Earnings Volatility and Market Valuation:
An Empirical Investigation *+
Ronnie Barnes
London Business School
Sussex Place, Regents Park
London
NW1 4SA
email: [email protected]
Tel: +44 (0) 20 7262 5050
First Draft: 23 October, 1999
This Draft: 8 January, 2001
* Preliminary: comments welcome. This work has been partially funded by a scholarship from the Institute ofChartered Accountants in England and Wales.+ This work constitutes part of my dissertation at LBS. Thanks are due to my supervisor Henri Servaes and the othermembers of my transfer committee (Shiva Shivakumar and Michel Habib). I have also benefited from discussions withDennis Oswald, Dick Brealey and Jan Mahrt-Smith and from the comments of workshop participants at the LBSInstitute of Finance and Accounting seminar, the EIASM Workshop on Performance Measurement, University ofIowa, Tulane University, Yale University, University of California at Irvine, University of California at Davis, Inseadand the 2000 American Accounting Association meetings. The usual disclaimer applies.
1
Abstract
This study investigates whether there is a systematic relationship between a firm’s market value (as
measured by its market to book ratio) and the variability of its quarterly earnings stream. Using data
from the Compustat full coverage, industrial and research quarterly files from 1973 to 1998 inclusive
I find that, after controlling for firm size, leverage, current profitability, the level of current
investment and sales growth, there is a significantly negative relationship between the market to
book ratio and earnings volatility (defined as the coefficient of variation of various earnings
measures). I further find that this negative relationship remains even after controlling for operating
cash flow volatility, indicating that “accounting-driven” earnings volatility does indeed have an
economic impact as has long been claimed. My results are robust to various specifications (including
fixed effects, the inclusion of industry dummies, the inclusion of year-quarter dummies and adjusting
all variables relative to the relevant industry year median) and to the inclusion of additional control
variables such as future cash flow volatility and future profitability. In addition to being statistically
significant, my results also have economic significance. For example, there is on average a 0.04
difference in market to book ratios between a firm whose earnings volatility is in the 5th percentile
and one in the 95th percentile. If we compare the 1st and 99th percentiles, we obtain a difference in
market to book ratios of 0.23, approximately 15% of the median.
2
1. Introduction and Motivation
This study is motivated primarily by extensive anecdotal evidence that corporate managers believe
there to be a systematic negative relationship between the volatility of a firm’s earnings stream and
its market valuation. Consider, as an example, the recent debate in the United States concerning the
appropriate method of accounting for derivatives. The release in June 1998 by the Financial
Accounting Standards Board of SFAS 133: Accounting for Derivative Instruments and Hedging
Activities represented the culmination of a six year program on the part of the board to unify the
accounting and disclosure requirements in this area and came a full two years after the first
appearance of the Exposure Draft which preceded the final standard. The relatively long gestation
period for the standard reflects the intense and often heated debate that surrounded the project.
Much of this debate concerned the impact of the (then) proposed Standard on volatility; some 44%
of the over 250 comment letters received noted that balance sheet volatility would increase, whilst
61% noted an increase in earnings volatility. The following comments were typical (see Smithson
(1997)):
“The proposal’s introduction of unnecessary financial statement volatility will cause some
companies to avoid effective risk management strategies and potentially erode shareholder value”
[Eli Lilly]
“Given the focus on earnings by analysts and shareholders, the earnings volatility potential
presented by fair value hedge accounting, as proposed, may have a material impact on market
valuation as well”
[Providian Bancorp]
“We do not believe that appropriate consideration has been given to the negative impact of
financial statement volatility on regulated industries”
[CoBank]
Since these comments were made within the context of a particular accounting issue, it might be
argued that the concerns expressed are specific to that issue and are not indicative of a more
widespread desire to minimise earnings volatility. Similarly, it might also be argued that a company
3
which submits a comment letter arguing against a particular accounting treatment is, almost by
definition, one whose management has a vested interest (perhaps because of compensation
considerations) in doing so and that these comments should therefore not be interpreted as
representative of the views of the corporate population in general. However, a casual review of the
business press would suggest that these arguments are of little merit and that delivering a steadily
increasing pattern of earnings growth is indeed an important objective for managers. For example,
the “Heard on the Street” column in the April 16-17, 1999 edition of the Wall Street Journal Europe
includes the following quote: ‘Here’s an exercise to give some CEOs a panic attack: Imagine what
would happen if you didn’t “manage,” or smooth out, your quarterly earnings.’ (McGough (1999)).
Consequently, the question as to whether or not a negative relationship between earnings volatility
and market value actually exists is one of considerable practical importance. The quote from Eli Lilly
suggests that, in the context of SFAS 133, companies may choose to forego potentially value-
enhancing risk management strategies in order to avoid the perceived drag on value that the resulting
increase in earnings volatility would create. If, however, such a relationship does not exist and the
market is indifferent to this increased volatility, then adoption of sub-optimal risk management
policies would needlessly destroy shareholder value.
In addition to this anecdotal evidence, the academic literature contains a large number of studies
which take as a given the fact that corporate managers dislike earnings volatility. The aim of these
studies is to rationalise this aversion and the resulting practice of income smoothing (loosely defined
as the adoption of operating strategies or accounting policies which minimise some measure of
earnings volatility). My paper differs from this line of research in that, rather than assuming that
earnings volatility has a negative valuation impact and attempting to explain the source of this
impact, I investigate whether such an impact does in fact exist. Additionally, I am not concerned
with the question of whether the earnings stream whose volatility we are measuring does or does not
reflect the effect of any income smoothing. In other words, I assume that if the motivation and
opportunity for income smoothing exist, these opportunities will be exercised by corporate
managements in an optimal manner and that the earnings volatility which remains is in some sense
“inherent” – it is the economic impact of this inherent volatility in which I am interested.
The remainder of the paper is structured as follows. A detailed review of the related literature is
contained in Section 2 whilst in Section 3, I provide a description of my hypotheses, research design,
4
sample selection and data. Section 4 contains a description of my methodology and empirical results
and I summarise and conclude in Section 5.
2. Review of Related Literature
This paper is related to two lines of existing research. As noted above, the papers in the first of these
areas take as a given a managerial aversion to earnings volatility and attempt to rationalise this
aversion and the resulting practice of income smoothing. For example, Hepworth (1953) claims that
rather than reporting the maximum level of profit possible, management’s objective might well be to
smooth reported profits over a number of years since owners will feel more confident toward a
company that reports stable earnings. Gordon (1964) suggests that, as far as the accounting rules
permit, management will smooth reported income since investors believe that this will permit a
higher dividend rate and lead to a higher stock price. Similarly, Beidleman (1973) suggests that by
smoothing, management are able to create a stable earnings stream which is “capable of supporting a
higher level of dividends than a more variable earnings prospect (p654)”1 whilst Barnea, Ronen and
Sadan (1975) hypothesise that management may smooth income in an attempt to enhance investors’
ability to predict future cash flows.
A potential link between earnings volatility and cost of capital has also been considered as an
explanation for income smoothing. Badrinath, Gay and Kale (1989) argue that institutional investors
normally avoid companies that experience large variations in earnings or firms that are perceived as
risky and tend to prefer companies with smoother earnings streams. If this reluctance of institutional
investors to invest in stocks with volatile earnings streams leads to a reduction in liquidity and a
consequent increase in the cost of capital, this motivation for income smoothing may indeed be well-
founded. Similarly, Bricker et al. (1995) present evidence that analysts associate earnings quality
with the ability of a company’s managers to manage earnings so as to avoid negative earnings
surprises, suggesting that a company’s analyst following is affected by its earnings volatility and that
income smoothing may be motivated by a desire to increase this following. This is certainly plausible
– the Wall Street Journal Europe article referred to earlier (McGough (1999)) contains the following
quote from an analyst in connection with a company which quite publicly declines to smooth its
1 If the level of dividends is used (as suggested in Miller and Rock (1985) by management as a means of signallingtheir inside information regarding the future prospects of the firm, it may well be the case that a high degree ofearnings variability prevents management from utilising this signalling mechanism in the most optimal way.
5
earnings: “I’m getting more nervous as the day gets closer. It’s like taking a final exam. Did I pass or
did I fail?” As above, if a reduced analyst following and the resulting increase in informational
asymmetry leads to an increase in the firm’s cost of capital, income smoothing may be a value-
maximising strategy. Finally, it is possible that if investors are able to estimate a firm’s cost of capital
only with error and that such estimates are based in part on the volatility of the firm’s earnings, an
increase in volatility may lead to an increase in the estimated cost of capital and a consequent loss of
value. Whilst there is evidence (see, for example, Beaver, Kettler and Scholes (1970), Rosenberg and
McKibben (1973), Lev and Kunitsky (1974) and Bowman (1979)) that earnings variability and
equity betas are positively related, it is questionable as to whether there is a causal link. For example,
in the context of SFAS 133, the disclosures required by the standard will enable investors to “strip
out” the source of any increased earnings volatility. If investors estimate the cost of capital using all
available information, these disclosures should guarantee that this increase in volatility does not in
fact lead to any increase in the estimated cost of capital.
Whilst related, the current paper has a somewhat different focus from that of these earlier studies,
and is more closely related to the line of research that addresses the general question of whether
earnings volatility has real economic effects. Among the earliest papers to address this question were
several ‘economic consequences studies’ that examined the stock price reactions to announcements
of accounting regulations. Specifically, these papers (which include Leftwich (1981) on accounting
for business combinations and Lev (1979), Collins, Rozeff and Dhaliwal (1981) and Lys (1984) on
oil and gas accounting)2 investigate whether the magnitude of the reaction of a particular firm’s
stock price to a mandated change in accounting policy that would potentially increase earnings
volatility is related to the perceived impact of the change on the probability that accounting-based
bond covenants will be violated and costs of financial distress (either direct or indirect) incurred.
Although the results of these studies (which typically use the debt to equity ratio as a proxy for the
extent to which such covenants are binding) appear to support this hypothesis, others have
questioned their interpretation (see, for example, Frost and Bernard (1989)). Nonetheless, these
studies are at least suggestive of a negative stock price reaction to anticipated increases in earnings
volatility that increase the likelihood that bond covenants will be violated. Similarly, Bartov (1993)
and Imhoff and Thomas (1988) provide evidence that firms adjust their real activities to avoid
2 See also Beattie et al. (1994)
6
volatility, and that the extent of these adjustments varies cross-sectionally with firms’ debt
constraints.
My paper also investigates the relationship between market valuation and earnings volatility and the
extent to which the strength of this relationship (if any) is affected by leverage. However, the
approach I adopt is considerably different and, I believe, has a number of advantages. In particular,
the economic consequences literature relies heavily on the event-study approach and, as a result,
suffers from certain methodological problems of which identification of the event date and
controlling for confounding events are perhaps the most significant. Further, this line of research
focuses on anticipated increases in earnings volatility resulting from a single source. However,
reported earnings (and the volatility thereof) result from the interaction of real operating decisions,
accounting regulations in their totality and managerial discretion in applying those regulations.
Consequently, I choose to estimate the association between market valuation and realised earnings
volatility (as a proxy for investors’ expectations of future volatility) within a multivariate regression
framework.
More recently, Minton and Schrand (1999) investigate the links between cash flow volatility,
discretionary investment and the costs of external financing. They find that “cash flow volatility is
associated with lower levels of investment in capital expenditures, R&D and advertising. Thus, firms
do not turn to external capital markets to fully cover cash flow short-falls (p423).” They further find
that this sensitivity of investment to cash flow volatility is more pronounced for firms which have
higher costs of accessing external capital markets (as proxied by debt YTM, S&P bond rating, equity
beta, stock price volatility, analyst following and dividend payout ratio). Finally, they provide
evidence that these proxies are positively3 related to earnings and cash flow volatility although the
relationships are often weak4.
Their interpretation of these results is that it is investors’ expectations of future cash flow volatility
that is important in determining the costs of accessing external capital markets and that historical
earnings volatility is relevant only in that it is a better predictor of this future cash flow volatility than
3 In the sense that firms with higher earnings and/or cash flow volatility have higher costs of accessing the debt andequity markets.4Specifically, S&P bond ratings and analyst following are significantly related to cash flow but not earnings volatility,the reverse is true for equity beta and dividend payout ratios whilst stock price volatility is significantly related to both.
7
is historical cash flow volatility5. However, they also note that the relationship between dividend
payout ratios and earnings volatility is consistent with the evidence in Smith and Warner (1979) who
observe that dividend restrictions in debt covenants are frequently based on accounting earnings
realisations.
Implicit in this last observation is the hypothesis that high earnings volatility can have economic
consequences, even when not accompanied by a correspondingly high cash flow volatility. This is of
particular interest when considering the implications of SFAS 133. If, as firms claim, derivatives are
primarily used for hedging purposes then an expected effect of SFAS 133 will be to cause cash flow
volatility decreasing transactions to generate an increase in earnings volatility6. Suppose that, as
Minton and Schrand suggest, current earnings volatility is important only as a predictor of future
cash flow volatility. Then, since the disclosure requirements of the standard are sufficiently detailed
that investors will be able to deduce that the increased earnings volatility has no negative
implications for cash flow volatility, the managerial concerns described above are invalid.
The results of Minton and Schrand indicate that both cash flow volatility and earnings volatility are
negatively related to the level of discretionary expenditure and positively related to the costs of
accessing external capital markets. In this paper, I examine whether these relationships translate into
a negative relationship between cash flow and earnings volatility and market valuations. Such a
relationship is not an automatic corollary of the Minton and Schrand results – it is not implausible
that since capital for discretionary investment is more readily available for firms with lower volatility,
such firms are more likely to invest in negative NPV projects.
Other papers which have explicitly considered the link between market values and earnings volatility
are Hunt, Moyer and Shevlin (1997) and Barth, Elliott and Finn (1999) (hereafter HMS and BEF
respectively). HMS hypothesise that in a regression of the market value of equity on net income, the
slope coefficient will be affected by the variability of net income, the variability of cash flow from
operations and the extent to which net income has been smoothed via non-discretionary and
5 In a footnote to their paper, Minton and Schrand describe results which suggest that this is in fact the case over shorthorizons but that for horizons beyond six years, the two predictors are essentially indistinguishable.6 This assumes that firms hedge cash flows rather than earnings. It might be interesting to investigate whether there isany systematic relationship between a firm’s use of derivatives and these two measures of volatility and for which therelationship (if any) is more pronounced. However, the inadequacy of pre-SFAS 133 disclosures may well make thisstudy difficult to implement.
8
discretionary accruals. They find that, in general, firms which smooth their income stream using
accruals have higher market values of equity (in the sense that the coefficient on net income is
higher) than do firms which do not undertake such smoothing; further, the impact is greater when
the smoothing is achieved using discretionary, rather than non-discretionary, accruals. They also find
that the coefficient is higher for firms with lower earnings and operating cash flow volatilities.
By contrast, BEF are primarily interested in whether the market values more highly firms which
deliver a steadily increasing stream of earnings. They also regress the market value of equity against
net income (and also the book value of equity) and, on the basis that earnings volatility is a proxy for
operating risk, hypothesise that the coefficient on net income will be decreasing in earnings volatility
– this is indeed the case.
Whilst similar in spirit, my study differs from these papers in terms of the research design employed.
Both HMS and BEF rely on specific valuation models (Modigliani and Miller (1966) and Ohlson
(1995)) which are valid only in extremely limited circumstances. For example, expressing the market
value of equity as a linear combination of current net income and book value is valid only if the
assumption of linear information dynamics without other information is valid. Moreover, there is
nothing within the theoretical development of these models which leads to a role for earnings
volatility. Consequently, neither HMS nor BEF are able to identify whether their empirical results
with respect to earnings volatility arise because such volatility is in itself important or because
earnings volatility is correlated with some relevant omitted variable. By contrast, my research design
(which uses a dependent variable which is widely accepted as a measure of firm value and then,
based on the results from prior research and economic reasoning, controls for other factors which
are likely to be associated with this variable) does allow me to address such issues.
Finally, using a similar research design to the one in the current paper, Shin and Stulz (2000)
investigate the relationship between firm value and risk. They find that firm value (as measured by
the market to book ratio) increases with systematic equity risk and falls with both unsystematic and
total equity risk.
Whereas the current study investigates the link between earnings volatility and market value, a
number of papers examine the relationship between this volatility and stock returns. For example,
Billings (1999) finds that a strategy of investing in stocks with low earnings volatility and financing
9
these positions by short selling high earnings volatility stocks generates consistently large, positive
returns.
Michelson et al (1995) divide a sample of S&P 500 firms into “smoothers” and “nonsmoothers and
find that the mean annualised continuously compounded return of smoothers over the 1982-1991
period is significantly lower than that of nonsmoothers. They also find that this result becomes
stronger as the criteria which a company has to satisfy to be classified as a smoother become stricter.
Finally, they find that smoothers have, on average, significantly lower equity betas (a result which is
consistent with Beaver, Kettler and Scholes (1970)) but are larger. They interpret this last result as
support for the hypothesis that larger firms have a greater incentive to smooth income than do
smaller firms since the former are subject to greater scrutiny by the government and general public; it
is also consistent with the observation that large firms are more likely to be followed by large
numbers of analysts who prefer to follow firms with stable earnings streams. However, in Michelson
et al (1999), the same authors find that when they use cumulative average abnormal returns
(measured against a market model benchmark), their principal result is reversed i.e. the cumulative
average abnormal return of smoothers is significantly higher than that of nonsmoothers and that this
result again becomes stronger as the criteria which a company has to satisfy to be classified as a
smoother become stricter.
The results of Billings (1999) and Michelson et al. (1995, 1999) appear to have two possible
interpretations. Firstly, earnings volatility is a risk factor (or is highly correlated with some other risk
factor) which is priced in equilibrium so that investors’ required returns are increasing in the level of
earnings volatility. Alternatively, the results may be suggestive of market inefficiency in that earnings
volatility is a firm-specific, fully diversifiable risk which investors incorrectly interpret as a priced,
systematic risk. However, care should be taken when interpreting these results. The fact that the
results of Michelson et al. (1995) are completely reversed when changing their dependent variable
from unadjusted to market-adjusted returns has to cast doubt on the robustness of the research
design, whilst there are several unexplained anomalies in the results of Billings (1999) – for example,
the relationship between equity beta and stock returns is indeterminate whilst the “size effect” (see,
for example, Fama and French (1992)) is reversed7.
10
3. Research Design and Sample Selection
The basic hypothesis that we wish to test in this paper is that there is a systematic negative
relationship between the volatility of a firm’s earnings stream and its market value. We choose to
address this question within a multivariate regression framework; specifically, we regress a measure
of market value against a measure of earnings volatility and a number of control variables which
prior research has found to be related to market value. In determining the details of the research
design, we faced three major issues: (i) measure of market value (ii) measure of earnings volatility
(iii) choice of control variables.
3.1 Measure of market value
In determining an appropriate measure of market value to use as our dependent variable, we were
motivated by the desire to find a measure which is directly comparable across firms without a need
for risk adjustment or normalisation. The obvious candidate was Tobin’s q, defined as the firm’s
market value (as a proxy for the present value of its future cash flows) divided by the replacement
cost of its tangible assets. Within the existing corporate finance literature, there are many examples
of studies which investigate the relationship between market value and some characteristic(s) of
interest using Tobin’s q as the dependent variable in a multivariate regression (for example, Lang and
Stulz (1994)8 and Servaes (1996) on corporate diversification, Allayannis and Weston (2000) on the
use of foreign currency derivatives and Hall et al. (2000) on patent citations). The intuition behind
the use of Tobin’s q in this way is as follows. Assuming that financial markets are efficient and that a
firm’s market value is therefore an unbiased estimate of the present value of its future cash flows, the
ratio of market value to the replacement cost of assets is a measure of the extent to which a firm’s
market value is derived from its intangible assets (for example, its investment opportunities or real
options). Now suppose, for example, that a high level of earnings variability increases the cost to a
particular firm of accessing external capital markets. It is then highly plausible that this will lead to
the firm being unable to exploit valuable investment opportunities and, as a direct corollary, a lower
market value. In this situation, we would expect there to be the hypothesised systematic negative
relationship between Tobin’s q and earnings volatility. Similarly, we would expect to find such a
7 Whilst beyond the scope of the current paper, an investigation of whether there is indeed a relationship betweenaverage realised stock returns and earnings volatility is an interesting topic for future research.8 On which the following discussion of the use and drawbacks of Tobin’s q draws heavily.
11
relationship if one or more of the other rationales for income smoothing discussed in Section 2 were
valid.
However, the major drawback to using Tobin’s q is that its calculation is computationally complex
and requires several detailed assumptions concerning, for example, the pattern of a firm’s future
acquisitions of plant and machinery, its depreciation policies, the difference between book value and
replacement cost in some “base year”, general changes in price levels, and the replacement cost of
inventories. Moreover, as Lang and Stulz (1994) and Perfect and Wiles (1994) (among others) have
shown, the results from regressions using the market to book ratio as the dependent variable are
often qualitatively very similar to those from regressions using Tobin’s q. For these reasons, we
choose to use the firm’s market to book ratio as the dependent variable in all of our regressions.
More specifically, the version of the market to book ratio we use is one which is based on the
“corporate finance” view of a firm as a set of net operating assets (including net working capital but
excluding cash) on which there are various claims (common stock, preference shares, minority
interests and long- and short-term debt (net of cash)). In view of the difficulty in obtaining market
values for most of these claims, we adopt a pragmatic approach (which is consistent with much of
the prior literature) and define the market to book ratio in terms of a numerator which is the sum of
the market value of common stock and the book values of preference shares, minority interests,
long-term debt, and debt in current liabilities, net of cash and short-term investments and a
denominator which is identical except that the market value of common stock is replaced by its book
value9.
3.2 Measure of earnings variability
9 It may be argued that the ratio should be computed using a lagged market value to reflect the fact that the book valueinformation relating to a particular date is not publicly available (and cannot therefore be impounded into shareprices) until some time later. However, there is a high degree of correlation between our chosen version of the ratioand an alternative version which does indeed use lagged market values. There is also a high degree of correlationbetween both of these versions and two other versions (one using contemporaneous market values, the other usinglagged market values) which are based on a view that liabilities such as accounts payable, income taxes payable anddeferred taxes are in fact elements of the firm’s financing. Further, the results from estimating our basic regression(see equation (1) in Section 4.1) using these alternative versions of the market to book ratio are qualitatively very closeto those from estimating the regression using our chosen version and are therefore not reported.Similarly, earlier studies (for example, Krishnan et al. (1999)) have used the market equity to book equity ratio as thedependent variable in regressions of this type. Given that we are using the market to book ratio as a proxy for Tobin’sq and for the reasons outlined in our discussion of Hunt, Moyer and Shevlin (1997) in Section 2, we feel that thisspecification would be inappropriate. However, when we estimate our regression using the market equity to bookequity ratio (which exhibits a correlation of around 60% with our chosen version of the market to book ratio) as thedependent variable, we obtain qualitatively similar results.
12
Consistent with Minton and Schrand (1999), we measure earnings volatility as the coefficient of
variation10 of some measure of earnings (net income, pretax income or income before extraordinary
items) over either the four or eight quarters preceding the end of the quarter at which we measure
the dependent variable. As discussed in more detail below, we exclude any firm-quarters for which
four or eight (as appropriate) contiguous observations of the relevant earnings variable are not
available.
3.3 Control variables
In addition to a measure of earnings volatility, our regressions also include as explanatory variables
firm size, leverage, current investment, sales growth and current profitability. Firm size is included
since numerous earlier studies have shown it to be related to various market-based variables (see, for
example, Fama and French (1992) for the relationship between size and stock returns) whilst
leverage is included as a proxy for the expected costs of financial distress. The level of current
investment is included as a proxy for the investment opportunity set - since our dependent variable
measures the extent to which a firm’s market value is derived from its investment opportunities or
real options, it is obviously important that we control for such opportunities. Similarly, sales growth
is included since it may capture aspects of the firm’s growth opportunities for which the level of
current investment is an inadequate proxy. Finally, we believe that a high level of current profitability
may indicate a strong competitive position which is value-enhancing.
Additionally, we include a measure of cash flow volatility in order to isolate the impact of
“accounting-driven” earnings volatility. In other words, earnings may vary due to variations in cash
flows or, even where cash flows are relatively stable, due to accruals such as depreciation, increases
in accounts receivable and so on. Since the focus of this study is on whether earnings volatility is
associated with market valuations, it is important that we control for cash flow volatility. Further,
given the results of Shin and Stulz (2000), cash flow volatility also serves as short-term measure of
firm risk.
3.4 Data
10 Defined as the standard deviation scaled by the absolute value of the mean. In addition to Minton and Schrand(1999), other studies which use the coefficient of variation as a measure of earnings (or cash-flow) volatility includeAlbrecht and Richardson (1990) and Michelson et al. (1995, 1999).
13
The data for this study is extracted from the Compustat full coverage, industrial and research
quarterly files from 1973 to 1998 inclusive. Additionally, we replace (wherever possible) any missing
value for research and development costs or advertising expenses with a value constructed from the
corresponding annual amount extracted from the relevant Compustat annual file11.
In order to ensure as far as possible the completeness and integrity of the data, we first attempt to
reconstruct any missing items from other, non-missing, items (for example, total current assets as the
sum of cash and short-term instruments, receivables, inventories and other current assets). We then
eliminate any observation for which one or more of the following data items is negative: cash and
short-term investments, accounts receivable, inventories, other current assets, total current assets,
property, plant and equipment, other assets, total assets, debt in current liabilities, accounts payable,
other current liabilities, other liabilities, long-term debt, preferred stock, sales, cost of goods sold,
selling, general and administrative expense, depreciation and amortization. Similarly, we eliminate
any observation for which either or both of total current liabilities and total liabilities is negative and
greater in absolute value than the sum of the absolute value of those component items which are
legitimately negative12. We also eliminate any observation where the difference between an item and
its components or between stockholders’ equity and total assets less total liabilities (which should by
definition be zero) exceeds (in absolute value) $1m. Where the value for capital expenditure,
research and development costs or advertising expenses is negative, we do not eliminate the
observation but treat the item as missing.
Finally, we eliminate any observation relating to a financial services company (SIC codes between
6000 and 6999) or a company which is not publicly traded, any observation with a missing value for
the number of common shares outstanding and/or the end of quarter share price and any observation
for which it was not possible to construct at least one of our measures of both earnings volatility and
cash flow volatility.
11 For example, suppose that a firm reports the following research and development costs: annual - 10m; quarter 1 -$3m; quarters 2,3 – missing; quarter 4 - $4m. We replace each of the missing observations for quarters 2 and 3 by$1/2(10 – 3 – 4)m = $1.5m.12 For example, total current liabilities may be negative because income taxes payable are negative; however, since itsother components are constrained to be non-negative, total current liabilities cannot be “more negative” than incometaxes payable.
14
The resulting sample consists of 283,498 observations relating to 11,662 firms; the number of
observations per firm ranged from 1 to 92 with a mean of 43. Descriptive statistics for key financial
statement items for the final sample are contained in Panel A of Table 1. The mean (median) book
value of equity is $298m ($31m) whilst for market value of equity, the corresponding amounts are
$724m and $54m; the mean (median) quarterly net income for the sample is $8m (zero). For all
items, the mean exceeds the median by a considerable margin. In Panel B of Table 1, we present the
time-series distribution of the sample – as expected, the number of firm observations per quarter
increases steadily through time13. There are very few observations until the fourth quarter of 1975 –
this is primarily the result of the difficulty (discussed in more detail below) in constructing a measure
of cash flow volatility.
4. Methodology and Empirical Results
4.1 Variable definition
The basic regression I estimate is
ititititititititit sgrowthinvestprofitlevsizecfvolearnvol q εββββββββ ++++++++= 76543210 (1)
where q denotes Tobin’s q, earnvol and cfvol are measures of the volatility of earnings and cash flow
respectively, size is (the natural logarithm of) total assets, and lev, profit, invest and sgrowth are
measures of leverage, current profitability, the level of current investment and sales growth
respectively. The subscripts i and t are used to denote firms and quarters respectively.
As discussed above, I use as a proxy for Tobin’s q the market to book ratio (MTB). MTB is defined
as the ratio of (a) a numerator which is the sum of the market value of common stock14 and the book
values of preference shares (item #55), minority interests (item #53), long-term debt (item #51), and
debt in current liabilities (item #45), net of cash and short-term investments (item #36) and (b) a
denominator which is identical except that the market value of common stock is replaced by its book
13 The low number of observations for the final three quarters of 1998 reflects the time at which data for the study wascollected.14 The product of the number of common shares outstanding (item #61 from the Compustat quarterly files) and theclosing share price (item #14).
15
value (item #59). In order to mitigate the impact of extreme values of the dependent variable on our
results, I winsorize at the 1% and the 99% levels - I then eliminate any observations where the ratio
is negative. Summary statistics on the resulting 269,427 observations are shown in Panel A of Table
2. The mean (median) market to book ratio for the sample is 3.42 (1.51) with a range of essentially
zero (due to rounding) to 55.70. The ratio has been increasing through time – the mean (median)
ratio in 1976 of 1.18 (0.93) increased to 3.27 (1.54) in 1986 and to 5.15 (2.16) in 1998. Similarly,
there is a wide variation across industries – SIC 2-digit code 28 (Chemicals & Allied Products) has a
mean (median) ratio of 8.31 (2.71) whilst that of 2-digit code 12 (Coal Mining) is only 1.13 (1.06).
As noted by Michelson et al. (1995, 1999), there is no consensus within the existing literature as to
which earnings number is the ultimate target of any smoothing activity. Consequently, I estimate my
regressions using three different bases for my measure of earnings volatility, earnvol – pretax income
(item #23), income before extraordinary items (item #8) and net income (item #69). More
specifically, I regress the market to book ratio at the end of quarter t against the coefficient of
variation of one of these earnings measures, calculated over quarter t-n+1 to quarter t inclusive
where n is equal to 4 or 8. i retain only those observations for which n quarters of data are available
for this calculation. Thus, for example, the regression of the market to book ratio against the
coefficient of variation of net income measured over 8 quarters will only include an observation for a
particular firm at the end of 1990 if, inter alia, net income is a non-missing data item for that firm for
each quarter of 1989 and 1990.
For consistency, I use as my measure of cash flow volatility, cfvol, the coefficient of variation of
operating cash flow calculated over the same time frame as earnvol. Since the implementation of
SFAS 95 “Statement of Cash Flows” in 1987, firms have been required to prepare a statement of
cash flows which shows the net cash flow from operating activities (OCF) on the face of the
statement (Compustat quarterly data item #108). Prior to SFAS 95, however, firms had some degree
of discretion over the format in which the statement was presented. In some formats, OCF was not
reported, in which case I adopt the approach of Biddle, Bowen and Wallace (1997) to construct a
measure of operating cash flow. If the firm used the “Cash Statement by Activity” format, funds
from operations – total (data item #82) was used. If either the “Working Capital Statement” or
“Cash Statement by Source and Use of Funds” was used, operating cash flow is estimated as funds
from operations - total, less the change in working capital (where working capital is defined as
current assets (excluding cash and short term investments) less current liabilities (excluding debt in
16
current liabilities)). Finally, where funds from operations - total was not available, this was proxied
by net income, adjusted for depreciation and deferred taxes15.
Summary statistics for the various specifications of earnvol and cfvol (after winsorization at the 1%
and the 99% levels) are shown in Panel B of Table 2. Panel C of Table 2 contains summary statistics
for the control variables size, lev, profit and sgrowth. Since size is the natural logarithm of total
assets, issues relating to outliers are less relevant here; hence I do not winsorize this variable since
doing so would leave my results qualitatively unchanged. Lev, my measure of firm leverage, is
defined as the book value of long-term debt divided by the sum of the market value of common
stock and the book values of preference shares, minority interests and long-term debt16. I use three
different versions of profit, my measure of current profitability – return on assets (defined as pretax
income plus interest expense (item #22)17 divided by total assets (item #44) less cash and short-term
investments)18, return on capital (pretax income plus interest expense divided by the denominator in
the market to book ratio) and return on equity (net income divided by the book value of common
stock). In all three cases, I use the denominator computed as at the end of the relevant quarter.
Alternative specifications based on denominators calculated using either start of quarter or the
average of end and start of quarter data are again highly correlated with my chosen specification19.
Finally, sgrowth is defined as the natural logarithm of one plus the percentage change in sales over
the previous 4 or 8 quarters, according to the time frame over which earnvol and cfvol are
calculated. As with size, I do not winsorize sgrowth – however, lev and profit are winsorized at the
1% and 99% levels.
15 As noted by Collins and Hribar (2000), any measure of operating cash flow which requires an estimation of accrualsusing consecutive balance sheets is subject to (potentially severe) estimation error due, for example, to mergers andacquisitions, divestitures and foreign currency translation. Consequently, our measure of cash flow volatility in thepre-1987 period is likely to be measured with error, leading to biased coefficient estimates. However, we haveestimated our basic regression using only the post-1987 sample and obtained qualitatively similar results to thoseobtained using the entire sample. Hence, we believe that any error in measuring operating cash flow has a negligibleimpact on our results.16 Given my definition of the market to book ratio, a more appropriate measure of leverage might be one where short-term debt, net of cash and short-term investments were included within both the numerator and the denominator.However, the correlation between this measure and the one I actually use is 67%; when I estimate my regression usingthis alternative measure, I find that my results are largely unchanged.17 In all of my measures of profitability, I actually multiply the numerator by four in order to annualise the measure.18 I exclude cash and short-term investments from the denominator since the income from these assets is excludedfrom the numerator – in other words, I am really calculating return on operating assets.19 Once again, estimation of alternative specifications of equation (1) confirms that my results are indeed largelyinsensitive to the particular version of profit I use.
17
My primary specification of invest is capital expenditure during the quarter although in certain
specifications of the regression I use research and development costs, advertising expenditure or the
sum of capital expenditure, research and development costs and advertising expenditure as
alternative measures. In all cases, I scale by total assets less cash and short-term investments at the
start of the quarter and winsorize at the 1% and 99% levels. Summary statistics for these variables
are shown in Panel D of Table 2.
A review of Panels A to D of Table 2 suggests a significant degree of skewness in a number of the
variables. Most notably, from Panel C, it can be seen that for all three versions of profit the minimum
observation is significantly greater in magnitude than the maximum observation whilst the mean
(which in all three cases is negative) is considerably lower than the median. For example, the
minimum return on assets is –361% whilst the maximum is only 84%; the mean and median
observations are –2% and 9% respectively. This significant negative skewness is consistent with the
positive skewness in the market to book ratio discussed above. Insofar as the measures of earnings
and cash flow volatility are concerned, all of the measures are positively skewed. Both the mean and
median observations are typically higher when the volatility measured is calculated using eight, rather
than four, quarters of data. Additionally, cash flows appear to be more volatile than (any of the
versions of) earnings, providing additional support for the belief that managers do indeed smooth out
earnings. Finally, all three measures of the level of current investment exhibit a moderate degree of
positive skewness, whilst the sales growth measures (as expected) exhibit little or no skewness.
4.2 Results
In Table 3, I present the results from a pooled cross-sectional/time-series OLS estimation of
equation (1) with invest defined as scaled capital expenditures and profit defined as return on assets.
As hypothesized, the coefficients20 on earnvol and cfvol are negative, indicating that higher earnings
and cash flow volatility are indeed associated with a lower market to book ratio. These coefficients
are also statistically significantly (at any reasonable level) different from zero. The coefficients on
size, lev and profit are also negative and statistically significantly different from zero whilst those on
invest and sgrowth are significantly positive. These results are independent of whether earnvol, cfvol
20 In all cases, the reported coefficients on earnvol and cfvol are (in order to aid comprehension of the results) 100times the actual coefficient i.e. a reported coefficient of –4.40 corresponds to an actual coefficient of –0.044.
18
and sgrowth are calculated using 4 or 8 quarters of data and of which particular specification of
earnvol is used.
Although the results in Table 3 are supportive of my principal hypothesis, the negative coefficient on
profit is somewhat troubling – it is difficult to construct a plausible explanation of why a higher level
of profitability should be associated with a lower market to book ratio. Based on the negative
skewness of all three measures of profit and the positive skewness of MTB, my conjecture is that
(even after winsorization) this result is driven by a small number of outlying observations with an
abnormally high negative current profitability and a correspondingly high market book ratio.
Consequently, I re-estimate equation (1) using WLS21 - the results are reported in Panel A of Table
4. I find that the coefficients on earnvol and cfvol remain significantly negative, as do those on lev.
However, the coefficients on size and profit are now significantly positive, whilst that on invest
remains significantly positive. These results are supportive of my conjecture that those in Table 3 are
disproportionately affected by a small number of outliers.
To illustrate the economic significance of our results, consider the first row of Panel A of Table 4.
As reported in Panel B of Table 2, the ranges for cvni4 and cvcfo4 are 0.04 to 29.73 and 0.09 to
73.91 respectively – the 5th/95th percentiles are 0.08/6.27 and 0.20/14.79 respectively. Consequently,
holding all else constant, a firm with an earnings volatility (as measured by cvni4) in the 95th
percentile will have a market to book ratio which is 0.04 lower than that of a firm whose earnings
volatility is in the 5th percentile – this represents 2.37% (1.05%) of the median (mean) market to
book ratio. The same analysis performed with respect to cash flow volatility also yields a difference
in the market to book ratio of 0.04. Comparing the 1st and 99th percentiles of cvni4 reveals a
difference in market to book ratios of 0.23, representing approximately 15% of the median.
A further econometric issue with the results reported in Panel A of Table 4 is that I am still
estimating equation (1) on a pooled cross-sectional/time-series basis with the implicit assumption
that the firm-quarter observations are independent. If (as is likely) this assumption is invalid, the
21 Implemented using the rreg (robust regression) procedure in Stata. This procedure begins by estimating theregression using OLS and excluding any observation for which Cook’s D statistic exceeds 1. Thereafter, the procedureoperates iteratively; firstly, it again estimates the regression using OLS on the remaining observations and calculatescase weights based on absolute residuals. It then estimates the regression using WLS with these case weights andcalculates a new set of weights based on the absolute residuals from this WLS regression. The iteration stops when themaximum change in weights falls below some prespecified tolerance level. (See Stata Reference Manual, Release 6,Volume 3, pp 253-58 for further details on the procedure).
19
estimated coefficients will be unbiased but the estimated standard errors and t-statistics will be
biased. To overcome this problem, I again estimate equation (1) using WLS but impose the
additional restriction that whilst observations are independent across firms, they are not necessarily
independent within firms22. The results are reported in Panel B of Table 4. As expected, the
additional restriction has the effect of reducing the t-statistics, in some cases considerably. However,
all coefficients remain significantly different from zero at any reasonable level of significance; in
particular, the coefficients on earnvol and cfvol are negative and significantly different from zero
with t-statistics ranging from –11 to –15.
4.3 Robustness
To test the robustness of the results in Table 4, I also estimate various alternative specifications of
equation (1), again using rreg to eliminate outliers and cluster to correct for autocorrelated
residuals. In the regressions reported in Panel A of Table 5, research and development costs and
advertising expenditures are included as separate control variables whereas in Panel B, invest is
defined as the sum of capital expenditures, research and development costs and advertising
expenditures. In both cases, the coefficients on earnvol and cfvol remain negative (and of the same
order of magnitude as in Table 4) with a somewhat lower level of significance due to a large
reduction in the number of observations. The remaining results are largely as expected – the
coefficients on all measures of current investment are positive and highly significant, with the
magnitude of the coefficients on research and development costs and advertising expenses in Panel A
typically higher than that of the coefficient on capital expenditures. This is plausible when I consider
that, although all of these variables are being used as proxies for the firm’s investment opportunity
set, the immediate write off of research and development costs and advertising expenses has the
effect of depressing the denominator in the dependent variable.
I also estimate a series of regressions (results not reported) which are essentially the same as those
reported in Panels A and B of Table 5 except that where research and development costs or
advertising expenditures are missing, I force the value to zero - the results are again qualitatively
unchanged.
22 Implemented using the cluster procedure within Stata (see Stata Reference Manual, Release 6, Volume 3, pp 158,
20
As noted earlier, Minton and Schrand (1999) suggest that earnings volatility might be important only
because investors are concerned about future cash flow volatility and this is better predicted using
current earnings volatility rather than current cash flow volatility. To allow for this possibility, I
estimate equation (1) with a measure of future cash flow volatility, futcfvol23, included as an
additional control variable. The results are reported in Table 6 and appear to suggest that earnings
volatility plays a role in the market’s valuation process above and beyond its use as a predictor of
future cash flow volatility. As conjectured, the coefficient on futcfvol is uniformly significantly
negative. However, the coefficients on earnvol and cfvol remain consistently significantly negative,
from which I conclude that earnings volatility is important in its own right and not just as a predictor
of future cash flow volatility.
A further explanation of the negative relationship between the market to book ratio and earnings
volatility is that (as suggested by Stein (1989)) it is higher quality firms which choose to smooth
earnings – in other words, it is not high earnings volatility which depresses the market to book ratio
but rather high quality (with quality proxied by the market to book ratio) which leads firms to
smooth earnings and show low earnings volatility. A plausible alternative proxy for quality is future
profitability; consequently, I estimate equation (1) with a measure of future profitability, futprof24,
included as an additional control variable. The results are shown in Table 7. As expected, the
coefficient on future profitability is significantly positive but, as above, my results with respect to
earnvol remain unchanged – the coefficient on this variable is consistently significantly negative, a
finding which tends to rule out the possibility that the market to book ratio and earnings volatility are
related only because the latter is a proxy for firm quality.
In Table 8, I summarize the results from estimating equation (1) with firm-specific fixed effects.
Once more, I consistently find negative coefficients on earnvol and cfvol which are significantly
different from zero. Other sensitivity checks I perform involve the inclusion of year-quarter dummies
(Table 9, Panel A), the inclusion of industry dummies at the two-digit SIC code level (Table 9, Panel
B) and the estimation of a specification in which all variables are adjusted relative to the relevant
industry-year median. In all of these specifications, although there are some differences in coefficient
estimates, the results are uniformly supportive of my hypothesis that there is a significant negative
178-79 for further details).23 Defined as the coefficient of variation of operating cash flow calculated over the four or eight (as appropriate)quarters starting with t+1.
21
association between a firm’s market to book ratio and the volatility of its earnings stream, even after
controlling for the volatility of its operating cash flows.
In Table 11, I present the results from estimating equation (1) with earnvol and cfvol calculated for
any given firm using the entire time-series of data (summary statistics for these specifications of the
two variables after winsorization at the 1% and 99% levels are reported in Table 10). In Panel A, I
treat earnvol and cfvol as intertemporal constants and use the entire panel of data to estimate the
regression whilst in Panel B, I recognise that (when defined in this way), earnvol and cfvol are only
known at the end of the period and so estimate the regression on a purely cross-sectional basis with
the dependent variable equal to the latest available market to book ratio. In all cases, the coefficients
on earnvol and cfvol are negative although the level of significance is considerably lower in the
specifications reported in Panel B. Similarly, the coefficient on cfvol is uniformly significantly
negative in Panel A whilst in Panel B, although still negative, it is insignificant at the 5% level. The
results with respect to the control variables are as expected.
5. Summary and Conclusions
In this study, I investigate whether there is a systematic relationship between a firm’s market value
and the variability of its quarterly earnings stream. Using data from the Compustat full coverage,
industrial and research quarterly files from 1973 to 1998 inclusive I compute firm market to book
ratios. On the grounds that it is a close proxy for Tobin’s q, (generally accepted as a measure of
value which is directly comparable across firms without a need for risk adjustment or normalisation),
I use this ratio as a measure of firm value. I define earnings volatility as the coefficient of variation of
various earnings measures, calculated over the previous 4 or 8 quarters, and find that, after
controlling for firm size, leverage, current profitability, the level of current investment and sales
growth, there is a significantly negative relationship between the market to book ratio and earnings
volatility. I further find that this negative relationship remains even after controlling for operating
cash flow volatility, indicating that “accounting-driven” earnings volatility does indeed have an
economic impact as has long been claimed. My results are robust to various specifications (including
fixed effects, the inclusion of industry dummies, the inclusion of year-quarter dummies and adjusting
24 Defined as the average return on assets calculated over the four or eight (as appropriate) quarters starting with t+1.
22
all variables relative to the relevant industry year median). They are also robust to the inclusion of
additional control variables such as future cash flow volatility and future profitability.
In addition to being statistically significant, my results also have economic significance. For example,
there is on average a 0.04 difference in market to book ratios between a firm whose earnings
volatility is in the 5th percentile and one in the 95th percentile. If we compare the 1st and 99th
percentiles, we obtain a difference in market to book ratios of 0.23, some 15% of the median.
Finally, my results have implications for corporate managements and accounting standard setters.
Given that highly volatile earnings streams do indeed translate into lower market values, it appears
that there is an opportunity to increase shareholder value by reducing earnings volatility. However,
whether this is possible is rather questionable. Given that managers have long accepted a negative
association between earnings volatility and market value, it is probable that the realised earnings
which firms report already reflect an “optimal” level of income smoothing and that any further
smoothing is either not simply not possible or, for whatever reason, suboptimal. Insofar as
accounting regulators are concerned, my results suggest that accounting standards (such as SFAS
133) which will increase earnings volatility may well lead managers to (rationally) alter their real
operating decisions in a way which might not, in itself, be shareholder value maximizing. This is not
to claim that regulators should not introduce volatility increasing standards, but rather to point out
that such standards do indeed have genuine economic consequences.
23
References
Albrecht, D.N. and F.M. Richardson. 1990. Income Smoothing by Economy Sector. Journal of
Business Finance and Accounting 17: 713-730
Allayanis, G. and J.P. Weston. 2000. The Use of Foreign Currency Derivatives and Firm Market
Value. Review of Financial Studies forthcoming
Badrinath, S.G., G.D. Gay and J.R. Kale. 1989. Patterns of Institutional Investment, Prudence, and
the Managerial ‘Safety-Net’ Hypothesis. Journal of Risk and Insurance 56: 605-629
Barnea, A., J. Ronen and S. Sadan. 1976. Classificatory Smoothing of Income with Extraordinary
Items. The Accounting Review 51: 110-122
Barth, M.E., J.A. Elliott and M.W. Finn. 1999. Market Rewards Associated with Patterns of
Increasing Earnings. Journal of Accounting Research 37: 387-414
Bartov, E. 1993. The Timing of Asset Sales and Earnings Manipulation. The Accounting Review 68:
840-855
Beattie V., S. Brown, D. Ewers, B. John, S. Manson, D. Thomas and M. Turner. 1994.
Extraordinary Items and Income Smoothing: A Positive Accounting Approach. Journal of Business
Finance and Accounting 21: 791-811
Beaver, W.H., P. Kettler and M. Scholes. 1970. The Association Between Market Determined and
Accounting Determined Risk Measures. The Accounting Review 45: 654-682
Beidleman, C. R. 1973. Income Smoothing: The Role of Management. The Accounting Review 48:
653-667
24
Biddle, G.C., R.M. Bowen, and J.S. Wallace. 1997. Does EVA beat Earnings? Evidence on
Associations with Stock Returns and Firm Values. Journal of Accounting and Economics 24: 301-
336
Billings, B. 1999. Is Earnings Variability Related to Future Stock Returns? Unpublished working
paper, Florida State University
Bowman, R. 1979. The Theoretical Relationship Between Systematic Risk and Financial
(Accounting) Variables. Journal of Finance 34: 617-630
Bricker, R., G. Previts, T. Robinson and S. Young. 1995. Financial Analyst Assessment of Company
Earnings Quality. Journal of Accounting, Auditing and Finance 10: 541-544
Collins, D.W., M.S. Rozeff and D.S. Dhaliwal. 1981. The Economic Determinants of the Market
Reaction to Proposed Mandatory Accounting Changes in the Oil and Gas Industry. Journal of
Accounting and Economics 3: 111-138
Collins, D.W., and S.P. Hribar. 2000. Errors in Estimating Accruals: Implications for Empirical
Research. Unpublished working paper, University of Iowa
Fama, E.F. and K.R. French. 1992. The Cross-Section of Expected Stock Returns. Journal of
Finance 47: 427-465
Frost, C. and V. Bernard. 1989. The Role of Debt Covenants in Assessing the Economic
Consequences of Limiting Capitalization of Exploration Costs. The Accounting Review 64: 788-808
Gordon, M.J. 1964. Postulates, Principles and Research in Accounting. The Accounting Review 39:
251-263
Hall, B.H., A.B. Jaffe and M.Trajtenberg. 2000. Market Value and Patent Citations: A First Look.
Unpublished working paper, National Bureau of Economic Research
Hepworth, S.R. 1953. Smoothing Periodic Income The Accounting Review 28: 32-39
25
Hunt, A., S. E. Moyer and T. Shevlin. 1997. Earnings Volatility, Earnings Management and Equity
Value. Unpublished Working Paper, University of Auckland, Seattle University and University of
Washington
Imhoff, G. and J. Thomas. 1988. Economic Consequences of Accounting Standards: The Lease
Disclosure Rule Change. Journal of Accounting and Economics 10: 277-310
Krishnan, M., S. Sankaraguruswamy and H-S. Shin. 1999. Skewness of Earnings and the
Believability Hypothesis: How Does the Financial Market Discount Accounting Earnings Numbers?
Unpublished Working Paper, Rutgers University, Georgetown University and Oxford University
Lang, L.H.P. and R.M. Stulz. 1994. Tobin’s q, Corporate Diversification, and Firm Performance.
Journal of Political Economy 102(6): 1248-1280
Leftwich, R. 1981. Evidence of The Impact of Mandatory Changes in Accounting Principles on
Corporate Loan Agreements. Journal of Accounting and Economics 3: 3-36
Lev, B. 1979. The Impact of Accounting Regulation on the Stock Market: The Case of Oil and Gas
Companies. The Accounting Review 54: 485-503
Lev, B. and S. Kunitsky. 1974. On the Association Between Smoothing Measures and the Risk of
Common Stocks. The Accounting Review 48: 259-270
Lys, T. 1984. Mandated Accounting Changes and Debt Covenants: The Case of Oil and Gas
Accounting. Journal of Accounting and Economics 6: 39-65
McGough, R. 1999. CEO Prefers to Take Rough with Smooth Regarding Earnings. Wall Street
Journal Europe, April 16-17
Michelson, S.E., J. Jordan-Wagner and C.W. Wootton. 1995. A Market Based Analysis of Income
Smoothing. Journal of Business Finance and Accounting 22: 1179-1193
26
Michelson, S.E., J. Jordan-Wagner and C.W. Wootton. 1999. Income Smoothing and Risk-Adjusted
Performance. Unpublished Working Paper, University of Central Florida and Eastern Illinois
University
Miller, M.H. and F. Modigliani. 1966. Some Estimates of the Cost of Capital to the Electric Utility
Industry, 1954-57. American Economic Review 56: 333-391
Miller, M.H. and K. Rock. 1985. Dividend Policy under Asymmetric Information. Journal of
Finance 40: 1031-1051
Minton, B.A. and C. Schrand. 1999. Does Cash Flow Volatility Affect Firm Value: Its Impact on
Discretionary Investment and the Costs of Debt and Equity Financing. Journal of Financial
Economics 54: 432-460
Ohlson, J.A. 1995. Earnings, Book Values, and Dividends in Equity Valuation. Contemporary
Accounting Research 11:661-688
Perfect, S.B. and K.W. Wiles. 1994. Alternative constructions of Tobin's q: An empirical
comparison. Journal Of Empirical Finance 1: 313-341
Rosenberg, B. and W. McKibben. 1973. The Prediction of Systematic and Specific Risk in Common
Stocks. Journal of Financial and Quantitative Analysis 8: 317-333
Servaes, H. 1996. The Value of Diversification During the Conglomerate Merger Wave. Journal of
Finance 51: 1201-1225
Shin, H-H. and R.M. Stulz. 2000. Firm Value, Risk, and Growth Opportunities. NBER Working
Paper Series, No. 7808
Smith, C.W. Jr. and J.B. Warner. 1979. On Financial Contracting: An Analysis of Bond Covenants.
Journal of Financial Economics 7: 117-161
27
Smithson, C.W. 1997. Managing Financial Risk - 1997 Yearbook: Chapter 7 (Accounting).
Available at http://www.cibcfp.school.com
Stein, J.C. 1989. Efficient Capital Markets, Inefficient Firms: A Model of Myopic Corporate
Behavior. Quarterly Journal of Economics 104: 655-669
28
Table 1
Panel A:
Descriptive Statistics on Sample Firm/Quarter Observations
The sample consists of 283,498 observations extracted from the Compustat full coverage, industrial and research quarterly files from 1973 to 1998
inclusive. An observation is included provided that none of the following items are negative: cash and short-term investments, accounts receivable,
inventories, other current assets, total current assets, property, plant and equipment, other assets, total assets, debt in current liabilities, accounts payable,
other current liabilities, other liabilities, long-term debt, preferred stock, sales, cost of goods sold, selling, general and administrative expense and
depreciation and amortization. Any observation for which either or both of total current liabilities and total liabilities is negative and greater in absolute
value than the sum of the absolute value of those component items which are legitimately negative is eliminated. Any observation where the difference
between an item and its components or between stockholders’ equity and total assets less total liabilities exceeds (in absolute value) $1m is eliminated.
Where the value for capital expenditure, research and development costs or advertising expenses is negative, the item is treated as missing. Any observation
relating to a financial services company (SIC codes between 6000 and 6999) or a company which is not publicly traded, any observation with a missing
value for the number of common shares outstanding and/or the end of quarter share price and any observation for which it was not possible to construct at
least one measure of both earnings volatility and cash flow volatility is eliminated. All amounts are expressed in $m.
Variable No. of
observations
Mean Median Standard
deviation
Minimum Maximum
Total assets 283,490 822 69 5,272 0 318,882
Total liabilities 283,483 524 31 4,131 0 283,297
Book value of equity 283,483 298 31 1,505 (4,822) 65,462
Market value of equity 283,498 724 54 4,173 0 295,679
Pretax income 274,084 14 1 103 (6,948) 5,215
Income before extraordinary items 274,673 8 0 63 (5,463) 2,681
Net income 274,658 8 0 71 (8,672) 3,291
Research and development costs 150,601 8 0 74 0 8,200
Capital expenditures 278,185 35 2 234 0 31,393
Advertising expense 94,163 8 0 38 0 931
29
Panel B:
Time-Series Distribution of Sample Firm/Quarter Observations
Quarter
Year 1 2 3 4 Total
1973 - - 1 1 2
1974 - - - 3 3
1975 5 6 9 423 443
1976 363 433 469 1,210 2,475
1977 1,171 1,309 1,408 1,480 5,368
1978 1,405 1,413 1,393 1,444 5,655
1979 1,401 1,410 1,398 1,435 5,644
1980 1,376 1,369 1,319 1,341 5,405
1981 1,159 1,127 1,088 1,174 4,548
1982 1,185 1,238 1,344 2,334 6,101
1983 2,211 2,253 2,225 2,849 9,538
1984 2,844 2,936 3,004 3,495 12,279
1985 3,478 3,519 3,524 3,660 14,181
1986 3,632 3,635 3,616 3,648 14,531
1987 3,662 3,715 3,745 3,811 14,933
1988 3,784 3,846 3,872 3,676 15,178
1989 3,658 3,656 3,664 3,862 14,840
1990 3,851 3,855 3,828 3,931 15,465
1991 3,921 3,915 3,918 3,901 15,655
1992 3,918 3,922 3,943 4,028 15,811
1993 4,092 4,169 4,245 4,310 16,816
1994 4,377 4,475 4,592 4,636 18,080
1995 4,696 4,731 4,704 4,742 18,873
1996 4,842 5,082 5,231 5,410 20,565
1997 5,528 5,671 5,887 5,676 22,762
1998 5,275 2,235 824 13 8,347
Total 283,498
30
Table 2
Summary Statistics
Panel A: Dependent Variable
Summary statistics on the dependent variable, MTB. MTB is the market to book ratio defined as the ratio of a numerator which is the sum of the market
value of common stock and the book values of preference shares, minority interests, long-term debt, and debt in current liabilities, net of cash and short-
term investments and a denominator which is identical except that the market value of common stock is replaced by its book value. The variable is
winsorized at the 1% and 99% levels.
Variable No of
observations
Minimum Maximum Mean Median
MTB 269,427 0.00 55.70 3.42 1.51
31
Panel B: Measures of Earnings Volatility and Cash Flow Volatility (based on four or eight
quarters of data)
Summary statistics on the independent variables earnvol and cfvol. Earnvol is defined as the coefficient of variation of either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe) calculated over either the preceding 4 or 8 quarters. Cfvol is the coefficient of variation of
operating cash flow calculated over either the preceding 4 or 8 quarters. Operating cash flow is defined as (depending on the statement of cash flows or
changes in financial position format used) either net cash flow from operating activities, funds from operations, funds from operations less changes in
working capital, or net income plus depreciation and deferred taxes less changes in working capital. All variables are winsorized at the 1% and 99% levels.
Variable No of
observations
Minimum Maximum Mean Median
earnvol
ni 4 267,684 0.04 29.73 1.70 0.55
ni 8 216,405 0.09 46.04 2.60 0.81
ibt 4 266,993 0.04 29.36 1.64 0.52
ibt 8 215,757 0.08 43.26 2.42 0.74
ibe 4 267,678 0.04 28.27 1.62 0.52
ibe 8 216,436 0.08 42.43 2.39 0.74
cfvol
4 283,484 0.09 73.91 3.84 1.18
8 229,571 0.17 97.10 4.95 1.51
32
Panel C: Control Variables (1)
Summary statistics on the control variables size, lev and profit. Size is the natural logarithm of total assets. Lev is defined as the book value of long-term
debt divided by the sum of the market value of common stock and the book values of preference shares, minority interests and long-term debt. Profit is
defined as either roa – return on assets (pretax income plus interest expense divided by total assets less cash and short-term investments), roc - return on
capital (pretax income plus interest expense divided by the denominator in the market to book ratio) and roe - return on equity (net income divided by the
book value of common stock); in all cases, the numerator is multiplied by four in order to annualise the measure. Lev, profit and sgrowth are winsorized at
the 1% and the 99% levels.
Variable No of
observations
Minimum Maximum Mean Median
size 283,464 -6.91 12.67 4.33 4.23
lev 276,661 0 0.84 0.20 0.13
roa 264,041 -3.61 0.84 -0.02 0.09
roc 254,168 -7.16 3.17 -0.02 0.13
roe 272,680 -4.90 2.99 -0.06 0.09
Panel D: Control Variables (2)
Summary statistics on the control variables invest and sgrowth. Invest is defined as either capex - capital expenditures, R&D – research and development
costs, or advert – advertising expenses, all scaled by total assets less cash and short-term investments. Sgrowth is defined as the log of one plus the
percentage increase in sales over either the preceding 4 or 8 quarters. All variables are winsorized at the 1% and the 99% levels.
Variable No of
observations
Minimum Maximum Mean Median
capex 277,683 0 0.45 0.05 0.03
rd 150,536 0 0.82 0.04 0.01
advert 94,133 0 0.13 0.01 0.01
sgrowth4 212,850 -8.71 8.32 0.09 0.09
sgrowth8 175,963 -9.71 8.58 0.18 0.17
33
Table 3
OLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditures and Sales Growth
Estimated coefficients (t-statistics) from OLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment and sales growth. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets
less cash and short-term investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 or 8
quarters ending with quarter t.
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(187,892)
4.44
(126.40)
-4.40
(-14.31)
-2.03
(-15.88)
-0.07
(-11.02)
-6.13
(-104.89)
-5.36
(-166.17)
7.15
(37.55)
1.23
(46.23)
ni 8
(146,584)
4.26
(109.73)
-2.98
(-14.07)
-1.61
(-14.85)
-0.08
(-10.47)
-5.81
(-92.22)
-5.33
(-141.00)
6.35
(29.41)
1.12
(48.71)
ibt 4
(187,828)
4.44
(126.44)
-4.78
(-15.33)
-2.03
(-15.86)
-0.07
(-11.06)
-6.12
(-104.76)
-5.37
(-166.24)
7.15
(37.52)
1.23
(46.19)
ibt 8
(146,489)
4.27
(109.85)
-3.65
(-16.05)
-1.60
(-14.84)
-0.08
(-10.61)
-5.80
(-91.93)
-5.34
(-141.32)
6.30
(29.17)
1.12
(48.67)
ibe 4
(187,906)
4.44
(126.44)
-4.73
(-14.76)
-2.02
(-15.82)
-0.08
(-11.15)
-6.13
(-104.84)
-5.36
(-166.17)
7.16
(37.59)
1.23
(46.23)
ibe 8
(146,600)
4.26
(109.71)
-3.28
(-14.24)
-1.60
(-14.85)
-0.08
(-10.61)
-5.81
(-92.26)
-5.33
(-141.02)
6.35
(29.44)
1.12
(48.72)
34
Table 4
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditures and Sales Growth
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment and sales growth. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets
less cash and short-term investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 or 8
quarters ending with quarter t
.
Panel A
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(187,892)
1.53
(309.75)
-0.76
(-17.66)
-0.28
(-15.55)
0.05
(54.81)
-1.44
(-174.84)
0.16
(34.58)
1.59
(59.09)
0.21
(56.97)
ni 8
(146,584)
1.45
(261.84)
-0.44
(-14.61)
-0.21
(-13.47)
0.06
(56.53)
-1.39
(-154.58)
0.21
(38.04)
1.37
(44.45)
0.22
(67.90)
ibt 4
(187,828)
1.53
(310.03)
-0.87
(-19.75)
-0.28
(-15.46)
0.05
(54.63)
-1.44
(-174.63)
0.15
(34.04)
1.59
(59.21)
0.21
(56.88)
ibt 8
(146,489)
1.45
(261.73)
-0.51
(-15.78)
-0.21
(-13.52)
0.06
(56.26)
-1.39
(-154.41)
0.20
(37.42)
1.37
(44.47)
0.22
(68.01)
ibe 4
(187,906)
1.53
(310.03)
-0.87
(-19.35)
-0.28
(-15.44)
0.05
(54.45)
-1.44
(-174.75)
0.15
(34.05)
1.59
(59.13)
0.21
(57.12)
ibe 8
(146,600)
1.45
(261.80)
-0.50
(-15.23)
-0.21
(-13.46)
0.06
(56.23)
-1.39
(-154.57)
0.20
(37.90)
1.37
(44.46)
0.22
(67.98)
35
Panel B (Robust Standard Errors)
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(169,852)
1.53
(93.59)
-0.76
(-13.78)
-0.28
(-14.23)
0.05
(16.36)
-1.44
(-62.93)
0.16
(10.54)
1.59
(25.44)
0.21
(30.00)
ni 8
(133,268)
1.45
(80.10)
-0.44
(-11.23)
-0.21
(-11.11)
0.06
(16.82)
-1.39
(-55.80)
0.21
(12.01)
1.37
(19.32)
0.22
(28.85)
ibt 4
(169,802)
1.53
(93.86)
-0.87
(-15.13)
-0.28
(-14.18)
0.05
(16.33)
-1.44
(-62.91)
0.15
(10.38)
1.59
(25.48)
0.21
(29.93)
ibt 8
(133,189)
1.45
(80.39)
-0.51
(-11.80)
-0.21
(-11.20)
0.06
(16.79)
-1.39
(-55.86)
0.20
(11.83)
1.37
(19.30)
0.22
(28.90)
ibe 4
(169,880)
1.53
(93.76)
-0.87
(-14.65)
-0.28
(-14.15)
0.05
(16.28)
-1.44
(-62.89)
0.15
(10.38)
1.59
(25.46)
0.21
(30.06)
ibe 8
(133,278)
1.45
(80.10)
-0.50
(-11.37)
-0.21
(-11.11)
0.06
(16.75)
-1.39
(-55.85)
0.20
(11.98)
1.37
(19.33)
0.22
(28.93)
36
Table 5
Panel A
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditures, Research and Development Costs,
Advertising Expenses and Sales Growth
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment and sales growth. The model is
itititititititititititsgrowthinvestinvestinvestprofitlevsizecfvolearnvol MTB εββββββββββ ++++++++++= 9876543210 321
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, invest1it, invest2it and invest3it are (respectively) the capital expenditures, the
research and development costs and the advertising expenses of firm i during quarter t (all scaled by total assets less cash and short-term investments at the
beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 or 8 quarters ending with quarter t.
constant earnvol cfvol size lev profit invest1 invest2 invest3 sgrowth
ni4
(35,270)
1.40
(37.43)
-0.83
(-6.04)
-0.32
(-7.66)
0.05
(6.94)
-1.43
(-27.23)
0.36
(10.79)
2.59
(17.10)
6.84
(18.93)
3.84
(6.46)
0.31
(13.46)
ni8
(27,275)
1.30
(31.30)
-0.50
(-5.40)
-0.25
(-6.95)
0.06
(8.07)
-1.39
(-24.04)
0.37
(9.59)
2.17
(12.68)
5.56
(13.12)
4.15
(6.40)
0.30
(11.59)
ibt4
(35,265)
1.40
(37.73)
-1.00
(-7.21)
-0.32
(-7.63)
0.05
(6.90)
-1.43
(-27.18)
0.36
(10.64)
2.58
(17.14)
6.81
(18.90)
3.85
(6.49)
0.31
(13.46)
ibt8
(27,245)
1.30
(31.51)
-0.57
(-5.44)
-0.25
(-6.95)
0.06
(8.03)
-1.38
(-24.04)
0.37
(9.58)
2.18
(12.71)
5.56
(13.19)
4.17
(6.43)
0.30
(11.47)
ibe4
(35,282)
1.40
(37.62)
-0.97
(-6.59)
-0.32
(-7.62)
0.05
(6.90)
-1.43
(-27.19)
0.36
(10.69)
2.59
(17.13)
6.83
(18.89)
3.86
(6.49)
0.31
(13.43)
ibe8
(27,287)
1.30
(31.41)
-0.63
(-5.80)
-0.25
(-6.93)
0.06
(8.00)
-1.38
(-24.04)
0.37
(9.54)
2.18
(12.70)
5.55
(13.08)
4.16
(6.41)
0.30
(11.57)
37
Panel B
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Total Current Investment and Sales Growth
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability and the level of current investment. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, sgrowthit is the growth in sales of firm i calculated over the 4 or 8 quarters
ending with quarter t, investit is the sum of the capital expenditures, the research and development expenditures and the advertising costs of firm i during
quarter t (scaled by total assets less cash and short-term investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm
i calculated over the 4 or 8 quarters ending with quarter t.
constant earnvol cfvol size lev profit invest sgrowth
ni4
(35,228)
1.46
(39.88)
-0.83
(-6.00)
-0.32
(-7.70)
0.05
(6.54)
-1.52
(-28.89)
0.28
(8.53)
3.28
(22.80)
0.31
(13.48)
ni8
(27,262)
1.36
(33.44)
-0.47
(-5.05)
-0.24
(-6.52)
0.06
(7.69)
-1.47
(-25.40)
0.32
(8.33)
2.81
(17.03)
0.30
(11.55)
ibt4
(35,229)
1.46
(40.18)
-1.00
(-7.14)
-0.32
(-7.68)
0.04
(6.50)
-1.52
(-28.83)
0.28
(8.39)
3.28
(22.83)
0.31
(13.48)
ibt8
(27,226)
1.36
(33.61)
-0.53
(-4.91)
-0.24
(-6.52)
0.06
(7.67)
-1.46
(-25.39)
0.32
(8.31)
2.81
(17.09)
0.30
(11.44)
ibe4
(35,244)
1.46
(40.07)
-0.98
(-6.55)
-0.32
(-7.66)
0.04
(6.50)
-1.52
(-28.84)
0.28
(8.43)
3.28
(22.82)
0.31
(13.44)
ibe8
(27,279)
1.36
(33.52)
-0.60
(-5.54)
-0.24
(-6.50)
0.06
(7.62)
-1.46
(-25.38)
0.32
(8.26)
2.81
(17.05)
0.30
(11.55)
38
Table 6
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditures, Future Cash Flow Volatility and Sales
Growth
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment, future cash flow volatility and sales growth. The model is
ititititititititititsgrowthfutcfvolinvestprofitlevsizecfvolearnvol MTB εβββββββββ +++++++++= 876543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets
less cash and short-term investments at the beginning of the quarter), futcfvolit is the coefficient of variation of operating cash flows (as defined in Table 2,
Panel B) of firm i calculated over the 4 or 8 quarters beginning with quarter t+1 and sgrowthit is the logarithmic growth in sales of firm i calculated over
the 4 or 8 quarters ending with quarter t.
constant earnvol cfvol size lev profit invest futcfvol sgrowth
ni4
(139,167)
1.49
(83.71)
-0.60
(-10.31)
-0.24
(-12.16)
0.05
(15.02)
-1.41
(-57.17)
0.34
(18.32)
1.62
(24.09)
-0.23
(-11.38)
0.24
(28.53)
ni8
(90,139)
1.38
(64.62)
-0.26
(-5.68)
-0.17
(-8.22)
0.06
(14.27)
-1.37
(-46.95)
0.54
(21.19)
1.41
(16.92)
-0.17
(-7.53)
0.24
(24.24)
ibt4
(139,164)
1.49
(83.95)
-0.70
(-11.76)
-0.24
(-12.10)
0.05
(15.00)
-1.41
(-57.14)
0.34
(18.18)
1.62
(24.14)
-0.23
(-11.34)
0.24
(28.56)
ibt8
(90,074)
1.38
(64.63)
-0.30
(-6.01)
-0.17
(-8.36)
0.06
(14.29)
-1.37
(-46.98)
0.54
(21.09)
1.41
(16.90)
-0.17
(-7.54)
0.24
(24.26)
ibe4
(139,186)
1.49
(83.79)
-0.69
(-11.03)
-0.24
(-12.10)
0.05
(14.95)
-1.41
(-57.12)
0.34
(18.18)
1.62
(24.09)
-0.23
(-11.30)
0.24
(28.59)
ibe8
(90,146)
1.38
(64.63)
-0.32
(-6.10)
-0.17
(-8.21)
0.06
(14.23)
-1.37
(-46.95)
0.54
(21.15)
1.41
(16.93)
-0.17
(-7.48)
0.24
(24.28)
39
Table 7
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditure, Future Profitability and Sales Growth
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment, future profitability and sales growth. The model is
ititititititititititsgrowthfutprofinvestprofitlevsizecfvolearnvol MTB εβββββββββ +++++++++= 876543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, sgrowthit is the growth in sales of firm i calculated over the 4 or 8 quarters
ending with quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets less cash and short-term investments at the
beginning of the quarter), futprofit is the average return on assets (as defined in Table 2, Panel C) of firm i calculated over the 4 or 8 quarters beginning
with quarter t+1.
constant earnvol cfvol size lev profit invest futprof sgrowth
ni4
(130,091)
1.44
(80.17)
-0.49
(-8.32)
-0.23
(-10.94)
0.05
(13.67)
-1.37
(-55.14)
0.17
(12.30)
1.69
(24.46)
0.55
(21.53)
0.22
(25.58)
ni8
(79,879)
1.31
(59.17)
-0.17
(-3.55)
-0.13
(-5.94)
0.05
(12.45)
-1.30
(-43.92)
0.28
(14.55)
1.52
(17.35)
0.95
(21.02)
0.22
(20.60)
ibt4
(130,115)
1.45
(80.39)
-0.60
(-9.89)
-0.23
(-10.87)
0.05
(13.62)
-1.37
(-55.10)
0.16
(12.05)
1.69
(24.49)
0.55
(21.49)
0.22
(25.61)
ibt8
(79,823)
1.31
(59.15)
-0.21
(-4.01)
-0.13
(-6.00)
0.05
(12.47)
-1.30
(-43.91)
0.28
(14.43)
1.52
(17.35)
0.94
(21.04)
0.22
(20.60)
ibe4
(130,101)
1.45
(80.31)
-0.58
(-9.16)
-0.23
(-10.87)
0.05
(13.62)
-1.37
(-55.10)
0.16
(12.12)
1.69
(24.47)
0.55
(21.56)
0.22
(25.66)
ibe8
(79,890)
1.31
(59.19)
-0.21
(-3.87)
-0.13
(-5.93)
0.05
(12.41)
-1.30
(-43.91)
0.28
(14.53)
1.52
(17.35)
0.94
(20.99)
0.22
(20.63)
40
Table 8
WLS Regressions (with firm-specific fixed effects) of Market to Book Ratio on Earnings
Volatility, Cash Flow Volatility, Size, Leverage, Current Profitability, Capital Expenditures
and Sales Growth
Estimated coefficients (t-statistics) from WLS regressions with firm-specific fixed effects of market to book ratio on earnings volatility, cash flow volatility,
size, leverage, current profitability, the level of current investment and sales growth. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets
less cash and short-term investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 or 8
quarters ending with quarter t.
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(169,852)
1.69
(117.32)
-0.64
(-16.16)
-0.13
(-7.74)
0.07
(20.97)
-1.70
(-136.49)
0.23
(30.25)
1.12
(35.64)
0.13
(32.27)
ni 8
(133,268)
1.57
(91.20)
-0.26
(-9.58)
-0.10
(-7.15)
0.08
(22.66)
-1.69
(-121.88)
0.23
(26.99)
1.00
(28.22)
0.13
(34.88)
ibt 4
(169,802)
1.69
(117.37)
-0.72
(-17.73)
-0.13
(-7.66)
0.07
(20.98)
-1.70
(-136.33)
0.23
(30.05)
1.12
(35.65)
0.13
(32.19)
ibt 8
(133,189)
1.56
(91.20)
-0.32
(-10.68)
-0.10
(-7.22)
0.08
(22.75)
-1.69
(-121.78)
0.23
(26.89)
1.00
(28.20)
0.13
(34.58)
ibe 4
(169,880)
1.69
(117.40)
-0.77
(-18.47)
-0.13
(-7.73)
0.07
(21.04)
-1.70
(-136.33)
0.23
(29.97)
1.12
(35.65)
0.13
(32.42)
ibe 8
(133,278)
1.56
(91.18)
-0.34
(-11.13)
-0.10
(-7.27)
0.08
(22.80)
-1.69
(-121.78)
0.23
(26.92)
1.00
(28.14)
0.13
(34.85)
41
Table 9
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditures and Sales Growth
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment and sales growth. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvolit is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated over the 4 or 8 quarters ending with quarter t, cfvolit is the coefficient of
variation of operating cash flows (as defined in Table 2, Panel B) of firm i calculated over the 4 or 8 quarters ending with quarter t, sizeit is the natural
logarithm of the total assets of firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is
the return on assets (as defined in Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets
less cash and short-term investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 or 8
quarters ending with quarter t.
Panel A (year-quarter dummies)
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(169,852)
1.81
(79.94)
-1.25
(-15.83)
-0.38
(-13.80)
0.04
(8.65)
-1.74
(-54.02)
0.18
(8.04)
1.76
(17.48)
0.24
(22.29)
ni 8
(133,268)
1.70
(67.67)
-0.74
(-13.46)
-0.26
(-10.09)
0.05
(10.03)
-1.67
(-47.88)
0.24
(8.94)
1.51
(13.09)
0.25
(21.35)
ibt 4
(169,802)
1.81
(80.19)
-1.36
(-16.98)
-0.37
(-13.72)
0.04
(8.62)
-1.73
(-54.00)
0.18
(7.86)
1.76
(17.48)
0.24
(22.24)
ibt 8
(133,189)
1.70
(67.90)
-0.83
(-13.67)
-0.26
(-10.23)
0.05
(10.01)
-1.67
(-47.85)
0.23
(8.81)
1.52
(13.09)
0.25
(21.31)
ibe 4
(169,880)
1.81
(80.16)
-1.38
(-16.55)
-0.37
(-13.76)
0.04
(8.56)
-1.73
(-54.00)
0.18
(7.86)
1.76
(17.48)
0.24
(22.34)
ibe 8
(133,278)
1.70
(67.74)
-0.84
(-13.63)
-0.26
(-10.14)
0.05
(9.97)
-1.67
(-47.88)
0.24
(8.92)
1.51
(13.12)
0.25
(21.38)
42
Panel B (industry dummies)
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(169,852)
1.76
(75.52)
-0.97
(-12.41)
-0.31
(-11.41)
0.05
(10.96)
-1.78
(-54.21)
0.13
(6.13)
1.82
(20.58)
0.23
(21.78)
ni 8
(133,268)
1.66
(64.29)
-0.58
(-10.62)
-0.22
(-8.76)
0.06
(11.83)
-1.70
(-47.75)
0.19
(7.47)
1.59
(15.89)
0.24
(21.24)
ibt 4
(169,802)
1.76
(75.73)
-1.10
(-13.65)
-0.31
(-11.36)
0.05
(10.95)
-1.77
(-54.16)
0.13
(5.96)
1.82
(20.59)
0.22
(21.72)
ibt 8
(133,189)
1.66
(64.47)
-0.65
(-10.67)
-0.23
(-8.93)
0.06
(11.85)
-1.70
(-47.68)
0.18
(7.36)
1.59
(15.94)
0.24
(21.19)
ibe 4
(169,880)
1.76
(75.71)
-1.10
(-13.09)
-0.31
(-11.38)
0.05
(10.89)
-1.78
(-54.19)
0.13
(5.95)
1.82
(20.60)
0.23
(21.83)
ibe 8
(133,278)
1.66
(64.33)
-0.66
(-10.63)
-0.23
(-8.82)
0.06
(11.79)
-1.70
(-47.76)
0.18
(7.44)
1.59
(15.93)
0.24
(21.25)
43
Panel C (adjusted to industry-year median)
constant earnvol cfvol size lev profit invest sgrowth
ni 4
(171,386)
0.04
(6.81)
-1.05
(-17.78)
-0.19
(-9.21)
0.03
(9.85)
-1.34
(-53.05)
0.33
(21.90)
1.41
(22.69)
0.20
(28.06)
ni 8
(134,672)
-0.01
(-1.58)
-0.66
(-15.11)
-0.12
(-6.13)
0.04
(10.64)
-1.33
(-47.25)
0.35
(19.96)
1.33
(18.41)
0.18
(24.57)
ibt 4
(171,396)
0.05
(6.89)
-1.16
(-18.27)
-0.19
(-9.12)
0.03
(9.79)
-1.34
(-53.02)
0.32
(21.77)
1.41
(22.71)
0.20
(28.10)
ibt 8
(134,589)
-0.01
(-1.51)
-0.77
(-15.56)
-0.13
(-6.26)
0.04
(10.60)
-1.33
(-47.16)
0.35
(19.75)
1.33
(18.39)
0.18
(24.53)
ibe 4
(171,390)
0.05
(7.67)
-1.11
(-17.44)
-0.19
(-9.12)
0.03
(9.82)
-1.34
(-53.00)
0.32
(21.80)
1.41
(22.78)
0.20
(28.10)
ibe 8
(134,703)
-0.01
(-1.48)
-0.76
(-15.11)
-0.12
(-6.13)
0.04
(10.52)
-1.33
(-47.20)
0.35
(19.90)
1.33
(18.39)
0.18
(24.63)
44
Table 10
Summary Statistics: Measures of Earnings Volatility and Cash Flow Volatility (using all
observations)
Summary statistics on the independent variables earnvol and cfvol. Earnvol is defined as the coefficient of variation of either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe) calculated using all available data. Cfvol is the coefficient of variation of operating cash flow
calculated using all available data. Operating cash flow is defined as (depending on the statement of cash flows or changes in financial position format
used) either net cash flow from operating activities, funds from operations, funds from operations less changes in working capital, or net income plus
depreciation and deferred taxes less changes in working capital. All variables are winsorized at the 1% and 99% levels.
Variable No of
observations
Minimum Maximum Mean Median
earnvol
ni 264,580 0.27 111.20 6.33 2.06
ibt 264,580 0.25 99.27 5.23 1.75
ibe 264,580 0.26 98.04 5.46 1.83
cfvol 265,044 0.39 134.82 7.17 2.24
45
Table 11
WLS Regressions of Market to Book Ratio on Earnings Volatility, Cash Flow Volatility, Size,
Leverage, Current Profitability, Capital Expenditures and Sales Growth
Panel A: Panel Regression
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment and sales growth. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvoli is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated using all available observations, cfvoli is the coefficient of variation of
operating cash flows (as defined in Table 2, Panel B) of firm i calculated using all available observations, sizeit is the natural logarithm of the total assets of
firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is the return on assets (as defined in
Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets less cash and short-term
investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 quarters ending with quarter t.
constant earnvol cfvol size lev profit invest sgrowth
ni
(171,556)
1.49
(85.39)
-0.17
(-4.71)
-0.11
(-4.01)
0.06
(16.87)
-1.41
(-60.90)
0.26
(17.04)
1.60
(25.12)
0.22
(30.10)
ibt
(171,544)
1.49
(84.65)
-0.14
(-3.43)
-0.11
(-4.13)
0.06
(16.80)
-1.41
(-61.10)
0.26
(17.06)
1.61
(25.21)
0.22
(30.05)
ibe
(171,559)
1.49
(85.39)
-0.23
(-5.71)
-0.10
(-3.98)
0.05
(16.65)
-1.41
(-61.17)
0.26
(16.92)
1.61
(25.23)
0.22
(30.04)
46
Panel B: Cross-Sectional Regression
Estimated coefficients (t-statistics) from WLS regressions of market to book ratio on earnings volatility, cash flow volatility, size, leverage, current
profitability, the level of current investment and sales growth. The model is
itititititititititsgrowthinvestprofitlevsizecfvolearnvol MTB εββββββββ ++++++++= 76543210
where MTBit is the market to book ratio of firm i at the end of quarter t, earnvoli is the coefficient of variation of earnings (either net income (ni), pretax
income (ibt) or income before extraordinary items (ibe)) of firm i calculated using all available observations, cfvoli is the coefficient of variation of
operating cash flows (as defined in Table 2, Panel B) of firm i calculated using all available observations, sizeit is the natural logarithm of the total assets of
firm i at the end of quarter t, levit is the leverage (as defined in Table 2, Panel C) of firm i at the end of quarter t, profitit is the return on assets (as defined in
Table 2, Panel C) of firm i for quarter t, investit is the capital expenditures of firm i during quarter t (scaled by total assets less cash and short-term
investments at the beginning of the quarter) and sgrowthit is the logarithmic growth in sales of firm i calculated over the 4 quarters ending with quarter t.
Unless otherwise stated, all variables are calculated for the last quarter for which all the relevant data for firm i is available.
constant earnvol cfvol size lev profit invest sgrowth
ni
(6,181)
1.67
(47.57)
-0.17
(-2.00)
-0.14
(-1.92)
0.09
(12.99)
-1.53
(-26.11)
0.27
(11.13)
1.22
(5.35)
0.17
(7.36)
ibt
(6,181)
1.67
(47.43)
-0.24
(-2.36)
-0.14
(-1.91)
0.09
(12.85)
-1.53
(-26.12)
0.27
(11.09)
1.22
(5.37)
0.17
(7.36)
ibe
(6,181)
1.67
(47.49)
-0.25
(-2.38)
-0.13
(-1.88)
0.09
(12.88)
-1.53
(-26.21)
0.27
(10.94)
1.22
(5.36)
0.17
(7.40)