The goal of the transportation planning process is to assist governments in providing an adequate transportation system at an acceptable cost. This involves modeling the behavior of the present system, estimating future travel demand, and estimating how changes in the system will affect travel behavior and the operation of the transportation system in the future.

The following approach described was first used in the 1950s in the Chicago Area Transportation Study (CATS).^{9} Variations of it were subsequently done in many other major metropolitan areas. The general approach has also been used in many smaller area studies.^{10}

Major transportation planning projects usually involve a four-step procedure for estimating travel movements. Then the merits of different possible changes in the transportation system can be evaluated.

Going through the steps requires building a geographic database. In general, the area in question, say, a metropolitan area, is divided into zones. In the Chicago study and many studies patterned after it, a rectangular grid is imposed on the region, and data are collected for each square in the grid. A typical grid might cover several thousand square miles. In other cases, particularly in smaller studies, the zones may be irregular, with shapes determined by terrain features, neighborhood boundaries, or other characteristics of the land-use pattern.

The data gathered for each zone include both population and economic information. The former includes such items as the number and type of housing units, number of residents, age structure of the population, household income, number of automobiles owned, and similar items. The economic information includes such items as the number of people employed in the area in various occupations and the number of square feet of floor space and land area devoted to retailing, wholesaling, manufacturing, office operations, and other activities. Each of these broad categories may be broken up into a number of subcategories. Sometimes, just total nonresidential floor space is used. It is often a good predictor of how many trips will be attracted to the zone, and it is much less costly to develop than the more detailed category-by-category data.

The Four-Step Process. Once the database is in place, any given transportation alternative can be evaluated. In general, a four-step process is used.

1. Estimating trip generation. Before deciding where people will go from a given point of origin, it is customary to estimate how many trips a given place will generate regardless of where those trips are destined. For estimating trip generation from a residential area, variables such as household income, number of persons in the household, number of vehicles owned by the household, and possibly population density might be used to estimate average trips per household per day. Usually the number of vehicular trips is positively related to the first three items and negatively related to the last. The reason for the last relationship is that trips that may be made on foot or by mass transit in a dense area are likely to be made by auto in a sparsely populated area, where trip distances are longer and parking and congestion problems minimal.

2. Estimating trip distribution. After trip generation has been resolved, the next issue is to distribute the trips. Suppose a given zone in a region contains 1,000 households, whose average size, vehicle ownership, and income are known. The total number of trips that they will make can thus be estimated, but the question is how the trips will be distributed among possible destinations. A variety of estimating methods have been developed over the years. The most commonly used is the gravity model, originally developed in the 1920s to analyze shopping patterns. (The original formulation was known as Reilly's Law of Retail Gravitation, and in the past gravity models were sometimes known as Reilly models.) The force of gravitation between two objects is proportional to the product of their masses and inversely proportional to the square of their distance. By analogy, the force of trip attraction between, say, a housing complex and an office complex would be proportional to the product of the number of households and the number of square feet of office floor space and inversely proportional to some function (perhaps the square or some value near the square) of the distance between them. In principle, then, one might estimate the relative number of trips made from origin A to destinations B and C by computing the relative force of attraction between A and B and between A and C. This process is illustrated in the box on page 248. For a region with a large number of zones, the database and the number of calculations are huge. Thus such a planning exercise was impossible before the computer.

Such an exercise could be done for an actual distribution of housing and floor space or for a hypothesized one. Distance might be taken as straight-line distance from the center of one zone to the center of another. Or it might be taken as actual road mileage, travel time, or some composite of both.

3. Estimating modal split. Where there is more than one mode of transportation available, say, automobile and bus, it is important to apportion by mode the trips distributed in the previous stage. Over the years considerable experience has been accumulated, and a number of mathematical estimating techniques worked out. In general, the two main criteria that determine which mode an individual takes are quality of service and cost. Quality of service is largely a matter of travel time. Quite frequently there is a clear trade-off between speed and cost. For example, in public transportation, commuter rail service is much faster than bus transport, and also substantially more expensive. Knowing the income distribution of the population using public transportation would thus help make estimates of how that population would divide between the two modes.

4. Predicting trip assignment. Once the choice of mode has been settled, the last issue is predicting how trips will be distributed between alternate routes from the same origin to the same destination. Again, the question is resolved by mathematical modeling. Consider that there are two routes, A and B, from zone X to zone Y. Imagine, also, that we begin by assuming that all traffic takes route A. As travelers shift from A to B, travel times on route A fall while those on route B lengthen. Mathematical models are used to predict when equilibrium will be reached.

In general these four modeling steps are used in the following way. First, the existing state of the transportation system is modeled mathematically, using the steps just described. Then the model is calibrated to produce results that correspond to the actual flow of traffic. The data used for calibration will come from measurements of traffic flow, for example, those made by counters that are tripped by the weight of a vehicle passing over a rubber hose. Once the model duplicates the observed travel behavior, alternative situations can be modeled. For example, a planner might assume an increase in the number of households in a given zone.

The model can be run again, and the region will show a slightly different pattern of trips. The transportation planners might then postulate changes in the road pattern to see how these will affect the pattern of trips.