Review of Accounting Studies, 3, 69–71 (1998)
1998 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Discussion of “Brand Values and Capital Market
JAMES A. OHLSON
Stern School of Business, New York University, N.Y.C., N.Y. 10012
The paper by Barth, Clement, Foster and Kasznik (1998) (BCFK henceforth) raises the
interesting questionwhether it makessense toaccount forinternally developed brand names,
besides putting them equal to zero. The issue is of obvious interest since nobody challenges
that brand names are economic assets. As the authors point out, the lack of ﬁnancial
statement recognition is generally perceived to be due to an expected low reliability. Would
investors really pay attention to such asset values if they were estimated and part of GAAP?
The question suggests that one ought to consider empirical evidence that bears on this
reliability hypothesis as it relates to brand names. The authors evaluate the reliability of
brand name values developed by Financial World (FW); they use a fairly standard empirical
approach in which prices and returns are dependent variables, and there are controls for
basic accounting variables such as earnings and book values.
What can we make out of the study that goes beyond the statistical ﬁndings? More
speciﬁcally, ﬁrst, does the study test the value-relevance of FW’s brand name estimates,
and, second, does this study have any policy implications?
II. On the Concept of Value-Relevance
For many years, empirical accounting research has generated an almost endless stream
of papers dealing with the “value-relevance” of various items found in ﬁnancial reports.
The typical study focuses on two sets of empirical evaluations, as does the BCFK paper.
First, regress market value on earnings, book value, and the item of interest (in this case
). Second, regress (market) returns on some earnings variable(s) and the item of
interest, all normalized by the start-of-period price. Tests of “value-relevance” then depend
on the statistical signiﬁcance of the coefﬁcient associated with the variable of interest. In
the BCFK study, both sets of tests yield statistical signiﬁcance associated with BRANDS
and the authors conclude that the assessment of brand values can be reliable and thus
I have always found this research paradigm problematic and, at best, somewhat contrived.
The issue that needs to be dealt with, but never is, concerns what we want to control for
in concept. Is it obvious that we ought to control for earnings book value in the price
regression? Why, at least as a ﬁrst cut, do researchers leave out variables of obvious
interest, such as measures of in prior periods’ earnings growth, analysts’ expectations of