This paper contributes to the growing literature on mean reversion in stock markets by examining a newly constructed Danish data set for the period 1922-95. Variance ratio tests clearly reject the random walk hypothesis at the 2-year horizon, that is, the riskiness of a 2-year investment is significantly less than twice the risk of a 1-year investment. Variance ratio tests for 3- and 4-year horizons are not significant under conventional significance levels, whereas autocorrelation tests of the joint hypothesis that there is departure from random walk at all horizons tend to reject the random walk hypothesis and support the mean reversion hypothesis.
