Do We Really Know That Financial Markets Are Efficient?

This paper examines the power of statistical tests commonly used to examine the efficiency of speculative markets. It shows that for markets with "long horizons" such as the stock markets, or the market for long term bonds, these tests have very low power. Market valuations can differ substantially and persistently from the rational expectation of the present value of cash flows without leaving statistically discernible traces in the pattern of ex-post returns. This observation also suggests that speculation is unlikely to insure rational valuations, since similar problems of identification plague both financial economists and would-be speculators.