Introduction

When I came to the Energy Research Group (ERG) in 1975, I thought that forecasting problems were in large part due to insufficient attention being paid to detail. The models used in forecasting exercises tended to be very aggregated and cover very small systems. I had the idea that forecasting problems could be overcome by paying more attention to this detail.

My idea, on what the forecasting problems were, were not very clear. However I was aware that serious mistakes had been made in the past. For example during the 1960's and early 1970's the electricity industry in Great Britain had made substantial over estimates of electricity demand as we (in ERG) found in our Electricity Industry study (Energy Research Group 1976). The conventional explanation given for the electricity industries past forecasting errors were:

poor economic performance,
cheap natural gas,
the price of the industries fuels (coal and oil) had been forced up by the miners and Arabs.
However we concluded that these reasons did not adequately explain the errors and that there were additional reasons which were:
an asymmetry in the penalties incurred by the industry between making under and over estimates of demand which lead to a tendency to make overestimates,
the development of an exponential mentality which assumed that growth was inevitable,
tying demand estimates to GDP then over estimating economic performance without regard to how the electricity could be used or by whom.

The first time that I personally found previous forecasting exercises inadequate was in a short study I did on goods vehicles. I was aiming to produce a population model of goods vehicles along the same lines as that developed by Chapman and Mortimer (1975). Basically they used their model to examine car registration data in terms of birth rates (new car registrations), death rates (scraping) and age. In their model there was an underlying assumption that demand for cars (the saturation level) is related to the countries population (which had been fairly static over the period they were investigating, 1950 to 1975). With the model they estimated a saturation level based upon the point at which the rate of change in the car population went through a peak.

However when applying their model to goods vehicles I encountered several problems, the most important of which was that there is no obvious static or saturation level of demand for goods vehicles (or road freight transport). There had also been a change in the relative distribution of goods vehicles by size. Finally I found a discontinuity in goods vehicle registration figures caused by the transfer of the Post Office from a government department to a public corporation in 1969.

The increasing demand for road freight transport suggested to me that there was a need for a model of goods transport demand. Those which had been developed for the United Kingdom by Tulpule (1969), Tanner (1974) and Sharp (1973) appeared to me to be totally inadequate. This led me into an investigation of freight transport statistics as a basis for a disaggregated freight transport forecasting model.

As a result of the work I did no freight transport statistics and subsequent work on future transport fuels and the compilation of data on transport for the Central Statistical Office's 1974 Input-Output tables, I have come to a deeper understanding of the problems inherent in forecasting and planning.

This thesis is in two parts. The first part starts with three chapters in which I describe the work I have done on freight transport, future transport fuels, and the 1974 input-output tables. In each of these chapters I draw out the problems which I encountered while trying to make improvements to the forecasting process. The first part concludes with a chapter in which I review the lessons I have learnt about forecasting.

In the second part I look at the forecasting process in detail and examine the problems involved. I then give a general description of the planning process and look at why forecasting appears to be necessary. Finally I examine the implications the forecasting problems have for planning.