Review of Accounting Studies, 3, 35–40 (1998)
1998 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Discussion of “Earnings Management and the
Graduate School of Business, Columbia University, New York, NY 10027
Arya, Glover, and Sunder (AGS) contribute to the earnings management literature along
two dimensions. First, they classify existing explanations for earnings manipulation, based
on the assumption of the revelation principle that is violated. Second, they introduce a
model where allowing a manager to manipulate earnings serves as a commitment device.
They show that both the owners and the manager can beneﬁt from earnings management (a
Pareto improvement). My discussion ﬁrst deals with the general phenomenon of earnings
management and then with the speciﬁcs of the AGS model.
The stated goal of ﬁnancial reporting is to provide information for decision making pur-
poses. Decision makers are not a homogeneous group, and in particular, include both
insiders and outsiders, who may have different informational needs (see Beaver, 1998). An
accounting system cannot disclose all possible relevant information. Financial statements
can contain only an aggregated summary of the large number of transactions that have ac-
tually occurred during the period. As a result, preparers must make discretionary choices,
deciding what kind of information is important or useful. Many users of accounting infor-
mation treat reported earnings as either a “sufﬁcient statistic” for a ﬁrm’s performance or,
at least, as relevant information for decision making. However, using one summary item,
like earnings, from the entire information system introduces a measurement problem; what
constitutes a “correct” earnings report? In other words, one would like to identify a metric
for earnings measurement that would be most relevant for decision makers. Heterogeneous
users and aggregated information make such a task impossible. Even in a simple world
where there are no conﬂicts of interests, reported earnings cannot coincide with any speciﬁc
user’s needs for any given decision.
In addition to the measurement problems discussed above, one should consider reporting
biases. In any real economic setting there exist conﬂicts of interests, either within the ﬁrm
or somewhere else in the market. For example, any publicly available information could
be used by those whose interests conﬂict with the interests of the disclosing ﬁrm, such as
competitors or regulators. As a result, a ﬁrm might use its accounting reporting system
strategically to inﬂuence users’ information sets.
Given the possibility of reporting biases, interpretation of ﬁnancial reporting becomes
more complicated. That is, even if there were an unambiguous superior method for earnings
measurement (or for any other accounting variable), ﬁnancial reports might not contain the
earnings’ realized value, because the provider prefers to disclose something else. This is
where the notion of earnings management is introduced. Earnings management is a process