes the largest and generally most liquid U.S. stocks (i.e., the top 1500,
less financial firms), it makes sense to consider equal-weighted returns in addition to the
previously reported value-weighted return results. In general the results are similar to those
in Panel A with the cash flow measures stronger than profitability-related measures. For the
return on asset measures in the upper part of the panel, based on high-low return differences,
we note a monotonic relationship, with all three of the DCFM measures dominating the
12
indirect method measure and all of the cash flow measures dominating the OP and GP
measures and the net income measure. The largest high-low monthly return difference of
0.69% (8.6% annually) is for the CFAFAT/TA variable. All of the cash flow measures are
significant (t-statistics from 2.47 to 3.65] while all of the profitability measures are not
statistically significant. We note the same monotonic relationship based on the information
ratio.
For the cash flow yield measures in the lower part of Panel B, the story is again very
similar. The CFAFAT/MVE and CFAF/MVE dominate the indirect method measure and
all dominate the OP and GP measures and the traditional earnings yield measure, although
the operating profit high-low return difference is significant while return differences based
on the gross profit measure is marginally significant. Again we note the same monotonic
relationship based on the information ratio.
Thus for our sample, before we adjust for risk factors (i.e., in our regression analysis
below), we don’t see the differences within the panels highlighted by Ball et al. (2014):
the total asset versus market value of equity denominator or deflator doesn’t appear to
substantively impact the results.
In un-tabulated results, we also perform difference of means t-tests of the P10-P1 high-low
return differences for various DCFM measures compared with indirect method measures and
profitability measures. For the value-weighted return results, the CFAFAT/TA sorted return
differences are significantly greater than the NI/TA returns; and both the CFAT/TA and
CFO/TA returns are significantly greater than both the CFIM/TA and NI/TA returns. For
the equal-weighted return results, each of the CFAFAT/TA, CFAT/TA and CFO/TA sorted
return differences are significantly greater than the both the GP/TA and NI/TA returns; and
each of the CFAFAT/MVE, CFAT/MVE and CFO/MVE returns are significantly greater
than the NI/MVE returns.
We also divide the test periods into four sub-periods, chosen as follows: sub-period 1,
October 1994 to February 2000 (65 months), coincides with the technology bubble period;
13
sub-period 2, March 2000 to July 2007 (89 months), coincides with the post-technology
bubble and pre-financial crisis period; sub-period 3, August 2007 to May 2009 (22 months),
coincides with the financial crisis period; and finally sub-period 4, June 2009 to December
2013 (55 months), coincides with the post-financial crisis period. We report these sub-period
results in Table 4 Panel A, B, C, and D, respectively.
In Panel A, the technology bubble, results are generally consistent with the overall results,
particularly for the cash flow return measures where the most significant measure is based
on the CFAFAT/TA measure followed by the CFAF/TA measure and then the third direct
method CFO/TA measure. The indirect method measure, CFIM/TA, is also significant but
has a lower information ratio, as do the OP/TA and GP/TA measures. The NI/TA measure
is not significant. None of the yield measure return differences are significantly positive but
the cash flow yields are positive while the OP/MVE is negative but not significant and the
GP/MVE and NI/MVE measures are significantly negative. In retrospect it is not surprising
that the yield measures, with stock price as the denominator, did not appear to be related
to future stock returns given the high valuations during this period.
In Panel B the results for the period between the technology bubble and the financial
crisis, while the CFAFAT/TA measure information ratio is not significant, the other two
direct method cash flow return measures, CFAF/TA and CFO/TA are significant. None
of the other return on asset measure return differences are significantly different from zero.
The yield measure return differences are all significant and the CFAF/TA measure slightly
dominates all of the other measures in terms of the information ratio.
In Panel C we see the results for the financial crisis period. For the cash flow return mea-
sures, CFAFAT/TA and CFAF/TA return differences are marginally significant although the
OP/TA and GP/TA measures dominate. The NI/TA measure is negative but not signifi-
cant. For the yield measures the CFAFAT/MVE measure has the largest information ratio
measure and is the only one that is significantly positive. What may be partially driving
some of the weaker results is the major failure of quantitative models in August 2007, as
14
described and explained in Khandani and Lo (2007). In addition, in retrospect given the
unprecedented events including the failure of Lehman Brothers and near-failures of other
major financial institutions, market participants may have overreacted and prices may have
diverged from true intrinsic values.
Finally, in Panel D, in the post-financial crisis period, we seemed to have returned to more
“normal” times. The CFAFAT/TA cash flow return measure has the highest information
ratio and the return difference is significant. Most of the other measure return differences are
not significant, perhaps due to the relative shortness of the sub-period, while the traditional
NI/TA measure is significantly negative. All of the direct method cash flow yield measures
as well as the indirect method measure return differences are statistically significant. The
OP/TA the GP/TA measure return differences are not significant, while the traditional
NI/TA measure return difference is negative and significant.
Overall we note that the yield measures, based on market values in the denominator, tend
to do better in “normal” markets such as sub-periods 2 and 4, while the return measures,
with book values in the denominator tend to do better in “abnormal” markets, particularly in
the frothy sub-period 1 technology bubble period. We also conclude that relying on cash flow
related metrics is a worthwhile endeavor and direct method measures, particularly AFAT,
are generally superior to the indirect method and the traditional measures generally have
very little predictive power.
It is worthwhile to highlight the differences, so far, between our results and Novy-Marx
(2013) who concludes that profitability measures are superior to cash flow measures as pre-
dictors of stock returns (we compare our regression results to his in the next section). First,
given our data limitations, we focus on a shorter and more recent period, 1994 to 2013, com-
pared to his 1963 to 2010 study. It is possible that the profitability measure has weakened
more recently. Second, our sample of stocks are different. He uses the universe of Compus-
tat stocks (excluding financials) while we focus on the largest 1500 stocks, which is a more
liquid and tradable sample. Third, our focus is on portfolio performance while he initially
15
emphasizes Fama and MacBeth (1973) regression results (his Table 1) with winsorized in-
dependent variables. As he points out, the methodology weights each observation equally
and thus puts tremendous weight on small or “nano/micro” stocks that make up two-thirds
of the sample but less than 6% by market capitalization weight (see also Fama and French
(2006)).16 Fourth, he utilizes an estimate of free cash flows that differs from our approach.
He defines free cash flow as net income plus depreciation less working capital change less
capital expenditures. Interestingly, in Appendix Table A2, his EBITDA-to-assets variable-
most similar to some of our direct cash flow method measures except it does not account for
capital expendituresperforms well and better than his free cash flow measure.
5 Portfolio Sorted Regressions
We adjust for risk in a more systematic manner by using the P1 through P10 portfolio returns
as well as the P10-P1 return spread (i.e., the portfolios sorted on either the return on asset
or yield variables) to run regressions against well-known risk factors. We use monthly data
for the entire October 1994 to December 2013 sample period. Our regression model is:
Pi,t+1 = αi+β1iMKTRFt+β2iSMBt+β3iHMLt+β4iUMDt+β5iLIQt+εi,t+1, t = 1, ..., N (1)
The dependent variables are the individual decile portfolios, P1 (low) through P10 (high),
as well as the P10-P1 portfolio return difference for the various return on asset and yield
sorted portfolio measures, i, in Table 3, for each monthly time series. Most of the independent
variables are based on Carhart’s (1997) extension of the Fama-French (1993) three-factor
model where MKTRF is the market risk premium (market return in excess of the risk-
free rate), SMB is the small minus big (size) factor, HML is the high minus low (book-
to-market) factor, and UMD is the up minus down (price momentum) factor. We also
16See also Fama and French (2008) for a discussion of the advantages and disadvantages of the Fama-MacBeth approach compared to sorting returns.
16
include LIQ, Pastor and Stambaugh’s (2003) traded liquidity factor, derived from dividing
U.S. common stocks into ten portfolios based on each stock’s sensitivity to the ‘non-traded’
liquidity innovation factor—the return on the portfolio of high liquidity betas less the return
on the portfolio of low liquidity betas.17 Once we control for all of these risk factors, the
intercept, α, is interpreted as the alpha or abnormal return. If alpha is positive and significant
it implies the other factors can’t be combined to form a mean-variance efficient portfolio and
it suggests how an investor can tilt their portfolio toward a more profitability strategy,
such as investing in portfolios that have recently generated high cash flows relative to their
total assets or market value of equity. We use White’s (1980) heteroscedasticity-consistent
standard errors. For comparison we present the average monthly return measured in excess
of the monthly Treasury bill (from Ken French’s website).
Table 5 summarizes the regression results. Panel A presents the regression results, with
value-weighted returns, for the P1 (low) portfolio sorted on the various return and yield
measures. Panel B presents regression results, with value-weighted returns, alphas-only, for
portfolios 2 through 9. Panel C presents the regression results, with value-weighted returns,
for the P10 (high) portfolio sorted on the various return and yield measures. Panel D presents
the regression results, with value-weighted returns, for the P10-P1 portfolio return difference
sorted on the various return and yield measures. Finally, Panel D presents the regression
results, with equal-weighted returns, for the P10-P1 portfolio return difference sorted on the
various return and yield measures.
Across the panels we see that the portfolio alphas (as well as the average returns) generally
tend to increase as we move from P1 through P10 for the various return on asset and
yield measures. In Panel A for the P1 (value-weighted return) results, most of the alphas
are negative but not significant; in Panel B (value-weighted return) all of the alphas are
significantly positive for portfolios P7 through P9; while in Panel C for the P10 (value-
17Data for the first four factors are available from Ken French’s website http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data library.html while the liquidity factor is available fromLubos Pastor’s website http://faculty.chicagobooth.edu/lubos.pastor/research/liq data 1962 2011.text.
17
weighted return) results the alphas are also all positive and significant. Consistent with our
results in Table 3, in both Panel D (value-weighted returns) and Panel E (equal-weighted
returns), all of the three DCFM cash flow based return and yield measure P10-P1 return
differences are significantly positive; the indirect measure return differences are positive and
in most cases significant; the OP return differences are significantly positive only for the yield
measure; and in most cases the cash flow based measures have higher alphas than for the
corresponding OP measure. The GP measure is significant with TA as the denominator and
with value-weighted returns but not equal-weighted returns, and is not significant with MVE
as the denominator and with value-weighted returns, but significant with equal-weighted
returns; in all cases with a lower alpha than for the corresponding OP. For the net income
measure, only with TA as the denominator and value-weighted returns are results significant.
For value-weighted return results, at least one of the DCFM variables result in larger alphas
than the indirect method variable. Among the DCFM variables, it is difficult to conclude
unequivocally which one is best. Based on value-weighted returns (Panel D) and TA in the
denominator, the alpha for CFO is 1.10%, which translates into an annual return difference
of 14.0%, even stronger than the Table 3 results. With MVE as the denominator, CFAF
has an alpha of 0.84%, which translates to an annual difference of 10.6%. Based on equal-
weighted returns (Panel E) and TA as the denominator, the CFAFAT alpha is 0.85%, which
translates to an annual difference of 10.7%. With MVE as the denominator, CFAF has an
alpha of 0.86%, which translates to an annual difference of 10.8%. (In section 7 we explore
what incremental information may lead to different results among the DCFM variables.)
Thus even controlling for well-known risk factors we find significant value from utilizing
information available from our “direct cash flow” statements.
We again contrast our P10-P1 results to those of Novy-Marx (2013) as reported in his
Table 2. Despite differences in methodologies described in the previous section as well as his
use of the traditional Fama-French three-factor model that excluded the liquidity factor in
our study, it is comforting to find that our alpha estimates for the gross profit-to-total assets
18
measure, 0.69% based on value-weighted returns and 0.33% based on equal-weighted returns,
bracket his estimate of 0.52%. (Curiously, for the gross profit-to-market value of equity
measure, we uncover an alpha of 0.48% based on value-weighted returns and 0.71% based on
equal-weighted returns, while Ball et al. (2014) find a negative but not significant alpha for
their longer sample period.) Thus we are not concluding that his measure performs poorly
given our different sample and design, but rather we conclude that cash-based measures
perform even better.
As noted by Fama and French (1992) and Ball et al. (2014), the relationship between
the various return on asset and yield measures and future stock returns can be explained
by either rational or irrational asset pricing stories. Mispricing may be explained by the
behaviors and biases of irrational investors or short-term trading frictions, or alternatively
there may be risk factors not captured by the five factors included in our regression model.
Persistence of returns, as examined in the next section, may be consistent with a rational
explanation if we assume that trading frictions shouldn’t persist over longer periods.
6 Robustness Checks
6.1 Extended Horizon Returns
In Table 6 we re-examine our earlier one-month-ahead value-weighted return results by ex-
amining a variety of extended horizon holding periods, including three months (Panel A), six
months (Panel B), and twelve months (Panel C). The results are generally consistent with the
one-month return results. In Panel A, the return differences are significantly positive in all of
the cash flow return measure cases as well as the operating profit, OP/TA, and gross profit,
GP/TA measures while the traditional return on asset measure, NI/TA, is not significant.
Return differences are highest for the DCFM measures followed by the GP/TA measure
and then the indirect method cash flow measures. The three-month return difference for
the CFAF/TA measure is 2.1% (annualized 8.7%). For the yield measures all of the return
19
differences are significantly positive with a similar order of magnitude for the CFAF/MVE,
CFIM/MVE, OP/MVE and GP/TA measures, while the earnings yield, NI/MVE, is only
marginally significant. The three-month return difference for the CFAF/MVE measure is
2.1% (annualized 8.6%). Among the return on asset measures, the information ratio is high-
est for the GP/TA measure given the much lower standard deviation of returns. Among the
yield measures the information ratio is much higher for the cash flow measures compared
with the profitability measures.
In Panel B, for the six-month horizon, the return differences for the return on asset
measures are again significantly positive in all cases except for the NI/TA measure. The
highest return difference is again for the CFAF/TA variable, with a six-month return of
3.4% (annualized 6.9%). Among the yield measures the CFIM/MVE and GP/MVE six-
month returns are almost identical, at around 3.8% (7.7% annualized). For the return on
asset measures, the information ratio is highest for the GP/TA measure followed by the
various cash flow measures, while for the yield measures the cash flow measures dominate
based on the information ratio.
Results in Panel C are similar to those in Panel B. For both the return on asset measures
and for the yield measures the DCFM measures resulted in superior information ratios
compared to the indirect measure. Most of the DCFM and indirect method return on asset
return differences are generally superior to the profitability measures, which in turn are
superior to the NI/TA and NI/MVE measures. Even for the 12-month horizon, the high-
low CFAF/MVE portfolio offered returns of almost 7.5%. Overall our Table 6 results are
consistent with our earlier results and also suggest, while still positive and significant, only
a small tapering of return differences for longer horizons.
6.2 Sector Analysis
We check to see whether our results are sector-specific. We perform a number of tests. First
we segregate our sample among the ten Global Industry Classification Standard (GICS) sec-
20
tors: energy, materials, industrials, consumer discretionary, consumer staples, health care,
financials, information technology, telecommunication services, and utilities (recall that fi-
nancials are not included in previous results but are included here since results are segmented
by industry). We estimate information ratios for each of our measures within each sector.
Based on un-tabulated results, while there is no consistent pattern across all sectors, we
can summarize our results as follows. For the measures with total assets as the denom-
inator, cash flow measures tend to do better than the various profitability and earnings
measures particularly in the energy and industrial sectors, but also in materials, consumer
discretionary, health care, and telecom, as well as financials. Information ratios are strong
across all measures for telecom firms, with the operating profitability measure doing the
best. Information ratios are low across all measures in consumer staples (except for NI/TA)
and utilities. For the measures with market value of equity as the denominator, cash flow
measures tend to do better than the various profitability and earnings measures particularly
in the industrial, consumer discretion, and technology sectors, but also in consumer staples
and telecom. Profitability measures tend to do better in the energy, materials, and health
care. Information ratios are low across all measures in utilities, and only NI/MVE does well
in the financial sector.
In Table 7, we present another set of robustness checks related to the nature of the sectors
in which stocks in the long-short portfolio are invested. It is possible that the composition
of the lowest return portfolio based on sorting, P1, might be very different from the com-
position of the highest return portfolio based on sorting, P10, in terms of the sectors in
which the stocks are invested. To control for this, we create sector neutral portfolios. We
repeat our sorting procedure within each of 20 of the 24 GICS industry groups (recall that
financials—banks, insurance companies, diversified financials, and REITs—are eliminated
from our sample). P1 (P10) is populated by the lowest (highest) performing stocks by decile
within each sector.
Table 7, Panel A presents the one-month-ahead value-weighted return results. The results
21
are generally consistent with those in Table 3, with some notable exceptions. For the return
on asset results, we see the pattern that cash flow measures tend to have higher return
differences and information ratios than profitability measures, except the NI/TA measures
does very well. Interestingly, neither of the OP/TA and GP/TA return differences are
significant, suggesting results in other studies that highlight these variables may be benefiting
from a long position in one particular industry and a short position in stocks in a different
industry. For the yield measures, the only marginally-significant results are for the GP/MVE
measure, while all of the other return differences are significant.
Table 7, Panel B presents a final sector robustness check. We present sector neutral
portfolio results for the regression analysis. The results are similar to those in Table 5,
with similar exceptions to those uncovered in Panel A. Neither of the OP/TA and GP/TA
alphas are significantly different from zero. For the yield measures the GP/MVE alpha is
also not significant. Thus it appears that our results are not being driven by sector specific
performance, with the exception of the gross profit results.
6.3 Growth, Size, Momentum, and Profitability
In this section of analysis we investigate the extent to which our results may be related to
four well-known empirical regularities whereby 1) value or high book-to-price stocks tend to
outperform growth or low book-to-price stocks; 2) small market capitalization stocks tend
to outperform large cap stocks; 3) high price momentum stocks tend to outperform (in
the short-run) low price momentum stocks; and most recently, as uncovered by Novy-Marx
(2013), 4) high gross profit-to-total asset stocks tend to outperform low gross profit-to-total
assets stocks. We perform four separate sets of analysis examining one-month-ahead returns.
We start by forming five portfolio quintiles sorted by their beginning-of-month book-to-price
(B/P) ratio. Then within each of those portfolios we form five sub-portfolios sorted on the
basis of the beginning-of-month cash flow metric, either a cash flow return metric or cash flow
yield metric. Consequently we have 25 portfolios. If our results are simply a manifestation
22
of the value-growth phenomenon then we would not expect to find any significant differences
in returns comparing the high (quintile 5) and low (quintile 1) cash flow metrics within a
particular B/P category. We perform similar analysis for size, momentum, and gross profit.
Table 8 compares various two-way sorts for one of our arbitrarily-chosen cash flow return
on asset measures, CFAFAT/TA, in Panel A, and one of our cash flow yield measures,
CFAFAT/MVE, in Panel B.
For the cash flow return on asset metric, Panel A, the high-low returns are significant
across most of the book-to-price quintiles except the third and fourth ones; across most of
the size quintiles expect the smallest one; across most of the momentum quintiles except the
fourth one; and across the second lowest GP/TA quintile. For the cash flow yield metric,
Panel B, the high-low returns are significant across most of the book-to-price quintiles except
the third and fourth ones; across all of the size quintiles; across most of the momentum
quintiles except the highest one; and across most of the GP/TA quintiles except the lowest
one. Consistent across both panels, results appear to be strongest for low momentum stocks.
Thus our analysis suggests that our results are not simply a confounding effect of other well-
known effects.
7 Cash Flow Components
In previous analyses we showed the superiority of various cash flow measures in predicting
the cross-section of future returns, but did not conclude whether one of the cash flow mea-
sures was better than others. In this final section of analysis we investigate the impact of
segregating various cash flow components in order to attempt to uncover why one cash flow
measure might be better than another at predicting stock returns.
As a preliminary investigation, we examine the correlations of the time-series of the
averages across firms of various cash flow measures and components: our direct method
cash flow from operations measure, CFODM; our indirect method cash flow from operations
23
measure, CFOIM; Novey-Marx’s (2013) free cash flow measure of net income plus deprecia-
tion/amortization, less working capital change, less capital expenditures, CFONM; financing
activities, FinAct (see Table 1 for definitions of this and other “activities” measures); tax
activities, TaxAct; other activities, OthAct; and capital expenditures, Capex. Each of the
variables is deflated by either total assets, TA or market value of equity, MVE. Results are
presented in Table 9, with Panel A showing the Pearson correlations and Panel B showing
the Spearman correlations. The correlation patterns are similar in both panels and so our
discussion below focuses on the Panel A results.
Not surprisingly, there is a high correlation among the CFODM/TA, CFOIM/TA and
CFONM/TA measures: 0.81 comparing the first two, dropping to 0.45 comparing the first
and third. Consistent with Ball et al. (2014) the correlations across variables with the same
numerator but MVE instead of TA in the denominator are quite low in many instances: for
CFODM 0.28, for CFOIM 0.32, and for CFONM 0.47. These results show the impact of the
book value of total assets that changes slowly over time compared to the market value of
equity that changes more frequently and generally by larger magnitudes.
Across the three cash flow measures that are deflated by TA, we see a consistent pattern
of correlations compared to FinAct (negative), TaxAct (positive), and Capex (positive).
The absolute value of the correlations with FinAct are generally low, below 0.23, while those
with Capex are generally slightly higher and those with TaxAct are higher still, above 0.48.
The absolute value of the correlations with OthAct is low compared to the CFODM and
CFOIM variables, but surprisingly high at 0.55 with the CFONM variable. Not surprisingly
the correlations tend to be lower with MVE in the denominator. These preliminary results
suggest a more rigorous investigation of the components of cash flows is warranted.
As such, in the spirit of Ball et al. (2014, Table 6) we perform Fama and MacBeth (1973)
cross-section regressions. For each month of our sample we regress the one-month ahead
returns for each stock on our net cash flow from operations measure, Net CF Ops (“A” in our
Table 1 template); Financing Activities as measured by interest expenses +/- other financing
24
income/expenses; Tax Activities as measured by taxes on the income statement +/- changes
in accounts payable taxes +/- changes in deferred taxes; other non-operating activities, Non-
Op Activities, as measured as discontinued operations/special charges +/- foreign exchange
gains/losses +/- pension gains/losses/contributions; and capital expenditures, Capex. We
deflate all of these variables by the book value of total assets. Similar to Ball et al. (2014) we
include a number of control variables: log(BVE/MVE) is the natural logarithm of the ratio
of the book value of equity to market value of equity; log(ME) is a size variable measured
as the natural logarithm of the market value of equity; r1,1 is the stock’s one-month-prior
return; and r12,2 is the stock’s prior year’s return skipping a month. We winsorize all
independent variables based on the first and 99th percentiles. We also perform separate
regressions that replace the Net CF Ops variable with either our measure of cash flows from
operations based on the indirect cash flow method, CFOIM, or Novey-Marx’s (2013) free cash
flow measure of net income plus depreciation/amortization, less working capital change, less
capital expenditures.
In Table 10 we present our results that include averages of the Fama and MacBeth (1973)
slope coefficients along with their t-values. Column (1) presents the regression of Net CF
Ops alone with the control variables. We see that the Net CF Ops coefficient estimate is
significant, with a t-statistic of 3.22. In the next four regressions we add one additional vari-
able that represents a segmentation of cash flows: column (2) presents Financing Activities
and control variables; column (3) presents Tax Activities and control variables; column (4)
presents Non-Op Activities and control variables; and column (5) presents Capex and control
variables. When we add these variables individually, the Net CF Ops coefficient estimate
remains significant but only the Capex variable is marginally significant, with a t-statistic
of -1.89. The next two regressions add combinations of variables: column (6) presents Fi-
nancing Activities, Tax Activities and control variables; and column (7) presents Financing
Activities, Tax Activities, Non-Op Activities and control variables. In these regressions the
Net CF Ops variable remains significant but none of the added variables are significant. Col-
25
umn (8) presents all variables. In addition to the Net CF Ops variable, the Tax Activities
coefficient estimate is negative and marginally significant with a t-statistic of -1.91, while
the Capex coefficient estimate is also negative and significant, with a t-statistic of -2.03.
Our finding no significance in the Financing Activities coefficient estimate is consistent with
Ball et al. (2014) who find no significance with their Interest coefficient estimate for their
“all-but-microcaps” sample. However, unlike Ball et al. (2014) we find our Tax Activities
coefficient estimate is significant while theirs is not. We would expect that as cash taxes
increase, there should be a negative correlation with the intrinsic value of equity holders.
Column (9) presents the regression of CFOIM and all variables. The CFOIM coefficient
estimate is significant as expected, with a t-statistic of 3.71. Consistent with our column (8)
results, the Capex coefficient estimate is significantly negative. The Tax Activities coefficient
estimate is still negative but not significant, while the Non-Op Activities coefficient estimate
is negative and marginally significant. Finally, column (10) presents free cash flows, FCF,
measured as in Novy-Marx (2013) as net income plus depreciation less change in working
capital less capital expenditures. Only the Non-Op Activities variables is significant. Overall
our results suggest that the best information for predicting the cross-section of returns for our
sample comes from the disaggregation of cash flows in a manner that captures the direct cash
flow method. In particular, there is incremental information in disaggregating tax activities
and capital expenditures.
8 Conclusions
Inspired by Novy-Marx (2013), much recent attention has been focused on the importance
of income statement related measures such as gross profit-to-total assets as predictors of the
cross-section of returns. We show that commonly used income statement related metrics
including return on assets and earnings yields are poor return predictors, and cash-based
measures are superior to operating profitability and gross profitability-to-total asset mea-
26
sures. We examine strategies that suggest superior risk-adjusted returns based on utilizing
transformed financial statements using our DCFM to estimate free cash flows. We create a
direct cash flow template by segregating operating cash flows from activities related to fi-
nancing, tax, non-operations, and investing. We argue that this segregation assists investors
to understand how value is being created and hence better estimate what is the true intrinsic
value of the firms equity. We find that DCFM return on asset and DCFM yield measures are
generally superior to the indirect cash flow method at predicting stock returns, which in turn
are generally superior to income statement measures based on profitability. When we con-
trol for well-known risk factors related to the market, market capitalization, book-to-market
ratios, price momentum, and liquidity using regression models, we find similar results. We
also test different return holding periods and we perform robustness checks that control for
sector performance, and find similar results. Through a double-sorting procedure we show
that our results are not simply manifestations of other well-known phenomena related to
value/growth, size, momentum, and profitability. Our Fama-MacBeth regressions suggest
the incremental importance, in particular, of information related to taxes and capital ex-
penditures, in addition to information on operating cash flows. Future research may provide
further explanations as to why those variables are important to forecasting stock returns.
Our conclusion—that there is incrementally better information by segregating cash flows
to account for both financing activities and taxes—lends support to the International Ac-
counting Standards Board (IASB) and the U.S. Financial Accounting Standard Boards
(FASB) initiatives to require reconstructed financial statements. From an investment per-
spective, investors may be able to derive better information about the future prospects—and
hence future stock returns—by relying on cash flows that disaggregate operating cash flows
from financing, tax, non-operating and non-recurring activities.
27
A Comparison between the Income Statement, the In-
direct and Direct Cash Flow Statements
Corporations can present their net cash flow from operating activities by using either the
direct or indirect method. Under each, net cash flows from operating activities will be the
same. However, the direct method presents the specific amounts of cash received and paid
for each significant business activity and the resulting net cash flow arising from operating
activities. We argue that using the structure of the income statement and the informational
content of the cash flow statement, investors are better able to isolate recurring cash flows
that are additive to net present value, segregate financing activities that are not and flag
one-off gains/losses.
The indirect method, on the other hand, derives net cash from operating activities with-
out separately presenting any of the operating cash receipts and payments. The indirect
approach merely reconciles information between the income statement and actual cash flows.
In this way, the indirect statement is not as intuitive as the direct method. In addition, it
is difficult to forecast corporate results that are presented ‘net’ of operating, financing and
tax effects.
To highlight some of the key concepts and illustrate how operating cash flows are reported
using both the direct and indirect methods, we begin with XYZ Corp.’s income statement
and balance sheet amounts:
Income Statement Amounts Balance Sheet AmountsSales $1000 Change in Accounts Receivable (A/R) +$200-Cost of Goods Sold 700 Change in Inventory +100-Operating Expenses 100 Change in Accounts Payable (A/P) Operations -50-Depreciation 10 Change in Accounts Payable (A/P) Taxes +90-Interest Expense 10+Interest Income 3-Corporate Taxes 35Net Income $148
Using the direct method, net operating cash flows are calculated using specific cash inflows
and outflows related to the income statement and balance sheet working capital accounts.
28
For example,
Cash Inflows:Sales $1000Change in A/R -200 (where A/R increases, cash has yet to be received)Gross Cash Inflows $800
Cash Outflows:Cost of Goods Sold -$700Operating Expenses -100Change in Inventory -100Change in A/P Operations -50 (where A/P decreases, cash was paid)Gross Cash Outflows -950Direct Business Cash Flows -$150
Financing:Interest Expense -$10Interest Income +3Financing Outflows -$7
Corporate Taxes:Income Statement Taxes -$35Changes in A/P Taxes +90Net Tax Cash Flows +$55
Net Cash From Operating Activities -$102
The information presented here depicts gross cash inflows and outflows in a straightforward
manner, along with an undistorted view of accrual revenues and costs versus cash inflows
and outflows. For example, while XYZ booked $1000 in sales, only $800 was received in cash
for the period, and the balance was sold on credit. Cost of goods sold, operating expenses
and working capital needs totaled $850. On an operating cash flow basis the firm suffered a
-$150 loss for the period. Recall that net income is reported as +$148. As a result of interest
expenses and a favourable tax affect, net operating cash flows were only -$102.
Using the indirect method, net operating cash flows are calculated using the differences
between net income and working capital changes. For example,
29
Net Income $148Add/Subtract Non-Cash Adjustments:
Depreciation +10Change in A/R -200Change in Inventory -100Change in A/P Operations -50Change in A/P Taxes +90
Net Cash from Operating Activities -$102
Both methods conclude that net cash from operating activities is -$102. However, the
direct method shows specific amounts of cash receipts and payments segregated by major
business activities. The indirect method merely reconciles net income to the operating cash
flows. Users of the indirect statement may not fully appreciate that business activities
(pre financing, pre-tax) generated -$150, while a tax benefit of $90 minimized the overall
cash outflows. Also, while the direct statements links working capital gains/losses to their
respective source accounts (accounts receivable to sales, accounts payables to cost of goods
sold) the indirect statement relates these working capital changes to net income. It leaves
the user with the cumbersome task of sorting and rearranging. Thus investors have three
different measures to extrapolate: net income $148, direct business cash flows of $-150, or
net operating cash flows of -$102.
30
References
American Institute of Certified Public Accountants, 1973, “Objectives of Financial
Statements,” Report of the Study Group on the Objectives of Financial Statements.
Ball, Ray, Joseph Gerakos, Juhani Linnainmaa, and Valeri Nikolaev, 2014, “Deflating
Gross Profitability,” Unpublished Working Paper, University of Chicago.
Beneish, Messod and Mark Vargus, 2002, “Insider Trading, Earnings Quality and Ac-
crual Mispricing,” Accounting Review 77:4, pp. 755-791.
Carhart, Mark, 1997, “On Persistence in Mutual Fund Performance,” Journal of Finance
52:1, pp. 5782.
Chan, Louis, Jason Karceski and Josef Lakonishok, 2003, “The Level and Persistence
of Growth Rates,” Journal of Finance 58:2, pp. 643-684.
DeFond, Mark and Chul Park, 2001, “The Reversal of Abnormal Accruals and the
Market Valuation of Earnings Surprises,” Accounting Review 76:3, pp. 375-404.
Esplin Adam, Max Hewitt, Marlene Plumlee, and Terri Lombardi Yohn, 2014, “Dis-
aggregating Operating and Financial Activities: Implications for Forecasts of Prof-
itability,” Review of Accounting Studies 19:1, pp. 328-362.
Fama, Eugene and Kenneth French, 1992, “The Cross-Section of Expected Returns,”
Journal of Finance 47:2, pp. 427-465.
Fama, Eugene and Kenneth French, 1993, “Common Risk Factors in the Returns on
Stocks and Bonds,” Journal of Financial Economics 33: 3, pp. 3-56.
Fama, Eugene and Kenneth French, 2006, “Profitability, Investment and Average Re-
turns,” Journal of Financial Economics 82, pp. 491-518.
31
Fama, Eugene and Kenneth French, 2008, “Dissecting Anomalies,” Journal of Finance
63:4, pp. 1653-1678.
Fama, Eugene and Kenneth French, 2013, “A Five-Factor Asset Pricing Model,” Un-
published Working Paper, University of Chicago.
Fama, Eugene and James MacBeth, 1973, “Risk, Return, and Equilibrium: Empirical
Test,” Journal of Political Economy 81:3, pp. 607-636.
Fairfield, Patricia and Terri Yohn, 2001, “Using Asset Turnover and Profit Margin to
Forecast Changes in Profitability,” Review of Accounting Studies 6, pp. 372-386.
Fairfield, Patricia, Karen Kitching, and Vicki Tang, V.W., 2009, “Are Special Items
Informative about Future Profit Margins?” Review of Accounting Studies 14, pp.
204-236.
Farshadfar, Shadi and Reza Monem, 2013, Further Evidence on the Usefulness of Direct
Method Cash Flow Components for Forecasting Future Cash Flows, The International
Journal of Accounting 48, pp. 111-133.
Financial Accounting Standards Board, 1987, “Statement of Financial Accounting Stan-
dards No. 95 – Statement of Cash Flows,” (November 1987).
Financial Accounting Standards Board, 2008a, “Original Pronouncements as Amended,
Statement of Financial Accounting Concepts No. 1, Objectives of Financial Reporting
by Business Enterprises.”
Financial Accounting Standards Board, 2008b, “Preliminary Views on Financial State-
ment Presentation,” Financial Accounting Series Discussion Paper No. 1630-100 (Oc-
tober 16, 2008).
Goyal, Mahendra, 2004, “A Survey of Popularity of The Direct Method of Cashflow
Reporting,” Journal of Applied Management Accounting Research 2:2, pp. 41-52.
32
Graham, John, Campbell Harvey and Shiva Rajgopal, 2005, “The Economic Implica-
tions of Corporate Financial Reporting,” Journal of Accounting and Economics 40:1,
pp. 3-73.
Hackel, Kenneth, Joshua Livnat, and AtuI Ra, 1991, Cash Flow and Security Analysis,
Homewood, Ill.: Business One-Irwin.
Hackel, Kenneth, Joshua Livnat, and AtuI Ra, 1994, “The Free Cash Flow/Small-Cap
Anomaly,” Financial Analysts Journal 50:5, pp. 33-42.
Healy, Paul and James Wahlen, 1999, “A Review of the Earnings Management Lit-
erature and Its Implications for Standard Setting,” Accounting Horizons 13:4 pp.
365-383.
Houge, Todd and Tim Loughran, 2000, ”Cash Flow Is King? Cognitive Errors by
Investors,” The Journal of Psychology and Financial Markets 1:3&4: pp. 161-175.
Kellogg, Irving, and Loren Kellogg, 1991, “Fraud, Window Dressing, and Negligence in
Financial Statements,” New York: McGraw-Hill.
Khandani, Amir and Andrew Lo, 2007, “What Happened to the Quants in August
2007?,” Journal of Investment Management 5:4, pp. 5-54.
Krishnan, Gopal and James Largay III, 2000, “The Predictive Ability of Direct Method
Cash Flow Information,” Journal of Business Finance & Accounting 27:1&2, pp. 215-
245.
Lakonishok, Josef, Andrei Shleifer, and Robert Vishny, 1994, “Contrarian Investment,
Extrapolation, and Risk,” Journal of Finance 49, pp. 1541-1579.
Lee, Charles, 2014, “Performance Measurement: An Investors Perspective,” Accounting
and Business Research 44:4, pp. 383-406.
33
Livnat, Joshua and Massimo Santicchia, 2006, “Cash Flows, Accruals, and Future Re-
turns,” Financial Analysts Journal 62:4, pp. 48-61.
Livnat, Joshua and German Lopez-Espinosa, 2008, “Quarterly Accruals or Cash Flows
in Portfolio Construction?” Financial Analysts Journal 64:3, pp. 67-79.
Lundholm Russell and Richard Sloan, 2006, Equity Valuation and Analysis with eVal
(2nd ed.), Boston, MA: McGraw-Hill Irwin.
Markham, Jerry, 2006, A Financial History of Modern U.S. Corporate Scandals: From
Enron to Reform, Armonk, NY: M.E. Sharpe Inc.
Nissim, Doron and Stephen Penman, 2001, “Ratio Analysis and Equity Valuation: From
Research to Practice,” Review of Accounting Studies 6, pp. 109-154.
Novy-Marx, Robert, 2013, “The Other Side of Value: The Gross Profitability Premium,”
Journal of Financial Economics 108:1, pp. 1-28.
OGlove, Thornton, 1987, T.L., 1987, “Quality of Earnings,” New York: The Free Press.
Orpurt, Steven and Yoonseok Zang, 2009, “Do Direct Cash Flow Disclosures Help
Predict Future Operating Cash Flows and Earnings?,” Accounting Review 84:3, pp.
893-935.
Pastor, Lubos and Robert Stambaugh, 2003. “Liquidity Risk and Expected Stock Re-
turns.” Journal of Political Economy 111:3, pp. 642-685.
Penman Stephen, 2013, Financial Statement Analysis & Security Valuation (5th ed.),
Boston, MA: McGraw-Hill Irwin.
Reilly, David, 2007, “Profit as We Know It could Be Lost with New Account-
ing Statements,” Wall Street Journal, May 12, 2007 http://online.wsj.com/article/
SB117893520139500814.html (accessed May 31, 2013).
34
Richardson, Scott, Richard Sloan, Mark Soliman, and Irem Tuna, 2005, “Accrual Relia-
bility, Earnings Persistence and Stock Prices,” Journal of Accounting and Economics
39:3, pp. 437485.
Siegel, Joel, 1991, “How to Analyze Businesses, Financial Statements, and the Quality
of Earnings,” 2nd Edition, Englewood Cliffs, NJ: Prentice-Hall.
Soliman, Mark, 2008, “The Use of DuPont Analysis by Market Participants,” The
Accounting Review 83:3, pp. 823-853.
Sloan, Richard, 1996, “Do Stock Prices Fully Reflect Information in Accruals and Cash
Flows about Future Earnings?” Accounting Review 71:3, pp. 289-315.
Thomas, Jacob, and Huai Zhang, 2002, “Inventory Changes and Future Returns.” Re-
view of Accounting Studies, vol. 7, nos. 2-3 (June-September): 163-187.
White, Halbert, 1980, “A Heteroskedasticity-Consistent Covariance Matrix Estimator
and a Direct Test for Heteroskedasticity,” Econometrica 48:4, pp. 817838.
Xie, Hong, 2001, “The Mispricing of Abnormal Accruals,” Accounting Review 76:3, pp.
357-373.
35
Table 1: Direct Cash Flow Template
This table presents a template for the estimation of various free cash flow direct measurespresented in the paper. FCF AFTA is measured as C - capex, FCF AF is measured as B -capex, and FCF O is measured as A - capex.
Operating ActivitiesCash Inflows:
Sales+/- change in accounts receivable+/- change in deferred revenues+/- change in other cash inflows from operations= Cash inflows to the firm
Cash Outflows:less: cost of goods soldless: selling general & administrative expenses+/- change in accounts payable from operations+/- change in inventories
= Net Cash Flows from Operations (A)
Financing Activitiesless: interest expense+/- other financing income/expenses
Net Cash Flows from Operations After Financing Activities (B)
Tax Activitiesless: taxes on income statement+/- change in accounts payable taxes+/- change in deferred taxes
= Net Cash Flows from Operations After Financing and Tax Activities (C)
Other Non-Operating Activities (Extraordinary Activities)
+/- discontinued operations/special charges+/- foreign exchange gains/losses+/- pension gains/losses/contributions
= Net Cash Flows from Operations After Financing, Tax, andExtraordinary Activities (D)
Investing Activitiesless capital expenditures (capex)
= Free Cash Flows to Equity Holders (FCF)
36
Table 2: Descriptive Statistics
This table presents summary statistics for variables used to create free cash flow estimates fornon-financial S&P 1500 stocks, quarterly from October 1992 to December 2013. The mean(Mean), minimum (Min) and maximum (Max) observations are presented, in $millions, alongwith various percentile cutoffs (P1, P50, and P99). Variables include operating activities-net cash flow (Xpressfeed mnemonic OANCF), net sales (SALE), accounts receivable - de-crease/increase (RECCH), cost of goods sold (COGS), selling, general and administrativeexpense (XSGA), depreciation and amortization - income statement (DP), depreciation andamortization - cash flow statement (DPC), funds from operations - other (FOPO), accountspayable and accrued liabilities - increase /decrease (APALCH), inventory - decrease/increase(INVCH), assets and liabilities - other - net change (AOLOCH), sale of property, plant andequipment and investments - gain/loss (SPPIV), interest and related expense - total (XINT),income taxes - total (TXT), income taxes - accrued - increase/decrease (TXACH), deferredtaxes - cash flow (TXDC), special items (SPI), discontinued operations (DO), extraordinaryitems (XI), capital expenditure (CAPX) and income before extraordinary items - availablefor common (IBCOM).
Variable Mean Min P1 P50 P99 MaxOANCF 513 -22,365 -133 81 7,609 61,236SALE 4,352 -219 6 914 58,689 476,294RECCH -30 -10,766 -767 -4 342 13,651COGS 2,992 -1 3 553 41,551 358,069XSGA 767 -283 4 160 10,782 91,129DP 216 -4 0 37 2,832 21,577DPC 229 -941 0 39 2,970 44,917FOPO 79 -112,954 -338 6 1,674 52,885APALCH 22 -5,797 -374 3 687 7,951INVCH -22 -55,024 -549 -1 263 58,188AOLOCH 0 -55,521 -635 0 634 38,540SPPIV -14 -14,535 -309 0 29 7,276XINT 98 -1 0 14 1,086 26,586TXT 129 -34,831 -161 19 2,092 39,217TXACH 3 -2,110 -157 0 180 9,454TXDC 8 -35,561 -314 0 475 7,895SPI -53 -52,740 -1,130 0 195 125,703DO 5 -14,316 -121 0 207 25,896XI -6 -54,235 -59 0 3 5,045CAPX 287 -1,053 0 40 4,018 37,985IBCOM 237 -62,834 -867 36 4,551 125,769
37
Tab
le3:
Ret
urn
san
dY
ield
Mea
sure
san
dO
ne-
Mon
th-A
hea
dR
eturn
s
This
table
mea
sure
sth
eav
erag
eon
e-m
onth
-ahea
dre
turn
(%)
for
the
enti
reO
ctob
er19
94to
Dec
emb
er20
13p
erio
don
por
tfol
ios
sort
edon
the
bas
isof
vari
ous
retu
rnon
asse
tan
dre
turn
onth
em
arke
tva
lue
ofeq
uit
yca
shflow
-bas
edan
dac
counti
ng-
bas
edm
easu
res
wit
hP
1th
elo
wes
tdec
ile
S&
P15
00st
ock
san
dP
10th
ehig
hes
tdec
ile
stock
s,an
dP
10-P
1is
the
retu
rndiff
eren
ce.
Std
Dev
isth
est
andar
ddev
iati
on,
T-s
tat
isth
et-
stat
isti
cof
the
sign
ifica
nce
ofP
1-P
10,
Min
isth
em
inim
um
retu
rn,
Max
isth
em
axim
um
retu
rn,
and
IRis
the
info
rmat
ion
rati
om
easu
red
asth
ean
nual
ized
P10
-P1
retu
rndiv
ided
by
the
annual
ized
stan
dar
ddev
iati
onof
the
retu
rndiff
eren
ce.
FC
FD
MA
FA
Tis
mea
sure
das
net
cash
flow
sfr
omop
erat
ions
afte
rfinan
cing
and
tax,
-ca
pex
;F
CF
DM
AF
ism
easu
red
asnet
cash
flow
sfr
omop
erat
ions
afte
rfinan
cing
acti
vit
ies
-ca
pex
;an
dF
CF
DM
Ois
mea
sure
das
net
cash
flow
sfr
omop
erat
ions
-ca
pex
.T
hes
edir
ect
met
hod
(DM
)ca
shflow
sar
eco
mpar
edw
ith
indir
ect
met
hod
(IM
)ca
shflow
s,F
CF
IMm
easu
red
asop
erat
ing
acti
vit
ies
net
cash
flow
(OA
NC
F)
-ca
pex
.F
orth
ere
turn
onas
set
mea
sure
s,th
eva
riab
leC
FA
FA
T/T
Ais
mea
sure
das
the
free
cash
flow
dir
ect
met
hod
mea
sure
FC
FD
MA
FA
Tdiv
ided
by
tota
las
sets
.T
he
vari
able
sC
FA
F/T
A,
CF
O/T
A,
and
CF
IM/T
Ahav
eas
the
num
erat
or,
FC
FD
MA
F,
FC
FD
MO
,an
dF
CF
IM,
resp
ecti
vely
.W
eco
mpar
eth
ese
mea
sure
sto
OP
/TA
,w
ith
the
num
erat
oras
oper
atin
gpro
fits
(as
esti
mat
edby
Bal
let
al.
(201
4),
subtr
acti
ng
from
reve
nue
cost
ofgo
ods
sold
and
sellin
g,ge
ner
al&
adm
inis
trat
ive
cost
sex
cludin
gre
sear
ch&
dev
elop
men
t);
GP
/TA
,th
egr
oss
pro
fit-
to-t
otal
asse
tsm
easu
re;
asw
ell
asth
etr
adit
ional
retu
rnon
asse
tsm
easu
reN
I/T
A,
whic
his
mea
sure
das
net
inco
me
orIB
CO
Mdiv
ided
by
tota
las
sets
.F
orth
em
arke
tva
lue
ofeq
uit
y-b
ased
mea
sure
s,th
eva
riab
les
hav
eth
esa
me
num
erat
ors
asth
ose
abov
ew
hile
the
den
omin
ator
isth
em
arke
tva
lue
ofeq
uit
y,M
VE
(inst
ead
ofto
tal
asse
ts).
Pan
elA
resu
lts
are
bas
edon
valu
e-w
eigh
ted
retu
rns
(for
non
-finan
cial
firm
son
ly);
Pan
elB
resu
lts
are
bas
edon
equal
-wei
ghte
dre
turn
s.
38
Panel
A:Value-W
eightedReturn
s
Measu
re
P1
P2
P3
P4
P5
P6
P7
P8
P9
P10
P10-P
1Std
Dev
T-sta
tM
inM
ax
IR(low)
(high)
CFAFAT/TA
0.517
0.547
0.688
0.745
0.931
0.759
1.079
1.135
1.058
1.159
0.642
4.555
[2.14]
-11.740
14.952
0.489
CFAF/TA
0.413
0.679
0.417
1.040
0.824
0.859
1.063
1.207
0.890
1.232
0.819
4.554
[2.73]
-12.906
15.300
0.623
CFO/TA
0.435
0.633
0.694
0.872
0.787
0.859
1.031
1.168
0.947
1.221
0.786
4.572
[2.61]
-12.089
16.011
0.595
CFIM
/TA
0.672
0.494
0.672
0.939
0.801
0.865
0.932
1.159
1.034
1.161
0.489
4.757
[1.56]
-15.334
13.541
0.356
OP/TA
0.630
0.816
0.681
1.058
0.770
0.786
0.885
0.878
0.995
1.118
0.488
4.743
[1.56]
-18.139
14.284
0.356
GP/TA
0.785
0.781
0.796
0.690
0.958
0.847
1.009
0.733
1.057
1.276
0.491
3.547
[2.10]
-8.589
11.545
0.480
NI/TA
1.055
0.710
0.918
0.638
0.976
0.940
0.942
0.858
0.781
1.110
0.056
4.847
[0.17]
-23.592
15.464
0.04
CFAFAT/MVE
0.760
0.556
0.497
0.560
0.691
1.223
1.061
1.122
1.311
1.476
0.716
3.935
[2.77]
-11.336
20.771
0.630
CFAF/MVE
0.728
0.354
0.532
0.437
0.909
1.129
1.170
1.090
1.437
1.494
0.766
3.837
[3.03]
-11.875
13.083
0.692
CFO/MVE
0.740
0.376
0.341
0.704
1.150
1.056
1.019
1.257
1.348
1.294
0.554
4.401
[1.91]
-13.778
20.451
0.436
CFIM
/MVE
0.762
0.606
0.410
0.640
0.597
0.976
1.146
1.407
1.213
1.490
0.729
4.086
[2.71]
-12.829
12.088
0.618
OP/MVE
0.611
0.833
0.807
0.891
0.816
0.931
1.127
1.118
1.474
1.310
0.699
6.381
[1.67]
-22.587
38.757
0.380
GP/MVE
0.792
0.646
0.829
0.974
1.068
1.023
1.151
1.296
1.268
1.373
0.581
6.427
[1.37]
-20.541
43.622
0.313
NI/MVE
0.916
0.663
0.804
0.870
0.846
0.870
0.909
1.015
1.248
1.341
0.425
5.437
[1.19]
-26.958
24.932
0.271
Panel
B:Equal-W
eightedReturn
s
Measu
re
P1
P2
P3
P4
P5
P6
P7
P8
P9
P10
P10-P
1Std
Dev
T-sta
tM
inM
ax
IR(low)
(high)
CFAFAT/TA
0.871
0.917
1.085
1.130
1.142
1.100
1.256
1.229
1.389
1.558
0.687
2.863
[3.65]
-10.284
10.230
0.831
CFAF/TA
0.867
0.998
1.069
1.170
1.031
1.126
1.262
1.340
1.329
1.485
0.618
3.168
[2.97]
-10.774
9.431
0.676
CFO/TA
0.874
1.007
1.097
1.135
1.038
1.179
1.293
1.225
1.374
1.454
0.579
3.276
[2.69]
-11.640
9.185
0.612
CFIM
/TA
0.825
1.053
1.113
1.171
1.077
1.201
1.246
1.293
1.289
1.419
0.594
3.647
[2.47]
-12.576
10.255
0.564
OP/TA
0.907
1.267
1.126
1.184
1.153
1.057
1.227
1.178
1.256
1.344
0.437
4.375
[1.52]
-26.002
11.593
0.346
GP/TA
1.045
0.996
1.060
1.099
1.227
1.092
1.186
1.294
1.343
1.301
0.257
3.340
[1.17]
-11.491
9.107
0.266
NI/TA
1.310
1.220
1.253
1.100
1.129
1.081
1.093
1.108
1.113
1.237
-0.073
4.710
[-0.23]
-21.944
12.430
-0.054
CFAFAT/MVE
0.938
0.868
0.912
0.916
1.005
1.167
1.244
1.340
1.464
1.814
0.876
3.076
[4.33]
-7.647
17.271
0.986
CFAF/MVE
0.929
0.855
0.847
0.867
1.057
1.167
1.273
1.319
1.523
1.832
0.903
3.311
[4.15]
-9.425
20.098
0.945
CFO/MVE
0.919
0.769
0.845
0.937
1.129
1.240
1.165
1.402
1.501
1.761
0.842
3.781
[3.38]
-11.021
22.062
0.771
CFIM
/MVE
0.919
0.953
0.868
0.891
1.004
1.057
1.285
1.499
1.397
1.804
0.885
3.483
[3.86]
-9.858
13.807
0.880
OP/MVE
0.778
0.876
0.880
1.062
1.026
1.172
1.230
1.474
1.604
1.594
0.816
6.054
[2.05]
-33.328
40.424
0.467
GP/MVE
0.804
0.725
0.960
0.973
1.222
1.181
1.279
1.375
1.490
1.631
0.827
6.767
[1.86]
-34.558
48.391
0.424
NI/MVE
1.325
1.063
0.960
1.024
0.986
1.068
1.087
1.120
1.483
1.523
0.198
4.358
[0.69]
-20.796
11.879
0.158
39
Table 4: Returns and Yield Measures and One-Month-Ahead Returns: Sub-Period Results
This table measures the average one-month-ahead value-weighted return (%) on portfolios sortedon the basis of various cash flow and yield measures with P1 the lowest decile S&P 1500 stocksand P10 the highest decile stocks, and P10-P1 is the return difference. The results are basedon value-weighted returns for non-financial firms only (see Table 3, Panel A for overall results).The various measures are defined in Table 3. Panel A covers October 1994 to February 2000, thetechnology bubble period; Panel B covers March 2000 to July 2007, the post-technology bubbleand pre-financial crisis period; Panel C covers August 2007 to May 2009, the financial crisis period;and Panel D covers June 2009 to December 2013, the post-financial crisis.
Panel A: Sub-period 1, October 1994 to February 2000 (65 months)
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA 1.398 3.066 1.669 4.788 [5.30] -10.505 12.545 1.207CFAF/TA 1.433 3.041 1.608 4.769 [5.12] -10.048 12.824 1.168CFO/TA 1.503 3.001 1.498 4.897 [4.65] -10.233 12.28 1.060CFIM/TA 1.627 3.039 1.412 4.766 [4.50] -8.919 13.376 1.026OP/TA 2.046 3.352 1.306 4.903 [4.05] -8.071 14.284 0.923GP/TA 2.049 3.062 1.013 4.138 [3.72] -8.589 11.545 0.848NI/TA 2.790 3.123 0.332 4.064 [1.24] -8.956 10.686 0.283
CFAFAT/MVE 1.442 1.81 0.368 3.877 [1.44] -11.336 9.193 0.329CFAF/MVE 1.546 1.805 0.259 3.895 [1.01] -11.875 8.992 0.231CFO/MVE 1.663 1.758 0.095 4.122 [0.35] -10.901 11.031 0.080CFIM/MVE 1.459 1.651 0.192 3.756 [0.78] -7.828 8.661 0.177OP/MVE 2.412 2.091 -0.321 5.938 [-0.82] -20.691 17.852 -0.187GP/MVE 2.711 1.701 -1.010 5.851 [-2.62] -20.541 12.533 -0.598NI/MVE 2.375 1.698 -0.677 4.609 [-2.23] -12.473 11.92 -0.509
Panel B: Sub-period 2, March 2000 to July 2007 (89 months)
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA -0.004 -0.043 -0.039 4.514 [-0.13] -11.740 14.953 -0.030CFAF/TA -0.609 0.117 0.726 4.737 [2.33] -12.906 15.300 0.531CFO/TA -0.543 0.142 0.685 4.523 [2.30] -8.031 16.011 0.525CFIM/TA 0.034 -0.018 -0.052 4.755 [-0.17] -15.334 13.541 -0.038OP/TA -0.216 -0.273 -0.057 4.485 [-0.19] -9.939 12.109 -0.044GP/TA 0.105 0.195 0.090 3.378 [0.40] -7.387 10.049 0.092NI/TA -0.399 -0.037 0.362 5.711 [0.96] -23.592 15.464 0.220
CFAFAT/MVE 0.385 1.078 0.693 3.497 [3.01] -8.402 11.313 0.687CFAF/MVE 0.156 1.300 1.144 3.547 [4.90] -6.733 12.210 1.117CFO/MVE 0.031 1.177 1.146 4.376 [3.98] -10.092 20.451 0.908CFIM/MVE 0.411 1.548 1.137 4.041 [4.28] -8.540 12.088 0.975OP/MVE -0.860 0.923 1.783 6.881 [3.94] -22.587 30.258 0.898GP/MVE -0.735 1.267 2.003 6.217 [4.90] -15.105 27.587 1.116NI/MVE -0.445 1.554 1.999 6.941 [4.38] -26.958 24.932 0.998
40
Panel C: Sub-period 3, August 2007 to May 2009 (22 months)
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA -0.622 -1.357 0.735 6.390 [1.75] -10.997 14.247 0.398CFAF/TA -0.594 -1.235 0.641 5.486 [1.78] -9.876 12.359 0.405CFO/TA -0.745 -1.175 0.430 6.044 [1.08] -12.089 12.361 0.247CFIM/TA -0.889 -1.477 0.589 6.361 [1.41] -11.750 12.467 0.321OP/TA -1.003 -2.543 1.540 7.097 [3.30] -18.139 11.874 0.752GP/TA -0.686 -1.842 1.155 3.333 [5.27] -3.054 10.506 1.201NI/TA -0.932 -0.652 -0.280 5.181 [-0.82] -12.697 12.461 -0.187
CFAFAT/MVE -0.271 -1.278 1.006 7.058 [2.17] -9.806 20.771 0.494CFAF/MVE -0.823 -1.242 0.419 6.247 [1.02] -10.851 13.083 0.232CFO/MVE -1.821 -0.973 -0.848 6.953 [-1.85] -13.778 13.532 -0.422CFIM/MVE -1.282 -1.270 -0.012 6.541 [-0.03] -12.829 11.692 -0.006OP/MVE -0.974 -1.567 0.594 10.178 [0.89] -10.228 38.757 0.202GP/MVE -0.913 -1.276 0.362 11.980 [0.46] -16.354 43.622 0.105NI/MVE -0.942 -0.758 -0.184 3.941 [-0.71] -7.057 5.961 -0.162
Panel D: Sub-period 4, June 2009 to December 2013 (55 months)
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA 1.565 1.069 0.496 3.172 [2.37] -6.604 7.676 0.541CFAF/TA 1.629 1.522 0.107 3.441 [0.47] -8.074 7.777 0.108CFO/TA 1.648 1.399 0.249 3.474 [1.09] -8.301 8.712 0.248CFIM/TA 1.668 1.436 0.232 3.912 [0.90] -7.219 9.160 0.206OP/TA 1.576 1.595 -0.018 3.600 [-0.08] -7.706 8.644 -0.018GP/TA 1.698 1.441 0.258 3.088 [1.27] -5.369 7.629 0.289NI/TA 1.405 2.039 -0.634 4.017 [-2.40] -9.750 9.287 -0.547
CFAFAT/MVE 2.425 1.378 1.047 2.921 [5.45] -5.848 5.998 1.242CFAF/MVE 2.366 1.474 0.892 2.904 [4.67] -6.822 8.478 1.064CFO/MVE 2.181 1.481 0.700 3.270 [3.26] -6.327 7.153 0.742CFIM/MVE 2.315 1.316 0.999 3.198 [4.75] -6.608 8.478 1.082OP/MVE 1.925 1.733 0.193 3.278 [0.89] -7.719 10.800 0.203GP/MVE 2.072 1.824 0.248 3.169 [1.19] -6.453 9.147 0.272NI/MVE 1.487 2.065 -0.577 3.077 [-2.85] -6.924 7.047 -0.650
41
Table 5: Returns and Yield Measures Regression Results
This table presents regression results for monthly data from October 1994 to December2013. Average returns are in excess of the one-month Treasury bill rate (from KenFrench’s website). The dependent variables are the P1 (lowest) to P10 (highest) portfoliovalue-weighted returns (for Panels A through D) or equal-weighted returns (Panel E) fornon-financial firms for the portfolios sorted on the basis of the various measures describedin Table 3. The independent variables are based on Carhart’s (1997) extension of theFama-French (1993) three-factor model where MKTRF is the market risk premium (marketreturn in excess of the risk-free rate), SMB is the small minus big (size) factor, HML isthe high minus low (market-to-book) factor, and UMD is the up minus down (momentum)factor; as well as LIQ, Pastor and Stambaugh’s (2003) traded liquidity factor—see equation(1). R2 is the adjusted R-square. T-statistics, reported in parentheses, are based on White’s(1980) heteroscedasticity-consistent standard errors are reported below estimates. Theintercept, α, is interpreted as the alpha or abnormal return.
Panel A: P1 (value-weighted returns)
Dependent AverageVariable Return α MKTRF SMB HML UMD LIQ R2
CFAFAT/TA 0.285 -0.0023 1.200 0.062 -0.221 -0.093 0.138 0.776[0.63] [-1.05] [22.66] [0.79] [-2.17] [-1.94] [2.32]
CFAF/TA 0.181 -0.0032 1.174 0.075 -0.246 -0.095 0.131 0.767[0.41] [-1.46] [22.46] [0.94] [-2.31] [-1.94] [2.21]
CFO/TA 0.203 -0.0029 1.188 0.059 -0.280 -0.105 0.13 0.774[0.45] [-1.28] [22.40] [0.74] [-2.73] [-2.23] [2.19]
CFIM/TA 0.440 -0.0023 1.285 0.167 -0.133 -0.048 0.179 0.781[0.93] [-0.97] [21.98] [2.07] [-1.74] [-1.05] [3.10]
OP/TA 0.398 -0.0003 1.272 -0.004 0.118 -0.164 -0.152 0.830[0.91] [-0.14] [25.64] [-0.08] [1.78] [-3.63] [-3.14]
GP/TA 0.553 0.002 0.968 -0.179 0.114 0.013 -0.041 0.806[1.79] [1.45] [26.37] [-4.77] [2.01] [0.41] [-1.05]
NI/TA 0.823 0.0038 1.265 0.293 -0.357 -0.235 -0.006 0.838[1.65] [1.73] [20.29] [3.72] [-4.50] [-4.11] [-0.10]
CFAFAT/MVE 0.528 -0.0002 1.068 0.016 0.190 -0.122 0.209 0.790[1.38] [-0.08] [22.47] [0.22] [2.83] [-3.18] [4.80]
CFAF/MVE 0.496 -0.0008 1.078 0.030 0.157 -0.109 0.226 0.790[1.28] [-0.39] [22.94] [0.40] [2.41] [-3.14] [5.13]
CFO/MVE 0.508 -0.0003 1.111 0.020 0.012 -0.105 0.194 0.769[1.25] [-0.14] [21.72] [0.25] [0.15] [-2.46] [3.88]
CFIM/MVE 0.530 -0.0013 1.189 0.060 0.232 -0.164 0.253 0.778[1.21] [-0.57] [20.15] [0.66] [3.35] [-3.85] [4.63]
OP/MVE 0.379 -0.0004 1.205 -0.026 -0.618 0.09 -0.009 0.846[0.85] [-0.25] [24.07] [-0.44] [-7.39] [2.06] [-0.17]
GP/MVE 0.560 0.0016 1.151 -0.126 -0.58 0.125 0.006 0.866[1.37] [1.07] [26.59] [-2.12] [-8.13] [3.35] [0.13]
NI/MVE 0.684 0.0024 1.265 0.305 -0.189 -0.336 0.010 0.830[1.35] [1.04] [20.76] [3.50] [-2.22] [-4.64] [0.16]
42
Panel B: P2 through P9 s (value-weighted returns)
DependentVariable P2 α P3 α P4 α P5 α P6 α P7 α P8 α P9 α
CFAFAT/TA -0.0018 0.0008 0.0009 0.0037 0.0012 0.0056 0.0053 0.0054[-1.18] [0.49] [0.65] [2.78] [1.01] [3.92] [3.68] [4.21]
CFAF/TA -0.0001 -0.0031 0.0044 0.0026 0.0022 0.0046 0.0063 0.0036[-0.07] [-2.33] [3.06] [1.97] [1.56] [3.78] [4.04] [2.87]
CFO/TA -0.0007 0.0002 0.0028 0.0021 0.0025 0.0043 0.0057 0.0046[-0.42] [0.11] [1.90] [1.85] [1.96] [3.67] [3.84] [3.77]
CFIM/TA -0.0019 0.0002 0.0026 0.0021 0.0022 0.0036 0.0062 0.0053[-1.14] [0.16] [2.00] [1.48] [1.36] [2.67] [4.77] [3.65]
OP/TA 0.0015 -0.0003 0.0038 0.001 0.0017 0.0029 0.003 0.0049[0.91] [-0.24] [2.42] [0.67] [1.26] [2.18] [2.62] [3.67]
GP/TA 0.0001 0.0005 0.0001 0.0039 0.0035 0.005 0.0027 0.0058[0.09] [0.39] [0.08] [2.82] [2.15] [3.28] [1.67] [4.74]
NI/TA -0.0003 0.0016 0.0000 0.003 0.0032 0.0041 0.003 0.0021[-0.20] [1.26] [-0.01] [2.33] [2.82] [3.45] [2.49] [1.95]
CFAFAT/MVE -0.0010 -0.0013 -0.0008 0.0016 0.0078 0.0054 0.0054 0.0066[-0.59] [-0.88] [-0.73] [1.49] [5.84] [3.67] [3.83] [3.77]
CFAF/MVE -0.0033 -0.0015 -0.0015 0.0042 0.0062 0.0067 0.0051 0.0089[-1.91] [-1.02] [-1.41] [3.20] [4.43] [4.41] [3.29] [5.19]
CFO/MVE -0.0034 -0.003 0.0017 0.0069 0.0051 0.0049 0.0067 0.0069[-2.07] [-2.15] [1.48] [5.26] [3.53] [3.61] [4.46] [4.30]
CFIM/MVE -0.0012 -0.0027 0.0002 0.0009 0.0048 0.0068 0.0092 0.0059[-0.69] [-2.17] [0.13] [0.70] [3.83] [4.56] [5.13] [4.11]
OP/MVE 0.0031 0.0024 0.0037 0.0032 0.0033 0.0053 0.0046 0.0085[2.29] [1.80] [2.81] [2.24] [2.11] [3.23] [2.77] [4.07]
GP/MVE 0.0006 0.0029 0.0042 0.0049 0.0047 0.0059 0.0068 0.0064[0.50] [2.42] [3.38] [3.79] [3.19] [3.96] [3.87] [3.89]
NI/MVE -0.0011 0.0009 0.0023 0.0032 0.0036 0.0042 0.0048 0.0073[-0.64] [0.67] [1.59] [1.92] [2.34] [3.30] [2.90] [4.20]
43
Panel C: P10 (value-weighted returns)
Dependent AverageVariable Return α MKTRF SMB HML UMD LIQ R2
CFAFAT/TA 0.927 0.0070 0.997 -0.203 -0.391 0.030 -0.051 0.827[2.68] [4.57] [25.69] [-3.29] [-6.69] [1.00] [-1.51]
CFAF/TA 1.000 0.0078 0.963 -0.236 -0.414 0.055 -0.022 0.828[2.98] [5.36] [25.42] [-4.02] [-7.62] [1.98] [-0.65]
CFO/TA 0.989 0.0075 0.968 -0.235 -0.383 0.062 -0.016 0.822[2.95] [5.05] [25.29] [-3.91] [-6.99] [2.34] [-0.47]
CFIM/TA 0.929 0.0071 0.982 -0.233 -0.472 0.078 -0.046 0.859[2.73] [5.28] [28.95] [-4.53] [-9.61] [2.23] [-1.40]
OP/TA 0.886 0.0066 1.021 -0.240 -0.543 0.110 -0.073 0.836[2.44] [4.35] [27.08] [-4.45] [-9.47] [4.22] [-2.07]
GP/TA 1.044 0.008 0.924 -0.085 -0.191 0.086 -0.108 0.773[3.30] [4.96] [22.65] [-1.70] [-3.21] [2.26] [-2.74]
NI/TA 0.878 0.0068 0.946 -0.257 -0.483 0.060 -0.018 0.854[2.65] [5.05] [29.30] [-5.65] [-10.16] [1.78] [-0.53]
CFAFAT/MVE 1.244 0.0081 0.977 0.214 0.546 -0.401 0.117 0.852[3.07] [4.93] [22.38] [3.78] [6.88] [-7.75] [2.50]
CFAF/MVE 1.262 0.0076 1.006 0.166 0.545 -0.356 0.18 0.845[3.12] [4.65] [22.79] [3.05] [6.95] [-8.40] [3.89]
CFO/MVE 1.062 0.0051 1.082 0.174 0.680 -0.325 0.098 0.861[2.54] [2.90] [21.32] [3.27] [8.59] [-7.31] [2.03]
CFIM/MVE 1.258 0.007 1.036 0.134 0.681 -0.286 0.133 0.859[3.18] [4.46] [26.44] [3.02] [10.74] [-6.82] [3.25]
OP/MVE 1.078 0.0053 1.116 0.172 0.515 -0.479 0.252 0.814[2.26] [2.16] [22.55] [1.89] [4.79] [-4.32] [4.34]
GP/MVE 1.141 0.0065 1.054 0.378 0.574 -0.511 0.158 0.816[2.40] [2.83] [18.57] [4.36] [5.35] [-4.89] [2.35]
NI/MVE 1.109 0.0061 1.009 0.029 0.400 -0.217 0.200 0.792[2.91] [3.32] [20.34] [0.52] [5.95] [-4.11] [4.48]
44
Panel D: P10-P1 (value-weighted returns)
Dependent AverageVariable Return α MKTRF SMB HML UMD LIQ R2
CFAFAT/TA 0.642 0.0071 -0.195 -0.339 -0.184 0.149 -0.179 0.206[2.14] [2.57] [-3.00] [-3.82] [-1.63] [2.84] [-2.65]
CFAF/TA 0.819 0.0093 -0.203 -0.265 -0.17 0.123 -0.189 0.161[2.73] [3.24] [-2.90] [-2.75] [-1.26] [2.10] [-2.59]
CFO/TA 0.786 0.011 -0.211 -0.311 -0.167 0.15 -0.153 0.185[2.61] [3.94] [-3.08] [-3.19] [-1.26] [2.41] [-2.17]
CFIM/TA 0.489 0.0104 -0.220 -0.294 -0.102 0.167 -0.147 0.193[1.56] [3.64] [-3.19] [-3.06] [-0.81] [2.89] [-2.07]
OP/TA 0.488 0.0094 -0.303 -0.399 -0.339 0.126 -0.225 0.277[1.56] [3.32] [-4.30] [-4.40] [-3.45] [1.99] [-3.31]
GP/TA 0.491 0.0069 -0.250 -0.235 -0.661 0.273 0.079 0.393[2.10] [2.81] [-4.08] [-2.99] [-6.73] [5.36] [1.25]
NI/TA 0.056 0.0059 -0.043 0.094 -0.305 0.073 -0.067 0.13[0.17] [2.61] [-0.71] [1.49] [-3.42] [1.52] [-1.08]
CFAFAT/MVE 0.716 0.0082 -0.090 0.198 0.356 -0.279 -0.092 0.263[2.77] [3.46] [-1.47] [2.15] [4.02] [-4.42] [-1.56]
CFAF/MVE 0.766 0.0084 -0.073 0.136 0.387 -0.247 -0.046 0.251[3.03] [3.66] [-1.19] [1.67] [4.06] [-4.97] [-0.76]
CFO/MVE 0.554 0.0054 -0.029 0.154 0.668 -0.22 -0.096 0.35[1.91] [2.16] [-0.43] [1.68] [5.69] [-3.75] [-1.65]
CFIM/MVE 0.729 0.0083 -0.153 0.074 0.449 -0.122 -0.120 0.226[2.71] [3.38] [-2.32] [0.86] [5.60] [-2.25] [-1.96]
OP/MVE 0.699 0.0057 -0.089 0.198 1.133 -0.568 0.260 0.62[1.67] [2.07] [-1.22] [2.14] [7.07] [-5.89] [3.34]
GP/MVE 0.581 0.0048 -0.097 0.504 1.155 -0.635 0.152 0.657[1.37] [1.85] [-1.41] [5.81] [7.59] [-6.78] [1.96]
NI/MVE 0.425 0.0037 -0.256 -0.276 0.589 0.120 0.190 0.321[1.19] [1.13] [-2.95] [-2.71] [4.50] [1.09] [2.24]
45
Panel E: P10-P1 (equal-weighted returns)
Dependent AverageVariable Return α MKTRF SMB HML UMD LIQ R2
CFAFAT/TA 0.687 0.0085 -0.140 -0.279 0.034 0.047 -0.059 0.247[3.65] [5.05] [-3.15] [-4.30] [0.52] [1.48] [-1.48]
CFAF/TA 0.618 0.0081 -0.175 -0.343 -0.023 0.111 -0.06 0.310[2.97] [4.48] [-3.62] [-4.72] [-0.33] [2.70] [-1.26]
CFO/TA 0.579 0.0075 -0.183 -0.344 0.032 0.124 -0.053 0.327[2.69] [4.09] [-3.75] [-4.57] [0.46] [2.99] [-1.14]
CFIM/TA 0.594 0.0088 -0.310 -0.385 -0.104 0.184 -0.089 0.469[2.47] [4.80] [-6.47] [-4.65] [-1.59] [4.29] [-1.81]
OP/TA 0.437 0.0071 -0.220 -0.383 -0.798 0.288 0.031 0.570[1.52] [3.47] [-4.31] [-5.24] [-9.43] [4.64] [0.62]
GP/TA 0.257 0.0033 0.0000 0.024 -0.129 -0.031 -0.019 0.02[1.17] [1.44] [0.00] [0.27] [-1.55] [-0.58] [-0.30]
NI/TA -0.073 0.0024 -0.349 -0.655 -0.412 0.307 0.031 0.564[-0.23] [1.11] [-6.59] [-7.66] [-4.57] [5.07] [0.68]
CFAFAT/MVE 0.876 0.0079 0.036 0.140 0.481 -0.08 -0.062 0.285[4.33] [4.26] [0.77] [1.88] [7.29] [-1.23] [-1.41]
CFAF/MVE 0.903 0.0086 0.007 0.094 0.52 -0.124 -0.072 0.328[4.15] [4.49] [0.13] [1.24] [7.17] [-1.81] [-1.44]
CFO/MVE 0.842 0.0079 -0.011 0.035 0.669 -0.154 -0.054 0.429[3.38] [3.86] [-0.20] [0.41] [9.01] [-2.10] [-1.09]
CFIM/MVE 0.885 0.0092 -0.119 -0.036 0.470 0.058 -0.158 0.332[3.86] [4.51] [-2.16] [-0.44] [6.87] [0.73] [-2.95]
OP/MVE 0.816 0.0077 0.028 -0.052 0.92 -0.566 0.166 0.594[2.05] [2.90] [0.45] [-0.43] [6.90] [-5.29] [2.59]
GP/MVE 0.827 0.0071 0.041 0.286 1.211 -0.632 0.099 0.641[1.86] [2.53] [0.68] [2.00] [9.04] [-5.45] [1.43]
NI/MVE 0.198 0.0036 -0.253 -0.430 0.219 -0.011 0.089 0.320[0.69] [1.42] [-4.27] [-4.07] [2.41] [-0.13] [1.66]
46
Table 6: Returns and Yield Measures and Various Return Horizons
This table measures the average one-month-ahead (Panel A), six-month-ahead (Panel B), andtwelve-month ahead (Panel C) return (%) on value-weighted portfolios of non-financial firms sortedon the basis of various measures are defined in Table 3 with P1 the lowest decile S&P 1500 stocksand P10 the highest decile stocks, and P10-P1 is the return difference. Std Dev is the standarddeviation, T-stat is the t-statistic of the significance of P10-P1, Min is the minimum return, Maxis the maximum return, and IR is the information ratio measured as annualized P10-P1 returndivided by the annualized standard deviation of the return difference.
Panel A: Three-Month Returns
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA 1.811 3.598 1.786 7.903 [3.44] -18.180 29.872 0.452CFAF/TA 1.584 3.691 2.107 8.066 [3.97] -19.829 30.423 0.522CFO/TA 1.684 3.717 2.033 8.057 [3.83] -20.919 30.188 0.505CFIM/TA 2.191 3.602 1.411 8.856 [2.42] -23.145 27.233 0.319OP/TA 2.198 3.462 1.265 8.708 [2.21] -27.720 21.504 0.291GP/TA 2.471 3.973 1.502 5.626 [4.06] -12.348 25.809 0.534NI/TA 3.186 3.429 0.244 8.299 [0.45] -24.221 20.945 0.059
CFAFAT/MVE 2.604 4.341 1.736 7.320 [3.61] -20.142 35.201 0.474CFAF/MVE 2.565 4.645 2.079 7.759 [4.07] -18.316 38.174 0.536CFO/MVE 2.456 4.018 1.563 8.171 [2.91] -20.556 29.747 0.382CFIM/MVE 2.423 4.512 2.088 7.460 [4.25] -17.380 26.080 0.560OP/MVE 2.253 4.188 1.935 11.426 [2.57] -33.748 64.878 0.339GP/MVE 2.253 4.188 1.935 11.426 [2.57] -33.748 64.878 0.339NI/MVE 2.875 4.027 1.152 9.771 [1.79] -31.252 35.465 0.236
Panel B: Six-Month Returns
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA 4.275 7.217 2.942 12.240 [3.65] -24.207 41.896 0.340CFAF/TA 3.866 7.265 3.400 12.061 [4.28] -22.853 41.202 0.399CFO/TA 4.091 7.337 3.246 12.277 [4.02] -24.132 39.851 0.374CFIM/TA 4.212 7.284 3.072 13.125 [3.56] -28.812 42.071 0.331OP/TA 4.833 6.994 2.161 12.762 [2.57] -35.584 31.882 0.240GP/TA 5.169 8.348 3.179 7.888 [6.13] -15.504 31.783 0.570NI/TA 6.578 7.008 0.430 12.107 [0.54] -34.101 28.370 0.050
CFAFAT/MVE 5.763 8.556 2.793 11.275 [3.77] -25.817 45.529 0.350CFAF/MVE 5.458 8.990 3.532 11.917 [4.50] -23.932 54.126 0.419CFO/MVE 5.112 8.121 3.009 12.669 [3.61] -31.090 41.945 0.336CFIM/MVE 5.207 8.983 3.776 12.079 [4.75] -37.005 47.482 0.442OP/MVE 5.074 8.695 3.621 17.105 [3.22] -48.616 90.347 0.299GP/MVE 5.097 8.879 3.782 19.824 [2.90] -49.186 90.897 0.270NI/MVE 5.750 7.890 2.139 15.036 [2.16] -45.316 59.371 0.201
47
Panel C: Twelve-Month Returns
Measure P1 P10 P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA 9.301 14.379 5.078 19.893 [3.88] -49.363 83.059 0.255CFAF/TA 8.722 14.242 5.520 19.313 [4.34] -44.934 70.266 0.286CFO/TA 9.234 14.344 5.110 20.000 [3.88] -48.936 76.418 0.256CFIM/TA 7.771 13.870 6.099 21.566 [4.30] -61.148 61.966 0.283OP/TA 10.454 13.604 3.150 18.530 [2.58] -54.246 62.154 0.170GP/TA 11.145 16.333 5.188 11.279 [6.99] -22.022 40.303 0.460NI/TA 11.459 13.653 2.194 19.942 [1.67] -75.912 45.506 0.110
CFAFAT/MVE 11.126 16.734 5.608 18.093 [4.71] -47.331 59.867 0.310CFAF/MVE 9.940 17.430 7.490 18.279 [6.23] -44.268 62.081 0.410CFO/MVE 9.775 16.176 6.401 20.650 [4.71] -90.129 64.469 0.310CFIM/MVE 11.125 16.945 5.820 18.539 [4.77] -87.021 63.190 0.314OP/MVE 10.568 16.660 6.092 22.725 [4.07] -38.039 109.377 0.268GP/MVE 10.428 16.124 5.696 30.276 [2.86] -56.879 136.297 0.188NI/MVE 11.214 15.311 4.097 22.912 [2.72] -54.225 76.405 0.179
48
Table 7: Returns and Yield Measures, Sector Neutral Results
Panel A measures the average one-month-ahead return (%) on value-weighted portfolios ofnon-financial firms sorted on the basis of various measures are defined in Table 3 with P1 thelowest decile S&P 1500 stocks and P10 the highest decile stocks, and P10-P1 is the returndifference. Portfolios are created by sorting within GICS sectors to ensure sector neutrality. StdDev is the standard deviation, T-stat is the t-statistic of the significance of P10-P1, Min is theminimum return, Max is the maximum return, and IR is the information ratio measured asannualized P10-P1 return divided by the annualized standard deviation of the return difference.Panel B presents regression results for monthly data from October 1994 to December 2013.Average returns are in excess of the one-month Treasury bill rate (from Ken French’s website).The dependent variables are the P10 (highest) less P1 (lowest) portfolio value-weighted returnsfor non-financial firms sorted on the basis of various measures described in Table 3. Portfolios arecreated by sorting within GICS sectors to ensure sector neutrality. The independent variables arebased on Carhart’s (1997) extension of the Fama-French (1993) three-factor model where MKTRFis the market risk premium (market return in excess of the risk-free rate), SMB is the small minusbig (size) factor, HML is the high minus low (market-to-book) factor, and UMD is the up minusdown (momentum) factor; as well as LIQ, Pastor and Stambaugh’s (2003) liquidity factor—seeequation (1). R2 is the adjusted R-square. T-statistics, reported in parentheses, are based onWhite’s (1980) heteroscedasticity-consistent standard errors are reported below estimates. Theintercept, α, is interpreted as the alpha or excess return.
Panel A: One-Month
Measure P10-P1 Std Dev T-stat Min Max IR
CFAFAT/TA 0.606 3.985 [2.31] -16.943 14.375 0.527CFAF/TA 0.661 4.233 [2.37] -17.556 14.866 0.541CFO/TA 0.693 4.104 [2.57] -17.547 13.238 0.585CFIM/TA 0.525 4.210 [1.90] -18.022 10.631 0.432OP/TA 0.340 4.272 [1.21] -15.177 18.118 0.276GP/TA 0.203 3.436 [0.90] -12.193 11.459 0.204NI/TA 0.644 4.033 [2.43] -10.512 12.615 0.553
CFAFAT/MVE 0.523 3.411 [2.33] -10.482 13.318 0.532CFAF/MVE 0.619 3.490 [2.70] -11.953 13.887 0.615CFO/MVE 0.774 3.618 [3.25] -10.206 11.792 0.741CFIM/MVE 0.504 3.461 [2.21] -9.476 13.487 0.504OP/MVE 0.868 5.448 [2.42] -15.708 30.744 0.552GP/MVE 0.592 5.025 [1.79] -15.730 17.600 0.408NI/MVE 0.567 3.768 [2.29] -9.457 15.645 0.521
49
Panel B: Regression Analysis
Dependent AverageVariable Return α MKTRF SMB HML UMD LIQ R2
CFAFAT/TA 0.606 0.0087 -0.150 -0.227 -0.475 0.102 -0.047 0.201[2.31] [3.56] [-2.23] [-1.92] [-4.44] [1.61] [-0.82]
CFAF/TA 0.661 0.0085 -0.119 -0.237 -0.452 0.185 -0.013 0.213[2.37] [3.37] [-1.69] [-2.03] [-3.96] [2.72] [-0.21]
CFO/TA 0.693 0.0091 -0.123 -0.247 -0.445 0.173 -0.041 0.217[2.57] [3.69] [-1.74] [-2.22] [-4.16] [2.63] [-0.65]
CFIM/TA 0.525 0.0091 -0.113 -0.432 -0.671 0.101 -0.132 0.328[1.90] [3.81] [-1.75] [-5.42] [-6.46] [1.59] [-1.99]
OP/TA 0.340 0.0039 -0.260 0.092 -0.432 0.195 0.190 0.323[1.21] [1.61] [-4.18] [0.95] [-4.18] [3.48] [3.06]
GP/TA 0.203 0.0028 -0.161 0.046 -0.273 0.162 0.002 0.212[0.90] [1.36] [-3.25] [0.77] [-3.30] [3.24] [0.04]
NI/TA 0.644 0.0088 -0.155 -0.470 -0.257 0.234 -0.103 0.340[2.43] [3.81] [-2.57] [-6.20] [-2.81] [4.05] [-1.72]
CFAFAT/MVE 0.523 0.0068 -0.168 -0.221 0.05 -0.029 0.026 0.140[2.33] [2.92] [-2.58] [-2.58] [0.57] [-0.46] [0.46]
CFAF/MVE 0.619 0.0078 -0.178 -0.203 0.156 -0.038 -0.004 0.180[2.70] [3.39] [-2.87] [-2.84] [2.02] [-0.69] [-0.06]
CFO/MVE 0.774 0.0087 -0.094 -0.115 0.23 -0.075 -0.038 0.126[3.25] [3.53] [-1.34] [-1.56] [2.38] [-1.37] [-0.55]
CFIM/MVE 0.504 0.0077 -0.195 -0.373 -0.006 -0.009 -0.076 0.270[2.21] [3.69] [-3.11] [-5.28] [-0.08] [-0.20] [-1.20]
OP/MVE 0.868 0.0065 0.070 0.370 0.838 -0.443 0.175 0.486[2.42] [2.40] [1.07] [4.16] [6.13] [-6.02] [2.02]
GP/MVE 0.592 0.003 0.037 0.446 0.964 -0.298 0.125 0.506[1.79] [1.16] [0.55] [5.44] [9.64] [-4.07] [1.50]
NI/MVE 0.567 0.0079 -0.100 -0.302 0.099 -0.024 -0.153 0.180[2.29] [3.30] [-1.42] [-3.91] [1.06] [-0.34] [-2.61]
50
Table 8: Returns and Yield Measures Related to Book/Price, Size, Momentum, and GrossProfit
This table compares various two-way sorts for cash flow returns, CFAFAT/TA, in Panel A, andcash flow yields, CFAFAT/MVE, in Panel B (see Table 3 for definitions). The measures are averagevalue-weighted one-month-ahead returns (%) of non-financial S&P 1500 firms sorted on the basisof each variable. 25 equal-sized portfolios are formed, first across five quintiles sorted on the basisof each variable, and then into quintiles again within each group, on the basis of book-to-price(B/P), market capitalization (size), price momentum (momentum), and gross profit-to-total assets(GP/TA). High - Low is the monthly return (%) difference between the highest and lowest cashflow returns or cash flow yield quintiles within a particular B/P, size, or momentum quintile. Thet-stat tests the significance of the high-low difference.
Panel A: Cash Flow Returns (CFAFAT/TA)
High-Low 2 3 4 High Low t-stat
Low B/P 0.316 0.518 0.962 1.034 1.196 0.880 [3.07]2 B/P 0.294 0.712 1.025 1.236 1.381 1.087 [3.59]3 B/P 0.916 1.111 0.707 1.000 1.201 0.285 [1.01]4 B/P 0.895 0.946 1.052 1.411 1.407 0.511 [1.54]High B/P 0.789 0.907 1.079 1.135 1.667 0.878 [2.59]
Small size 1.376 1.397 1.299 1.566 1.816 0.440 [1.53]2 size 0.883 1.257 1.130 1.283 1.533 0.650 [3.13]3 size 0.722 0.991 1.265 1.153 1.381 0.659 [3.26]4 size 0.762 0.995 1.123 1.226 1.438 0.676 [3.58]Large size 0.470 0.814 0.819 0.969 1.188 0.718 [3.28]
Low momentum 0.114 0.518 0.973 0.923 1.341 1.227 [3.16]2 momentum 0.672 0.821 0.986 0.932 1.165 0.493 [1.79]3 momentum 0.453 0.613 0.794 1.124 1.012 0.559 [2.07]4 momentum 0.844 0.881 1.000 1.087 0.995 0.151 [0.59]High momentum 0.873 1.046 1.081 1.221 1.467 0.593 [1.71]
Low GP/TA 0.771 0.648 0.827 1.059 0.825 0.054 [0.19]2 GP/TA 0.235 0.819 0.747 0.880 1.021 0.785 [2.31]3 GP/TA 0.619 0.743 1.101 1.153 1.077 0.458 [1.28]4 GP/TA 1.035 0.763 1.227 0.788 1.048 0.013 [0.04]High GP/TA 0.941 0.819 1.426 1.347 1.282 0.341 [1.04]
51
Panel B: Cash Flow Returns (CFAFAT/MVE)
High-Low 2 3 4 High Low t-stat
Low B/P 0.300 0.470 0.836 1.021 1.260 0.960 [2.93]2 B/P 0.378 0.560 0.985 1.381 1.216 0.838 [2.65]3 B/P 0.837 1.105 0.965 0.834 1.169 0.332 [1.29]4 B/P 1.109 0.803 1.064 1.347 1.418 0.309 [1.02]High B/P 0.818 0.741 1.278 1.414 1.468 0.650 [1.80]
Small size 1.123 1.361 1.403 1.495 2.018 0.896 [3.33]2 size 0.919 0.964 1.191 1.387 1.573 0.654 [3.47]3 size 0.806 0.749 1.090 1.284 1.531 0.725 [3.57]4 size 0.814 0.907 1.117 1.295 1.399 0.585 [2.65]Large size 0.396 0.676 0.878 1.249 1.222 0.826 [3.55]
Low momentum 0.326 0.437 0.886 1.189 1.734 1.408 [3.60]2 momentum 0.847 0.746 0.955 1.122 1.306 0.460 [1.64]3 momentum 0.477 0.353 0.999 1.172 1.384 0.907 [3.81]4 momentum 0.707 0.677 0.870 1.222 1.235 0.528 [2.09]High momentum 1.085 0.938 1.178 1.186 1.176 0.091 [0.32]
Low GP/TA 0.880 0.570 0.665 0.944 1.067 0.187 [0.65]2 GP/TA 0.566 0.583 0.609 0.970 1.409 0.843 [2.29]3 GP/TA 0.706 0.469 1.288 0.727 1.400 0.693 [1.84]4 GP/TA 0.873 0.380 0.801 1.313 1.709 0.836 [2.38]High GP/TA 0.876 0.843 1.394 1.530 1.494 0.618 [1.84]
52
Tab
le9:
Cor
rela
tion
s
Th
ista
ble
pre
sents
corr
elat
ion
sof
the
tim
e-se
ries
ofth
eav
erag
esac
ross
firm
sof
vari
ous
cash
flow
mea
sure
san
dco
mp
on
ents
:ou
rd
irec
tm
eth
od
cash
flow
from
oper
atio
ns
mea
sure
,C
FO
DM
;ou
rin
dir
ect
met
hod
cash
flow
from
oper
atio
ns
mea
sure
,C
FO
IM;
Nov
ey-M
arx
’s(2
013)
free
cash
flow
mea
sure
ofn
etin
com
ep
lus
dep
reci
atio
n/a
mor
tiza
tion
,le
ssw
orkin
gca
pit
alch
an
ge,
less
cap
ital
exp
end
itu
res,
CF
ON
M;
fin
anci
ng
acti
vit
ies,
Fin
Act
;ta
xac
tivit
ies,
Tax
Act
;ot
her
acti
vit
ies,
Oth
Act
;an
dca
pit
alex
pen
dit
ure
s,C
ap
ex.
Each
of
the
vari
able
sis
defl
ated
by
eith
erto
tal
asse
ts,
TA
orm
arke
tva
lue
ofeq
uit
y,M
VE
.
Pan
elA
:P
ears
onC
orre
lati
ons
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(1)
CF
OD
M/T
A1.
00(2
)C
FO
IM/T
A0.
811.
00(3
)C
FO
NM
/TA
0.45
0.66
1.00
(4)
Fin
Act
/TA
-0.1
3-0
.22
-0.2
11.
00(5
)T
axA
ct/T
A0.
490.
500.
62-0
.27
1.00
(6)
Oth
Act
/TA
-0.2
30.
120.
55-0
.02
0.24
1.00
(7)
Cap
ex/T
A0.
220.
290.
280.
040.
140.
071.
00(8
)C
FO
DM
/MV
E0.
280.
08-0
.21
0.27
-0.1
7-0
.31
0.04
1.00
(9)
CF
OIM
/MV
E0.
220.
320.
020.
21-0
.11
-0.0
10.
140.
741.
00(1
0)C
FO
NM
/MV
E-0
.07
0.12
0.47
0.12
0.10
0.53
0.14
-0.1
70.
201.
00(1
1)F
inA
ct/M
VE
-0.1
7-0
.25
-0.2
80.
60-0
.29
-0.0
6-0
.03
0.59
0.43
0.03
1.00
(12)
Tax
Act
/MV
E0.
050.
090.
250.
010.
520.
280.
05-0
.02
0.12
0.43
-0.0
31.0
0(1
3)O
thA
ct/M
VE
-0.1
20.
120.
45-0
.08
0.23
0.70
0.07
-0.6
1-0
.20
0.67
-0.3
00.3
11.0
0(1
4)C
apex
/MV
E-0
.08
-0.0
7-0
.11
0.32
-0.2
2-0
.01
0.52
0.50
0.52
0.20
0.57
0.0
4-0
.21
1.0
0
53
Pan
elB
:S
pea
rman
Cor
rela
tion
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(1)
CF
OD
M/T
A1.
00(2
)C
FO
IM/T
A0.
831.
00(3
)C
FO
NM
/TA
0.58
0.70
1.00
(4)
Fin
Act
/TA
-0.1
8-0
.26
-0.3
01.
00(5
)T
axA
ct/T
A0.
530.
520.
71-0
.28
1.00
(6)
Oth
Act
/TA
-0.1
10.
090.
34-0
.03
0.24
1.00
(7)
Cap
ex/T
A0.
240.
280.
350.
070.
180.
121.
00(8
)C
FO
DM
/MV
E0.
290.
13-0
.16
0.40
-0.1
4-0
.14
0.11
1.00
(9)
CF
OIM
/MV
E0.
220.
33-0
.02
0.30
-0.1
10.
030.
170.
831.
00(1
0)C
FO
NM
/MV
E-0
.07
0.04
0.24
0.30
0.05
0.34
0.23
0.53
0.61
1.00
(11)
Fin
Act
/MV
E-0
.33
-0.4
1-0
.48
0.89
-0.4
4-0
.04
0.01
0.57
0.47
0.45
1.00
(12)
Tax
Act
/MV
E0.
120.
110.
260.
110.
600.
250.
100.
350.
340.
570.
181.0
0(1
3)O
thA
ct/M
VE
-0.0
50.
150.
40-0
.08
0.30
0.98
0.13
-0.2
1-0
.03
0.27
-0.1
30.2
21.0
0(1
4)C
apex
/MV
E-0
.15
-0.1
5-0
.17
0.42
-0.2
60.
040.
650.
540.
540.
560.
590.2
4-0
.03
1.0
0
54
Table 10: Fama-MacBeth Regressions for Cash Flow Components
This table presents average Fama and MacBeth (1973) slope coefficients and their t-valuesfrom cross-sectional regressions. The dependent variables are the one-month ahead returnsfor each stock. The independent variables are net cash flow from operations measure (A inTable 1), Net CF Ops; net cash flow from operations measured by the indirect cash flowmethod, CFOIM; free cash flows measured as net income plus depreciation less change inworking capital less capital expenditures, FCF; Fin Act, financing activities as measured byinterest expenses +/- other financing income/expenses; Tax Act, tax activities as measuredby taxes on the income statement +/- changes in accounts payable taxes +/- changes indeferred taxes; other non-operating activities, Non-Op Act, non-operating activities asmeasured as discontinued operations/special charges +/- foreign exchange gains/losses+/- pension gains/losses/contributions; and Capex, capital expenditures. We deflate allof these variables by the book value of total assets. Other control variables include: log(BVE/MVE), the natural logarithm of the ratio of the book value of equity to market valueof equity; log(ME), the natural logarithm of the market value of equity; r1,1 is the stock’sone-month-prior return; and r12,2 is the stock’s prior year’s return skipping a month. Wewinsorize all independent variables based on the first and 99th percentiles. Adjusted R2 isthe adjusted R-square.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)Net CF Ops 0.0153 0.0141 0.0171 0.0133 0.0181 0.0167 0.0151 0.0194
[3.22] [2.83] [3.64] [2.58] [3.69] [3.53] [2.87] [3.60]CFOIM 0.0257
[3.71]FCF 0.0077
[0.93]Fin Act -0.0367 -0.0578 -0.0602 -0.0577 -0.0403 -0.0512
[-0.74] [-1.07] [-1.13] [-1.08] [-0.72] [-0.92]Tax Act -0.0133 -0.0248 -0.0243 -0.0286 -0.0223 -0.0052
[-0.93] [-1.51] [-1.60] [-1.91] [-1.49] [-0.36]Non-Op Act -0.0088 -0.0046 0.0025 -0.0157 -0.0233
[-0.82] [-0.45] [0.25] [-1.69] [-2.17]Capex -0.0186 -0.02 -0.0215 -0.0065
[-1.89] [-2.03] [-2.18] [-0.55]log(BE/ME) 0.002 0.0019 0.0019 0.0018 0.0021 0.0015 0.0013 0.0013 0.0013 0.0013
[1.71] [1.70] [1.52] [1.55] [1.73] [1.31] [1.15] [1.19] [1.10] [1.13]log(ME) 0.0008 -0.0008 -0.0008 -0.0008 -0.0008 -0.0008 -0.0008 -0.0008 -0.0009 -0.0007
[-1.36] [-1.32] [-1.38] [-1.37] [-1.38] [-1.35] [-1.37] [-1.44] [-1.52] [-1.21]r1,1 -0.024 -0.023 -0.024 -0.025 -0.024 -0.024 -0.024 -0.025 -0.025 -0.025
[-3.09] [-2.96] [-3.18] [-3.19] [-3.17] [-3.09] [-3.17] [-3.28] [-3.30] [-3.24]r12,2 -0.0009 -0.0007 -0.0011 -0.0009 -0.001 -0.001 -0.001 -0.0014 -0.0014 -0.001
[-0.24] [-0.19] [-0.30] [-0.25] [-0.29] [-0.28] [-0.29] [-0.38] [-0.41] [-0.28]Adjusted R2 5.64% 6.17% 5.91% 5.98% 6.19% 6.49% 6.73% 7.29% 7.31% 7.31%
55