The Variance of Property Returns: Some Problems of Time‐Weighted Measures

Suggests that the use of the geometric mean as a measure of average return on investment presents problems for estimating the variance as a measure of risk. Notes that the use of a measure based on the arithmetic mean seems an uncomfortable compromise. Shows that measures based on the geometric mean are also systematically biased in the case of log normal returns. Concludes that this can have major consequences for investment decision‐making and portfolio selection.