Testing for Short Termism in the UK Stock Market: A Reply

Abstract

In their useful comment Satchell and Damant and (hereafter S and D) raise two issues which, they argue, could account for the apparent evidence for short termism that is excess discounting of expected cash flows that accrue further in the future presented in my paper in this JOURNAL (November I 993). Their main point, which I think addresses the central issue in interpreting my own (and indeed most other) evidence of market inefficiency, concerns the term structure of risk premia. My comments focus almost exclusively on this issue. But a second point raised by S and D is worth addressing first. They argue that since the results reported in my paper are based on analysis of 477 companies which reported in each year (I977-90) survivorship bias will have an impact on the results. I am not convinced of this. The results in my paper point to the existence of excess discounting of longer-term cash flows relative to shorterterm cash flows where all cash flows accrue to companies that in fact survived; there is prima facie evidence of short termism when we confine the sample of companies to those that made it through the I980s. Evidence of short termism which came from comparing a group of companies that did survive with a random sample of companies (including some that went on to fail) would not be convincing precisely because survivorship bias would mean that companies with the longest record of dividends the survivors were those most likely to have had good luck and whose ex ante valuation (share price) would most seem to undervalue their ex post profits; there would then be a spurious correlation between companies whose share price appeared low, relative to future dividends, and those for which data on dividend flows for the longest period ahead were available. But by using data only on survivors this problem is sidestepped. But the main point of S and D's note is that evidence of excess discounting may simply reflect a non-horizontal term structure of risk premia risk premia on cash flows accruing further in the future may be higher than the risk premia on near term cash flows. The really useful thing about their note is that it uses an explicit (and widely used) model of asset pricing to derive an analytic solution for the term structure of risk premia. This is helpful because the observation that evidence of excess discounting may simply reflect appropriate adjustments for risk in itself is rather empty; it is only in the context of a correctly specified asset pricing model that the plausibility of the claim that varying risk premia account for the pattern of share prices can be assessed. By using a model of asset pricing firmly based on maximising behaviour to derive the term structure of risk premia S and D help make the issue of whether to interpret my results as saying something about market efficiency amenable to