Econometrics of the Basu Asymmetric Timeliness Coefficient

and Accounting Conservatism

Ray Ball
University of Chicago Booth School of Business
5807 South Woodlawn Avenue, Chicago, IL 60637
(773) 834-5941
ray.ball@chicagobooth.edu

S.P. Kothari
MIT Sloan School of Management
50 Memorial Drive, E52-325 Cambridge, MA 02142-1261
(617) 253-0994
kothari@mit.edu

Valeri Nikolaev
University of Chicago Booth School of Business
5807 South Woodlawn Avenue, Chicago, IL 60637
(773) 834-4116
valeri.nikolaev@chicagobooth.edu

First draft: May 2006

Current version: 18 April 2010

We acknowledge the helpful comments of an anonymous referee, Jerry Zimmerman (the

editor), Sudipta Basu, John Core, Andrei Kovrijnykh, and Karl Muller.
 Abstract

A substantial literature investigates conditional conservatism, and how it depends on market,

political and institutional variables. Studies typically assume the Basu (1997) asymmetric
timeliness coefficient (incremental slope on negative returns in a piecewise-linear regression of
accounting income on stock returns) is a valid conditional conservatism measure. We analyze its
validity, in a model with accounting income incorporating information components with
different lags, and adding noise. This model demonstrates that asymmetric timeliness varies
with firms information environments, including the relative importance of growth option
expectation revisions (proxied by market/book). We conclude that much criticism of the Basu
regression misconstrues researchers objectives.

Keywords: Capital markets; Conditional conservatism; Timely loss recognition; Basu model;
Returns-earnings regressions; Earnings response coefficients
 Econometrics of the Basu Asymmetric Timeliness Coefficient
and Accounting Conservatism

1. Introduction

Conservatism has been a central accounting principle for centuries (Watts and Zimmerman,

1986; Basu, 1997; Watts, 2003a). Basu (1997, p. 7) defines conservatism as the accountants

tendency to require a higher degree of verification to recognize good news as gains than to recognize

bad news as losses, a definition that is consistent with the adage anticipate no profits but anticipate

all losses. Basu therefore expects accounting income to be timelier in incorporating negative shocks

to firm value (bad news) than positive shocks (good news). He tests this expectation using a

measure of asymmetric timeliness constructed as follows. In an efficient market, stock return reflects

all public information affecting the value of equity that arrives during the period, and therefore

provides a valid measure of economic shocks to firm value. Consequently, the contemporaneous

return-earnings relation is expected to differ between periods of positive stock return (a proxy for

good news) and periods of negative return (a proxy for bad news). In a piecewise-linear regression of

accounting income on fiscal-period stock return, the incremental coefficient on negative return is

assumed to be a valid measure of this accounting asymmetry.1 Basu predicts and finds that the

incremental coefficient on negative stock return indeed is positive.2

Basu and many since refers to the incremental coefficient on negative return as the

asymmetric timeliness coefficient, and employs it as a valid measure of conservatism. Under this

definition of conservatism, accounting income is contingent on the sign of the shock to firm value, so

Ball and Shivakumar (2005) and Beaver and Ryan (2005) term it conditional conservatism. This

1
We use the term validity as shorthand for what commonly is described as construct validity, which Peter (1981, p.
134) notes generally is used to refer to the vertical correspondence between a construct which is at an unobservable,
conceptual level and a purported measure of it which is at an operational level. The notion is developed more fully in the
classic Cook and Campbell (1979) text and following editions.
2
The model and its origins are described in Basu (2009).
 2

contrasts with defining conservatism as the practice of reporting unconditionally low accounting

outcomes for earnings or book value of equity. The key difference between these concepts is that

conditional conservatism carries information, and hence can be useful in contracting.3

In a comparatively short period of time, the Basu (1997) piecewise-linear regression of

accounting income on stock returns has become one of the principal models of the financial

accounting literature. The models formulation of accounting conservatism provided a breakthrough

in our understanding of financial reporting practice. It therefore is not surprising that, following its

introduction in Basu (1997) and its application in an international context in Ball, Kothari, and Robin

(2000), the asymmetric timeliness coefficient has been used extensively in the accounting literature.4

Researchers have estimated asymmetric timeliness to investigate conditional conservatism in

accounting (i) as a function of countries market, political, legal and taxation regimes;5 (ii) over time,

due to regime change;6 (iii) across fiscal quarters;7 (iv) as a function of litigation, auditors legal

liability, earnings management or auditor malfeasance;8 (v) as a function of the different demands on

financial reporting of public companies, private companies and not-for-profit organizations, and of

family ownership;9 (vi) to study the effect of conditional conservatism on the equity cost of capital;10

3
See also Ball, Kothari, and Robin (2000, fn. 15), Basu (2005, Section 2) and Ball and Shivakumar (2005, 2008a). Other
things equal, conditional conservatism implies unconditional conservatism, but the converse does not hold. Conditional
conservatism is the stricter concept, in the sense that it requires a correlation with real shocks to firm value.
4
As of 26 February 2010, Basu (1997) has 1058 citations in Google Scholar and 205 citations in the Social Sciences
Citation Index, making it one of the most highly referenced papers in the modern accounting literature.
5
Ball, Kothari, and Robin (2000), Pope and Walker (1999), Giner and Rees (2001), Cao and Lee (2002), Ball, Robin, and
Wu (2000, 2003), Huijgen and Lubberink (2005), Jindrichovska and Kuo (2004), Tazawa (2003), Peek, Buijink, and
Coppens (2004), Chandra, Wasley, and Waymire (2004), Basu, Huang, Mitsudome and Weintrop (2005), Garca Lara and
Mora (2005), Brown, He and Teitel (2006), Bushman and Piotroski (2006), Gassen, Fuelbier and Sellhorn (2006),
Grambovas, Giner and Christodoulou (2006), Kwon, Yin and Han (2006), Mak, Strong, and Walker (2006), and Ball,
Robin and Sadka (2007).
6
Basu (1997), Ball, Kothari, and Robin (2000), Givoly and Hayn (2000), Holthausen and Watts (2001), Ryan and
Zarowin (2003), Sivakumar and Waymire (2003), Raonic, McLeay, and Asimakopoulous (2004), Grambovas, Giner and
Christodoulou (2006).
7
Basu, Hwang, and Jan (2001a).
8
Basu (1997), Ball, Kothari, and Robin (2000), Basu, Hwang, and Jan (2001b), Gul, Srinidhi, and Sheih (2002), Chaney
and Philipich (2003), Kelley, Shores, and Tong (2004), Garca Lara, Garca Osma and Mora (2005), Ruddock, Taylor,
and Taylor (2006), Krishnan (2005a, 2005b, 2007), and Qiang (2007).
9
Ball and Shivakumar (2005, 2008a), Burgstahler, Hail and Leuz (2006), Wang (2006), and Barragato and Basu (2007).
 3

(vii) to study the role of debt;11 (viii) to examine the role of accounting accruals;12 (ix) to study

management compensation, independent directors, and corporate governance;13 (x) to study the effect

of conditional conservatism on corporate investment and acquisitions;14 and (xi) as a function of

environmental uncertainty and the investment cycle.15 These studies analyze asymmetric

conservatism from a variety of perspectives, including costly contracting and agency theory, legal

institutions, and political and tax influences on financial reporting. The concept of asymmetric

earnings timeliness also has been utilized to increase our understanding of phenomena such as the

shape of the earnings distribution and the properties of analysts earnings forecast errors (Gu and Wu,

2003), the payoff function for analysts (Basu and Markov, 2004), and the properties of accounting

accruals (Ball and Shivakumar, 2006). The range and importance of these applications is testimony to

the pervasiveness of conservatism as a property of financial reporting, and also to researchers

confidence in the validity of their estimates of it. These studies regularly report differences in

conditional conservatism that are consistent with a variety of plausible hypotheses, which provides

added confidence in the validity of the estimates.

Despite its importance and popularity, the measure of asymmetric timeliness of earnings has

been challenged on several grounds. We attribute this largely to ambiguity arising from the absence

of a formal econometric analysis of the properties of its piecewise-linear regression specification.

That is, the Basu model has been assumed to be well-specified and to produce valid measures of

conditional conservatism, based largely on its intuitive appeal, or by informal analysis.16 The need

10
Francis, LaFond, Olsson, and Schipper (2004).
11
Ball, Bushman and Vasvari (2007), Ball, Robin and Sadka (2008), Wittenberg-Moerman (2008), Callen et al. (2009),
Hammermeister and Werner (2009), and Nikolaev (2009).
12
Basu (1997, Table 2), Ball, Kothari and Robin (2000, Table 6), and Ball and Shivakumar (2005, 2006).
13
Lubberink and Huijgen (2001), Beekes, Pope, and Young (2004), Cuijpers, Moers and Peek (2005), Leone, Wu and
Zimmerman (2006), Ahmed and Duellman (2007), and Garcia Lara, Garcia Osama and Penalva (2009).
14
Bushman, Piotroski and Smith (2007), Francis and Martin (2010).
15
Khan and Watts (2009).
16
Basu (1997) offers no formal proof. Increasingly formal analyses are offered over time in Ball, Kothari and Robin
(2000), Pope and Walker (1999), Ryan and Zarowin (2003), and Beaver and Ryan (2005).
 4

for a formal econometric analysis is heightened by several claims that the Basu asymmetric

timeliness coefficient is not a valid measure of conservatism, and indeed that it can be observed even

in the absence of conditional conservatism. The primary goal of this paper thus is to clarify the

econometrics of the asymmetric timeliness methodology and to stimulate its application in different

contexts. The more specific objectives of our study are threefold.

First, before we begin a discussion of the properties of the Basu model, we discuss the typical

research objective of the studies that employ it, which we believe to be a major source of

misconception in the literature. We therefore provide the more formal econometric analysis needed to

resolve these concerns about model specification. Specifically, we consider the case where returns are

endogenous to earnings and when the joint return-earnings distribution is asymmetric.

Second, and more importantly, we propose a model of the relation between accounting

income and economic income that is based on the salient properties of income recognition as it is

practiced in accounting. We then use the model to derive and analyze the earnings asymmetric

timeliness coefficient. The model distinguishes four components of information that financial

reporting rules and practices cause to be incorporated in accounting income at different points in

time. One information component is incorporated contemporaneously. This earnings component

could include news about present-period cash flows (i.e., the realized cash flow for the period versus

expected). It also could include news about future cash flows that is verifiable at low cost and can be

incorporated in income through accruals, such as news about accounts receivable collectability or

inventory levels. The second information component is incorporated contemporaneously or with a

lag, depending on its sign or magnitude. This earnings component is the source of conditional

conservatism, which arises because negative news about future cash flows is subject to a lower

accounting verification threshold than positive news (Basu, 1997, p.4). It could include news about

the present value of future-period cash flows from booked assets, including purchased assets in
 5

place and purchased intangible assets such as patents and goodwill, and also could include news

about the present value of future-period obligations such as lawsuit settlements. The third information

component in our model always is incorporated with a lag, such as news about unbooked rents or

growth options. The first three information components are assumed to affect security prices, even

when they are not recognized in current earnings (i.e., when disclosed via other information

channels). The model also has accounting income reflecting noise that reverses over time, due to

reversals of accounting errors. For example, miscounting inventory, or estimating uncollectible

accounts receivable with error, affects accounting income in successive periods with opposite signs.

We believe this is the first model of the relation between accounting and economic incomes, or

earnings and returns, to incorporate the salient properties of accounting recognition rules and

practices.

In this framework, the information components that are contemporaneously incorporated in

accounting income might cause returns, or vice versa. Nevertheless, for reasons outlined below we

are not concerned about causality, as distinct from association, so the direction or directions of the

earnings-returns relation are not at issue.

This framework allows a more formal derivation and improves intuitive understanding of the

Basu measure of conservatism, and of how it relates to and interacts with other attributes of a firms

information system. In particular, our analysis shows that Basu coefficient is not a structural (causal)

parameter that needs to be identified but rather is a function of more fundamental properties of the

accounting information system. Several criticisms of the Basu measure, in our view, are attributable

to a lack of understanding of this result.

Third, we extend and formalize the Roychowdhury and Watts (2007) analysis of the market-

to-book ratio. Our analysis develops a conjecture in Ball, Kothari and Robin (2000, p. 48), that

earnings timeliness is decreasing in the importance of information about growth options. We use our
 6

earnings-returns model to demonstrate that the asymmetrically timely loss recognition coefficient

decreases in the variance of information about growth options, holding other things equal, because

accounting recognition lags growth option news. Intuitively, the market-to-book ratio is likely to

proxy for the proportion of price changes associated with growth options, so a negative relation

between market-to-book ratios and asymmetric timeliness coefficients is expected. We view the

negative relation between the market-to-book ratio and the Basu asymmetric timeliness coefficient as

a property of income recognition in accounting, which is different from the perspective offered by

Roychowdhury and Watts (2007), even though the approaches make similar predictions. The model

also offers a number of additional empirical implications.

Section 2 begins with some observations about the implications of the research objective for

the econometric model relating accounting income and stock return, and proceeds with a more formal

econometric analysis and justification for model specification in Basu (1997). Section 3 outlines a

simple model of the income-return relation, an essential feature of which is the well-known

phenomenon that accounting income reflects considerable news about economic fundamentals with a

lag: that is, prices lead earnings. Section 3.3 derives the Basu asymmetric timeliness coefficient as a

function of information attributes. In section 3.4 we show that a negative correlation between the

market-to-book ratio and the asymmetric timeliness coefficient is expected empirically, and discuss a

number of empirical implications. Section 4 discusses some testable implications of conditional

conservatism that do not involve regressions of accounting income on returns. Section 5 reviews

challenges that have been raised to the validity of the Basu asymmetric timeliness coefficient. A short

summary and our conclusions appear in section 6.

 7

2. Econometrics of the Basu Regression

In this section, we clarify the econometrics of the Basu (1997) model. We start with a

discussion of whether estimating a regression of earnings on returns is appropriate, and whether this

gives rise to econometric issues. We then discuss whether partitioning on stock return, which itself is

driven in part by earnings information, is appropriate. Subsequently, we discuss whether there are

restrictive assumptions that are unlikely to be satisfied in practice, and in particular we show that

asymmetry in earnings and return distributions (skewness) does not invalidate the Basu (1997)

coefficient as a measure of conditional conservatism.

2.1 Is a regression of earnings on returns misspecified?

We start with a general observation that the appropriate econometric model specification

relating accounting income and stock returns depends on the researchers objective. When the

research objective is to explain a certain property of accounting income, as actually reported by firms,

the appropriate explained (or dependent) variable is accounting income, not stock return. In Basu

(1997), the property of accounting income being investigated is asymmetrically timely incorporation

of economic gains and losses in accounting income, so it is appropriate to specify accounting income

as the dependent variable and stock return (the proxy for economic income) as the independent

variable. 17

More formally, the objective of researchers using the model in Basu (1997) generally is to

understand the expected timeliness with which earnings reflect the information that becomes public

over a certain interval. Over a firms life, each dollar of information in stock returns translates into a

dollar of earnings, and vice versa. Over shorter time periods (e.g., years), however, there is no one-

to-one mapping between earnings and returns and thus, for modeling purposes, they can be viewed as

17
Stock return is the natural dependent variable when studying information effects on prices (e.g., when estimating
market efficiency or the effect of earnings announcements). Even then, one might be cautious in specifying earnings news
as the independent variable because it is measured with error, due to error in measuring expected earnings.
 8

random variables sharing a joint distribution. The question of interest is how timely accounting

earnings are in reflecting the information available to the market, and whether this timeliness differs

depending on the direction of economic news. Econometrically, the question of earnings timeliness

translates into the question of the conditional expectation of earnings (It), given a stock returns

realization (Rt); in other words, what portion of a given dollar of economic income over a period is, in

expectation, reflected contemporaneously in accounting earnings. Answering this question, as we

show, requires minimum restrictive assumptions, for example about the shape of the distribution, or

about the extent to which stock returns are caused by accounting income or, vice versa, about the

extent to which the accounting system directly reflects information provided to accountants by stock

market events. The latter distinction is, in fact, irrelevant as the identification of causal links is not an

objective, while the conditional expectation is.

One misconception is that the research objective in Basu (1997) and following studies is to

identify causality running from earnings to returns by estimating the inverse of the earnings response

coefficient. Dietrich et al. (2007) assume linearity in the return-earnings relation and propose the

following structural model:18

Rt = I t + t (2.1)

where, for simplicity, the error term t is independent of It and all variables have zero mean. Given

these assumptions, the OLS estimator of is unbiased. Reversing regression (2.1) makes it

impossible to identify (the inverse of) this coefficient, because the error term t in the OLS regression:

1 1
It = Rt t = Rt + t (2.2)

18
See also Beaver et al. (1997) and Beaver et al. (2008).
 9

is not independent of Rt and thus the OLS estimate of is a biased estimate of 1/ .19 Can one

conclude, based on this, that the regression of returns on earnings is misspecified if estimating 1/ is

not the research objective?

As discussed earlier, the goal of a Basu (1997) regression is to identify the conditional

expectation of earnings It given a return realization Rt, E ( I t | Rt ) . It is well known that:

E ( I | R) = arg min E ( I ( R))2 (2.3)

(R)

Given the functional form (.) of the conditional expectation (which should be guided by theory, as

we discuss later), minimizing the sum of least squared deviations from the conditional mean (i.e.,

running OLS when the function is linear in the parameters) must provide an unbiased estimate of the

conditional expectation function.20 Therefore, keeping the previously stated research objective in

mind, the least squares regression of earnings on returns indeed is appropriate. Such a regression

does not aim to or allow the researcher to identify the separate causal effects of earnings on

returns and vice versa, but it allows identifying the portion of the aggregated information in stock

returns that in expectation enters earnings in a given period.

We reiterate that when the research objective is to understand or estimate a property of

accounting income, namely, timeliness or asymmetric timeliness, then accounting income is the

appropriate dependent (i.e., explained) variable, and the fact that stock returns are in part caused by

accounting income simply is irrelevant. This is because we are interested in how the information

19
Dietrich et al. (2007) call this sample-variance-ratio bias.
20
Dietrich et al. (2007) assume a linear relation between returns and earnings, i.e., they assume away conservatism. Given
the linearity assumption, ( Rt ) = E ( I t | Rt ) = Rt . The solution to the least squares minimization problem defined
above is given by * = E ( Rt I t ) / E ( Rt ) . At the same time we can write I t = E ( I t | Rt ) + t = Rt + t where the
2

error term is mean independent of R (to see this, note that E ( | R ) = E ( I E ( R | I ) | R ) = 0 ). This implies a
weaker condition E ( R ) = E ( RE ( | R )) = 0 . Therefore, E ( R( I R )) = 0 and the true expectation parameter is
given by = E ( Rt I t ) / E ( Rt ) . In other words, least squares regression correctly identifies the conditional expectation
2

and thus the objective of the reverse regression analysis is met.

 10

about economic gains and losses is incorporated (or reflected) in accounting income, regardless of

whether the source of new information is accounting income itself. Moreover, in this research

context, models designed to estimate the separate causal effects of returns on earnings, and of

earnings on returns, are both unnecessary and misleading. 21

2.2 Does partitioning on returns result in a truncation bias?

Perhaps a more serious criticism of the model estimated in Basu (1997) is that partitioning

returns based on their sign may not be appropriate as it causes truncation bias due to the endogenous

nature of returns. For example, Dietrich et al. (2007) claim that such partitioning will result in

evidence of conservatism when in fact none is present. We revisit this claim, adopting for this

purpose those authors assumption of linearity in the return-earnings relation to rule out the presence

of conditional (i.e., news dependent) conservatism. Linearity implies the following functional form:

E ( I t | Rt ) = Rt . Further, consider partitioning the regression of earnings on returns on the sign of Rt

and thus estimating two separate regressions:

I t = 0 Rt + t | Rt 0 (Good news regression), (2.4a)

I t = 1Rt + t | Rt < 0 (Bad news regression) (2.4b)

The structural model in (2.1) assumed by Dietrich et al. indeed implies that E ( I tt | Rt > 0) 0

and E ( I tt | Rt < 0) 0 . Does this introduce sample truncation bias, and is this bias a function of the

shape of the earnings/returns distribution? To answer this question we need to evaluate the following

probability limits:

21
Short-window event studies such as Foster, Olsen and Shevlin (1984) and Bernard and Thomas (1990) show at least some
causality running from earnings to returns. However, we have known since Ball and Brown (1968) that this effect is small:
accounting income is primarily anticipated. Ball and Shivakumar (2008b) report that the average quarterly earnings
announcement accounts for approximately 2% only of annual stock return variance. Thus, even if the Dietrich et al. (2007)
claim that return endogeneity biases the Basu model slopes were valid, their conclusion that Basu (1997) asymmetric
timeliness coefficients are due to such bias would be highly unlikely. Return endogeneity simply is not important enough
in practice to be a substantial issue. Ryan (2006) offers a similar assessment.
 11

cov( Rt , I t | Rt 0) cov( Rt , I t | Rt < 0)

plim(0 )= and plim(1 )= (2.5)
var( Rt | Rt 0) var( Rt | Rt < 0)

It may seem that these probability limits are arbitrary constants and, depending on the shape of the

distribution, will not be the same (Dietrich et al. 2007). However, a closer look concludes otherwise.

First, we perform the following decomposition, which can be applied to any random variable:

I t = E ( I t | Rt ) + t . This implies E ( | R ) = E ( I E ( R | I ) | R ) = 0 , or t is mean independent of Rt .

Substituting this decomposition into the probability limits in (2.5) yields:

cov( Rt , E ( I t | Rt ) + t | Rt 0) cov( Rt , E ( I t | Rt ) | Rt 0)
plim(0 )= = , (2.6a)
var( Rt | Rt 0) var( Rt | Rt 0)

cov( Rt , E ( I t | Rt ) + t | Rt < 0) cov( Rt , E ( I t | Rt ) | Rt < 0)

plim(1 )= = . (2.6b)
var( Rt | Rt < 0) var( Rt | Rt < 0)

Recall that E ( I t | Rt ) = Rt , an assumption implying the absence of conservatism, and thus:

cov( Rt , Rt | Rt 0)
plim(0 )= = , and plim(1 ) = (2.7)
var( Rt | Rt 0)

Eq. (2.7) implies that conditional conservatism is identified correctly, and that it will not be

observed due to any bias arising from partitioning on Rt. That is, the Basu regression would

appropriately indicate the true absence of conservatism. The intuition for why sample truncation bias

does not arise in these circumstances is straightforward. Although the explanatory variable may be

endogenous, truncation here is with respect to the conditioning variable, and not the variable that is

being conditioned. It therefore does not introduce a mechanical negative correlation with the error

term, which in turn would have lead to truncation bias.22 The assumption E ( I tt | Rt > 0) 0 , while

true, is not relevant, but what is relevant is that E ( Rt | Rt > 0) = 0 . This, however, is easy to see

22
Truncation is depicted incorrectly in Dietrich et al. (2007, Figure 2), where the regression of earnings on returns
minimizes the horizontal (squared) deviations from the regression line, not the vertical deviations. After one notes this, it
immediately becomes clear from the graph that no truncation bias takes place.
 12

since: E ( Rtt | Rt > 0) = E ( E ( Rtt | Rt ) | Rt > 0) (smaller information set dominates) and

thus: E ( Rtt | Rt > 0) = E ( Rt E (t | Rt ) | Rt > 0) = 0 .

Note that we did not make any assumptions about the shape of the distribution. Therefore, the

result above implies that the Basu (1997) methodology is not biased by asymmetry/skewness in the

distributions of It and/or Rt. In the absence of conservatism, asymmetric timeliness defined as 0 1

equals zero in expectation, independent of skewness.

It also is helpful to examine the case when earnings actually are conditionally conservative. In

this case one should expect that, due to asymmetry in accounting rules and practices, the conditional

expectation function depends on the type and sign of economic news. One simple representation of

this is a piecewise-linear expectation function E ( I t | Rt ) = ( + 1{ Rt >0} ) Rt (this assumption might seem

ad hoc, but our analysis in the following section implies it is warranted). It simply implies that

earnings are expected to reflect economic news differentially, depending on their sign (consistent

with earnings being conditionally conservative). In this case, it follows that:

cov( Rt , E ( I t | Rt ) | Rt 0) cov( R,( + 1( R<0) ) Rt | Rt 0)

plim(0 )= = = , (2.8a)
var( Rt | Rt 0) var( Rt | Rt 0)

cov( Rt , ( + 1( R<0) ) Rt | Rt < 0)

plim(1 )= = + (2.8b)
var( Rt | Rt < 0)

Thus, the methodology in Basu (1997) correctly identifies asymmetric timeliness . We conclude

that this methodology is valid for estimating conditional conservatism, despite concerns expressed

that partitioning on returns causes truncation bias.

2.3 Is a piecewise linear regression of earnings on returns misspecified?

Dietrich et al. (2007, p. 97) go so far as to conclude that all results in the literature using

estimates of asymmetric timeliness are attributable to biased test statistics, and thus cannot be
 13

interpreted as evidence of accounting conservatism. The basis for this conclusion is the alleged

problems addressed in the two previous subsections: (i) accounting income has a causal effect on

returns, which allegedly biases the OLS estimate of conditional conservatism; and (ii) the piecewise

linear regression partitions on an endogenous variable which, in conjunction with asymmetry in

return-earnings distributions also allegedly leads to bias. We reach the opposite conclusion, and

believe the Dietrich et al. (2007) view is based on faulty econometric analysis.

3. A framework for interpreting regressions of earnings on returns

In this section, we develop a framework in which, as a result of accounting rules and

practices, information components are incorporated in accounting income differentially, depending

on their type. The model is based on realistic and intuitive assumptions about the incorporation of

information about economic value into accounting income, and demonstrates that under these

assumptions the Basu (1997) piecewise linear regression of accounting income on stock returns

provides valid estimates of conditional conservatism. The central assumptions are that accounting

recognition rules and practices cause earnings to lag price changes generally, and that conditional

conservatism in the form of different verifiability thresholds for gains and losses causes the expected

lag for price decreases to be less than for price increases. We formalize accounting recognition lags in

a simple one-lag model. We then use this model as a framework to show that Basu model parameters

validly capture the underlying construct of asymmetrically timely recognition of economic gains and

losses.

To facilitate tractability we assume throughout that capital markets are informationally

efficient and discount rates are constant through time. We believe violation of these assumptions

would not be central in typical studies using the Basu model. First, most studies use relatively long-

horizon returns (one year or one quarter) which, as argued in Ball et al. (2000), substantially mitigates
 14

concerns about the informational efficiency of security prices. Second, while changes in expected

rates of return through time naturally affect return variability (Fama, 1990), our primary objective is

to understand the covariability of firm-level earnings and returns due to firm-level information about

future cash flows.23 Therefore, we assume discount rates are constant, which means all return

variability stems from cash flow information.

The analysis is performed using continuously compounded growth rates (i.e., log growth

rates) in income and stock prices, for analytical tractability and simplicity. However, the intuition

from the analysis is equally applicable using other, conventional measures of growth rates or levels of

income and prices. Under the maintained hypotheses of market efficiency and constant expected

rates of return, stock prices follow a random walk, and thus growth in prices is permanent and

returns are serially uncorrelated (Bachelier 1900, Samuelson 1965, Fama, 1970; Campbell, Lo, and

MacKinlay, 1997). We therefore assume growth in price and all its components are permanent.

3.1 A model of the relation between earnings and returns

We propose what we believe to be the first model of the relation between accounting and

economic incomes to incorporate the salient properties of accounting recognition rules and practices.

The origins of the model can be traced back to Ball and Brown (1968), Beaver, Lambert, and Morse

(1980), Fama (1990), Kothari and Sloan (1992), Basu (1997), Kothari (2001), and others. We first

present a general form of the model and subsequently offer simpler cases.

Since we assume capital markets are informationally efficient, prices reflect all publicly

available information in a timely fashion. In contrast, accounting rules and practice emphasize

verifiability, objectivity, and conservatism, as reflected for example in the historical cost principle

and revenue recognition rules, and hence accounting income in any period incorporates some but not

23
Ball and Brown (1968), Basu (1997), Ball et al. (2000) and Ball et al. (2003) control for market-wide returns when
constructing firm-level news proxies. These excess returns are zero-sum in cross-section and hence in principle they
are diversifiable and riskless, somewhat finessing the issue of variation in expected returns.
 15

all the information that becomes publicly available during the period. The consequence is that

accounting income incorporates some information with a lag: that is, prices lead earnings. 24

Information in our model can be made public via earnings or any other information channel.

We do not implicitly or explicitly assume that accounting has no informational role in price formation

(or that equity prices would be the same in the absence of earnings announcements). Instead, we

assume that to the extent the market responds to reported earnings, it understands its components. In

particular, we assume the market distinguishes the components that reflect information made public

in prior periods from those that reflect current-period information.

Note that asymmetric accounting conservatism should also apply to the way private

information revealed by the manager is incorporated in earnings (e.g., private information about bad

debt expense). If the private information that a manager intends to communicate to the market in a

particular period does not meet a certain verifiability standard, that information must be

communicated via channels other than earnings. If the verification standard is asymmetric, private

bad news is more likely to be conveyed through earnings than private bad news, and private bad news

is more likely to be conveyed through other channels such as management forecasts.

The model decomposes the total revision in security price (i.e., stock return) into three

components, which are incorporated in accounting income differently:

Rt = xt + yt + gt (3.1)

It = xt + wt yt + (1 wt-1)yt-1 + gt-1 + t - t-1 (3.2)

where subscripts t and t1 refer to time periods.

Rt = security return expressed as the continuously compounded growth rate in price, i.e., the
natural logarithm of the firms economic income deflated by beginning-of-period price;

24
The fact that prices lead earnings is clear from the graphs in Ball and Brown (1968), Foster, Olsen and Shevlin (1984) and
Bernard and Thomas (1990), from longer-horizon studies such as Beaver, Lambert and Morse (1980) and Kothari and Sloan
(1992), and from even a casual reading of the financial press.
 16

It = accounting income, expressed as the continuously compounded growth rate in earnings;

xt = portion of the total growth rate in security price Rt that invariably is contemporaneously
captured in accounting income, It;

yt = portion of the total growth rate in the security price that is not contemporaneously captured in
It unless required by conservative accounting;

wt = is an indicator variable that takes the value of one when conservative accounting rules and
practices lead to recognition of y in the current period; and

gt = portion of the total growth rate in the security price that never is contemporaneously captured
in It, but always is incorporated with a lag;

t = noise in accounting earnings that reverses the next period.

For analytical tractability, we assume xt, yt, gt, and t are stationary, symmetric,25 and time-

independent (i.e., serially uncorrelated) 26 random variables whose variances are denoted by x2, y2,

g2 and 2, respectively, and:

corr(xt, yt) = xy > 0, corr(xt, gt) = xg > 0, and corr(yt, gt) = yg > 0 (3.3)

In this model, financial reporting rules and practices lead to accounting income always

contemporaneously incorporating the information component xt, regardless of its sign or magnitude

(i.e., without conditional conservatism). The intuition is that this component represents the least

costly source of information to verify, and thus it invariably is recognized in the same period as it

affects returns. This type of information could include current-period news about current-period cash

flow, such as learning the actual realizations of current period revenues and expenses in comparison

with their expectations. It also could include news about future cash flows that is verifiable at low

cost, and that is incorporated symmetrically in accounting income via working capital accruals.

Symmetrically low-cost verification typically applies to current operating-cycle information, such as

25
We discuss the case of asymmetric distributions in Section 2, and show that asymmetry does not invalidate the
specification in Basu (1997).
26
As expected returns are assumed to be constant, market efficiency requires returns and thus their components to be
independent over time.
 17

accounts receivable, accounts payable and inventory information. For example, accruals are used to

adjust current-period cash flow for both increases and decreases in closing inventory relative to

opening inventory, because they are approximately equally low in cost to verify. Similarly, cash

collections from customers are adjusted for both increases and decreases in accounts receivable.

These working capital accruals incorporate into accounting income current information about future

cash flows. For example, other things equal an increase in closing inventory is information that less

cash will be spent on purchasing inventory in future periods, and it is verifiable at low cost regardless

of its sign. In some limited circumstances, symmetrically low-cost verification can apply to long-

cycle information as well, an example being index funds, in which gains and losses on long term

investments in traded stocks are symmetrically low-cost to verify and in practice are accounted on a

daily basis.

The second component of stock return, yt, is incorporated in accounting income either

contemporaneously or with a lag, depending on the accounting operator wt. The intuition here is that

yt represents information that is costly to verify and, because the verification threshold is lower for

negative than for positive news (Basu 1997, page 4), this information is incorporated asymmetrically.

Such information could include the current-period revision in the expectation of unrealized future-

period cash flows from booked assets, including purchased assets in place and purchased intangible

assets such as patents and goodwill. This component also could include news about the present value

of future-period cash outflows, such as lawsuit settlements. Using accrual accounting to bring

forward shocks to expected future cash flows is known as timely gain and loss recognition.

Conditional conservatism implies that w is more likely to be triggered by bad news (adverse shocks

to expected future cash flows) than good news, and hence that loss recognition generally is timelier

than gain recognition. Thus, w can be thought of as an accounting write-down indicator (specified

more fully below).

 18

In the event that timely recognition is not triggered and thus the return component yt is not

contemporaneously incorporated in income, it is incorporated with a lag, the intuition being that some

revisions in expectations of future cash flows are not reflected in accounting income until the actual

cash flow realizations occur. Conditional conservatism implies that incorporation with a lag is more

likely for good news than bad. The effect of asymmetrically delayed incorporation of information is

that, in a two-period model, the total effect on current period income of revisions in cash flow

expectations is wt yt + (1 wt-1)yt-1.

In practice, the extent of the timely recognition asymmetry is determined by a number of

factors. Identification of these factors is the principal objective of the extensive literature surveyed in

our Introduction. The factors studied include economic incentives, debt and compensation

contracting, governance, GAAP, regulation, and taxes (e.g., Watts and Zimmerman 1986, and Watts

2003a,b), in addition to properties of information such as verifiability. Since our objective is not to

provide an equilibrium model of the extent of conditional conservatism, but rather to investigate the

validity of the Basu (1997) measure of conditional conservatism used in this literature, our model

takes the extent of the asymmetry as a given, and studies the properties of its measurement. In our

setting, timely recognition is triggered when y is below an exogenous threshold c.27

The third component of stock return, gt-1, invariably is incorporated in accounting income

with a lag. One source of this component would be revisions in the value of growth options. Because

they by definition are not booked as assets on balance sheets, shocks to firms growth opportunities

ultimately are incorporated in earnings only when the associated cash flows are (or are not) realized.

A related source of this component would be revisions in expectations of future monopoly rents

(Roychowdhury and Watts, 2007).

27
Alternatively, write-downs can occur with probability p which can be a function of y, i.e., Pr(w=1| y) = p(y)).
 19

The model also has accounting income incorporating uncorrelated noise that reverses over

time. One source of this earnings component would be accounting errors arising from imperfect

accounting accruals. Errors that reverse over time include miscounting inventory, which affects

current and future cost of goods sold and hence earnings with opposing signs. Other examples are

errors in estimating uncollectible accounts receivable, errors in forecasting deferred tax liabilities, the

effects of using historical-cost interest rates on debt, and errors in estimating assets useful lives.

Because accounting errors reverse over time, in our two-period model the error term is reversed out

in the following period.28 For tractability we assume accounting error is uncorrelated with other

variables, and thus we ignore earnings management or smoothing, which could produce a non-

random error component that is negatively correlated with other components.

In this formulation, yt-1 and gt-1 are the two sources of delayed recognition of economic

income in accounting earnings. They generate the anticipated or stale component of earnings growth

whose value consequences are reflected in price prior to the period in which they are incorporated in

accounting income. They are uncorrelated with current period return Rt, which is influenced only by

information arriving contemporaneously. We view this lagged recognition of some components of

stock return as a natural feature of the income recognition process in accounting, which can be

measured and investigated (for example, as a function of managers incentives, or countries

economic and political institutions). As will become evident, we do not view or model the

recognition lag as a measurement error issue, but as a property of accounting income that the

researcher wishes to estimate.

While yt-1 and gt-1 are assumed to be uncorrelated with xt, yt, and gt, and, therefore, with Rt,

there are economic reasons to expect a positive correlation among xt, yt, and gt. Recall that all three

28
Knowledge of t-1 would help predict earnings, but not returns. We assume the market unravels the current accounting
error t and does not react to it. This assumption can be relaxed without altering our conclusions.
 20

components are a consequence of economic news affecting investors cash flow expectations.

However, only the news generating the stock price growth rate component xt, and possibly also yt

(i.e., when wt = 1), is incorporated in accounting income contemporaneously, whereas news

generating the return component gt finds its way into accounting income in the following period.

Some news is likely to be common to the three components. For example, an unexpected increase in

sales in the current period could affect current earnings and hence xt, but simultaneously it could

affect the expectation of sales growth in the next period and hence the growth rate yt. The news then

is partly reflected in both current returns and earnings (xt or yt), and in current returns and future

earnings (gt or yt). The positive correlation between xt and yt provides the basis for the piecewise-

linear accruals-cash flow models in Ball and Shivakumar (2006).

Despite its simplifying assumptions, we believe the earnings process modeled in (3.2)

captures the important properties of asymmetric accounting recognition rules and practices, and how

they affect reported earnings. We next use this framework to investigate the properties of the Basu

measure and develop empirical implications for the benefit of future research.

3.2 Timeliness of accounting income in a symmetric return-earnings relation

We start with a baseline case where timely gain or loss recognition does not take place (i.e., wt

= 0), and hence accounting treats y and g equivalently, and examine the econometric properties of

regressing accounting income on security return:

It = + Rt + t (3.4)

The regression slope reflects the extent to which accounting income contemporaneously captures

economic income. Let denote the OLS estimate. It is well-known that:

plim = cov(It, Rt)/var(Rt)

= cov(xt + yt-1 + gt-1 + t - t-1, xt + yt + gt)/var(xt + yt + gt)

 21

= (var(xt)+cov(xt, yt + gt))/ (var(xt)+2cov(xt, yt + gt)+var(yt + gt)). (3.5)

Equation (3.5) obtains because the lagged income components, yt-1 and gt-1, do not correlate

with xt, yt, and gt. In equation (3.5), the estimated slope coefficient from a regression of earnings on

contemporaneous returns increases in the timeliness of earnings, var(xt)/var(yt + gt), which is the ratio

of the information in returns that accounting income incorporates in a timely fashion to the

information it incorporates with a lag. Equivalently, as timeliness increases the contribution to

current accounting income It of stale cash flow news, yt-1 and gt-1, diminishes. In this model of the

accounting process, accounting income is 100% timely with respect to economic income if var(yt)

and var(gt) both are zero, in which case the estimate of converges to one and accounting income

perfectly correlates with stock returns.

Our analysis incorporates what we believe to be a realistic scenario, that xt, yt, and gt are

positively correlated. While we would like to continue the analysis under that assumption, the

algebra becomes tedious, so we relegate derivations allowing correlation to the appendix.

3.3 Econometrics of the Basu asymmetric timeliness coefficient and conservatism

Basu (1997) estimates the asymmetric timeliness coefficient from the regression model:

I t = 1 + 2 Dt + 1Rt + 2 Dt Rt + t (3.6)

where Dt = 0 if Rt 0, Dt = 1 if Rt < 0, and 2 is the asymmetric timeliness coefficient. 2 then is the

incremental coefficient on negative return (the proxy for negative economic income), and is predicted

to be positive because conditionally conservative accounting incorporates negative economic income

into accounting income sooner than it incorporates positive economic income.

Let 2 denote the OLS estimate of 2 from equation (3.6). Then we have:

plim2 = 2 1 , (3.7)
 22

cov( I t , Rt | Rt 0) cov( I t , Rt | Rt < 0)

where 1 = and 2 = .
var( Rt | Rt 0) var( Rt | Rt < 0)

Our model (3.2) of accounting income recognition assumes that timely recognition is

triggered when the unobserved information component y falls below a threshold value, such as zero.

However, the Basu regression model (3.6) conditions the regression slope on total return R, which the

researcher observes. We now show that the incremental regression coefficient 2 = 0 if earnings is

not conditionally conservative, and 2 > 0 if earnings is conditionally conservative.

3.3.1 Basu regression coefficients in the absence of conditional conservatism. In the

absence of conditional conservatism, accountants symmetrically do not recognize any of the

information in yt in the current period t, even if it indicates an adverse shock to future cash flows, and

thus wt always is 0. It is easy to see that the assumption of symmetry in the distributions of xt, yt, and

gt implies that var(Rt |Rt 0) = var(Rt |Rt < 0), and cov(It, Rt |Rt 0) = cov(It, Rt |Rt < 0). It

immediately follows that plim 2 = 0. Thus, under the null hypothesis of no conditional conservatism,

the Basu coefficient is zero, which indeed should be the case if the model is well-specified and the

coefficient is a valid measure of conditional conservatism.

3.3.2 Basu regression coefficients with conditionally conservative accounting. We next

derive the estimate for the 2 coefficient under conditionally conservative accounting, and then show

that it behaves in a predictable fashion as a function of the firms information environment. We begin

by considering a simpler case in which the information component g is zero. Then:

cov( xt + wt yt + (1 wt 1 ) yt 1 + t t 1 , xt + yt | Rt 0)
1 = (3.8a)
var( Rt | Rt 0)

cov( xt , xt + yt | Rt 0) + cov( wt yt , xt + yt | Rt 0)
= (3.8b)
var( Rt | Rt 0)
 23

Equation (3.8b) follows from the assumption that all components of stock returns (and thus total

return) are independent over time. Similarly:

cov( xt , xt + yt | Rt < 0) + cov( wt yt , xt + yt | Rt < 0)

2 = (3.9)
var( Rt | Rt < 0)

Now recall that (in this section) we make the assumption that the distributions of earnings and

returns are symmetric. While this is a simplifying assumption, it is necessary to keep the analysis

tractable. As we discuss in Section 2, asymmetry in the distributions of return and earnings does not

per se bias the analysis in Basu (1997). The symmetry assumption implies var(Rt |Rt 0) = var(Rt |Rt

< 0) and thus we have:

cov( wt yt , xt + yt | Rt < 0) cov( wt yt , xt + yt | Rt 0)

plim2 = (3.10)
var( Rt | Rt 0)

Henceforth we drop the subscript t for expositional ease, and all variables are measured at time t

unless explicitly indicated by the subscript t1.

We model conditional conservatism as recognition of y in the current period only when it is

sufficiently bad news. We assume w = 1 if y < c and zero otherwise, where c is a threshold below

which current period recognition is required (e.g., through impairment). In an unrealistic limiting

case where c = + , timely recognition of y always occurs and there is symmetric recognition of

good and bad news. Thus, plim 2 = 0. The same result follows in the limiting case of c = , where

timely recognition of y never occurs. We set the threshold c to zero to be in line with adage

anticipate no profits, but anticipate all losses and to reconcile with the empirical formulation of the

Basu (1997) regression. In practice, c could be expected to be negative and close to zero. To the best

of our knowledge, this assumption is not restrictive and does not bias the results.

To show that 2 is positive under these assumptions, it suffices to show that:

cov(wy, x + y | R < 0) cov(wy, x + y | R 0) > 0 (3.11)

 24

Equation (3.11) can be rewritten as:

0 y 0 +
= y ( x, y | R < 0)dxdy y y ( x, y | R 0)dxdy +
2 2

 
 

1 2
(3.12)
0 y 0 + 0 +
+ xy ( x, y | R < 0)dxdy y xy ( x, y | R 0)dxdy + E ( x + y | R 0) x ( x, y | y < 0)dxdy
 


 
 

3 4

where ( x, y) is a bivariate normal density.

The integrals can be evaluated explicitly, and under normality (see Appendix for derivation):

(
= 2 1 y x 1 xy2 1 + arctan( ) 1 + 2 , ) (3.13)

y + x xy2
where = (0, +) .
(1 xy2 )0.5 x

It further can be shown that for any positive , the expression for is positive (to see this consider

the intervals (0,1] , (1, 3] , and ( 3, +) ). This result implies the Basu incremental coefficient on

negative returns reflects the existence of conditional conservatism in accounting income.29

These results demonstrate that the Basu coefficient 2 is not driven by a single structural

parameter, but rather is a function of conditional accounting conservatism and other attributes of the

information environment, including information about growth options and rents, discussed next.

3.4 Conditional conservatism, market-to-book, and empirical implications

Collins and Kothari (1989) and Easton and Zmijewski (1989) document that the market-to-

book ratio correlates with return-earnings regression slope coefficients, though they do not study

asymmetric timelines. Ball, Kothari and Robin (2000, p. 48) conjecture that the proportion of

growth options relative to assets in place influences Basu asymmetric timeliness coefficients. This

conjecture is based on the rule of thumb that market value of equity can be viewed as the sum of

29
See also Pope and Walker (1999), Ryan and Zarowin (2003) and Beaver and Ryan (2005).
 25

assets in place and the value of growth opportunities (Brealey et al., 2006, pp. 72-76; Ross et al.,

2005, p. 120). Stock return then reflects news about the values of both assets in place and growth

opportunities. Pae et al. (2005), Givoly et al. (2007) and Roychowdhury and Watts (2007) confirm

the conjecture empirically, reporting evidence that the asymmetric timeliness coefficient is negatively

correlated with the market-to-book ratio, a widely used proxy for the importance of growth options.

We now examine the conjecture econometrically.

Consider the case where the (variance of) the information component g in economic income is

non-zero. Current-period accounting income then includes a component that represents a linkage to

past positive shocks to the firms growth opportunities, as in equation (3.2) where shocks to growth

opportunities show up in accounting income with a simple one-period lag.

cov( y , x ) + cov( y , g )
Define z x + g , z2 var( z ), and corr( z , y ) = .
var( y ) var( x + g )

In this set up with innovations to growth opportunities, the Basu asymmetric timeliness coefficient is

given by (see appendix for derivation):

cov(wy, x + y + g | R < 0) cov( wy, x + y + g | R 0)

plim2 =
var( R | R 0)

=
2 y z 1 2
( 2) var( R)
(1 + arctan( )
1 1 )
1 + 12 > 0 (3.14)

y +z
where 1 = (0, + ) .
(1 2 ) 0.5 z

Several limiting cases are of interest. To avoid cumbersome notation, here we loosely refer to

plim 2 as 2 . As the correlation coefficient between z and y approaches 1, we have:

y
lim 1 2 = (3.15)
y +z
 26

This result is intuitive and suggests the asymmetric timeliness coefficient depends on the fraction of

the variance of y in the total variance of returns, as one would expect when the components are

perfectly correlated. 30 As one also would expect, lim y 2 = 1 , and lim y 0 2 = 0 . In the first case

the variation in y subsumes all other components, while in the second there is no role of conditionally

conservative accounting. Finally, we have lim z

2 = 0 , and lim z 0
2 = 1 . These mirror the

previous results, and also are intuitive. The limiting cases suggest the Basu specification captures the

extent of accounting conservatism in a meaningful and intuitive way.

We now show that the Basu asymmetric timeliness coefficient declines in the variance of

growth opportunities, and thus is expected to be negatively correlated with the market-to-book ratio.

The derivative of 2 with respect to z is:

2
=
2 y 1 2
z ( 2)( y + 2 y z + z )
2 2 2 (
( y z y2 z y ) arctan( ) y z 1 2 , )
(3.16)

where = 1 + arctan( ) 1 + 2 > 0 . This expression is negative for economically meaningful

values of the parameters and the expression for 2 is symmetric in x and g . Since z is

increasing in both x and g , it follows that, holding the correlation coefficient constant, 2 is

decreasing in x and g . This result is intuitive, in that the x and g components of revision in price

are treated symmetrically in financial reporting (fully incorporated and fully ignored, respectively),

whereas the y component is not (incorporated only if below a threshold).

30
Equation (3.14) seems to suggest that 2 approaches zero as approaches 1. This, however, is not the case since
1 tends to plus infinity in this case.
 27

Figure 1 graphs the asymmetric timeliness coefficient as a function of g and y . Other

parameters are fixed at the following levels: xz = xg = 0.3 and x = 0.2 .31 As expected, the

coefficient increases in the variance of y (reaching 1 in the limit), that is as timely loss recognition

becomes more important. It also follows from Figure 1 that asymmetrically timely loss recognition

decreases in the variance of g (information about growth options), reaching zero in the limit. This

also is intuitive. Since changes in growth expectations g are not captured contemporaneously in

accounting income, in a regression of accounting income on stock returns the variability due to

changing growth prospects dampens the coefficient on returns, consistent with the conjecture in Ball,

Kothari and Robin (2000).

Figure 1: 2 coefficient as a function of g and y

31
Analogous results obtain when other parameters are fixed at different levels, including correlation coefficients of zero.
For illustration, we therefore report the results only under one set of parameters.
 28

Empirical implications. As Figure 1 shows, the incremental coefficient on negative returns

decreases in the relative variance of shocks to growth opportunities. This variance, var(g), is expected

to be increasing in the magnitude of those opportunities. Because growth opportunities are less likely

to be recorded on firms books than other assets, firms with higher market-to-book ratios are likely to

experience relatively larger shocks, positive or negative, to their growth opportunities. Indeed,

empirically return variability increases in market-to-book, controlling for firm size.32 Consequently,

the market-to-book ratio acts as a proxy for the extent to which revision in price is associated with

revision in booked versus unbooked items, but does not per se determine the Basu slope. This

implies that market-to-book need not be used as a control variable (depending on the question at

stake) but rather can be used as an instrument for the degree of conservatism, in line with Khan and

Watts (2009).

Another implication of the graph above is that the relation between the Basu slope and the

market-to-book is non-linear, suggesting that different attributes of the information environment

interact with each other. This prediction is straightforward to test empirically. Finally, designing a

powerful test to detect conditional conservatism can be a challenge for high market-to-book firms,

because timely loss recognition is harder to detect for high growth firms and firms with larger

portions of unbooked intangible assets. This, however, does not mean that conditional conservatism

is absent, or that Basu model is mis-specified. To detect conditional conservatism in such companies,

it is helpful to identify settings or an economic event where firms experience substantial shocks to

assets in place (rather than growth opportunities), since timely loss recognition is increasing in the

variance of y (despite frequent shocks to growth options).

The result that asymmetric timeliness is decreasing in x also is intuitive in our setup and

represents another empirical prediction. This implies that companies with short operating cycles and

32
For example, Lewellen (1999) and Ali, Hwang and Trombley (2003).
 29

short investment cycles, or companies that have a substantial realized component of economic

income (such as mature companies) should exhibit lower levels of timely loss recognition. This

reconciles with the evidence reported in Khan and Watts (2009), and also with their predictions about

conditional conservatism increasing in environmental uncertainty, which we expect to be inversely

related to the variance of x and positively related to the variance of y (due to different verification

requirements for earnings components that are more uncertain and more costly to verify).

The model also implies that asymmetric timeliness declines as one extends the earnings

horizon. In line with this result, Basu (1997, 5.1) shows that the timeliness asymmetry declines with

the length of the accounting period (he studies periods of one to four years) and Roychowdhury and

Watts (2007) show that the negative association between growth opportunities and the asymmetric

timeliness coefficient diminishes as the return horizon is extended back. Intuitively, as the return

period is extended back, current-period earnings reflects more of the growth opportunities at the

beginning of the period. Further, shocks to growth opportunities are more likely to flow through the

financials over longer horizons. One can think of these as increasing the variance of x relative to the

variance of the other components of I, and hence the asymmetric timeliness coefficient is attenuated.

This is an implication of the more general proposition that, under clean surplus accounting,

economic and accounting incomes converge as the observation interval increases.

Finally, we note that Pae et al. (2005), Givoly et al. (2007) and Roychowdhury and Watts

(2007) cast the negative correlation between market/book and the asymmetric timeliness coefficient

as a measurement error problem, whereas our analysis views it is a fundamental property of income

recognition in accounting interacting with properties of firms. We return to this important distinction

in Section 5. We argue that for the purposes of studying actual reporting behavior by firms, the true

level of conditional conservatism in reported earnings is a decreasing function of market/book.

Equivalently, viewing this as a measurement error problem is equivalent to viewing accounting

 30

income as actually reported by firms to be an error-laden version of some non-existent alternative

notion of income.

3.5 Asymmetric timeliness and correlation among components of earnings innovations

The literature is silent on how asymmetric timeliness is affected by complementarities or

correlations among the information components that drive returns but are reflected in earnings in

different periods. We use equation (3.14) to plot 2 as a function of the correlations xg and yg (we

do not separately consider yx since it is easy to see that its effect is the same as that of yg ). Figure

2 presents the results when other parameters are fixed at levels: xy = 0.3 and x = y = g = 0.2 .

Figure 2: 2 coefficient as a function of xg and yg

As the figure shows, asymmetric timeliness increases in yg , the correlation between y and g,

and it decreases in xg , the correlation between x and g. Intuitively, the first result obtains because in

our model y exhibits timely recognition, but the researcher observes total return Rt, which also
 31

contains other information components. As the correlation between the return components y and g

strengthens, the Basu models negative-return variable improves as an indicator of the negative news

incorporated in accounting income contemporaneously, and hence the asymmetric timeliness

coefficient increases. Similarly, the second result obtains because variation in the sum of the

symmetrically-recognized x and g components becomes more pronounced due to their covariability,

and the conditional conservatism asymmetry becomes less salient. Thus, the correlations and

interactions between different information components are expected to have an effect on timely loss

recognition empirically.

3.6 Effect of recognition lags on accounting income variability relative to returns

A large body of literature in finance and economics starting with Shiller (1981) argues that the

volatility of stock returns is large relative to that of dividends and accounting income.33 One

interpretation of this result is that prices have excessive volatility compared to underlying

fundamentals, due for example to investors overreacting to economic news or due to frequent

changes in discount rates. We show that the volatility of stock returns is expected to be large relative

to that of accounting income due to the way accounting income is calculated. The lack of timeliness

of accounting income makes it a function of both current and lagged stock returns, and hence the time

series of accounting income is smoother than that of returns.

In our model, the components of the firms total economic growth rate that are incorporated in

growth in accounting income in a timely and lagged fashion are xt and (yt-1 and gt-1), respectively. A

quantitative measure of timeliness of accounting income therefore is the relative variances of xt and

(yt-1+gt-1). The higher the variance of xt, other things equal, the timelier is accounting income in

33
See also Campbell and Shiller (1988), Fama (1990), Campbell (1991), Campbell and Ammer (1993), Kothari and
Shanken (1992), Collins et al. (1994), Shiller (2000), Vuolteenaho (2002), and Hecht and Vuolteenaho (2006).
 32

capturing the information in economic income.34 Recall that xt is intended to capture information that

is symmetrically low-cost to verify, such as actual realizations of current-period cash flows and

working capital accruals, and that consequentially it is incorporated in income regardless of its sign

or magnitude. Because yt-1 is simply the lagged value of yt and the process is stationary across time,

their variances are equal. Moreover, all information affecting the stock price in the current period

(i.e., xt, yt, and gt) is not correlated with the stale components yt-1 and gt-1 of earnings growth. To

simplify the presentation, and without loss of generality, we restrict the variances of growth options

gt and accounting error t to zero. The variance of accounting income then is:

var(It) = var(xt + wtyt + (1 - w t-1)yt-1)

= var(xt) + 2cov(xt, wtyt) + var(wtyt) + var((1-wt-1)yt-1)

= var(xt) + cov(xt, yt) + 2var(wtyt). (3.17)

The last result follows because cov(xt, wtyt) = E(wtytxt) = E(ytxt)/2 = cov(xt, yt)/2, and var(wtyt-1) =

var((1-wt-1)yt-1), assuming symmetry and stationarity.

The variability of returns also is a function of the variances of its components, xt and yt:

var(Rt) = var(xt + yt)

= var(xt) + 2cov(xt, yt) + var(yt). (3.18)

Comparing equations (3.17) and (3.18), stock return is more variable than accounting income

because it incorporates the positively-correlated shocks xt and yt, whereas accounting income

incorporates xt, but (unless there is timely loss recognition) incorporates the other as an uncorrelated

shock yt-1 from the previous period. The sum of the uncorrelated shocks xt and yt-1 to accounting

income is less variable than the sum of the positively-correlated shocks xt and yt to stock return. In

34
This is due to our casting differences between accounting and economic income as a property of income recognition
rules in accounting, and not as a measurement error issue. Consequently, differences in the amount of information
conveyed in the two variables directly map into differences in their true variances. In other words, the variances need not
be interpreted as indicators of measurement error.
 33

the absence of a positive correlation between the xt and yt components, it is straightforward to show

that the variability of returns exceeds that of accounting income.

This result is clearer in the reasonable case where the loss recognition threshold c = 0, which

implies var(wtyt) = var(yt)/2 - E(yt |yt < 0)2. Then:

var(It) = var(xt) + cov(xt, yt) + var(yt) - 2E(yt |yt < 0)2

= var(Rt) - cov(xt, yt) - 2E(yt |yt < 0)2 (3.19)

A consequence of accounting income being less timely than stock returns in reflecting

economic income, due largely to historical cost accounting, therefore is a smoother time series for

earnings than returns. The effect is evident in our two-period model, and would be magnified in a

more realistic multi-period analysis.

The difference between dividend and return volatilities is expected to be even greater than is

evident for earnings. Corporations smooth dividends as a function of past accounting income

(relevant evidence dates back to Lintner, 1956, and more recently Ball et al., 2000, and Sadka, 2007).

To the extent that dividends are a weighted sum of current and past accounting income, they will

exhibit an even smoother time series than accounting income, and hence than stock returns.

Finally, changes in expectations about growth options will affect stock prices, and will thus

generate volatility, but some of these changes will never affect accounting earnings or dividends. For

example, an increase in growth prospects in one year one can be offset entirely by a decrease in the

following year. Since the changes are offsetting, they will not affect accounting earnings but will

affect the volatility of stock prices.

Empirical implications. In general, the literature on allegedly excess volatility in stock

returns does not explore the implications of prices leading earnings and dividends for their relative

volatilities, and consequently tends to over-estimate excess volatility. The implicit assumption in

the excess volatility literature that income and dividend volatilities provide a valid benchmark for
 34

return volatility is not consistent with the economic roles and fundamental properties of accounting

income and dividend distributions. The time series of accounting income and dividends are smoother

than prices because they convey the information in price revision with a lag. An interesting empirical

question, in our view, is the extent to which the volatility of stock returns relative to the volatility of

earnings is explained by the properties of accounting earnings, such as timeliness and conservatism,

as distinct from extant explanations such as investor irrationality or information-processing biases,

and time-varying expected returns.

4. Using Other Measures of Conditional Conservatism to Confirm Validity

Our analysis of the incorporation of information about economic value into accounting income

concludes that the Basu (1997) asymmetric timeliness coefficient from a piecewise linear regression

of accounting income on stock returns provides valid estimates of conditional conservatism. An

alternative approach to establishing validity is to compare the predictions based on the asymmetric

timeliness coefficient with those of alternative measures of conditional conservatism.

Starting with the time-series measure of Basu (1997), the literature has developed alternative

measures of conditional conservatism that do not involve regressions of accounting income on stock

returns, in large part to address concerns about the validity of returns-based asymmetric timeliness

coefficients.35 At least three alternative research designs use proxies for positive and negative

economic news based on variables other than market returns. The results based on these measures are

similar to those from returns-based measures, and also reconcile in a predictable way with many

economic hypotheses, thereby providing useful evidence on the validity of the returns-based

asymmetric timeliness measure. We discuss their strengths and weaknesses this section.

35
See discussion in Ball, Kothari and Robin (2000, p. 48) and Ball, Robin and Wu (2003, p.255).
 35

Basu (1997, Table 3) conducts a piecewise linear regression of change in earnings on prior-

period change in earnings, and shows that negative earnings changes are more transitory than positive

changes (i.e., more likely to revert). This asymmetry measure is based on the notion that economic

income is uncorrelated over time, which implies successive changes in economic income are

negatively correlated. Conditional conservatism therefore implies that changes in accounting

earnings are more likely to revert when they are incorporating negative economic income than when

incorporating positive economic income, and hence negative earnings changes are more likely to

exhibit negative first-order autocorrelation. In this time-series measure, the implied proxy for the sign

of economic income is change in earnings, not stock return.36

The time-series asymmetry measure avoids problems one might ascribe to using stock returns

as a proxy for economic income, but it has two potential deficiencies of its own. First, because it does

not condition on returns, it only measures whether negative earnings changes are more transitory than

positive changes, and does not measure timeliness. For example, similar results would be obtained if

all firms earnings delayed the recognition of gains and losses by an arbitrarily long amount of time.

Second, the time-series asymmetry measure does not distinguish between true economic income and

noise in accounting income, both of which are transitory. 37 Despite these potential deficiencies,

Basus (1997, Table 3) results from this alternative asymmetry measure are remarkably similar to

those from his returns-based measure. Subsequent studies reach much the same conclusion, both in a

range of countries,38 in private and public firms,39 and in the context of auditor liability.40

36
It is important that this variable is specified as earnings changes, not levels as in Hayn (1995), primarily because the
construct for which it proxies is new information. Another advantage of a changes specification is that it is substantially
less susceptible than a levels specification to survivorship-induced reversion bias, because a decrease in earnings does not
imply the firm has negative earnings.
37
Random errors in balance sheet amounts affect earnings in successive periods with opposing signs, inducing negative
serial correlation in earnings levels. Balance sheet amounts such as accounts receivable valuations, marketable security
valuations, derivatives valuations, loss provisions and inventory counts inevitably are not perfectly accurate, and
subsequent events cause them to be revised. For example, under-forecasting future loan losses over-estimates the balance
sheet value of receivables, adding positive error to current-period earnings and negative error to future earnings.
38
Ball and Robin (1999) and Ball, Robin and Wu (2003, Table 4).
 36

Similar observations can be made about viewing the extent of left skew in accounting income

(relative to the skew in stock returns and cash flows) as an indicator of conditional conservatism.

Conditional conservatism produces left skew in accounting income by incorporating bad news in

larger amounts and lower frequency than good news, which is more likely to be spread out over time.

However, left skew also could be generated, for example, by write-offs against earnings that are

uncorrelated with stock returns, at any lag. Nevertheless, predictions that relative skew is a function

of variables that plausibly influence conditional conservatism are borne out in Basu (1997), Ball et al.

(2000), Givoly and Hayn (2000), Ball et al. (2003), and Krishnan (2005).

Ball and Shivakumar (2005, 2006) propose a measure of conditional conservatism in the

relation between accruals and cash flows, which is potentially useful in a setting where stock returns

are not available. The research designs utilizes a piecewise-linear relation of accruals on cash flows

(depending on their sign), and generally it is concluded that accruals function in predictable ways to

incorporate losses in a more timely fashion than gains. Besides providing insight into the role of

accrual accounting, the results have the added advantage of utilizing non-price proxies for the sign of

economic income. Potential disadvantages of this measure are similar to those of the asymmetric

earnings persistence tests. For example, cash flow changes are noisy proxies of economic income and

thus the presence of conditional conservatism is harder to detect with this measure. On the benefit

side, this measure facilitates studying the reporting properties of private companies, since it does not

require the presence of market prices.

A related validity concern arises from evidence that cash flow itself is a non-linear function of

returns, as reported in Basu (1997, Table 2) and Ball, Kothari and Robin (2000, Table 6), potentially

fueling the criticism discussed above that there could be some known or unknown biases in Basu

39
Ball and Shivakumar (2005).
40
Basu, Hwang and Jan (2001a,b), and Krishnan (2005a,b).
 37

coefficient estimation. We make three points. First, most of the asymmetric timeliness of earnings

nevertheless arises from accruals, not cash flow. The coefficient on negative returns is lower for cash

flow than for earnings (Basu, 1997, Table 2; Ball, Kothari and Robin, 2000, Table 6). Second, it is

unlikely that the efficient response of operating cash flow to news is completely symmetric

(Papadakis, 2007). The optimal response to bad news might be to increase current-period cash

spending (e.g., increased advertising in response to a fall in demand; increased use of consultants and

investment bankers in response to strategic threats; or increased maintenance, repair and warranty

costs in response to a plant or product failure), but the optimal response to good news might not be to

symmetrically decrease current-period cash spending (e.g., decreased advertising in response to

increased demand). Third, there are errors in separating earnings into its cash flow and accruals

components, especially using balance sheet data (Hribar and Collins, 2002). When researchers

estimate accruals as earnings less cash flows, the effect is that errors in one component of earnings

are mirrored by errors in the other. The implication is that the measured properties of cash flows and

accruals, including asymmetric timeliness, are imperfectly distinguished.

In sum, research using alternative measures provides corroborating evidence of the existence

of conditional conservatism. When used in combination with the more familiar returns-based

measure, they produce similar results. When used in contexts where stock returns are unavailable,

notably private companies, they generally behave as conditional conservatism is predicted to behave.

They provide reassuring evidence of the validity of the returns-based asymmetric timeliness

coefficient as a measure of conditional conservatism.

5. Conservatism, Market-to-book and the Asymmetric Timely Coefficient

Feltham and Ohlson (1995, 1996), Beaver and Ryan (2000 and 2005), Easton (2001), Easton

and Pae (2004), Givoly and Hayn (2000), Dietrich et al. (2007) Givoly, Hayn and Natarajan (2007),
 38

and others model the entire difference between the market and book values of equity as resulting

from accounting conservatism. Following Roychowdhury and Watts (2007), we view only a portion

of the difference between the market and book values of equity as resulting from accounting

conservatism (specifically, the difference between the net asset value and the book value of net

assets). Similar views are held in Basu (1997), Holthausen and Watts (2001), Watts (2003a and

2003b), Ball and Shivakumar (2006), Roychowdhury and Watts (2007), and others. Because the

objective of financial reporting is not to value the firm or its equity (FASB, 1978, 41), some

components of market value would not be recorded in book value even in the absence of

conservatism. For example, operating synergies among assets are central to the existence of the firm

(Coase, 1937), but they are not recorded on balance sheets because the accounting system is

transactions-based and therefore focuses on individual assets and not on aggregate firm value. It is

this transactions-based property of accounting not conservatism that causes synergies to be

omitted from book values until they are realized over time as net revenues from the firms operating

activities.41 Thus, we view the market-to-book ratio as a proxy for conditional conservatism that is at

best noisy and at worst misleading.42

Several studies criticize the asymmetric timeliness as a proxy for conditional conservatism by

making an observation that there is a negative relation between the Basu asymmetric timeliness

41
The accounting focus on prices paid for individual assets in arms length transactions (historical costs) presumably is
because actual transactions are considerably less costly to verify than the private information of managers about future
cash flows that would be required to calculate and report firm values.
42
Of course, one could define conservatism as the book/market ratio, and hence encompass all sources of difference
between book and market values, but that would be at the expense of discarding correspondence with the meaning of
conservatism in accounting. For example, this would make conservatism the reason accounting is transactions-based and
not valuation-based, whose economic origin is symmetrically high costs of verifying private information (see prior
footnote). Conservatism then also would encompass symmetric random errors in book values (e.g., arising from errors in
counting closing inventory, or in accruals generally), whose economic origin is the cost of operating an accounting
system, not conservatism. Conservatism would be a symmetric function of any lags in incorporating information into
earnings, whose source also lies in accounting costs. Oddly, conservatism then would be a function of changes in
expected returns (discount rates) that affect market prices, including both increases and decreases in expected returns
symmetrically. More importantly, the correspondence between conservatism and the long-standing dictum anticipate all
losses and no gains would be severed. Yet this is precisely what the Dietrich et al. model does.
 39

coefficient and the market-to-book ratio.43 Pae et al. (2005), Givoly et al. (2007) and Roychowdhury

and Watts (2007) cast the negative correlation between market/book and the asymmetric timeliness

coefficient as a measurement problem, the implication being that the coefficient is an inherently

flawed conservatism measure. For example, Givoly et al. (2007, p. 67) conclude that because the

asymmetric timeliness coefficient is negatively correlated with other measures of conservatism

(notably, market/book), it is insufficient to assess all dimensions of conservatism. The negative

correlation has lent credence to criticisms that (i) the Basu asymmetric timeliness coefficient is

econometrically biased, and/or (ii) it picks up something other than accounting conservatism, such as

earnings persistence (Price, 2005).

We do not view the negative correlation with book-to-market as a problem. In our model, the

negative correlation of conditional conservatism and growth options (for which book-to-market is a

proxy) is a natural result of the accounting income recognition practices. Our results confirm the

original conjecture of Ball, Kothari and Robin (2000, p. 48) that asymmetric earnings timeliness is a

function of the relative importance of news about unbooked growth options.

Our analysis most closely relates to Roychowdhury and Watts (2007), who argue that the

negative relation between the market-to-book ratio and the Basu asymmetric timeliness coefficient

occurs because both measure conservatism with error, and that this error diminishes as the

measurement horizon is increased. The rationale is that a high market-to-book ratio might indicate

the presence of rents (or growth opportunities) which serve as a buffer against having to record

subsequent losses, thus lowering asymmetric timeliness. Cross-sectional variation in the beginning-

of-period market-to-book ratio therefore is likely to add correlated error when measuring conditional

conservatism. Over longer horizons, the market-to-book ratio likely is a less noisy measure of

43
See Gigler and Hemmer (2001), Richardson and Tinaikar (2004), Beaver and Ryan (2005), Price (2005), Pae,
Thornton, and Welker (2005), Givoly, Hayn, and Natarajan (2007), and Roychowdhury and Watts (2007).
 40

conditional conservatism, for example because rents present in the beginning of the period tend to

disappear over time due to competition.

Our analysis has several distinctions. First, we do not rely on the presence of a buffer that

dampens loss recognition in subsequent periods. Despite its appeal, this argument relies on the

assumption that subsequent bad news is treated as a reversal of unbooked rents, which need not be

the case (for example, that the presence of growth options reduces the likelihood of accounting

impairment of spoiled inventory). It may well be that bad news affects the value of booked assets or

affects proportionally both the value of rents and the value of separable assets, in which case a

negative relation between the proportion of rents in the book-to-market ratio and asymmetric

timeliness does not arise. Our second distinction is that the absence of recognition of rents is not

important. Roychowdhury and Watts (2007, p. 12), argue that since increases or decreases in rents

(or unverifiable assets) are not recognized by accountants, this reduces asymmetric timeliness. While

this argument holds, our analysis shows that even in a hypothetical case where rents are recorded by

accountants, their presence can lower the asymmetric timeliness coefficient. The key is their

symmetric treatment by the accounting system. Our analysis suggests the asymmetric timeliness

coefficient decreases in the proportion of return components that are either symmetrically recognized

or symmetrically deferred.

Because the market-to-book ratio is widely considered to be a proxy for accounting

conservatism, evidence of a negative association has been taken by some as challenging the

asymmetric timeliness coefficients validity as a measure of conservatism. For the purposes of

studying actual reporting behavior by firms, we conclude that the true level of conditional

conservatism in reported earnings is a decreasing function of market/book.

 41

6. Summary and Conclusions

Our prior belief is that conditional conservatism is a true property of accounting income, not

merely a statistical artifact. As Basu (1997) notes, the adage anticipate no profits but anticipate all

losses is centuries old, so there are strong priors in favor of an asymmetry existing in fact. The

Financial Accounting Standards Boards Accounting Standards Codification (FASB, 2009) lists

impairment rules for debt and equity securities, property, plant and equipment, goodwill, intangibles,

receivables, beneficial interests, servicing assets, loans, deferred costs, exchange memberships, life

settlement contracts, joint ventures, capital leases, non-refundable fees, and cumulative foreign

exchange translation adjustments for foreign subsidiaries, among other things. Furthermore, asset

impairment charges and restructuring charges were practiced well before formal standards such as

SFAS No. 121 (issued 1995, since superseded) required them, but upward revaluation of long-term

assets is comparatively rare. More importantly, the economic incentives of financial statement

preparers are biased toward timely loss recognition relative to gain recognition, particularly in

litigious jurisdictions (Ball et al., 2000; Ball et al., 2003).

Our priors that conditional conservatism is a true property of accounting income are

reinforced by the variety of studies (briefly summarized in our introductory section) in which

estimated asymmetric timeliness coefficients reveal predictable associations with economic, legal and

political institutional variables, at the firm, industry, and jurisdictional levels. The variety and

consistency of the predicted associations further suggests that estimated asymmetric timeliness

coefficients capture a true property of accounting income, not some statistical artifact. In addition,

asymmetric timeliness estimates based on alternative research designs discussed in the previous

section generally produce similar results to the Basu regression coefficients, and vary in predictable

ways with its economic determinants, such as listing status (public/private) and listing jurisdiction.
 42

These research designs provide additional evidence that the more conventional price-based estimates

indeed capture conditional conservatism.

In sum, since asymmetrically timely loss recognition is a basic property of GAAP, exhibits

predictable associations with economic, legal and political institutional variables, and behaves

consistently under alternative measures, claims that go so far as to conclude that the results of prior

studies employing the asymmetric timeliness measure are attributable to biases, and cannot be

interpreted as evidence of accounting conservatism, can be discounted. The claims have developed

some credence, nevertheless.

We address these claims by presenting a model of accounting income that is based on salient

properties of income recognition in accounting, and use it to analyze the econometrics of the Basu

asymmetric timeliness coefficient. The analysis addresses conceptual and econometric challenges to

the coefficients validity as a measure of conditional conservatism. We show that the Basu measure

is unbiased under the null hypothesis of zero asymmetry, and that under the alternative hypothesis it

captures conditional conservatism as formulated in our model. We also demonstrate that any

differential persistence in economic gains and losses (as opposed to differential persistence in the

gains and losses that are recognized in accounting) does not explain the Basu coefficient. We

extend the Roychowdhury and Watts (2007) analysis and demonstrate econometrically that a

negative relation between market-to-book ratio and the asymmetric timeliness coefficient is expected

as a consequence of how changes in the markets expectations concerning unbooked growth

options are reflected in earnings. The market-to-book ratio does not per se determine asymmetric

timeliness; its effect arises because it is correlated with the amount of price revision associated with

revisions in unbooked components such as growth option expectations.

We trust that the analysis will clarify the econometrics of the asymmetric timeliness

methodology, and stimulate its application to even more contexts.

 43

Appendix:

Derivation of the expression for the asymmetric timeliness coefficient

We derive the asymmetric timeliness coefficient for a general case where all three

components of return, x, y, and g, are considered. The cases when some of these components are

omitted follow immediately. We omit subscripts t when all random variables are contemporaneous.

We start by noting that:

cov( xt + wt yt + (1 wt 1 ) yt 1 + g t 1 + t t 1 , xt + yt + g t | Rt 0)
1 =
var( R | R 0)
cov( xt + wt yt , xt + yt + g t | Rt 0)
= (A1)
var( R | R 0)
cov( x, x + y + g | R 0) + cov( wy , x + y + g | R 0)
=
var( R | R 0)

and

cov( x, x + y + g | R < 0) + cov( wy , x + y + g | R < 0)

2 = (A2)
var( R | R < 0)

Exploiting symmetry we have:

plim2 = 2 1 =
cov( wy, x + y + g | R < 0) cov( wy, x + y + g | R 0)
= (A3)
var( R | R < 0) var( R | R 0)
cov( wy, x + y + g | R < 0) cov( wy, x + y + g | R 0)
=
var( R | R 0)

where w = 1( y < 0) is an indicator variable taking a value of one in case y < 0 and zero otherwise.

Define z x + g , z2 var( z ), and

cov( y, x) + cov( y, g ) xy x + yg g
corr( z , y ) = = .
var( y ) var( x + g ) x2 + 2 xg x g + g2

To show that is 2 = 0 positive it suffices to show that > 0 , where:

cov( wy , z + y | R < 0) cov( wy , z + y | R > 0) > 0.

 44

Note further that:

= E ( wy 2 + wyz | R < 0) E ( wy 2 + wyz | R 0) E ( wy | R < 0) E ( z + y | R < 0) + E ( wy | R 0) E ( z + y | R 0)

0 y 0 + 0 +
= ( y 2 + zy ) ( z , y | R < 0) dzdy ( y 2 + zy ) ( z , y | R 0) dzdy + E ( z + y | R 0) y ( z , y | y < 0)dzdy
y  

0 y 0 +
= y ( z, y | R < 0)dzdy y ( z , y | R 0)dzdy +
2 2

 

y
 

1 2
0 y 0 + 0 +
+ zy ( z , y | R < 0)dzdy y zy ( z , y | R 0)dzdy + E ( z + y | R 0) y ( z , y | y < 0)dzdy
 

 
 

3 4

(A4)

1 1 z 2 2 zy y 2
where ( z , y ) = exp + is bivariate normal
2 z y (1 2 ) 0.5 2(1 2 ) z2 z y y2

(0) 2 z2 + 2 z y + y2
density and E ( z + y | R 0) = E ( R | R 0) = var( R) = .
1 (0) 2

The integrals can be evaluated explicitly under the normality assumption, first noting that symmetry

implies Pr(R < 0) = 0.5.

1
0
1 z 2 2 zy y 2
=
2 z y (1 2 ) 0.5 Pr( R < 0)
y exp + dzdy
2(1 2 ) z2 z y y2

0 y z z y
2

1 y 1 y2 dzdy
= y 2(1 2 ) (1 ) (1 2 ) 0.5 z y + (1 ) 2
2 2
exp
z (1 2 ) 0.5 y (A5)

y y z z y
= u= ,v= , dy = y du , dz = z (1 2 ) 0.5 dv
y (1 2 ) 0.5 z y
y 0
1 2

1 2 2 y
=


u exp u
2
du 2
exp v dv =
2

Further, making the same transformation, and integrating by parts it can be shown that:
 45

y 0 y y z z y
2
2
y2 1 2 y
1 = y2 2(1 2 ) (1 ) (1 2 )0.5 z y +
2
exp (1 ) dzdy
z (1 2 )0.5 2

y

y y z z y y +z
= u= ,v= , dy = du , dz = (1 2 0.5
) dv , u u
y (1 2 )0.5 z y
y z
(1 2 )0.5 z
y2 0 u
1 2 2
y 2

=
u exp 2

2
( v + u )

dvdu =
+ 1 + 2
(A6)

y +z
where = / 2 arctan( ) , and = (0, + ) .
(1 2 ) 0.5 z

By analogy (making the same substitution as above) we have:

y2 0
1 2 2
y 2

2 = u u exp 2 ( v + u ) dvdu = 1 + 2
2

(A7)

Using integration by parts, it can be further shown that:

1
0 y 1 z2 zy 2 y2 2 y
2
3 =
z y (1 2 )0.5 zy exp
2(1 2 ) z2
2

+
2
+ (1 )
2

dzdy
z y y y
y y z z y y +z
= u= ,v= , dy = y du , dz = z (1 2 )0.5 dv, u u
y (1 ) z y
2 0.5
(1 2 )0.5 z
z y 0 u
1
= u (v 1 2 + u ) exp ( v 2 + u 2 ) dvdu
2
z y 1
= 1 2 + + ,
1+ 2
1 + 2 (A8)

and, by analogy,

z y 0
1
4 = u (v 1 2 + u ) exp ( v 2 + u 2 ) dvdu
u 2
z y 1
= 1 1 + 2 + 1 + 2 .
2

(A9)

Now we can use these results to evaluate the expression for :

 46

0 y 0 + 0 +
=

( y 2 + zy ) f ( z , y ) dzdy
y
( y 2 + zy ) f ( z , y ) dzdy + E ( z + y | R > 0)

y f ( x, y ) dzdy

2 z2 + 2 z y + y2 y2 y
2

= 1 2 + 3 4 + = + 2

2 1+ 1 + 2
z y 1 z y 1
+ 1 2 + + 2
1 2 +
1+ 2
1 + 1+ 2
1 + 2
2 y 2 z + 2 z y + y
2 2

.
2 2 (A10)

Rearranging this expression yields:

y 2 2 z 1 2
= ( y + z )( 2 + ) + 2 z 1 2 1 + 2
1+ 2
1+ 2


2 y z 1 2
=
(1 + arctan( ) )
1 + 2 > 0.
(A11)

Now the asymmetric timeliness coefficient can be computed as follows:

2 =

=
2 y z 1 2
var( R | R > 0) ( 2) var( R )
(
1 + arctan( ) 1 + 2 > 0 ) (A12)

2
where we use the result that var( z + y | R > 0) = var( R ) E ( R | R > 0) 2 = var( R ) 1 .

 47

References

Ahmed, A., Duellman, S., 2007, Accounting conservatism and board of director characteristics: An
empirical analysis, Journal of Accounting & Economics 43, 411-437.

Ali, A., Hwang, L., Trombley, M., 2003, Arbitrage risk and the book-to-market anomaly, Journal of
Financial Economics 69, 355373.

Bachelier, L., 1900, Thorie de la spculation, Annales Scientifiques de lcole Normale Suprieure
3, 21-86.

Ball, R., Brown, P., 1968, An empirical evaluation of accounting income numbers, Journal of
Accounting Research 6, 159-177.

Ball, R., Kothari, S., Robin, A., 2000, The effect of international institutional factors on properties of
accounting earnings, Journal of Accounting & Economics 29, 1-51.

Ball, R., Robin, A., 1999, Time-series properties of accounting earnings: International evidence,
unpublished, University of Rochester and Rochester Institute of Technology, May.

Ball, R., Robin, A., Sadka, G., 2008, Is financial reporting shaped by equity markets or by debt
markets? An international study of timeliness and conservatism, Review of Accounting Studies
13, 168205.

Ball, R., Robin, A., Wu, J., 2000, Accounting standards, the institutional environment and issuer
incentives: Effect on timely loss recognition in China, Asia Pacific Journal of Accounting and
Economics 7, 71-96.

Ball, R., Robin, A., Wu, J., 2003, Incentives versus standards: Properties of accounting income in
four East Asian countries, Journal of Accounting & Economics 36, 235-270.

Ball, R., Shivakumar, L., 2005, Earnings quality in U.K. private firms: comparative loss recognition
timeliness, Journal of Accounting & Economics 39, 83-128.

Ball, R., Shivakumar, L., 2006, The role of accruals in asymmetrically timely gain and loss
recognition, Journal of Accounting Research 42, 207-242.

Ball, R., Shivakumar, L., 2008a, Earnings quality at initial public offerings: Managerial opportunism,
or public-firm conservatism? Journal of Accounting & Economics 45, 324-349.

Ball, R., Shivakumar, L., 2008b, How much new information is there in earnings? Journal of
Accounting Research 46, 975-1016.

Ball, R., Watts, R., 1972, Some time series properties of accounting income, Journal of Finance 27,
663-682.
 48

Ball, R., Bushman, R. and Vasvari, F., 2007, The debt-contracting value of accounting information
and loan syndicate structure, Journal of Accounting Research 46, 247-287.

Barragato, C. A., Basu, S., 2007, Properties of accounting earnings in not-for-profit organizations,
unpublished, Emory University and Long Island University.

Basu, S., 1997, The conservatism principle and the asymmetric timeliness of earnings, Journal of
Accounting & Economics 24, 3-37.

Basu, S., 2005, Discussion of Conditional and unconditional conservatism: Concepts and
modeling, Review of Accounting Studies 10, 311321.

Basu, S., 2009, Conservatism research: Historical development and future prospects, China Journal of
Accounting Research 2, 1-20.

Basu, S., Hwang, L., Jan, C., 1998, International variation in accounting measurement rules and
analysts forecast errors, Journal of Business Finance and Accounting 25, 1207-1247.

Basu, S., Hwang, L., Jan, C., 2001a, Auditor conservatism and quarterly earnings, Working Paper,
Baruch College, CUNY.

Basu, S., Hwang, L., Jan, C., 2001b, Differences in conservatism between big eight and non-big eight
auditors, Working Paper, Baruch College, CUNY.

Basu, S., Huang, A., Mitsudome, T., Weintrop, J., 2005, Timeliness and conservatism of Taiwanese
earnings, Asia-Pacific Journal of Accounting and Economics 12, 113-134.

Basu, S., Markov, S., 2004. Loss function assumptions in rational expectations tests on financial
analysts earnings forecasts, Journal of Accounting and Economics 38, 171-203.

Beaver, W., Lambert, R., Morse, D., 1980, The information content of security prices, Journal of
Accounting & Economics 2, 3-28.

Beaver, W., Lambert, R. and Ryan, S., 1987, The information content of security prices, A second
look. Journal of Accounting and Economics 9, 139158.

Beaver, W., Landsman, W., Owens, E., Asymmetry in Earnings Timeliness and Persistence: A
Simultaneous Equations Approach, Stanford University (March 2008).

Beaver, W., M. McAnally and C. Stinson, 1997, The information content of earnings and prices: A
simultaneous equations approach, Journal of Accounting and Economics 23, 53-81.

Beaver, W., Ryan, S., 2000, Biases and lags in book value and their effects on the ability of the book-
to-market ratio to predict book return on equity, Journal of Accounting Research 38, 127-148.

Beaver, W., Ryan, S., 2005, Conditional and unconditional conservatism: Concepts and modeling,
Review of Accounting Studies 10, 269-276.
 49

Beekes, W., Pope, P., Young, S., 2004, The link between earnings timeliness, earnings conservatism
and board composition: Evidence from the UK, Corporate Governance: An International
Review 12, 47-59.

Bernard, V., Thomas, J., 1990, Evidence that stock prices do not fully reflect the implications of current
earnings for future earnings, Journal of Accounting and Economics 13, 305-340.

Brealey, R., Myers, S., Allen, F., 2006, Principles of Corporate Finance, 8th edition, McGraw-Hill
Irwin.

Brown, W. D., Jr., He, H., Teitel, K., 2006. Conditional conservatism and the value relevance of
accounting earnings: An international study, European Accounting Review 15, 605-626.

Burgstahler, D. C., Hail, L., Leuz, C., 2006. The importance of reporting incentives: Earnings
management in European private and public firms, The Accounting Review 81, 983-1016.

Bushman, R., Piotroski, J., 2006, Financial incentives for conservative accounting: The influence of
legal and political institutions, Journal of Accounting & Economics 42, 107-148.

Bushman, R., Piotroski, J., Smith, A., 2007, Capital allocation and timely accounting recognition of
economic losses, working paper, The University of North Carolina at Chapel Hill.

Callen, J., Chen, F., Dou, Y., Xin, B., 2009, Information asymmetry and the debt contracting demand
for accounting conservatism, working paper, University of Toronto.

Campbell, J., 1991, A variance decomposition of stock returns. Economic Journal 101, 157-179.

Campbell, J., Ammer, J., 1993, What moves the stock and bond markets? A variance decomposition
for long-term asset returns, Journal of Finance 48, 3-37.

Campbell, J., Lo, A., MacKinlay, C., 1997, The Econometrics of Financial Markets. Princeton
University Press, Princeton, NJ.

Campbell, J., Shiller, R., 1988, The dividend-price ratio and expectations of future dividends and
discount factors, Review of Financial Studies 1, 567-592.

Cao, Y., Lee, C., 2002, Asymmetry in earnings-returns relations: The case of China. Working paper,
University of Rochester.

Chandra U., Wasley, C., Waymire, G., 2004, Income conservatism in the U.S. technology sector.
Working Paper, University of Rochester.

Chaney, P., Philipich, K., 2003, The price-earnings relation in troubled times: The case of Arthur
Andersen. Working paper, Vanderbilt University and The Ohio State University.

Collins, Daniel W., S.P. Kothari, 1989, An analysis of intertemporal and cross-sectional determinants of
earnings response coefficients, Journal of Accounting and Economics, 11, 143-181.
 50

Collins, D., Kothari, S., Shanken, J., Sloan, R., 1994, Lack of timeliness versus noise as explanations
for low contemporaneous return-earnings association, Journal of Accounting & Economics
18, 289-324.

Cook, T. D. and Campbell, D.T., 1979, Quasi- Experimentation: Design and Analysis Issues for Field
Settings. Chicago: Rand-McNally.

Cuijpers, R., Moers, F., Peek, E., 2005, Firm-level reporting incentives and accounting properties:
Stakeholder-versus shareholder-oriented governance structures. Working paper, Universiteit
Maastricht.

Dietrich, D., Muller, K., Riedl, E., 2007, Asymmetric timeliness tests of accounting conservatism,
Review of Accounting Studies 12, 95-124.

Easton, P., 2001, Discussion of when capital follows profitability: Non-linear information dynamics,
Review of Accounting Studies 6, 267-274.

Easton, P., Pae, J., 2004, Accounting conservatism and the relation between returns and accounting
data, Review of Accounting Studies 9, 495-521.

Easton, Peter E., and Mark E. Zmijewski, 1989, Cross-sectional variation in the stock market response to
accounting earnings announcements, Journal of Accounting and Economics 11, 117-141.

Fama, E., 1970, Efficient capital markets: A review of theory and empirical work, Journal of Finance
25, 383-417.

Fama, E., 1990, Stock returns, expected returns, and real activity, Journal of Finance 45, 1089-1108.

Feltham, J., Ohlson, J., 1995, Valuation and clean surplus accounting for operating and financial
activities, Contemporary Accounting Research 11, 689-731.

Feltham, J., Ohlson, J., 1996, Uncertainty resolution and the theory of depreciation measurement,
Journal of Accounting Research 34, 209-234.

Financial Accounting Standards Board (FASB), 1978. Concepts Statement No. 1: Objectives of
Financial Reporting by Business Enterprises. Norwalk, Connecticut: Financial Accounting
Standards Board.

Financial Accounting Standards Board (FASB), 2009. Accounting Standards Codification.

Norwalk, Connecticut: Financial Accounting Standards Board. Website:
http://asc.fasb.org/home (visited 15 April, 2009).

Foster, G., Olsen, C., Shevlin, T., 1984, Earnings releases, anomalies, and the behavior of securities
returns, The Accounting Review 59, 574-603.

Francis, J., LaFond, R., Olsson, P., Schipper, K., 2004, Costs of capital and earnings attributes, The
Accounting Review 79, 967-1010.
 51

Francis, J., Martin, X., 2010, Acquisition profitability and timely loss recognition, Journal of
Accounting and Economics 49, 161-178.

Garca Lara, J.M., Mora, A., 2005, Balance sheet versus earnings conservatism in Europe, European
Accounting Review 13, 261-292.

Garca Lara, J., Garca Osma, B., Mora, A., 2005, The effect of earnings management on the
asymmetric timeliness of earnings, Journal of Business Finance & Accounting 32, 691-726.

Garcia Lara, J., Garcia Osama, B. and Penalva, F., 2009, Accounting conservatism and corporate
governance, Review of Accounting Studies 14, 161201.

Gassen, J., Fuelbier, R., Sellhorn, T., 2006. International differences in conditional conservatism -
The role of unconditional conservatism and income smoothing, European Accounting Review
15, 527-564.

Gigler, F., Hemmer, T., 2001, Conservatism, optimal disclosure policy, and the timeliness of
financial reports, The Accounting Review 76, 471-493.

Giner, B., Rees, W., 2001, On the asymmetric recognition of good and bad news in France, Germany
and the United Kingdom, Journal of Business Finance and Accounting 28, 1285-1331.

Givoly, D., Hayn, C., 2000, The changing time series properties of earnings, cash flows, and
accruals: Has financial reporting become more conservative? Journal of Accounting &
Economics 29, 287-320.

Givoly, D., Hayn, C., Natarajan, A., 2007, Measuring reporting conservatism, The Accounting
Review 82, 65-106.

Grambovas, C.A., Giner, B., Christodoulou, D., 2006, Earnings conservatism: Panel data evidence
from the European Union and the United States, Abacus 42, 354378.

Gu, Z., Wu, J.S., 2003. Earnings skewness and analyst forecast bias, Journal of Accounting and
Economics 35, 5-29.

Guay, W., Kothari, S., Watts, R., 1996, A market-based evaluation of discretionary-accrual models,
Journal of Accounting Research (Supplement) 34, 83-115.

Gul, F., Srinidhi, B., Shieh, T., 2002, The Asian financial crisis, accounting conservatism and audit
fees: Evidence from Hong Kong, Working Paper, City University of Hong Kong.

Hammermeister, J., Werner, J., 2009, Conditional conservatism, debt markets and financial structure:
Further evidence from the United Kingdom, Working paper, University of Bremen.

Hayn, C., 1995, The information content of losses. Journal of Accounting and Economics 20, 123
153.
 52

Hecht, P., Vuolteenaho, T., 2006, Explaining returns with cash flow properties, Review of Financial
Studies 19, 159-194.

Holthausen R., Watts, R., 2001, The relevance of the value-relevance literature for financial
accounting standard setting, Journal of Accounting & Economics 31, 3-75.

Hribar, P., Collins, D.W., 2002, Errors in estimating accruals: implications for empirical research,
Journal of Accounting Research 40, 105134.

Huijgen, C., Lubberink, M., 2005, Earnings conservatism, litigation, and contracting: The case of
cross-listed firms, Journal of Business Finance and Accounting 32, 1275-1309.

Jindrichovska, I., Kuo, D., 2004, Asymmetric timeliness of earnings in the emerging market of the
Czech Market. In: M. Mejstrik (Ed.) Cultivation of Financial Markets in the Czech Republic
(Prague. The Karolinum Press), 341-364.

Kelley, S., Shores, D., Tong, Y., 2004, Independence in appearance, earnings conservatism, and
prediction of future of cash flows, Working Paper, University of Washington.

Khan, M., Watts, R., 2009, Estimation and empirical properties of a firm-year measure of accounting
conservatism, Journal of Accounting & Economics, forthcoming.

Kothari, S., 2001, Capital markets research in accounting, Journal of Accounting & Economics 31,
105-231.

Kothari, S., Shanken, J., 1992, Stock return variation and expected dividends: A time series and
cross-sectional analysis, Journal of Financial Economics 31, 177-210.

Kothari, S., Sloan, R., 1992, Information in prices about future earnings, Journal of Accounting &
Economics 15, 143171.

Krishnan, G., Labelle, R., 2005a, Did Houston clients of Arthur Andersen recognize publicly
available bad news in a timely fashion? Contemporary Accounting Research 22, 165-204.

Krishnan, G., 2005b, The association between big 6 auditor industry expertise and asymmetric
timeliness of earnings, Journal of Accounting Auditing & Finance 20, 209-228.

Krishnan, G., 2007, Did earnings conservatism increase for former Andersen clients? Journal of
Accounting, Auditing and Finance 22, 141-163.

Kwon, S., Yin, J., Han, J., 2006, The effect of differential accounting conservatism on the over-
valuation of high-tech firms relative to low-tech firms, Review of Quantitative Finance and
Accounting 27, 143-173.

Leone, A., Wu, J., Zimmerman, J., 2006, Asymmetric sensitivity of CEO cash compensation to stock
returns, Journal of Accounting & Economics 42, 167-192.

Lewellen, J., 1999, The time-series relations among expected return, risk, and book-to-market,
Journal of Financial Economics 54, 5-43.
 53

Lintner, J., 1956, Distribution of incomes of corporations among dividends, retained earnings, and
taxes, American Economic Review 46, 97-113.

Little, I.M.D., 1962, Higgledy-piggledy growth, Bulletin of Oxford Institute of Statistics 24, 387-412.

Lubberink, M., Huijgen C, 2001, A wealth-based explanation for earnings conservatism, European
Finance Review 5, 323-349.

Maddala, G., 2001, Introduction to Econometrics, Third Edition, John Wiley & Sons, Ltd., New
York, NY.

Mak, C., Strong, N., Walker, M., 2006, Earnings conservatism and corporate refocusing activities,
working paper, Manchester Business School, University of Manchester.

Nikolaev, V., 2009. Debt covenants and accounting conservatism, Journal of Accounting Research,
forthcoming.

Pae, J., Thornton, D., Welker, M., 2005, The link between accounting conservatism and the price-to-
book ratio, Contemporary Accounting Research 22, 693-717.

Papadakis, G., 2007, Investment dynamics and the timeliness properties of accounting numbers, PhD
thesis, MIT Sloan School of Management.

Peek, E, Buijink, W., Coppens, L., 2004, Accounting conservatism in the European Union: The
influence of institutions and incentive, Working Paper, University of Maastricht.

Peter, J. P., 1981, Construct Validity: A review of basic issues and marketing practices, Journal of
Marketing Research 18, 133145.

Pope, P., Walker, M., 1999, International differences in the timeliness, conservatism and
classification of earnings, Journal of Accounting Research 37, 53-87.

Price, R., 2005, Accounting conservatism and the asymmetry in current and lagged returns, working
paper, Rice University.

Qiang, X., 2007. The effects of contracting, litigation, regulation, and tax costs on conditional and
unconditional conservatism: Cross-sectional evidence at the firm level, The Accounting
Review 82, 759-796.

Raonic, I., McLeay, S., Asimakopoulos, I., 2004, The timeliness of income recognition by European
companies: An analysis of institutional and market complexity, Journal of Business Finance
and Accounting 31, 115-148

Richardson, G., Tinaikar, S., 2004, Accounting based valuation models: What have we learned?
Accounting and Finance 44, 223-255.

Ross, S., Westerfield, R., Jaffe, J., 2005, Corporate Finance, Seventh edition, McGraw-Hill Irwin.
 54

Roychowdhury, S., Watts, R., 2007, Asymmetric timeliness of earnings, market-to-book and
conservatism in financial reporting, Journal of Accounting & Economics 44, 2-31.

Ruddock, C., Taylor, S., Taylor, S., 2006, Non-audit services and earnings conservatism: Is auditor
independence impaired? Contemporary Accounting Research 23, 701-746.

Ryan, S.G., 2006, Identifying conditional conservatism, European Accounting Review 15, 511-525.

Ryan, S.G., Zarowin, P., 2003, Why has the contemporaneous linear returns-earnings relation
declined? The Accounting Review 78, 523-553.

Samuelson, P. A., 1965, Proof that properly anticipated prices fluctuate randomly, Industrial
Management Review 6, 4149.

Shiller, R., 1981, Do stock prices move too much to be justified by subsequent changes in dividends?,
American Economic Review 71, 421-36.

Shiller, R., 2000, Irrational Exuberance, Princeton University Press.

Sivakumar, K., Waymire, G., 2003, Enforceable accounting rules and income measurement by early
20th century railroads, Journal of Accounting Research 41, 397-432.

Tazawa, M., 2003, The timeliness of earnings and accruals under conservatism in Japan, Working
Paper, Nagoya City University.

Vuolteenaho, T., 2002, What drives firm-level stock returns, Journal of Finance 57, 233-264.

Wang, D., 2006, Founding family ownership and earnings quality, Journal of Accounting Research
44, 619656.

Watts, R., 2003a, Conservatism in accounting Part I: Explanations and implications, Accounting
Horizons 17, 207-221.

Watts, R., 2003b, Conservatism in accounting Part II: Evidence and research opportunities,
Accounting Horizons 17, 287-301.

Watts, R., Zimmerman, J., 1986, Positive Accounting Theory, Prentice-Hall.

Wittenberg-Moerman, R., 2008. The role of information asymmetry and financial reporting quality in
debt contracting: Evidence from the secondary loan market, Journal of Accounting &
Economics 46, 240-260.