Forecasting of Statistics
Nearly two thousand years have passed since a census decreed by Caesar Augustus become part of the greatest story ever told. Many things have changed in the intervening years. The hotel industry worries more about overbuilding than overcrowding, and if they had to meet an unexpected influx, few inns would have a manager to accommodate the weary guests. Now it is the census taker that does the traveling in the fond hope that a highly mobile population will stay long enough to get a good sampling. Methods of gathering, recording, and evaluating information have presumably been improved a great deal. And where then it was the modest purpose of Rome to obtain a simple head count as an adequate basis for levying taxes, now batteries of complicated statistical series furnished by governmental agencies and private organizations are eagerly scanned and interpreted by sages and seers to get a clue to future events. The Bible does not tell us how the Roman census takers made out, and as regards our more immediate concern, the reliability of present day economic forecasting, there are considerable differences of opinion. They were aired at the celebration of the 125th anniversary of the American Statistical Association. There was the thought that business forecasting might well be on its way from an art to a science, and some speakers talked about newfangled computers and high-falutin mathematical system in terms of excitement and endearment which we, at least in our younger years when these things mattered, would have associated more readily with the description of a fair maiden. But others pointed to the deplorable record of highly esteemed forecasts and forecasters with a batting average below that of the Mets, and the President-elect of the Association cautioned that “high powered statistical methods are usually in order where the facts are crude and inadequate, the exact contrary of what crude and inadequate statisticians assume.” We left his birthday party somewhere between hope and despair and with the conviction, not really newly acquired, that proper statistical methods applied to ascertainable facts have their merits in economic forecasting as long as neither forecaster nor public is deluded into mistaking the delineation of probabilities and trends for a prediction of certainties of mathematical exactitude.