This thesis examines the predictability of asset prices for an Australian
investor. Evidence supporting the mean reversion alternative to the random walk
hypothesis is presented, with a discussion of potential models, both linear and nonlinear.
The normality and homoscedasticity assumptions are investigated and their
use in asset models is validated. A study of fund performance is carried out and value
is found to be added by timing asset allocation but not by stock selection, though
there is no correlation between past and present rankings of managers. The difficulty
of proving mean reversion or reversion to trend, other than for large deviations or
extremes, and the actual performance by managers, implies a strategy of allocation at
these extremes. That is, managers should adhere to their policy portfolios and let
markets run short term; making appropriate large strategic moves when markets have
moved to extremes.