英语论文网

equivalence result and identify a restricted set
of accounting measures that meet a certain `axiomatic' property, as follows. When considering
candidate earnings numbers with the intention of predicting ¯rm value, we require that as
the chosen earnings number increases, this rationally results in higher estimates of future ¯rm
value. We thus propose that this simple monotonicity property should be satis¯ed by candidate
earnings measures, on the grounds that investors will question any measurement methods of an
earnings number for which current higher earnings, can mean lower ¯rm value in the future.
Initially, one may suspect that satisfaction of this seemingly quite mild axiomatic condition
will not be particularly discriminating and that many earnings measures will satisfy the axiom.
However, interestingly, we ¯nd that the established earning measure used in the literature
(residual income) fails to satisfy the axiom and show how an alternative income measure based
upon the established q-theory of income does satisfy the axiom. Clearly, with one simple axiom
we cannot provide a way to discriminate between all possible earnings measures. However, we
suggest that unlike the FO approach which provides no discrimination, our analytical approach
is amenable to testing the satisfaction of additional well-speci¯ed axiomatic requirements and
so o®ers the ability to re¯ne the number of candidate earnings numbers that satisfy a chosen
set of axioms1.
In addition to providing a theoretical means to discriminate between alternative earnings
measures, our approach also contributes to empirical issues. In particular, since our approach is
based upon multi-period optimization, we are able to derive comparative statics results which
explain in a constructive way, why for instance, earnings response is non linear. In particular
we show why for ¯rms that are expanding aggressively, the earnings response coe±cient may be
quite low. Since our model is based upon optimizing behaviour we believe we may o®er a superior
explanation for the role of earnings in estimating future value in such settings than those
researchers who simply conclude that a low earnings response coe±cient may be interpreted
1We believe this to be of some importance at this time, given the active debate concerning the overall
desirability of comprehensive and other earnings measures.
2
as evidence of the lack of usefulness of earnings numbers and rush to explore the explanatory
power of non ¯nancial performance measures.
1.1 Real Options and the Feltham-Ohlson Model
In our model management need to evaluate real options embedded within typical investment
decisions. We review an established model in section two which derives the q-theory of investment
in such a setting. In section three we introduce a new investment model in which
real options naturally arise and can be solved for optimally. This analysis allows us to make
precise statements about expected ¯rm future value and leads us naturally to think about an
alternative measure of income based upon q-theory. In section four we then consider how an
investor could utilize alternative income measures to forecast future ¯rm value. We show that
estimates based on residual income are subject to `hysteresis e®ects', and expected future ¯rm
value can tak