4. Step 4: Make demand forecasts


4.1 Decide on the unit of demand.

4.2 Segment the market.

4.3 Ascertain the base for projection.

4.4 Make Base Case forecasts.

4.5 Make Project Case forecasts - diverted traffic.

4.6 Make Project Case forecasts - generated traffic.

4.7 Develop pricing assumptions (if applicable).

Benefits from transport initiatives are usually strongly related to forecast infrastructure utilisation levels. So demand forecasts play a critical role in appraisal of initiatives.

Additional discussion of demand forecasting methods is available in Part T1 Travel Demand Modelling and in the mode specific guidance.

4.1 Decide on the unit of demand

Possible units of demand include:

  • Passenger, cyclist, walker, vehicle or train numbers
  • Freight tonnes or containers
  • Passenger, cyclist, walker, vehicle, or train-kilometres or freight tonne-kilometres.

All of these are expressed per period of time (day, week, year).

If forecasts are in passenger numbers, freight tonnes or freight containers, they need to be converted into vehicle or train numbers at some stage of the appraisal in order to estimate operating costs and congestion impacts. Be transparent about the conversion factors used.

4.2 Segment the market

The level of accuracy achievable in the demand forecasts will depend, in part, on the extent to which the analyst segments the market. Classifications include trip purpose (people only), time (peak/off-peak), commodity (freight only), transport mode, load type (freight only, bulk, non-bulk), vehicle or train type and origin–destination pair.

The level of market segmentation will depend on data availability, the nature of the initiative and whether the analysis is a rapid or detailed CBA.

4.3 Ascertain the base for projection

Recent demand data is needed to serve as a starting point for the projections. Consistent data are required for trend analysis and aggregating across transport modes.

4.4 Make Base Case forecasts

Make demand forecasts for both the Base and Project Cases. In the absence of generated and diverted traffic, they will be identical.

There are three broad categories of forecasting methods:

  • Simple extrapolation of past trends
  • Extrapolation by relating the forecast variable to one or more explanatory variables (usually through an econometric model based on economic theory)
  • Application of informed judgment based on available evidence, including guidance from scenario analysis.

These methods are not mutually exclusive and will often be used in combination. The choice of technique will depend on data availability, the amount of effort chosen to apply to forecasting, and the extent to which extrapolation of past trends is considered warranted.

Extrapolation cannot be the sole forecasting technique employed when there is a likelihood of changes that bear no relationship to the past. Projections may have to be adjusted to allow for one-off events such as implementation of other transport initiatives or development of new industries or residential areas. This requires judgment. When preparing appraisals, provide details about the judgments applied in making forecasts.

For projections based on population, the Australian Bureau of Statistics (ABS) is a useful source of population forecasts (medium series), including at the statistical local area level[1]. Some jurisdictions prefer to use their own population forecasts.

4.5 Make Project Case forecasts - diverted traffic

The ATAP Guidelines define diverted traffic as freight, passengers or vehicles that switch from one mode, route, time of day, origin or destination to another as the result of an initiative.

To estimate diverted traffic, obtain an estimate of the maximum volume of traffic that could potentially divert for each market segment. Then estimate the proportion of the potentially divertible traffic likely to divert. Accuracy will be greater the more the market is segmented, because the propensity to divert depends on characteristics such as origin–destination and time-sensitivity.

The simplest way to estimate the proportion of traffic likely to divert is to nominate a percentage using judgment. Preferably, past experience with similar initiatives and market knowledge will inform those judgments.

If a small number of shippers are responsible for a considerable percentage of the divertible freight, ask them directly about their probable responses to the proposed change.

A simple quantitative method is to use the concept of cross elasticity. Where there are quality improvements such as time savings that can impact on the amount of diverting traffic, the price could be expressed as a ‘generalised cost’ (see Box 4 for an explanation). Logit models are a more sophisticated technique for predicting impacts of initiatives on mode or route shares. The logit model splits up the market between modes and routes according to how they compare in terms of price, time taken and other quality attributes.

For rapid CBAs and many detailed CBAs, values for elasticities based on studies by others may need to be assumed, including overseas studies. Since surveys are very expensive to undertake, they are only justified for detailed CBAs of very large initiatives or programs of related initiatives.

Levels of traffic diversion may increase over time as transport users have time to adjust. The literature sometimes distinguishes between short-run and long-run elasticities. If this applies to the initiative, consider allowing for a ‘ramp up’ period during which the level of diversion builds up gradually to the long-run level.

4.6 Make Project Case forecasts - generated traffic

The ATAP Guidelines define generated traffic as altogether new demand resulting from an initiative.[2] If the new traffic comes from a specific source such as a new industrial development or land use change that is expected to occur as a consequence of the initiative, collect information about the source to estimate levels of generated traffic. Where the sources of generated traffic are more diverse, a price elasticity of demand could serve as the basis of estimation. The comment about a ‘ramp up’ period for diverted traffic applies equally to generated traffic. Higher long-run elasticities are a simple way to allow for land-use change caused by an initiative.

Where there is considerable uncertainty about levels of generated traffic, a sensitivity test can be undertaken by estimating traffic levels with and without the generated traffic. The risk analysis stage is an opportunity to explicitly consider the uncertainty associated with generated traffic.

4.6.1 Induced demand

Induced demand is the sum of generated and diverted traffic. Part T1, Section 3.4 of the ATAP Guidelines discusses the modelling of ‘induced demand’, which it defines as the impacts of transport improvements in terms of the resulting changes in route choice, the time of day travel occurs, mode choice, trip distribution (that is, choice of trip destination), trip generation (that is, the number of trips undertaken), land use changes and the location decisions of both households and businesses. Chapters 7 and 8 below discuss the role of induced demand in user benefit estimation.

4.7 Develop pricing assumptions (if applicable)

Where infrastructure use is charged, make assumptions about prices. Price levels affect levels of demand (existing, diverted and generated). Factors affecting prices include the demand curve, costs and the objective of price setting (profit maximisation, economic efficiency maximisation, revenue target or cost-recovery target). Constraints relating to the forms of charging available also need to be taken into account (such as the extent to which charges can vary with time of day).

Prices may have to be estimated simultaneously with developing demand forecasts because the demand curve is needed to estimate the price, and then the price is needed to estimate the quantity demanded.

Prices are critical for estimating revenues for financial analyses.

[1] The central scenario of a CBA should aim to estimate the most likely or ‘expected value’ CBA results, for which the ABS medium series population forecasts would seem most appropriate. There is a tendency to use the highest population forecasts available, which can be a source of ‘optimism bias’. Where ABS’s high series population forecasts are employed for the central scenario, sensitivity tests using the medium and low forecasts should be undertaken.

[2] Economists usually refer to all additional traffic using the infrastructure in the Project Case, compared with the Base Case, as ‘generated traffic’ regardless of the source. Generated traffic in the economist's use of the term, may be divided into ‘diverted traffic’ and ‘new traffic’ depending on its source. However, transport planners use the terms ‘generated traffic’ to refer to traffic that is new altogether - as in the 'trip generation' phase of transport modelling. Because the ATAP Guidelines are aimed at a broader audience than economists, the transport planners’ terminology has been adopted.