Some distributional properties of monthly stock returns in Sweden 1919-1990

This paper examines the distributional properties of a newly constructed dataset oj monthly returns on the Swedish stock market. The standard assumptions that stock returns are log-normally distributed, serially independent, non-seasonal and homoscedastic are all rejected by data. Swedish stock returns are more likely to belong to a peaked and fat-tailed distribution, with positive first order autocorrelation, strong seasonality and changing volatility over time. These results are well in line with what has been reported from other national stock markets. Our major conclusion is that, given the failure of data to meet the usual distributional assumptions in finance, it may be worthwhile to pay more attention to modeling both the return generating process and the volatility generating process for the market index, instead of simply assuming a strict random walk model.