The risk premium on ordinary shares is investigated, by studying the total returns on ordinary shares, and on both long-term and short-term fixed-interest investments over the period 1919 to 1994, and by analysing the various components of that return. The total returns on ordinary shares exceeded those on fixed-interest investments by over 5% p.a. on a geometric mean basis and by over 7% p.a. on an arithmetic mean basis, but it is argued that these figures are misleading, because most of the difference can be accounted for by the fact that price inflation turned out to be about 4.5% p.a. over the period, whereas investors had been expecting zero inflation.

Quotations from contemporary authors are brought forward to demonstrate what contemporary attitudes were. Simulations are used along with the Wilkie stochastic asset model to show what the results would be if investors make various assumptions about the future, but the true model turns out to be different from what they expected. The differences between geometric means of the data and arithmetic means are shown to correspond to differences between using medians or means of the distribution of future returns, and it is suggested that, for discounting purposes, medians are the better measure.