The VAR methodology of Campbell and Shiller !(1989) is employed under
four different assumptions regarding equilibrium expected returns to a
ssess the efficiency of the UI; stock market. In our first model. equi
librium expected (real) returns are assumed to be constant, while in t
he second model. excess returns are assumed to be constant. The next t
wo models assume that equilibrium returns depend upon a time-varying r
isk premium which varies with tile conditional expectation of the retu
rn variance (i.e. the CAPM). Our results yield evidence of short-termi
sm, even when the key assumption of a time-invariant discount rate is
relaxed.