Problems of Transport Planning Methodology

The discussion in this chapter will be confined to transport planning; its integration with land use planning will be dealt with in the next chapter. In the first part of the chapter general shortcomings of the transport planning process will be considered and in the second part the transport model used in most transport studies will be examined.

The very limited range of explicitly stated transport policies which are used when drawing up a transport plan, and the implicit assumption that demand is independent of transport provision are the root cause of most criticisms of transport plans. The lack of policies has resulted in cost-benefit analysis being the only evaluation which is made. Consequently, on the whole, only those factors which affect the cost-benefit analysis are considered and built into transport models. Criticisms of these shortcomings should not be levelled at the models but rather at the policies (or lack of policies) which led to their development.

The tendency in Transport Studies has been to produce a plan for a target date, some time 20 to 30 years ahead, without first giving any consideration to operational improvements which could be made in the immediate future. This, plus the conviction that all road traffic demand must be met, the grant structure, and the feeling that urban motorways are intrinsically "modern" and "progressive" and of high status relative to rail and bus, have produced a tendency for inflexible, capital intensive, road biased solutions to be put forward.

The transport models used in transport planning have in the past been considered as neutral and objective tools. However they can never be completely comprehensive, and to the extent that some factors are left out or are not represented very well, any resultant plan will be distorted. For example the transport model used in the Greater London Development Plan has been criticized because it "..... assumed that underlying travel habits do not change, when the whole point of the exercise ought to have been to look for ways of changing travel habits" (Robertson 1973). In the same way "these studies often assume that the basic relationship derived from survey and analysis of the present conditions provide a suitable basis for the 'vertical' projection of travel demands over a considerable period of time, often 20 years or more, yet it is frequently stated that these basic relationships cannot be transferred 'horizontally' from one area to another". (Jamieson et. al. 1967).

It is now time to turn to specific aspects of the transport model. Many of these problems or deficiencies are due to the previously mentioned limited transport policies and the assumption that demand is not related to provision.

Forecasting

To produce estimates of future transport patterns the transport models require forecasts of a very precise nature of land use and socio-economic details. Forecasts must be made of the spatial distribution of "(housing, employment, shopping, etc.) ... also [of] economic growth (industrial production, household income, retail expenditure, vehicle ownership) and [of] social behaviour (working hours, shopping hours, the valuation of time, peoples resistance to travel over increasing distances)" (Solesbury and Townsend 1970). Not only must an average value be estimated for all these, but many must be estimated for each of many zones. There are very considerable problems involved in this type of activity.

First there is the problem of accuracy. The probable accuracy of any socio-economic forecast over even five years is likely to be low, yet forecasts have to be made over a period of twenty years and they have to be subdivided into many zones. as yet no evaluation of the likely level of accuracy of these forecasts has been made, let alone has any study been made of the resultant level of accuracy of the output of the transport model.

Secondly there is the problem of trend forecasting. When a secular tread is noticed and predictions are made that the trend will continue in the future, provision is often made for the predicted behaviour. However once the provision is made, it is often the provision. Itself, which produces the predicted result. In other words the prediction is self-fulfilling. To forecast an increase in traffic (or a decrease) and make provision for such a change will in many cases ensure the change actually occurs.

Thirdly there is a problem because transport models are not based upon causal relationships, but rather on regression analyses of the present situation. It is assumed that the relationships between the land use and socio-economic variables and the transport patterns found at the present will be the same in the future, so that predictions of these future transport patterns can be made. However since there is no evidence of a causal relationship between the explanatory variables and the transport pattern, or even that the important variables are included in the regression analysis, there is no guarantee that the relationships will hold in the future.

For example, one of the correlations usually found is that between socio-economic characteristics, such as car ownership or the number of shopping trips made per week, and income level. This correlation is then built into the transport model. If it is forecast, that in the future the population from a given area will have a higher income level, it is assumed they will behave in the same way as people who have that level of income at the present. However it may be that many aspects of behaviour are 'controlled' by class rather than income level, so the projected behaviour is not valid. (The correlation between income and behaviour in this case is due to the correlations between income and class, and class and behaviour. Changing one, that between income and class may not alter the other).

Finally, regardless of how accurate the transport model may be made, the output can be no more accurate than the forecasts used as an input. To the extent that social and economic forecasts will always be uncertain, so too will the transport predictions.

Generation

Stephen Plowden has criticized transport planners for confusing transport demand and transport usage (Plowden 1967). That is the assumption that trips will be made regardless of provision, which is one of the implicit assumptions in the transport model. Present travel patterns or transport "usage" are taken to represent present "demand". It is assumed that if the variables which "determine" present "usage" (and these do not include the level of provision of transport facilities) are forecast for some time in the future the future "demand" can be forecast. What is lacking is any consideration of the effect which the level of provision has upon the level of "demand" for, or more correctly, usage of, the transport system. The probable reason for ignoring this effect is that it is easy to recognise its existence but that it is very difficult to measure.

Distribution/Modal Split

Most transport models either make a pre or post distribution modal split (some transport models, such as that used in the SELNEC study, treat the modal split and distribution as a composite process). That is the mode used to make a trip is either decided before the destination of the trip is decided, in all cases, or after, in all cases. Neither pre-distribution nor post-distribution modal split is particularly realistic in all cases.

There are also problems which relate to modal split alone. All models, of necessity, have to limit the number of alternative modes which are considered. However the exclusion of a mode from the model such as bicycles, makes the consideration of the use of that mode impossible throughout the whole transport planning process. Person trips are generally forecast on a household basis and consequently assume that if there is a car owner in the household, all members of the household will use it. However the car may not be available all the time or to every member of the household. For example, a wife will not be able to use the car to go shopping while her husband is at work (if he drives to work), teenagers will probably want to travel independently of their parents most of the time even when a car is available[1] .

Finally some transport models have been based upon considerations of the time distance between points only. That is the time taken to make a trip has been used as the only variable which determines the choice of destination and choice of mode. The use of time only explicitly excludes the possibility of considering the effect of, for instance, changes in price, such as a new fare-structure or parking charges. Recently the concept of a 'generalised cost' which is a composite function of time and cost has been introduced to overcome this problem. However this is still prone to difficulties due to the use of average values. (These are dealt with later).

Assignment

The major problem in the assignment procedure is that in a complex network there is generally more than one route between any two points. To simplify the choice of route or "assignment" the usual procedure is to assume that all trips between any two zones travel on the "shortest" route which variously can be defined as that which has minimum distance, time, cost or generalised cost. Here and throughout this discussion only the distance function will be considered, but any of the other three functions (time, cost or generalised cost) could be used instead. If two routes are of very similar "length" then all the trip makers are unlikely to perceive the same route as shortest and there will be a division of trips between the two routes. However a multiple route assignment, using some type of diversion curve procedure to divide the trips between two zones, between routes of similar "length", is computationally extremely complex, and it is almost impossible to spot or trace any errors in the computations. This is the reason for the use of the "all or nothing" assignment procedure, although it has acknowledged shortcomings. This is another area where the use of average values can cause problems.

Finally there are several more general points on transport models. First there is very seldom any detailed consideration in a transport model of the different travel patterns occuring at different times of day. The model will generally be of the conditions either for the peak hour or for the greater part of the day probably 16 or 24 hours. Loadings on routes for other periods (i.e. for 24 hours when the peak hour is modelled or vice versa) will usually be derived by multiplying the loadings by some average factor. This procedure will not be very accurate, since the variation from hour to hour, in load on any link, will vary between links. That is - there may be a very large difference between the average flow and peak-flow on one link, whereas there may be little or no variation on another. Second - there is no consideration of the dynamic effects of changes to the road pattern. That is no-one has an instantaneous awareness of changes in a network, so there will be a time after the opening or closing of any link in a network while the flows on each link find a new equilibrium level. At present no-one knows how long these transient effects last but they are probably of a reasonably short duration and of small overall effect.

Finally there is considerable variation in the level of accuracy of different stages of the modelling process. For example the calibration of a model will be taken to an accuracy of 5% or less whereas the forecasts on which any estimates of future transport patterns depend, will have a probable accuracy which is far lower than this.

In conclusion the transport planning process is at present based upon a very limited range of policies and so the evaluation of alternatives is extremely limited. It is this limited evaluation which explains some of the limitations of transport models. Others are explained by the difficulty of measuring the size of the effect of provision on level of usage, and the lack of evidence that the relationships built into the model are in fact casual relationships. Finally, since the accuracy of the output can be no greater than that of the input, and a large part of the input consists of socio-economic forecasts which are not likely to be very accurate, the accuracy of the output from the transport model will not be very high.

[1] In evidence presented to the House of Commons Expenditure Committee (1972 Vol. II p. 237) Dr. Meyer Hillman made a critical analysis of the assumptions of household car use.

About | Contents
Previous: The Inadequacies Of Present Transport Policies | This: Problems of Transport Planning Methodology | Next: Problems of Land Use and Transport
Copyright © Michael Baker 1974,2005. All Rights Reserved.