The use of predetermined variables to represent public information and
time-variation has produced new insights about asset pricing models,
but the literature on mutual fund performance has not exploited these
insights. This paper advocates conditional performance evaluation in w
hich the relevant expectations are conditioned on public information v
ariables. We modify several classical performance measures to this end
and find that the predetermined variables are both statistically and
economically significant. Conditioning on public information controls
for biases in traditional market timing models and makes the average p
erformance of the mutual funds in our sample look better.