Review of Accounting Studies, 4, 259–264 (1999)
1999 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Discussion of “Asset Valuation and Performance
Measurement in a Dynamic Agency Setting”
RICHARD A. LAMBERT
The Wharton School, University of Pennsylvania
One of the most encouraging things about this paper is that the authors are able to formulate
and solve a genuine multiperiod model to address performance measurement issues. In
many respects, we need a multiperiod model to have an accounting problem to talk about.
Insingleperiodmodels,cashﬂowandaccrualaccountingnumbers are identical; multiperiod
models are therefore essential to the study of accounting performance measures.
the obvious importance, not much work has been done on multiperiod models in the agency
literature. The reason is, of course, tractability problems. In most multiperiod models,
numerous technical issues arise that are often tangential to the accounting or performance
measurement issues that we would like to focus on as accounting researchers. For example,
even with models where everything seems to be independent over time, we have to worry
about borrowing and lending, wealth effects, randomization, how the contact parameters in
one period depend on realizations from prior periods, the form of the contract, the ability
to commit to long term contracts, etc. Even information signals that would seem to be
informationally “meaningless” sometimes play an important role in helping to randomize
actions, or in coordinating the actions of different parties.
Some papers have addressed multiperiod issues in an ad hoc fashion by analyzing models
that are really one period models, but in which the true outcome is not observed till beyond
the contracting horizon. In these models, some of the agent’s actions may have only a
“short term” effect that is properly captured by that period’s accounting earnings, whereas
other actions may have longer term effects that the current period accounting number does
not capture. These papers can analyze performance measurement issues, the congruity of a
performance measure with the realoutcome, the motivational effects of short term measures,
etc., butthe models are not truly multiperiod. See Bushman and Indjejikian (1993), Feltham
and Xie (1994), Datar et al. (1999), and Feltham and Wu (1999) for examples of single
period models of performance measurement where the agent is responsible for multiple
Recently there has been renewed interest in multiperiod models that has been spurred in
part by technical breakthroughs, and in part by an increased willingness by researchers to
be more restrictive (less general) in their modeling. In some cases, researchers have placed
exogenous limitations on the contract form (e.g., linear contracts), or exogenous restrictions
on the form of the agent’s utility function (e.g., the negative exponential utility function).
There have been two branches of recent multiperiod papers that employ models and
technology that are similar in many respects, yet yield very different conclusions regarding