Herman Stekler has hypothesized that longerterm predictions have smaller variability, relative to actual changes, than do nearer-term forecasts (in a paper presented to the Fourth International Symposium on Forecasting, 1984). The authors have examined a set of OECD forecasts and found support for Stekler's hypothesis [Economics Letters, Vol. 19, 1985, pp. 14143]. This note further tests the hypothesis using data from the ASA-NBER business outlook survey. The survey is conducted among members of the Business and Economics section of the American Statistical Association who forecast on a regular basis. The data used here are those available through Citibase and cover the period from the second quarter of 1970 to the last quarter of 1984. Citibase reports the median forecasts to avoid the effect of the occasional extreme forecasts made by some of the 40 to 50 panel members. As the measure of the variability of predicted changes to the variability of the actual changes, the authors use the ratio of the standard deviations. Thus, define FASDi, i = 2,3,4, as the ratio of the standard deviation of Fi to the standard deviation of the actual percentage change series, calculated over the 59 quarter sample period. Estimates of FASD2, FASDa and FASD4 (in that order) are as follows: nominal GNP (.44, .45, .42); real GNP (.60, .47, .32); implicit price deflator for GNP (.76, .70, .69); corporate profits after taxes (.45, .35,. 25); unemployment rate (.54, .50, .31); index of industrial production (.51, .41, .33); and new private housing starts (.57, .52, .58). There are various ways of testing the hypothesis. The authors give two. The null hypothesis is that for any forecast variable it is equally likely that FASD3 will be greater or less than FASD2 and equally likely that FASD4 will be greater or less than FASD3. Thus, the probability of observing FASD4 ~ FASD3 FASD2, which is predicted by the hypothesis, is 0.5 times 0.5 equals 0.25. The authors have seven variables and for five of them have observed the hypothesized pattern. Using the binomial distribuiton, one can see that the probability of observing the pattern by chance is .0115. Alternatively, one may pool all pairs. The hypothesis is that FASD4 ~ FASD3 and FASD3 ~ FASD2. The anticipated signs are observed for t2 of the 14 pairs and (again using the binominal distribution) the probability is .0065. Using either test, the hypothesis is not rejected.