Forecasting may not only provide inaccurate estimates, but it may also support flaws leading to incorrect interpretations. When forecasting is used to support investment decisions for infrastructure developments, these flaws can be far-reaching in consequences such as fewer returns than expected and longer amortization. The goal is obviously to be able to provide capacity in line with the expected traffic level. There are three conventional flaws in forecasting, particularly when such an exercise takes place within a standard cyclical growth pattern that includes introduction, acceleration, peak growth, and maturity:
- A. Under-estimation bias. Forecasting is usually using a reference year from which future trends are extrapolated. Over the short term, there is likely to be limited differences between the forecasted and actual traffic, leading to the impression that the trend is accurate. However, as the growth enters a phase of acceleration a negative forecasting gap emerges where the level of activity is higher than expected. These biases are common at the beginning of business cycles and are the outcome of a lack of understanding of the growth potential, mostly because of uncertain market conditions. When the forecasting gap becomes too acute, congestion and capacity shortages take place. New forecasting attempts are usually made, which often leads to an overestimation bias.
- B. Over-estimation bias. During phases of acceleration and peak growth, positive outcomes are expected in the long term concerning the continuation of this growth. However, as growth enters a phase of maturity, a positive gap emerges, leading to less activity than expected. The usage of compounded annual growth forecast methodologies further exaggerates the bias since they assume an exponential growth. The over-estimation bias is usually the most far-reaching in consequence since it often leads to an over-investment in capacity. It is supported by the enthusiastic perception by managers and planners that future outcomes and growth are highly positive.
- C. “Return to normal” bias. On occasion, forecasts are made in a low growth context but assuming a growth trend that took place during peak growth. Such an approach is more a cognitive dissonance issue with a refusal to see the new context because it is contradictory to institutional or stakeholder interests.
The identified flaws are difficult to avoid since, as always, forecasting involves assessing future trends based on incomplete information and uncertain market developments. Yet, recognizing that there are biases may lead to a more pragmatic view of how forecasting fits transportation infrastructure development.