4 Forecasting: benefits, practices, value, and limitations162
Mr. Buffett said his advice for the cash left to his wife was that 10 per cent should go to short-term government bonds and 90 per cent into a very low-cost S&P 500 index fund.
The purpose of this unique article is to provide an encyclopedic knowledge about the various aspects of forecasting. In this article, there are more than 140 sections and subsections, with more than 2,100 references, written by 80 of some of the best-known forecasting researchers and practitioners in the world, making it into a selective, encyclopedic piece covering, into a single source, a great deal of the available knowledge about the theory and practice of forecasting. We hope that this article will serve as an easy-to-use reference source. We aim to convert it into an online resource that will be regularly updated as new information becomes available.
But some people argue if there is any value in attempting to predict the future and if forecasting is any different than fortune telling, given the large numbers of mistaken forecasts made in the past, including our inability to accurately predict the progression of COVID-19 and its economic and human consequences? What is, therefore, the usefulness of a paper like the present one when crystal balling is not possible, and uncertainty reigns? It is the aim of this concluding article to set the record straight, explaining the benefits and practical value of forecasting while reporting its limitations too.
The Myriad of Forecasts: All planning and the great majority of decisions we make require forecasting. Deciding what time to get up in the morning, not to be late for work implies a correct prediction of the commuting time to go to the office. Determining what to study is another decision requiring elaborate predictions about the demand for future jobs decades away. In the business world, firms must decide/forecast how many products to manufacture, the price they should be sold, how much money to spend on advertising and promotion, how much and in what type of new technologies to invests and a plethora of other future-oriented decisions requiring both predictions and assessing their inevitable uncertainty. Whether we like it or not, we have no choice but making these forecasts to benefit as much as possible from their value, knowing perfectly well that all predictions are uncertain while some may turn out to be wrong.
The Pervasiveness of Uncertainty: Apart from some areas of hard sciences, all other forecasts are uncertain and must be accompanied with a measure of its magnitude, expressed as a prediction interval, or as a probability distribution around the most likely forecast. Although the value and usage of forecasts is clear, that of uncertainty is not. Worse, it becomes an unwelcome source of anxiety whose usefulness is misunderstood. Executives want to know the exact sales of their firm for next month to set up their production schedule. Instead, they are given prediction intervals (PIs) around such forecast and told that most of the time, sales will be within this interval, assuming the fluctuations follow some distributional assumptions. They argue that forecasting must decrease, not amplify, uncertainty and that the PIs are too wide and ‘uninformative’ to be used for making practical business decisions. The trouble is that these PIs are based on past fluctuations and present the best estimation of future uncertainty, even if they seem too wide. Worse, empirical research has shown that they are too narrow, underestimating uncertainty often considerably.
Assessing Uncertainty and Dealing with its Implied Risks: Uncertainty entails risks, requiring action to minimise their negative consequences. There are two kinds of uncertainty that can be illustrated by a commuting example. The first relates to fluctuations in the commuting time under normal driving conditions when there are no serious accidents, road works or major snowstorms. Such fluctuations are small and can be captured by a normal curve that allows to balance the risk of arriving earlier or later than the desired time. In the opposite case, uncertainty is fat-tailed and hard to estimate, as delays can be substantial depending upon the seriousness of the accident or that of the snowstorm while the risk of being early to work is smaller than being late. Moreover, such risk is substantially different when going to the airport to catch a flight, requiring starting much earlier than the average time it takes to go to the airport to minimise the risk of missing the flight.
More Accurate Ways of Forecasting and Assessing Uncertainty: Extensive empirical research, including forecasting competitions, has shown that systematic approaches improve the accuracy of forecasting and the correct assessment of uncertainty resulting in substantial benefits when compared to ad-hoc judgmental alternatives (S. Makridakis et al., 2020). The biggest advantage of such approaches is their ability to identify and estimate, in a mathematically optimal manner, past patterns and relationships that are subsequently extrapolated to predict their continuation, avoiding the over optimism and wishful thinking associated with judgmental approaches. At the same time, it must be clear that the accuracy of the forecasts and the correctness of uncertainty will depend on the established patterns/relationship not changing much during the forecasting period.
Using Benchmarks to Evaluate the Value of Forecasting: The accuracy of the forecasts and the correct assessment of uncertainty must be judged not on their own but in comparison to some simple, readily available benchmarks. In stock market forecasts, for instance, the accuracy of predictions is compared to that of today’s price used as the forecast for future periods. Empirical comparisons have shown that such a benchmark beats the great majority of professional forecasters, hence Buffet’s advice in the epigram for his wife to invest in a low-cost index fund that selects stocks randomly. In weather forecasting, meteorologists are judged by the improvement of their forecasts over the naive prediction that tomorrow’s weather will be the same as today.
Concluding remark: Accepting the advantages and limitations of systematic forecasting methods and most importantly avoiding any exaggerated expectations of what it can achieve is critical. Such methods do not possess any prophetic powers, they simply extrapolate established patterns and relationships to predict the future and assess its uncertainty. Their biggest advantage is their objectivity and ability for optimal extrapolation. Their biggest disadvantages are: (i) the patterns and the relationships must remain fairly constant during the forecasting phase for the forecasts to be accurate, and (ii) uncertainty must not be fat-tailed so it can be measured quantitatively.
This subsection was written by Spyros Makridakis (last update on 22-Oct-2021).↩︎