Review of Quantitative Finance and Accounting, 19: 273–290, 2002
2002 Kluwer Academic Publishers. Manufactured in The Netherlands.
Testing for Income Smoothing Using the Backing
Out Method: A Review of Speciﬁcation Issues
STEVE C. LIM
Texas Christian University
Abstract. This study investigates potentially spurious correlation in prior studies of income smoothing which
use a research method that we call backing out.IfE is reported income, Y a discretionary earnings component
hypothesized to smooth income, and T the smoothing target, the backing out method consists of regressing Y
on E − Y − T . A negative regression coefﬁcient is interpreted as evidence that Y is being managed to smooth
earnings. We argue that the negative regression coefﬁcient may simply reﬂect the positive correlation between E
and T , which may or may not be the result of manipulating Y .IfE is not manipulated by Y , the negative regression
coefﬁcient reﬂects measurement errors in E − Y as an estimate of unmanaged earnings.
We replicate the backing out method of prior studies and use each of the accrual and cash ﬂow components
of earnings in the Statement of Cash Flows for Y . We ﬁnd that the regression coefﬁcient, hypothesized to be
negative because of smoothing with Y , is always negative and its magnitude depends on the variances of Y and
E − T (smoothing error) independent of the nature of Y (discretionary or nondiscretionary). We conclude that the
negative coefﬁcient on E − Y − T is equally consistent with three possibilities: (1) managers smooth earnings with
Y , (2) managers smooth earnings with something other than Y , and (3) managers do not smooth earnings at all.
Key words: income smoothing, speciﬁcation issues, backing out method
JEL Classiﬁcation: M41
This study analyzes speciﬁcation issues in prior studies of income smoothing that use a
popular research design we call backing out. Although caveats about this design have been
raised in earlier papers, the technique continues to be used. Our objective is to review
and clarify the issues regarding the backing out method. Given the practical importance of
identifying the instrument of earnings management, it is useful to have an explicit statement
about what conclusions follow when the null hypothesis of no income smoothing is rejected
using the backing out method.
Researchers frequently hypothesize that a particular component of earnings such as gains
from asset sales (a representative approach) or an aggregate of discretionary accruals (a port-
folio approach) is being managed to smooth income.
Researchers then look for evidence
Address correspondence to: Steve C. Lim, M. J. Neeley School of Business, Texas Christian University,
TCU Box 298530, Fort Worth, Texas 76129.