Journal of Accounting and Economics 75 (2023) 101535

Contents lists available at ScienceDirect

Journal of Accounting and Economics

journal homepage: www.journals.elsevier.com/
journal-of-accounting-and-economics

Accounting comparability and relative performance

evaluation by capital markets*
Sang Wu a, *, Wenjie Xue b
a
Columbia University, USA
b
National University of Singapore, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: This paper examines how accounting comparability affects the monitoring role and the
Received 24 February 2020 risk allocation role of capital markets. We develop the statistical and informational
Received in revised form 12 July 2022 properties of accounting reports under varying degrees of comparability. A perfectly
Accepted 28 July 2022
comparable accounting information system enables investors to perfectly infer the dif-
Available online 11 August 2022
ference between any two firms' future cash flows although investors remain uncertain
about either firm's cash flow. Comparability alleviates entrepreneurs' moral hazard
JEL classification:
problem by strengthening the price response to the relative accounting performance, but
G12
G14
can induce excessive price risk as well as residual systematic cash flow risk. Unlike the
G18 investors (users) who earn their surplus by bearing the residual systematic risk, the en-
M41 trepreneurs (preparers) do not find perfect comparability desirable. Hence, a standard
M48 setter would mandate higher comparability than preferred by preparers, but not perfect
comparability.
Keywords:
© 2022 Elsevier B.V. All rights reserved.
Accounting comparability
Standard setting
Measurement error
Information externality

1. Introduction

Over the past 40 years, reducing diverse practices and inconsistent guidance is the most frequently cited reason by the
Financial Accounting Standards Board (FASB) to take on a project, and more than half of the standards are intended to enhance
comparability (Jiang et al., 2018). Despite that, the current standards still allow for diverse practices and inconsistent
treatment for similar transactions. FASB board members who are former CFOs or controllers are less likely than those with
user backgrounds to dissent because a standard allows for exceptions or gives the management accounting alternatives. As
noted by practitioners and researchers, while the users of financial information emphasize on accounting comparability, the
preparers often dispute its importance (see, e.g., Van Riper, 1994; De Franco, Kothari and Verdi, 2011; Jiang et al., 2018; Kurt,

*
We are deeply indebted to Carlos Corona and Pierre Liang for their guidance and encouragement. We thank Jeremy Bertomeu, Paul Fischer, Pingyang
Gao (reviewer), Henry Friedman, Michelle Hanlon (editor), Lin Nan, Bryan Routledge, Gaoqing Zhang, and participants at the Carnegie Mellon University
workshop and 2019 Junior Accounting Theory Conference for helpful discussions and comments. Wenjie Xue acknowledges the grant S.No.02.2019.133
from National University of Singapore.
* Corresponding author.
E-mail addresses: [email protected] (S. Wu), [email protected] (W. Xue).

https://doi.org/10.1016/j.jacceco.2022.101535
0165-4101/© 2022 Elsevier B.V. All rights reserved.
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

2020). The natural question is: What drives the different attitudes toward comparability by the two major constituents of
financial reporting? Why do we still observe diverse practices despite regulatory emphasis on comparability?
One explanation for the existence of diverse accounting practices is that it allows for flexibility and is thus preferred by
preparers who may benefit from opportunistic reporting. However, preparers do not always benefit from more discretion
when investors rationally correct for the opportunistic reporting on average (see, e.g., Guttman et al., 2006; Gao and Zhang,
2019). In this paper, we show that the information externalities of comparable financial reports can potentially explain why
preparers are less enthusiastic about adopting common accounting practices, even in the absence of opportunistic reporting.
Indeed, a too-high level of comparability can induce excessive price risk and risk premium borne by the preparers. Thus, we
provide a different rationale for diverse accounting practices which is that permitting alternative accounting methods for the
same economics phenomenon diminishes comparability among reports (FASB, 2018, QC25; IASB, 2018, 2.29).
We first develop the statistical and informational properties of comparability as they are decision-relevant to economic
agents. Comparability is viewed as a property of the relationship between the accounting information of at least two firms
(FASB, 2018, QC21; IASB, 2018, 2.25). Suppose an accounting information system measures the fundamental value of each firm
with errors, we model comparability as the correlation between the measurement errors of any pair of firms under a
reporting regime. This proposed measure is theoretically consistent with the notion that adopting a common standard
features a larger common measurement error (Dye and Sridhar, 2008; Zhang, 2013), as the correlation between measurement
errors is positively associated with the proportion of the common error among reports. On the one hand, higher comparability
renders the difference between two reports more informative about the difference between their reporting firms' funda-
mentals, and hence improves the users’ knowledge about the firm-specific fundamentals relative to other firms. On the other
hand, it impairs their knowledge about the average fundamental and thus the aggregate economy due to a larger common
measurement error.12
To study the economic consequences of accounting comparability, we build our model on the overlapping generations
setting in Dye (1990), Dye and Sridhar (2007) and Gao (2010), and extend it to multiple firms. Each firm is initially owned and
managed by a risk-averse entrepreneur. The cash flow to each firm is jointly determined by an unobservable effort by the
entrepreneur, an economy-wide shock common to all firms, and an idiosyncratic shock specific to each firm. After privately
exerting the effort, the entrepreneur publicly issues an accounting report which is a noisy signal about the cash flow. En-
trepreneurs then sell their firms to risk-averse investors and consume the proceeds due to life-cycle considerations. At the last
date, cash flows are realized, and investors consume the cash flows generated by the firms.
In the presence of information externalities, investors efficiently use other firms' reports to price any firm in the market.
Their inference about the cash flows given their conjectures of the entrepreneurs' efforts can be decomposed into two tasks.
In the first task, the investors infer each firm's idiosyncratic cash flow shock by comparing its report with the other reports. In
the second task, the investors learn from all reports to infer the cash flow shock common to all firms. Comparability improves
the first inference but impairs the second one. This informational property leads to various efficiency implications.
Comparability alleviates the moral hazard problem arising from the unobservability of the entrepreneurs' efforts. An
entrepreneur internalizes the return to his effort to the extent that it affects the firm price through the accounting report. As
the investors' inference about the common shock is disciplined by the reports of the other firms, the entrepreneur is
incentivized to work hard only to differentiate his report from the others’ reports. Due to the noisiness of accounting signals,
the rate at which his effort increases the market price is lower than the rate at which it increases the firm value. Higher
comparability alleviates this moral hazard problem by strengthening the price response to the relative accounting
performance.
Despite its positive effect on the monitoring role of capital markets, perfect comparability is not considered desirable by
the entrepreneurs as it induces excessive price risk and risk premium. An undiversified entrepreneur who consumes the
proceeds from selling his firm bears the risk associated with the volatility of the firm price. This price risk has both an
idiosyncratic and a systematic component. Each component of the price risk is positively associated with the extent to which
the investors use the reports to infer its corresponding cash flow shock. Although comparability impairs the investors'
inference about the common shock, it can increase the price risk due to the investors' enhanced inference about the idio-
syncratic shock when the level of comparability is sufficiently high. The entrepreneurs sell at a lower price due to the risk
premium which is determined by the diversified investors’ residual uncertainty about the common shock. Higher compa-
rability increases the risk premium by increasing such uncertainty about the aggregate economy.
However, maximizing comparability makes the diversified investors better off because they earn their surplus by bearing
the residual systematic risk (Gao, 2010; Bertomeu and Cheynel, 2016). The risk premium equals their marginal cost of bearing
the systematic risk which is higher than the average cost, and their surplus increases in the residual systematic risk. The above

1
The former informational effect manifests itself through the increased price response to a firm's own report. From this perspective, comparability is an
enhancing characteristic of useful financial information as described by the conceptual frameworks in FASB (2018, QC19) and IASB (2018, 2.23).
2
We can further illustrate this informational trade-off using an analogy in which students' ability is assessed through exams. The students can be taking
the same exam, or assigned to distinct and independent exams. The grades in the former case are more “comparable” because students are evaluated under
the same criterion, and hence the ranking of students is more informative of their relative ability. On the other hand, the grades in the latter case are more
informative about the average ability of the class because the noisiness induced by independent exams is diversified away in the average grade.

2
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

results combined suggest that the preparers prefer a lower level of comparability than the users. As noted by a former FASB
member Van Riper (1994):
“Though its importance is often disputed by preparers of financial information, comparability of financial information
has long been a major concern of analysts and legislators.”
As lower comparability is associated with diverse accounting practices, our model can potentially explain why preparers
are more tolerant of diverse practices relative to users and even tend to lobby for a lower level of standardization (Jiang et al.,
2018). If the standard setter incorporates the preferences of both preparers and users of financial reports, the observed
standards should reflect both of their preferences (Watts and Zimmerman, 1978). Hence, the standard setter would mandate a
higher level of comparability than that would otherwise be chosen by the preparers, but not perfect comparability because of
the excessive costs on the preparers. In other words, we should still observe diverse accounting practices despite regulatory
emphasis on comparability.3 The model also predicts that the mandated level of comparability would be higher when the
idiosyncratic cash flows are less volatile. That is, we should observe standards allowing for more diverse practices in an
economy with highly heterogeneous firms. Moreover, our main results continue to hold qualitatively even when extending
our model to allow for firm-wise heterogeneity in the level of comparability.
This paper contributes to the current literature in several aspects. First, it formally defines comparability and investigates
the economic consequences of increasing comparability from an informational perspective. Accounting has been intensively
studied as an information function (Butterworth, 1972; Liang and Zhang, 2008). Current standard setters follow the
conceptual-framework approach to standard setting by first defining desirable qualitative characteristics of financial infor-
mation and then evaluating alternative standards against them (Bullen and Crook, 2005). However, the common notion of
comparability is theoretically ambiguous (Sunder, 2010). A key feature of our theoretical definition is that we disentangle
comparability from the informativeness of a stand-alone report which has been studied extensively since Feltham (1972). We
examine how comparability affects the usefulness of an accounting report given other firms' reports while controlling for the
informativeness of a stand-alone report. Although comparability enhances the usefulness of an accounting report given other
firms’ reports, we show that the standard that takes into account both constituents would not feature perfect comparability.
Nevertheless, we provide a justification for the regulatory emphasis on comparability, as preparers would otherwise choose
an even less comparable reporting system than what the standard setter would choose.
Second, this paper is related to the study of the monitoring role of capital markets in addressing agency issues (e.g.,
Holmstrom and Tirole, 1993; Dye and Sridhar, 2004, 2007). Efficient pricing by the market serves as a disciplinary mechanism
for those with control rights. Consequently, how much information is available to the market affects the efficiency of their
actions. We extend this line of inquiry by examining the information externalities of financial reports in the sense of Dye
(1990) and Dye and Sridhar (2008). Comparability can facilitate the investors' learning of firm value through comparison
between firms, thus incentivizing the entrepreneurs to exert efforts to differentiate themselves from the others. This
mechanism also formally establishes the intuition in Core et al. (2003) that equity incentives render an implicit relative
performance evaluation scheme. More recently, Jennings et al. (2020) formally document that a firm's price depends on the
relative ranking of its earnings in its peer group.
Finally, this paper addresses the risk allocation effect of disclosure policies (e.g., Dye, 1990; Dye and Sridhar, 2004, 2007;
Gao, 2010). Disclosure resolves uncertainty early, increasing the price risk while reducing the residual cash flow risk
(Hirshleifer, 1971). In the situation where the current shareholders have to sell to the next generation of shareholders, more
precise disclosure is efficient if and only if the new generation is relatively more risk averse than the old generation. Our
information structure differs from the standard one in that the level of comparability does not affect the precision of a stand-
alone report, but rather affects risk allocation through the information externalities. Due to the investors’ diversification of the
idiosyncratic risks, only the systematic risk is allocated between the two generations. A higher level of comparability shifts
more systematic risk to the investors. Nonetheless, even when the entrepreneurs are relatively more risk averse than the
investors, it is not efficient to allocate all the systematic risk to the investors as it also induces excessive idiosyncratic price risk
borne by the undiversified entrepreneurs.
The rest of the paper is organized as follows. Section 2 develops the model and discusses the measure of comparability in
more details. Section 3 characterizes the equilibrium given an exogenous level of comparability. Section 4 studies the welfare
implication of increasing comparability on both constituents of financial reporting. Section 5 discusses the empirical im-
plications. Section 6 extends our main setting and enables firm-wise heterogeneity in the level of comparability. Section 7
concludes.

2. Model

This section describes the basic setting of the economic model and the informational properties of comparability. Since the
notion of comparability can only be applied to more than one firm, we model a continuum of firms, with each firm indexed by

3
There can be many other reasons for users to prefer higher comparability. For example, comparability can perhaps facilitate more efficient information
processing by investors. These possible forces not captured by our model do not contradict our results. In practice, there are also many other stakeholders of
financial reporting not captured by our analysis. Arguably, preparers and users are among the most important ones.

3
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

i 2 [0, 1].4 Each firm is owned and managed by an entrepreneur with the same index as the firm he owns for simplicity. There
is also a continuum of investors indexed by j 2 [0, 1] who buy the firms from entrepreneurs after the release of accounting
signals, but before the cash flows are realized.

2.1. Setup

Firm i's cash flow e

qi is determined jointly by (i) the effort ai exerted by its entrepreneur, (ii) the economy-wide shock he
common to all firms, and (iii) an idiosyncratic shock edi specific to each firm. That is, given the effort ai,
   
qi ¼ ai þ eh þ edi ; with eh  N 0; t1
e e 1
h ; di  N 0; td ; (1)

where th > 0 and td > 0 are the precisions of the two cash flow shocks, respectively. Both the price and the return of the risk-
free asset are normalized to be 1.
The output of the accounting system is a noisy signal about the firm's cash flow. It carries both a common error and an
idiosyncratic error. The common error comes from the use of a common set of measurement methods required by the ac-
counting standards, and the idiosyncratic error is due to diverse practices. Formally, conditional on the cash flow, firm i's
accounting signal is
   
esi ¼ qi þ eε þ exi ; with eε  N 0; t1
ε ;e
xi  N 0; t1
x ; (2)

where eε and exi are the common error and the idiosyncratic error, with corresponding precisions tε > 0 and t > 0, respectively.
R1 x
Since the firms are otherwise symmetric, we may also use the economy-wide average signal e s ≡ 0 esi di ¼ a þ h
e þe
ε as a
R1
sufficient statistic of all (other) firms' accounting reports, where a ≡ 0 ai di is the average effort. To disentangle comparability
from the informativeness of a stand-alone report, we restrict the precision of an individual signal about its reporting firm
(hereafter, reporting precision) to be a constant t, with 0 < t < ∞ to ensure usefulness of accounting signals, i.e.,

t1 ≡ t1 1
ε þ tx : (3)

In equilibrium, the investors rationally infer the unobservable effort exerted by the entrepreneurs. However, they are
unable to perfectly deduce the underlying cash flows of the firms, as the cash flows are subject to the common shock and the
idiosyncratic shocks as well. The noisy accounting reports thus help them to imperfectly estimate the extent to which the cash
flow shocks have affected the terminal cash flows.
Entrepreneurs and investors are assumed to be risk averse with constant absolute risk aversion (CARA). Entrepreneur i's
utility function is given by ui ðw ei , where w
e i Þ ¼  r1 erE w e i denotes his consumption financed by the proceeds from selling his
E
firm. Investor j's utility function is given by vj ðw ej , where w
e j Þ ¼  r1 erI w e j denotes her consumption financed by the cash flows
I
generated by the portfolio she buys from the entrepreneurs.
The timeline of the model consists of four dates, as depicted in Fig. 1. At date t ¼ 1, the entrepreneur of each firm i 2 [0, 1]
chooses an effort ai and personally bears a quadratic cost 12a2i . The level of the effort is private information to the entrepreneur who
made the choice. At date t ¼ 2, the accounting system of each firm produces a public signal about the cash flow that will be
generated by the firm. Investors then use all the accounting signals to make portfolio decisions. At date t ¼ 3, the trading takes place.
Entrepreneurs sell the firms to investors and consume the price of their firms. At the last date t ¼ 4, cash flows of all firms are
realized, and investors consume the cash flows generated by their portfolios.
All random variables are independent from each other and all parameters are common knowledge unless specified
otherwise.

2.2. Measure of comparability

Our theoretical measure of comparability is based on its statistical and informational properties. Wang (2014) defines
comparability as the correlation between the measurement processes of two firms’ accounting earnings. Because the mea-
surement errors capture the informational properties of the measurement process in our informational framework, the
correlation between the measurement processes of the two firms, i and i0 , is
 
  Cov eε þ exi ; eε þ exi0 t1
c ≡ Corr esi  e
qi ; esi0  eqi0 ¼ h    i1 ¼ 1
ε
: (4)
e e 2 tε þ t1
Var eε þ xi Var eε þ xi0 x

4
The extent of the information externalities increases in the number of firms in a finite-firm setting (Dye and Sridhar, 2008), so one can view this
assumption as a limiting case of the finite-firm setting.

4
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

Fig. 1. The timeline of events.

In other words, we measure comparability as the proportion of variance from the common measurement error, c 2 [0, 1].
Thus, controlling for the precision of an individual report as t, we have

t1 1 1 1
ε ¼ ct ; tx ¼ ð1  cÞt : (5)

According to the FASB (2018, QC21) and the IASB (2018, 2.25), comparability enables users to identify and understand
differences among firms. Taking the difference between the accounting reports of firm i and i0 renders a noisy signal of the
difference between their fundamentals:
   
qi  eqi0 þ exi  exi0 :
esi  esi0 ¼ e (6)

The common measurement error is cancelled out during the comparison, and the variance of the noise is 2t1 x which
decreases in c. That is, the amount of information conveyed through the comparison of accounting signals is positively
associated with comparability.
This measure is also consistent with the notion that financial reports are more comparable through the application of a
common standard because applying a common standard can be associated with a larger common measurement error (Dye
and Sridhar, 2008; Zhang, 2013). However, it should be noted that comparability is not equivalent to uniformity. In our setting,
comparability not only increases common measurement error, but also reduces the idiosyncratic measurement error so that
“like things look alike and different things look different” (FASB, 2018, QC23; IASB, 2018, 2.27).
Limitations of the measure. First, we do not attempt to model how accounting standards map transactions into financial
reports. Therefore, the proposed measure of comparability can neither be used to study the frictions in the measurement
process nor to provide any operational guidance on evaluating alternative accounting rules (Gao, 2013). Specific accounting
rules differ in multiple dimensions in terms of all qualitative characteristics. One qualitative characteristic may have to be
diminished to maximize another qualitative characteristic (FASB, 2018, QC34; IASB, 2018, 2.38). Choosing one accounting rule
over another may increase comparability but impair other qualitative characteristics, which reduces the informativeness of a
report about its reporting firm. In our theoretical inquiry, we hold constant the precision of each stand-alone report in order
to focus on the information-externality effects.
Second, our measure does not capture firm-wise heterogeneity in the level of comparability. The level of comparability
between any two firms' reports may be affected by their reporting firms’ characteristics, including to what extent their
transactions are regulated by the standard or if they are in the same industry (De Franco et al., 2011). In other words, c can only
be interpreted as a regime-level measure of comparability among all reports. Studying a setting with heterogenous exposures
to correlated measurement errors requires carefully specifying the distribution of the exposures, which has nontrivial im-
plications for the way investors use the reports. Nevertheless, in section 6, we simplify the analysis by considering an
extension with only two groups of firms, with firms in each group only subject to a group-wide common measurement error
besides idiosyncratic errors. Our main results remain qualitatively robust in this extended setting.
Third, our measure of comparability captures the notion that applying a common standard enhances comparability in a
reduced-form manner of modelling the common measurement error. It omits other aspects of adopting a common standard.
For example, applying a common disclosure regulation to all firms can economize on the information processing costs of users
confronted with multiple reports (Gao et al., 2019), reduce the distortion created by lobbying (Friedman and Heinle, 2016),
and impose rigidity on financial reporting thus reducing opportunistic reporting (Dye and Sridhar, 2008).

3. Equilibrium analysis

The equilibrium in this model consists of the effort choices simultaneously made by the entrepreneur of each firm and the
pricing rules applied by the investors in the capital market. We characterize a perfect Bayesian equilibrium in which all
players make optimal decisions that maximize their utility given all their available information as well as their rational
expectations regarding the strategic behavior of the other players, utilizing Bayes’ rule to make inferences and update their

5
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

a i the conjecture of the players about the effort level a*i and Pi : R2 /R the pricing equation,
beliefs. Formally, denoting by b
respectively, any perfect Bayesian equilibrium must satisfy the following conditions for any i 2 [0, 1] and for any si ; s2R:

i. a*i 2arg max E½Pi   r2E VarðPi Þ  12a2i ;

Ii

ii. Pi ðsi ; sÞ ¼ E½e

qi jsi ; s  rI Varðehjsi ; sÞ; and

a i ¼ a*i .
iii. b

Condition (i) pertains to the efforts simultaneously chosen by the entrepreneurs. Each entrepreneur chooses the level of the
effort that maximizes his expected utility, given his rational expectations about the other entrepreneurs' efforts and the investors'
pricing rule. Condition (ii) describes the pricing rule applied by the investors following firms' issuance of accounting reports.5 Here,
the risk-averse investors are only compensated for bearing nondiversifiable risks, and the risk premium equals their marginal cost
of bearing such risk. Condition (iii) ensures that all players have rational expectations regarding each other's behavior.
We derive the interrelated equilibrium outcomes of entrepreneurs’ efforts and the capital market prices using backward
induction.
Lemma 1. Each investor holds an identical proportion of the market portfolio and is only compensated for bearing non-
diversifiable risks. With conjectured levels of effort b
a i and b
a for firm i and the economy-wide average, respectively, the pricing
equation applied by the investors is:
tx tε 1
Pi ðsi ; sÞ ¼ b
ai þ ðs  sÞ þ a Þ  rI
ðs  b : (7)
td þ tx i th þ tε th þ tε

In equilibrium, investors’ conjecture coincides with the actual effort exerted by the entrepreneurs:

tx
a i ¼ a*i ¼
b : (8)
td þ tx

In the presence of cash flow uncertainties, investors cannot precisely infer each firm’s cash flow despite their rational
expectations about the entrepreneurs’ efforts. They thus find the accounting reports useful for learning about the two cash
flow shocks. When pricing a firm, the inference is decomposed into two tasks. First, they use the relative accounting per-
formance, i.e., the difference between the firm-specific signal and the average signal, to learn about the idiosyncratic shock to
the firm. Second, they use the average signal to learn about the common shock to all firms. Essentially, the average report
plays two informational roles - it not only is informative about the aggregate economy, but also teases out the common
measurement error in the individual firm’s report. The latter role is similar to the efficient use of information in relative
performance evaluation schemes in Holmstrom (1982).
The equilibrium effort exerted by an entrepreneur is always lower than the efficient level, that is, there is a moral hazard
problem.6 Because the effort is unobservable to the investors, the entrepreneur is incentivized to exert effort only to the
extent it increases the firm price through the accounting report. Although the investors cannot distinguish cash flow shocks
from the effect of the effort on the firm value, the inference about the common shock is disciplined by the average report.
Hence, the entrepreneur can only jam their inference about the idiosyncratic shock. As can be seen from equation (7), the
marginal benefit of effort to the entrepreneur is the price coefficient on the relative performance. In other words, the
entrepreneur works hard to differentiate his own report from the average report. The economic return of one additional unit
of effort in improving future cash flows is 1, but the price coefficient on the relative accounting performance is always lower
than 1 due to the imperfection of the accounting signals. This incentive effect of the pricing equation is consistent with the
conjecture in Core et al. (2003) that equity incentives of managers render an implicit relative performance evaluation scheme.
Improving comparability has different effects on investors’ posterior uncertainty about the two cash flow shocks. An
increase in comparability leads to a simultaneous increase of the common error and decrease of the idiosyncratic error in
accounting signals (see equations (5)), and thus improves investors’ knowledge about the idiosyncratic shock while impairing
their knowledge about the common shock. More precisely, the posterior variances of the two shocks are, respectively,

1
Varðe
di jsi ; sÞ ¼ Varðedi jsi  sÞ ¼ ; (9)
td þ tx

5
The pricing equation is taken as given in this definition of equilibrium for succinctness. In fact, since there is no asymmetric information among in-
dividual investors, it is directly solved by no-arbitrage and market clearing. A similar characterization can be also found in Gao (2019). The complete proof is
provided in the appendix.
6
The efficient level of effort is the one that maximizes the net return of the costly effort. Denote it as aFB FB ¼ arg max a  1a2 ¼ 1. This efficient
i , then ai i 2 i
ai
outcome can also be achieved if there is no information asymmetry regarding the level of effort.

6
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

1
Varðh ejs  b
ejsi ; sÞ ¼ Varðh aÞ ¼ : (10)
th þ tε

Higher comparability enables the investors to learn more about a specific firm relative to the other firms, but less about the
overall fundamental of the aggregate economy. As a result, firm price depends more on the relative accounting performance
and less on the average accounting performance following an increase in comparability. This prediction is consistent with the
standard setter’s view of relevance being a fundamental characteristic and comparability being an enhancing characteristic
because for the same level of reporting precision of a stand-alone firm, comparability increases the usefulness of the focal
firm’s report given other firms’ reports as reflected by its increased earnings response coefficient.
Proposition 1. When both constituents are risk neutral, perfect comparability is uniquely Pareto-optimal, i.e., c*rn ¼ 1.
In a risk-neutral world, both entrepreneurs and investors would unanimously agree on implementing a perfectly com-
parable accounting standard. Indeed, in this case, the investors break even on expectation and are indifferent about the level
of comparability. The entrepreneurs prefer perfect comparability because it resolves the moral hazard problem. Higher
comparability increases the entrepreneurs’ perceived marginal benefit from exerting effort through the strengthened price
response to the relative accounting performance, and thus encourages the entrepreneur to work harder. Following
Holmstrom and Tirole (1993), we refer to this as the market monitoring effect of comparability. This incentive effect derives
from the market’s efficient use of all other firms’ reports as comparability does not affect the precision of a stand-alone report.
In a risk-averse world, however, the information externalities of the financial reports also have effects on the risk borne by the
entrepreneurs and the investors. In the next section, we show that risk aversion is a key driver of our main results that are
consistent with the observed different attitudes toward the desirability of comparability by the different constituents of
financial reporting.

4. Main results

In equilibrium, the risk premium in firm prices equals the diversified investors' marginal cost of bearing the residual
systematic cash flow risk and is higher than the average cost. The gap between the two is the source of the investors' surplus.
Thus, the investors' net certainty equivalent from bearing the risk is half of the risk premium. With rational expectations,
investors’ conjectured levels of effort coincide with the actual ones chosen by the entrepreneurs, which are in turn incor-
porated into the prices. As a result, an entrepreneur internalizes the net return of his effort in addition to the risk premium.
Moreover, the risk-averse entrepreneurs also bear the price risk associated with the volatility of price which contains both an
idiosyncratic and a systematic component since they cannot diversify. Formally, the ex-ante certainty equivalents of a
representative investor and a representative entrepreneur are respectively:

rI 1
CEI ¼ ; (11)
2 th þ tε
 2  
tx 1 tx r 1 tx 1 tε 1
CEE ¼   E þ  rI : (12)
td þ tx 2 td þ tx 2 td td þ tx th th þ tε th þ tε
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflffl{zfflfflfflfflfflffl}
net return of effort price risk risk premium

Comparability impairs the investors’ knowledge about the common cash flow shock, leading to a higher residual sys-
tematic cash flow risk. Since investors earn a higher surplus by bearing more systematic risk, their surplus is maximized with
perfect comparability.7 That is, investors prefer the highest level of comparability, i.e., c*I ¼ 1.
Entrepreneurs' preference for comparability is jointly determined by its effects on the three terms in equation (12). Despite
its positive market monitoring effect on net return of effort, higher comparability results in a less precise inference about the
common shock and a higher risk premium, which makes the entrepreneurs worse off. We refer to this as the risk premium
effect. On the other hand, the price risk effect associated with the ex-ante volatility of the price can go in either direction. To see
this, note that the more responsive the price to information, the more volatile it is ex-ante. As comparability strengthens the
price response to the relative accounting performance and weakens the price response to the average accounting perfor-
mance, the price volatility first decreases and then increases in comparability, and is minimized at an intermediate level of
comparability. To summarize, the three effects each has a different inclination to the entrepreneurs’ preferred level of
comparability c*E :

i. the market monitoring effect drives c*E toward 1;

ii. the price risk effect drives c*E toward thtþd td ; and
iii. the risk premium effect drives c*E toward 0.

7
For a more detailed discussion on risk premium and investors' surplus, see Gao (2010) and Bertomeu and Cheynel (2016).

7
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

Proposition 2. When both constituents are risk averse, the investors prefer perfect comparability, i.e., c*I ¼ 1, and the entre-
preneurs prefer a strictly lower level of comparability, i.e., 0  c*E < 1.
As comparability has different implications on the idiosyncratic risk and the systematic risk, the two risk-averse con-
stituents desire different levels of comparability. While the investors prefer perfect comparability, the entrepreneurs prefer a
strictly lower level of comparability. From the entrepreneurs’ perspective, despite its positive effect on the net return of effort,
perfect comparability induces excessive price risk as well as risk premium. As shown by Gao (2010), resolving uncertainty
allocates risk from investors to entrepreneurs. Although the systematic risk is allocated between the two generations, perfect
comparability maximizes the risk premium without minimizing the price risk. Increasing comparability not only allocates
more systematic risk to the diversified investors through the increased uncertainty about the common shock and increases
the risk premium, but also increases the idiosyncratic risk borne by the undiversified entrepreneurs through the decreased
uncertainty about the idiosyncratic shock. Moreover, the entrepreneurs may not desire any comparability at all when the risk
premium effect dominates in the presence of sufficiently risk-averse investors.
The different preferences for comparability by the entrepreneurs and the investors coincides with the fact that the
importance of comparability is often disputed by the preparers of financial information despite the users’ desire for
comparability (Van Riper, 1994; Jiang et al., 2018). To capture in a parsimonious way that the standard setter is influenced by
both constituents of financial reporting (Watts and Zimmerman, 1978), we assume the standard setter aims to maximize W ≡
(1  u)CEE þ uCEI, where u 2 (0, 1) is the weight placed on the investors’ surplus, and denote the optimal level of
comparability chosen by the standard setter as c*.
Proposition 3. The optimal comparability c* is strictly higher than the one that would otherwise  be chosen by the entrepreneurs
whenever it is non-zero, i.e., c* > c*E if c*E 2ð0;1Þ. Moreover, there exists a threshold ub 2 23; 1 such that c* is strictly lower than the
perfect level, i.e., c* < 1, if and only if u < u
b . The interior c* increases (decreases) in the weight the standard setter places on the
investors (entrepreneurs).
The optimal level of comparability chosen by the standard setter is greater than that desired by the entrepreneurs because
the standard setter caters for the investors’ preference for perfect comparability. In addition, as long as the standard setter
places a sufficient weight on the entrepreneurs, the optimal policy would not feature perfect comparability as it induces
excessive price risk as well as residual systematic cash flow risk. Indeed, the relative weight placed on the two constituents
determines not only the relative importance but also the direction of the effect of the residual systematic risk on the optimal
comparability. Since the investors’ net certainty equivalent from bearing the risk is half of the risk premium (see equations
(11) and (12)), the effect of the residual systematic risk drives the optimal comparability toward the minimum level 0 when
u < 23, but toward the maximum level 1 when u > 23. In other words, the standard setter’s inclination on the risk premium is
more aligned with the entrepreneurs (investors) in the former (latter) case. Moreover, even when the effect of the residual
systematic risk vanishes at u ¼ 23, the price risk effect will still render the optimal comparability lower than 1. Thus, the
threshold u b above which the optimal comparability is perfect has to be strictly higher than 23.
Taken together, the above results potentially explain why we still observe diverse accounting practices despite regulatory
emphasis on comparability. Without the regulatory intervention, the level of comparability implemented by the preparers
would likely not internalize the users’ preference for perfect comparability and hence be lower than the optimal level from
the standard setter’s perspective. To better understand the role played by different agents’ risk attitudes in generating our
results, we consider the following two special cases.
Corollary 1. When either constituent is risk neutral, we have

i. rE > rI ¼ 0: investors are indifferent, and c*E ¼ c* 2ð0; 1Þ; and

ii. rI > rE ¼ 0: c*E < c* < c*I ¼ 1 when c*E > 0 and u is sufficiently low.

When the investors are risk neutral, they are not compensated for bearing risks. As a result, the break-even investors are
indifferent to the level of comparability, and the standard setter’s optimal policy coincides with the one preferred by the risk-
averse entrepreneurs. Since the existence of the price risk effect diminishes the desirability of perfect comparability, the
standard setter would mandate an interior level of comparability. On the other hand, when the investors are risk averse, they
prefer perfect comparability because it maximizes their surplus gained by bearing the residual systematic cash flow risk.
However, the risk-neutral entrepreneurs prefer a lower level of comparability even without the price risk effect because they
sell at a lower price due to the risk premium. The standard setter’s optimal policy is thus determined by trading off both
constituents’ preferences. When the entrepreneurs prefer some level of comparability and the standard setter places a
sufficient weight on the entrepreneurs, the optimal comparability admits an interior solution and is higher than that would
otherwise be chosen by the entrepreneurs.
To summarize, investors’ risk aversion is necessary for our model to speak to their preference for the level of comparability.
In particular, to generate the joint empirical regularity that (i) preparers of financial information prefer a lower level of
comparability than users, and (ii) perfect comparability is not an observed policy, the assumption that the investors are risk
averse is indispensable. In other words, the higher risk premium due to the higher systematic risk left in the market by a
higher level of comparability is the key friction that prevents the observed policy from featuring perfect comparability when
the standard setter places sufficient weights on both constituents.

8
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

Next, we explore how the primitive variables of the model affect the optimal policy that would be chosen by the standard
setter.
Corollary 2. The optimal level of comparability c*, whenever it admits an interior solution,

*
(i) decreases in investors' degree of risk aversion rI, i.e., dcdrI
< 0 if and only if u < 23;
*
1u th t2d t2d ð2th tÞ
(ii) increases in entrepreneurs' risk aversion rE, i.e., ddcr > 0, if and only if u < 23, rI > 2 23 u th td þtðth þtd Þ and rE > th td þtðth þtd Þ;
E *
23u
increases in the common cash flow volatility t1 h i.e., dth < 0 *if and only if rE > 1u rI ;
(iii) dc

(iv) decreases in the idiosyncratic cash flow volatility t1 d , i.e., dc > 0; and
dtd
*
< 0, if u < 23 and rE > 23 u
(v) decreases in the reporting precision t, i.e., dc
dt 1u rI .

Recall that the relative weight placed on the two constituents determines to which direction the residual systematic risk
drives the optimal comparability. Generally, investors’ risk aversion tends to decrease the optimal level of comparability
because the associated risk premium effect drives the optimal comparability closer to the minimum level. However, when the
weight placed on the investors is excessively high, i.e., u > 23, investors’ surplus takes precedence, making the effect of the
residual systematic risk drive the optimal comparability closer to the maximum level instead. Hence, the optimal compa-
rability increases in the investors’ risk aversion in this case.
The effect of entrepreneurs’ risk aversion depends on the original level of comparability. Recall that the price risk borne by
the entrepreneurs is minimized at c ¼ thtþd td . When the weight placed on the investors is not excessively high and both
constituents are relatively risk averse, the market monitoring effect is dominated by the price risk effect and risk premium
effect, leading the optimal comparability to lie in 0; thtþd td . As the entrepreneurs become more risk averse, it increases  the
optimal comparability toward its upper bound thtþd td . Otherwise, when the optimal level of comparability lies in thtþd td ; 1 , the
optimal comparability moves closer toward its lower bound thtþd td as the entrepreneurs become more risk averse.
As discussed earlier, a higher level of comparability allocates more systematic risk to the investors. When the investors are
relatively more risk tolerant than the entrepreneurs, they are efficiently allocated more risks. Hence, in this case, an increase
in the common cash flow volatility t1 h leads to an increase in the optimal level of comparability. Similarly, when the en-
trepreneurs are relatively more risk tolerant than the investors, they are efficiently allocated more risk with a decreased level
of comparability.
Observe from equations (11) and (12) that the idiosyncratic volatility affects the optimal comparability only through the
net return of effort and the idiosyncratic component of the price risk. As the idiosyncratic shock becomes more volatile, that
is, as t1
d increases, firm price is more responsive to the relative accounting performance, which results in higher effort by the
entrepreneurs but also more price risk. Hence, the optimal policy leans more toward reducing the idiosyncratic component of
the price risk than resolving the moral hazard problem by introducing more idiosyncratic measurement error with a lower
level of comparability.
As for the effect of an improvement in the reporting precision on the optimal comparability, we first consider the special
case in which the two constituents have the same risk attitude and are equally weighted by the standard setter. In this case,
the standard setter’s optimal choice of comparability is independent of both the common cash flow shock and the common
measurement error due to the risk allocation role of comparability. Consequently, when the reporting quality improves, the
optimal allocation of the reporting precision to the idiosyncratic measurement error stays the same and it is more efficient to
devote all the efforts to reducing the common measurement error. Generally, when the weight placed on the investors is not
excessively high, increasing the relative risk aversion of the entrepreneurs will only strengthen the idiosyncratic part of the
price risk effect. As a result, the negative effect of an improved reporting precision on the optimal level of comparability will
even be more salient.

5. Empirical implications

Besides rationalizing the existence of diverse accounting practices despite regulatory emphasis on comparability, our
analysis generates various additional empirical implications. The pricing equation (7) suggests that firms are priced relatively
to other firms given their relative accounting performance. More recently, Jennings et al. (2020) document that a firm's price
responds positively to an improvement in the relative ranking of its earnings among its peers. Our theory thus predicts that
such price response to the relative accounting performance is more pronounced in regimes with a higher level of
standardization.
Rewriting the pricing equation in Lemma 1, we have

tε b 1
Pi ðsi ; sÞ ¼ b
a i þ asi þ bs  a  rI ; where
th þ tε th þ tε
(13)
tx tε tx
a¼ ; b¼  :
td þ tx th þ tε td þ tx

The price coefficient a captures the extent to which a firm's price relies on its own accounting signal, and is henceforth
referred to as the direct earnings response coefficient (the direct ERC), whereas the price coefficient b captures the extent to

9
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

which a firm's price relies on the average accounting signal (or, equivalently, the peer firms' accounting signals), and is
henceforth referred to as the cross earnings response coefficient (the cross ERC).
Corollary 3. The direct ERC increases in comparability and the cross ERC decreases in comparability. Moreover, the cross ERC is
positive (negative) if comparability is lower (higher) than thtþd td .
Higher comparability increases the direct ERC due to the increased informativeness of the relative accounting perfor-
mance. It enables the market to learn more about the idiosyncratic characteristic of each firm relative to the average firm
(market portfolio). As a result, the price of the firm responds more aggressively to the firm’s report when the reports are more
comparable. On the other hand, achieving higher comparability impairs investors’ knowledge of the common shock by
introducing more common error. For example, a high average report can stem from either a favorable common shock or too
much false positive due to the common measurement error. When there is more common error, the average report is used
relatively more in correcting for the common error than in updating about the common shock. The cross ERC thus decreases
in comparability and can even be negative when comparability is sufficiently high.
We can exploit the above feature of the pricing equation to construct a regime-level empirical proxy for comparability. The
cross ERC estimated using a cross section of firms is negatively associated with the level of comparability. A limitation of this
proxy is that it does not allow for firm-wise variations as in De Franco et al. (2011) and Fang et al. (2022). However, an advantage
is that it does not require time-series data assuming that comparability is time-invariant. Hence, this proxy can be used to
compare comparability across regimes at a particular time, or to study time-series variations of comparability in a particular
regime over time, for example, around policy changes. In addition, this proxy can be adopted to test some of the predictions in
Corollary 2. For example, the model predicts that the optimal level of comparability decreases in the idiosyncratic cash flow
volatility, implying that markets with more heterogeneous firms tend to feature more diverse accounting practices.8

6. Extension

In this section we extend the baseline model to allow for firm-wise heterogeneity in reporting comparability. For example,
financial reports of firms in the same industry may be more comparable than those in different industries. To simplify the
analysis, we assume there are two groups of firms and firms in each group are subject to a group-specific common mea-
surement error in addition to firm-specific idiosyncratic measurement errors. Indeed, we restrict the cross-group compa-
rability to be zero while allowing the within-group comparability levels to vary across the two groups. We obtain the results
corroborating the robustness of the analysis in the baseline setting, and discuss the potential of testing our predictions using
firm-level empirical measures of comparability such as the ones constructed by De Franco et al. (2011) and Fang et al. (2022).9
Specifically, in this extended setting, the future cash flow of firm i in group k 2 {1, 2} is

e
q ¼ ak;i þ h ek þ e
eþh dk;i ; (14)
k;i

e  N ð0; t1
where ak,i is the effort chosen by the entrepreneur, h h Þ is the economy-wide common cash flow shock, h ek  N ð0;
t1 e 1
hk Þ is the within-group common cash flow shock, and dk;i  N ð0; tdk Þ is the firm-specific idiosyncratic shock. Conditional on
the cash flow, firm i's accounting signal is

esk;i ¼ e
qk;i þ eεk þ exk;i ; (15)

εk  N ð0; t1
where e e 1
εk Þ is the within-group common error and xk;i  N ð0; tx Þ is the firm-specific idiosyncratic error. As in the
k

baseline model, controlling for the reporting precision, i.e., t1

k
≡ t1 1
εk þ tx , we measure within-group comparability as ck ¼
k
t1
εk
1 . Without loss of generality, we assume that each group is of a half measure.
εk þtx
t1
k

Lemma 2. With conjectured levels of effort b

a k;i and b
a k for firm i and group-k average, respectively, the pricing equation applied by
the investors is:

  t xk  
Pk;i sk;i ; sk ; sk0 ¼ b
a k;i þ s  s þ bkk ðsk  b a k0 Þ  rI mu Shjs m;
a k Þ þ bkk0 ðsk0  b (16)
tdk þ txk k;i k
where bkk, bkk’, Sh|s, and m are constants given in the appendix.
In equilibrium, investors’ conjecture coincides with the actual effort exerted by the entrepreneurs:

8
Dye and Sridhar (2008) make a similar prediction that when the level of heterogeneity is high, a flexible regime with opportunistic reporting and no
common measurement error dominates a rigid regime with a common measurement error but no opportunistic reporting. We abstract from the reporting
discretion issue but compare regimes with a spectrum of proportions of the common error with respect to the total reporting noise.
9
Those firm-level empirical measures of comparability are constructed for each firm-year against the other firms in an identified peer group.

10
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

txk
a k;i ¼ a*k;i ¼
b : (17)
tdk þ txk
When inferring the future cash flow of firm i in group k, investors first use the difference between the firm-specific signal
and the group average signal sk;i  sk to infer about the firm’s idiosyncratic shock edk;i , and then use the average signal of each
group sk and sk0 to learn about the common shocks h eþh ek . Both group average signals are useful in the second task because of
the economy-wide correlation induced by the common cash flow shock e h. Lastly, the risk premium is determined by the
investors’ residual uncertainties about the undiversifiable common shocks fh e þhe1 ; e
h þhe2 g, Sh|s, and the mass of each group in
the market portfolio m.
Rewriting the pricing equation (16) following equation (13), the direct ERC and cross ERC for firms in group k are
respectively:

txk t xk
ak ¼ ; b ¼ bkk  : (18)
tdk þ txk k tdk þ txk
Since the average signal of the other group sk0 also affects the price of a firm in group k, from an econometrics point of view,
the reports of other groups also need to be controlled in order to have unbiased estimates of the direct ERC and cross ERC.
Moreover, the direct ERC ak increases in comparability ck and the cross ERC bk decreases in comparability ck. At the same time,
however, the cross ERC of firms in the other group bk0 increases in comparability ck. In other words, the difference ak  ak0
monotonically increases in ck  ck’, and the difference bk  bk0 monotonically decreases in ck  ck’. Hence, the predictions in
Corollary 3 can be tested using firm-level empirical measures of comparability. For example, Chen et al. (2020) show that the
empirical measure constructed by De Franco et al. (2011) is positively associated with the direct ERC.10
The certainty equivalents of a representative investor and a representative entrepreneur of each group are respectively:

rI
CEI ¼ mu Shjs m; (19)
2
   
1 * 2 rE 1 *
CEE;k ¼ a*k  a  a þ Sh  Shjs  rI mu Shjs m; (20)
2 k 2 tdk k k;k

Fig. 2. The Effects of Idiosyncratic Cash Flow Volatility on Comparability.

Note: Fig. 2(a) plots the effects of the group-1 idiosyncratic cash flow shock volatility, i.e., t1
d1 , on different constituents’ preferred and the optimal levels of
comparability, holding constant the group-2 idiosyncratic cash flow shock volatility. Fig. 2(b) plots the effect on the cross-sectional difference in the two groups’
optimal levels of comparability. Other parameter values are u ¼ 0.5, rE ¼ rI ¼ 0.5, th ¼ th1 ¼ th2 ¼ 1, td2 ¼ 1:5, and t1 ¼ t2 ¼ 1.

10
Variations in the firm-level empirical measures are mostly cross-sectional. Hence, we need to show that the cross-sectional difference ck  ck’ affects the
cross-sectional differences ak  ak0 and bk  bk0 in the same manner as c affects a and b in Corollary 3.

11
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

where Sh denotes the prior uncertainty about the common shocks fh e þe h1 ; he þhe2 g. Assuming that the standard setter
 
aims to maximize W≡12 ð1  uÞ CEE;1 þ CEE;2 þ uCEI , let fc*I;1 ; c*I;2 g, fc*E;1 ; c*E;2 g, and fc*1 ; c*2 g be the pair of comparability
that respectively maximizes CEI, CEE,1 þ CEE,2, and W.11 Consistent with our main results in Propositions 2 e 3, the
investors always prefer perfect comparability for both groups of firms, whereas the entrepreneurs prefer a strictly
lower level of comparability. Consequently, the optimal levels of comparability chosen by the standard setter lie in
between.
We also illustrate the robustness of our results in this extension using numerical examples as presented in Fig. 2 due to
computational complexity. We focus on the effect of idiosyncratic cash flow volatility because it is the only unambiguous and
monotone comparative statics in Corollary 2. As can be seen from Fig. 2(a), although group-1 idiosyncratic cash flow volatility
affects the entrepreneurs’ preferred comparability c*E;k and the optimal policy c*k for both groups, its effects on group-1's tend
to be in the opposite direction than those on group-2's and also much more significant in magnitude. Combining it with the
observation from Fig. 2(b), both the group-1 optimal comparability c*1 and the cross-group difference in the optimal
comparability c*1  c*2 decrease in group-1 idiosyncratic cash flow volatility t1 d1 . Indeed, the same holds symmetrically for
group-2 idiosyncratic cash flow volatility. Hence, the empirical predictions provided by Corollary 2(iv) can potentially be
tested using firm-level empirical measures of comparability as well.

7. Conclusion

This paper proposes a theoretical framework that highlights the information externalities of comparable accounting re-
ports and the resulting trade-off in learning about similarities and differences between firms. Such an informational trade-off
leads to subsequent economic trade-offs and implies different preferences for accounting comparability by different agents,
depending on whether they are the users or the preparers of financial information. As a result, although preferred by the
users, perfect comparability is not what standard setters should attempt to achieve because it imposes excessive costs on the
preparers. The model thus explains why we still observe diverse accounting practices despite regulatory emphasis on
comparability. As accounting policies are influenced by constituents with different interests, the analysis of the “optimal level
of comparability”, which internalizes the interests of both entrepreneurs (preparers) and investors (users) in the model,
provides empirical predictions about the determinants of accounting policies from a positive perspective (Watts and
Zimmerman, 1978).
In addition, the pricing equation that reflects the efficient use of accounting information suggests a structural approach to
construct a regime-level empirical proxy for comparability. Our model predicts that, controlling for a firm's own report, the
firm's price response to the average report in a reporting regime decreases in the level of comparability and can even become
negative with sufficiently high comparability. This property of the pricing equation allows the price coefficient on the average
report estimated with a cross section of firms to proxy for the level of comparability in this reporting regime. Intuitively,
higher comparability renders the firm price more responsive to the relative accounting performance, thus resulting in the
other firms' positive reports being used more as bad news when pricing a firm. This empirical proxy can be used to compare
comparability across regimes at a particular time, or to study the time-series variations of comparability in a particular regime
over time, for example, around policy changes.12
Our study is subject to some caveats. We work with comparability as an informational property while assuming away
a more practical problem: how standards should be set to achieve this property. Analytical research can potentially add to
understanding this practical problem. For example, most accounting measurement methods are discrete in the sense that
continuous accounting inputs are truncated at a recognition threshold (Gao and Jiang, 2020). It is thus an open question
whether using the same threshold for all firms (uniformity) enhances or undermines comparability. We also assume
away opportunistic reporting by the preparers, although the lack of comparability can be endogenously caused by
opportunistic reporting. Dye and Sridhar (2008) compare two polar cases: one with common measurement error but
without reporting discretion (the rigid regime) versus one without common measurement error but with reporting
discretion (the flexible regime). Future research can extend the analysis to examine how more flexibility and opportu-
nistic reporting cause lower correlation of measurement errors and lower comparability. Moreover, we assume a stylized
economy in which firms do not have strategic interactions with others. In reality, firms could be substitutes because they
are competitors, or complements because of technology and demand spillovers.

Appendix

Proof of the Pricing Equation. The price of each firm given all available information I ≡fsi gi2½0;1 follows a strict factor
structure. As shown by Al-Najjar (1998), strict factor structure and no-arbitrage imply that for almost every asset (or a
portfolio of assets) i with payoff e
qi ,

11
Here, the entrepreneurs' preferred pair of comparability is defined as the one that maximizes their total surpluses rather than the Nash-equilibrium
outcome because we consider the entrepreneurs as a single political constituent who influences the standard setter as a coalition. Nevertheless, solving for
the pair of comparability as a Nash-equilibrium outcome yields qualitatively similar results.
12
We thank Frank Zhou for pointing it out at the 2019 Junior Accounting Theory Conference.

12
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

E½e
qi jI  ¼ Pi þ F

where Pi is the price of asset i, the constant F is the factor price, and by assumption, the loading on the factor F is 1 for all assets
in our setting. Given the above relationship between price and return, the payoff of holding any portfolio is a (weakly) mean
preserving spread of the payoff to the market portfolio.13 The price of market portfolio Pm must clear the market such that

ejI   Pm
a þ E½h
¼ 1;
rI VarðhejI Þ

which gives the risk factor F ¼ rI Varðh

ejI Þ. Hence, the risk-averse investors are only compensated for bearing non-diversifiable
risks, and the price of firm i is

Pi ¼ E½e
qi jI   F ¼ E½eqi jI   rI VarðhejI Þ:

Proof of Lemma 1. Given investors’ conjectured levels of effort b

a i and b
a for firm i and the economy-wide average,
respectively,

Pi ðsi ; sÞ ¼ E½e
qi jI   rI VarðhejI Þ

a i þ E½e
¼b di jsi  s þ E½ehjs  ba   rI Varðhejs  aÞ
tx tε 1
¼b
ai þ ðs  aÞ þ a Þ  rI
ða  b :
td þ tx i th þ tε th þ tε

a i ¼ a*i for all firms, where

In equilibrium, b

rE 1
a*i ¼ arg max E½Pi   VarðPi Þ  a2i
ai 2 2
 
tx 1 r 1 tε 1 tx 1
¼b
ai þ ðai  aÞ  rI  E þ  a2i
td þ tx th þ tε 2 th th þ tε td td þ tx 2
tx
¼ : ,
td þ tx

Proof of Proposition 1. The proof follows immediately after plugging in rI ¼ rE ¼ 0 to equations (11) and (12). ,
Proof of Proposition 2. From equations (11) and (12), we have
 2
dCEI 1 rI tε
¼ > 0; (A1)
dc t 2 th þ tε

dCEE vCEE dtε vCEE dtx

¼ þ
dc vtx dc vtε dc
8 9
>
> >
>
>
> " >
>
> 2  2 #  2 > (A2)
1< t2d t2x rE tx tε tε =
¼      rI ;
>
t> td þ tx 3 2 td þ tx th þ tε th þ tε >
>
>
> |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} >
>
>
: |fflfflfflfflfflffl
ffl {zfflfflfflfflfflffl
ffl} >
price risk effect risk premium effect ;
market monitoring effect
 
where tε ¼ tε(c) and tx ¼ tx(c) are defined in equation (5). It follows that sgn dCE
dc
E
¼ sgnðFðcÞÞ where

ð1  cÞt2d 2
ð1  cÞtd þ t
FðcÞ ≡ 2  rE  ð2rI  rE Þ :
ð1  cÞtd þ t cth þ t

h i
ð1cÞt2 ð1cÞtd þt 2
Let F1 ðcÞ ≡ 2 ð1cÞt þd t  rE and F2 ðcÞ≡ð2rI  rE Þ cth þt , then
d

13
The argument resembles the two-fund separation theorem in Ross (1978) and Connor (1984).

13
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

t þ t2
2t2d lim F2 ¼ ð2rI  rE Þ d
lim F1 ¼ r c/0 t
c/0 td þ t E  2
lim F1 ¼ rE
t
lim F2 ¼ ð2rI  rE Þ :
c/1 c/1 th þ t
2ðt þtÞ3 r 2t2 t2
As a result, we have lim F1 > lim F2 if and only if rE > t ðdt þtÞðIt þ2tÞd , and lim F1 < lim F2 for any rE, rI > 0. Hence, c ¼ 1 is
c/0 c/0 d d d c/1 c/1
not desirable for the entrepreneurs. Next, we characterize the conditions when the entrepreneurs prefer an interior level of
comparability. Observe that

vF1 tt2d
¼ 2 < 0;
vc ½ð1  cÞtd þ t2
v2 F1 ct3d
2
¼ 4 < 0;
vc ½ð1  cÞtd þ t3
 
vF2 ð1  cÞtd þ t vF2
¼ 2ðrE  2rI Þ ½t h t d þ tð th þ td Þ 0 sgn ¼ sgnðrE  2rI Þ;
vc ðcth þ tÞ3 vc
!
v2 F2 t ðcth þ tÞ þ 3th ½ð1  cÞtd þ t v2 F2
¼ 2ð2rI  rE Þ d ½ th t d þ t ðth þ td Þ 0 sgn ¼ sgnð2rI  rE Þ:
vc2 ðcth þ tÞ4 vc2

2ðt þtÞ3 r 2t2 t2

When rE  2rI, c*E 2ð0; 1Þ if rE > t ðdt þtÞðIt þ2tÞd ; otherwise c*E ¼ 0.
d d d
When rE < 2rI.

2ðt þtÞ3 r 2t2 t2

i. If rE > t ðdt þtÞðIt þ2tÞd : then c*E 2ð0; 1Þ; and
d d d

2ðt þtÞ3 r 2t2 t2

ii. If rE  t ðdt þtÞðIt þ2tÞd : if F1(c)  F2(c) cc 2 [0, 1], then c*E ¼ 0; and if F1(c) ¼ F2(c) has two interior solutions 0 < c1E <
d d d
c2E < 1 where c1E is the local minimum and c2E is the local maximum, then c*E ¼ 0 or c2E . ,

Proof of Proposition 3. Given the standard setter's objective function, let

dW dCEE dCEI
LðcÞ ≡ ¼ ð1  uÞ þu : (A3)
dc dc dc

Whenever c*E is interior, evaluating the first-order derivative at c ¼ c*E yields

  dCEI
L c*E ¼ u > 0;
dc c¼c*E

implying c* > c*E since CEI is monotonically increasing in c.

Next we derive the condition under which c* admits an interior solution. Rewriting the certainty equivalents of the two
constituents as functions of c, suppose to the contrary that c* ¼ 1, then for any c2ðc*E ; 1Þ, we have

ð1  uÞCEE ð1Þ þ uCEI ð1Þ > ð1  uÞCEE ðcÞ þ uCEI ðcÞ

CEE ðcÞ  CEE ð1Þ
⇔ u>
CEE ðcÞ  CEE ð1Þ þ CEI ð1Þ  CEI ðcÞ
CEE ðcÞ  CEE ð1Þ
⇔ uu
b ≡ sup
c2ðcE ;1Þ
* CE E ðcÞ  CEE ð1Þ þ CEI ð1Þ  CEI ðcÞ

since CEE(c)  CEE (1) > 0 and CEI (1)  CEI(c) > 0 by the proof of Proposition 2.
Further note that by L’Ho^ pital's rule, we have

CEE ðcÞ  CEE ð1Þ CE0E ðcÞ CE0E ð1Þ

lim ¼ lim 0 0 ¼ :
c/1 CEE ðcÞ  CEE ð1Þ þ CEI ð1Þ  CEI ðcÞ c/1 CEE ðcÞ  CEI ðcÞ CE0E ð1Þ  CE0I ð1Þ

Since CE0E ð1Þ < 0 and CE0I ð1Þ > 0, we conclude that u
b 2ð0; 1Þ.
To see that u b > 23, we plug in u ¼ 23 and evaluate L(1), which yields Lð1Þju¼2 < 0.
3
Now we proceed to examine how the interior c* changes with respect to u. When c* 2 (0, 1), we have

14
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

vL dCEI dCEE
¼  >0
vu c¼c* dc c¼c* dc c¼c*

dCEI dCEE
since dc c¼c*
> 0 and dc c¼c*
< 0. Further note that vL
vc c¼c* < 0 by local concavity. Therefore, by implicit function theorem,

dc* vL
u
¼  vvL > 0: ,
du vc
c¼c*

Proof of Corollary 1. The proof follows immediately from the proofs of Propositions 2 and 3. ,
Proof of Corollary 2. Plugging in equations (11) and (12) to the standard setter's objective function, we have
( 2   )
1 tx rE 1 tx
tx 1 tε 2  3u 1
W ¼ ð1  uÞ   þ  r :
td þ tx 2 td þ tx 2 td td þ tx th th þ tε 2ð1  uÞ I th þ tε

 
Comparing to expression (12) and following the proof of Proposition 2, we have sgn dW
dc
¼ sgnðGðcÞÞ, where
 
ð1  cÞt2d 2  3u ð1  cÞtd þ t 2
GðcÞ ≡ 2  rE  rI  rE ;
ð1  cÞtd þ t 1u cth þ t

and that when c* 2 (0, 1), if rE  23 u

1u rI , then c* is the only critical point; otherwise, c* is the only critical point or c* ¼ c2
where 0 < c1 < c2 < 1 are the two critical points. By implicit function theorem,

vG    
dc* dc* vG
¼  vx
vG
0 sgn ¼ sgn for x2frI ; rE ; th ; td ; tg;
dx vc
dx vx c¼c*
c¼c*

since vvcG c¼c*

< 0 by local concavity. In particular,
 *  
vG 2  3u ð1  c* Þtd þ t 2 dc 2
¼ 0 sgn ¼ sgn u  ;
vrI c¼c* 1u c* th þ t drI 3
 
vG th þ td td ð1  c* Þtd þ t
¼ *  c* þ1 :
vrE c¼c* c th þ t th þ td c* th þ t

    
u th t2d
Note that sgn Gðc* Þ  G thtþd td 1u rI  2 th td þtðth þtd Þ , then.
¼ sgn 23

i. If u  23, then c* > thtþd td ;

uÞ t t2 uÞ t t2
ii. If u < 23, then when rI < 2ð1 * td
23u th td þtðth þtd Þ, we have c > th þtd . Otherwise, plugging
h d
in rI ¼ 2ð1 h d
23u th td þtðth þtd Þ , we have
     
vG t ðt þ t Þ 4
ðð tð t þ t Þþt t Þr þ t 2
ðt 2 t ÞÞ t2 ð2t tÞ
vc c¼ d ¼  , implying sgn thtþd td  c* ¼  sgn vvcG c¼ td ¼ sgn rE  tðthdþt hÞþth t .
h d h d h d E d h
t 4
th þtd ðtðth þtd Þþth td Þ th þtd d d

* uÞ 2ð1uÞ t t2 * t t2 *
To conclude, if u  23 , then ddcr < 0; if u < 23, then when rI < 2ð1
23u th td þtðth þtd Þ , drE < 0; when rI > 23u th td þtðth þtd Þ , drE < 0 for
h d dc h d dc
E
td ð2th tÞ
2 * td ð2th tÞ
2
rE < tðth þt Þþth t and dr > 0 for rE > tðth þt Þþth t .
d
dc
d d d
E

   *  
vG 2  3u ð1  c* Þtd þ t 2 c* dc 2  3u
¼2 rI  rE 0 sgn ¼ sgn r  rE ;
vth c¼c* 1u c* th þ t c* th þ t dth 1u I
(   )
vG * ð1  c* Þtd þ 2t 2  3u ð1  c* Þtd þ t dc* 2  3u
¼ 2ð1  c Þ td þ rE  rI 0 > 0 when rE  r:
vtd c¼c*
*
½ð1  c Þtd þ t 2 1  u *
ðc th þ tÞ 2 dt d 1u I

 *  
When rE < 23 u
1u rI , sgn dt
dc ¼ sgn vvtd dW since v dW < 0.
d dc c¼c* vc dc c¼c*

15
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

( " # )
v dW t2d ½2ð1  c* Þtd  t td th 2  3u th
¼ ð1  uÞt  rE þ þ r <0
vc dc c¼c* ½ð1  c* Þtd þ t4 ½ð1  c* Þtd þ t3 ðc* th þ tÞ3 1  u I ðc* th þ tÞ3

2  3u th t2 ½2ð1  c* Þtd  t th t2 ½2ð1  c* Þtd  t

r þ d rE þ d
1  u I ðc* th þ tÞ3 ½ð1  c* Þtd þ t4 *
ðc th þ tÞ ½ð1  c* Þtd þ t4
3
0 rE > td th > td th
* 3
þ * *
þ *
½ð1  c Þtd þ t ðc th þ tÞ3 ½ð1  c Þtd þ t3
ðc th þ tÞ3
2ð1  c* Þtd  t
0 rE > td
ð1  c* Þtd þ t
v dW tð1  c* Þ ttd ð1  c* Þ
¼ ð1  uÞ *
ð½ð1  c* Þtd þ trE  td ½ð1  c* Þtd  2tÞ > ð1  uÞ >0
vtd dc c¼c* ½ð1  c Þtd þ t4
½ð1  c* Þtd þ t3
dc* 2  3u
0 >0 when rE < r:
dtd 1u I
 
vG 2  3u ð1  c* Þtd þ t * ð1  c* Þt2d
¼ 2 rE  rI ½c ðt h þ t Þ  t   2
vt c¼c* 1u ðc* th þ tÞ3
d d
½ð1  c* Þtd þ t2

 2  3u ð1  c* Þtd þ t 2
c* ðth þ td Þ  td
rE  r 2 þ 1  rE
FOC 1u I c* th þ t c* th þ t
¼ :
ð1  c* Þtd þ t

h i h i
td þt 2 cðt þt Þt
Note that ð1cÞ
*
cth þt 2 hcth þd t d þ 1  1 with the upper bound achieved at c ¼ thtþd td . Hence, dc dt
< 0 if u < 23 and
23 u
rE > 1u rI . ,
Proof of Corollary 3. The proof follows immediately after plugging in equation (5). ,
Proof of Lemma 2. Recall that the price of each firm given all available information I ≡fsk;i gi2½0;1;k2f1;2g follows a strict
2
factor structure. The risk premium is determined by the non-diversifiable risk of market portfolio, e h þ 12 ðeh1 þhe2 Þ. Denote Sh
and Ss as the 2  2 variance-covariance matrix of fh e þe e e
h1 ; eh þhe2 g and fs1 ; s2 g, respectively, and Shs as the 2  2 covariance
matrix of the two sets of variables. Then given investors’ conjectured levels of effort b a k;i and b
a k for firm i and the group k
average, respectively,
h i  
  1
Pk;i sk;i ; sk ; sk0 ¼ E eqk;i I  rI Var eh þ ðhe1 þ he2 Þ I
2
txk  
¼b
a k;i þ s  s þ bkk ðsk  b a k0 Þ  rI mu Shjs m;
a k Þ þ bkk0 ðsk0  b
tdk þ txk k;i k

where m is a 2  1 vector with 12 in each element that denotes the mass of each group, and the coefficients on the average
signals and the residual uncertainties about the undiversifiable common shocks fh e þh
e1 ; e
h þeh2 g are:
!
  1 tε1 ððth þ th1 Þðth2 þ tε2 Þ þ th2 tε2 Þ th1 th2 tε2
bij 22 ≡ Shs S1
s ¼ ;
B th1 th2 tε1 tε2 ððth þ th2 Þðth1 þ tε1 Þ þ th1 tε1 Þ
!
1 ðth þ th1 Þðth2 þ tε2 Þ þ th2 tε2 th1 th2
Shjs ≡ Sh  Shs S1 u
s Shs ¼ ;
B th1 th2 ðth þ th2 Þðth1 þ tε1 Þ þ th1 tε1

where B ¼ ðth þ th1 þ th2 Þtε1 tε2 þ th1 th2 ðtε1 þ tε2 Þ þ th ðth1 th2 þ th1 tε2 þ th2 tε1 Þ.
The rest of the proof follows in a similar way to the proof of Lemma 1 and is hence omitted.

References

Al-Najjar, Nabil I., 1998. Factor analysis and arbitrage pricing in large asset economies. J. Econ. Theor. 78 (2), 231e262.
Bertomeu, Jeremy, Cheynel, Edwige, 2016. Disclosure and the cost of capital: a survey of the theoretical literature. Abacus 52 (2), 221e258.
Bullen, Halsey G., Crook, Kimberley, 2005. A New Conceptual Framework Project: Revisiting The Concepts (Financial Accounting Standards Board).
Butterworth, John E., 1972. The accounting system as an information function. J. Account. Res. 10 (1), 1e27.
Chen, Bingyi, Kurt, Ahmet C., Wang, Irene Guannan, 2020. Accounting comparability and the value relevance of earnings and book value. J. Corp. Account.
Finance 31 (4), 82e98.
Connor, Gregory, 1984. A unified beta pricing theory. J. Econ. Theor. 34 (1e2), 13e31.

16
S. Wu, W. Xue Journal of Accounting and Economics 75 (2023) 101535

Core, John E., Guay, Wayne R., Larcker, David F., 2003. ‘Executive Equity Compensation and Incentives: A survey.’ Federal Reserve Bank Of New York Economic
Policy Review, pp. 27e50.
De Franco, Gus, Kothari, Sagar P., Verdi, Rodrigo S., 2011. The benefits of financial statement comparability. J. Account. Res. 49 (4), 895e931.
Dye, Ronald A., 1990. Mandatory versus voluntary disclosures: The cases of financial and real externalities. Account. Rev. 65 (1), 1e24.
Dye, Ronald A., Sridhar, Sri S., 2004. Reliability-relevance trade-offs and the efficiency of aggregation. J. Account. Res. 42 (1), 51e88.
Dye, Ronald A., Sridhar, Sri S., 2007. The allocational effects of the precision of accounting estimates. J. Account. Res. 45 (4), 731e769.
Dye, Ronald A., Sridhar, Sri S., 2008. A positive theory of flexibility in accounting standards. J. Account. Econ. 46 (2e3), 312e333.
Fang, Vivian W., Iselin, Michael, Zhang, Gaoqing, 2022. Consistency as a means to comparability: Theory and evidence. Manag. Sci. 68 (6), 4279e4300.
FASB, 2018. Conceptual Framework for Financial Reporting. Financial Accounting Foundation.
Feltham, Gerald A., 1972. Information Evaluation. American Accounting Association).
Friedman, Henry L., Heinle, Mirko S., 2016. Lobbying and uniform disclosure regulation. J. Account. Res. 54 (3), 863e893.
Gao, Pingyang, 2010. Disclosure quality, cost of capital, and investor welfare. Account. Rev. 85 (1), 1e29.
Gao, Pingyang, 2013. A measurement approach to conservatism and earnings management. J. Account. Econ. 55 (2e3), 251e268.
Gao, Pingyang, 2019. Idiosyncratic information, moral hazard, and the cost of capital. Contemp. Account. Res. 36 (4), 2178e2206.
Gao, Pingyang, Zhang, Gaoqing, 2019. Accounting manipulation, peer pressure, and internal control. Account. Rev. 94 (1), 127e151.
Gao, Pingyang, Jiang, Xu, 2020. The economic consequences of discrete recognition and continuous measurement. J. Account. Econ. 69 (1), 101250.
Gao, Pingyang, Jiang, Xu, Zhang, Gaoqing, 2019. Firm value and market liquidity around the adoption of common accounting standards. J. Account. Econ. 68
(1), 101220.
Guttman, Ilan, Kadan, Ohad, Kandel, Eugene, 2006. A rational expectations theory of kinks in financial reporting. Account. Rev. 81 (4), 811e848.
Hirshleifer, Jack, 1971. The private and social value of information and the reward to inventive activity. Am. Econ. Rev. 61 (4), 561e574.
Holmstrom, Bengt, 1982. Moral hazard in teams. Bell J. Econ. 13 (2), 324e340.
Holmstrom, Bengt, Tirole, Jean, 1993. Market liquidity and performance monitoring. J. Polit. Econ. 101 (4), 678e709.
IASB, 2018. Conceptual Framework for Financial Reporting. IFRS Foundation.
Jennings, Jared, Seo, Hojun, Soliman, Mark, 2020. The market's reaction to changes in relative relative performance ranking. Rev. Account. Stud. 25,
672e725.
Jiang, John Xuefeng, Wang, Isabel Yanyan, Wangerin, Daniel D., 2018. How does the FASB make decisions? A descriptive study of agenda-setting and the role
of individual board members. Account. Org. Soc. 71, 30e46.
Kurt, Ahmet C., 2020. ‘Don't Overlook Accounting Comparability. CFA Institute Blogs. Available at: https://blogs.cfainstitute.org/investor/2020/08/18/dont-
overlook-accounting-comparability/.
Liang, Pierre Jinghong, Zhang, Xiaojun, 2008. Information economics and accounting measurements: a blueprint for scholarly research. China Accounting
Review 6 (1), 107e107.
Ross, Stephen A., 1978. Mutual fund separation in financial theory e the separating distributions. J. Econ. Theor. 17 (2), 254e286.
Sunder, Shyam, 2010. Adverse effects of uniform written reporting standards on accounting practice, education, and research. J. Account. Publ. Pol. 29 (2),
99e114.
Van Riper, Robert, 1994. Setting Standards for Financial Reporting: FASB and the Struggle for Control of a Critical Process. Quorum Books Westport, CT).
Wang, Clare, 2014. Accounting standards harmonization and financial statement comparability: Evidence from transnational information transfer. J. Ac-
count. Res. 52 (4), 955e992.
Watts, Ross L., Zimmerman, Jerold L., 1978. Towards a positive theory of the determination of accounting standards. Account. Rev. 53 (1), 112e134.
Zhang, Guochang, 2013. Accounting standards, cost of capital, resource allocation, and welfare in a large economy. Account. Rev. 88 (4), 1459e1488.

17