Dick Nelson and Don Shakow

Integrated Transportation Research

122 Northwest 50th Street, Seattle, Washington 98107-3419

Phone/Fax (206) 781-0915---74211.1102@Compuserve.com


Submitted for presentation at:

Transportation Research Board, 76th Annual Meeting, January 12-16, 1997

Washington, D.C.


A search for congestion mitigation strategies for metropolitan regions based on economic criteria does not require an estimate of the total social benefits derived from the transportation infrastructure. The assumption that societal benefits are constant is adequate for sketch planning exercises designed to identify cost-effective strategies that address urban congestion and its environmental impacts. Least-cost transportation planning, as it has been derived from energy sector models, is therefore a more practical real-world regional transportation planning methodology than is traditional benefit-cost analysis.

Key words: Least-cost planning, benefit-cost analysis, integrated transportation planning, metropolitan transportation planning.



Renewed interest in the application of economic criteria to urban transportation planning and decision-making has generated a number of studies directed at designing new methods and models, and comparisons of these with existing methods and models.

A new least-cost sketch planning method has been devised to assess the costs and benefits of alternative solutions which range from capacity enhancements to demand management (1). The method employs a computer model of a metropolitan transportation system that is capable of searching across a large number of alternatives for the set of synergistic alternatives that provide a maximum amount of mobility/accessibility at the lowest investment cost. The model applies full cost accounting. It accounts for both direct capital and operating costs and indirect social and environmental costs.

Full cost accounting has been aided by recent efforts to establish more precisely the incremental social costs of transportation. A major contribution to this growing body of work is an extensive study which is forthcoming in a series of papers published by the University of California Transportation Center. Delucchi has summarized this work (2).

A recent report (3) to the USDOT prepared by COMSIS Corporation, Parsons, Brinkerhoff, Quade and Douglas, and ECONorthwest suggests that:

(1) Least-cost planning (LCP) is a subset of benefit-cost analysis (BCA); and (2) LCP in transportation must incorporate an explicit accounting of benefits (in the form of consumers' surplus) and that the electricity model which attempts to reduce costs relative to a fixed benefit standard is inappropriate for transportation. The authors of this paper take issue with these and other assumptions in the COMSIS report (hereinafter abbreviated as COMSIS). We develop our critique in the following manner. First we summarize the problems faced by regional planners that require a new planning paradigm. We then outline those criteria that, in our view, should serve as the basis for selecting an appropriate transportation planning methodology. We then outline alternative methodologies. Finally, we review these methodologies in light of our preferred criteria and draw conclusions regarding an appropriate methodology that differ in some essential respects from those advanced in COMSIS.


Traffic congestion is a problem that continues to severely challenge metropolitan area transportation planners and decision makers. Political reality demands that they respond to frustrated travelers who experience long delays in their weekday commute and, increasingly, during off peak and weekend trips. To ignore the problem would be tantamount to political suicide.

Yet a magic cure for congestion seems to be as illusive as one for the common cold. Urban regions invest large sums in increased roadway capacity only to see it fill in a relatively short time (4). The opposite problem is experienced by urban areas that invest in major transit improvements such as rail systems -- not enough new transit riders. Although before and after rider surveys are rare, when comparisons are made it is found that most of the riders of the new rail service were riders of the old bus service, and relatively few have switched from the auto mode (5). Sometimes this result is predicted using a regional travel demand model before investments are made (6).

The problem is compounded by the lack of comprehensive transportation demand management programs. Despite numerous "toolboxes" that describe a wide range of available incentives and disincentives to solo driving, few regions have systematically developed TDM as an alternative or even a complement to capacity, and then gained the public and private support needed for its implementation. Effective TDM efforts are isolated examples of what could be accomplished. Capacity building, whether roadways or transit, remains the tangible and familiar option.

Planners, however, are not immune from the troubling predictions of increasing hours of congestion that their demand models deliver. Much of their effort is spent in working to increase the transit and non-motorized friendliness of urban space. Yet the limited effect of land use reforms and transit on already developed urban patterns is clear (7). For this reason planners seem to be collectively holding out for an acceptable pricing scheme that would allocate roadway capacity more efficiently and put to use the often underutilized capacity of existing public transit systems. Given time, failed capacity remedies, and increased commuter impatience, that bet may eventually pay off. The question, however, is what metropolitan regions do before the sea change occurs.


A useful planning methodology should necessarily be capable of weighing the costs of a congestion remedy, whether in the form of expanded capacity or a strategy to increase the efficiency of the existing system or manage travel demand, against the benefits produced. The most efficacious remedy will most likely take the form of an integrated set of measures that provide synergistic benefits. Benefits will be measured by the reduction in travel time and excess fuel consumed, both for personal and commercial travel, reduction in accidents stemming from congested conditions, and a reduction in the health and environmental damage associated with air pollution. A full accounting of the external environmental costs would also include the global warming impacts resulting from excess fuel consumption.

This is not the basis on which major transportation investment decisions are usually made. Citizens are usually asked to decide without being given alternatives from which to choose, and without a clear understanding of the benefits in relation to the costs. Up or down decisions are required in the face of considerable uncertainty and ignorance. Proposals are most often presented in the context of election campaigns in which media consultants play a prominent role.

An important implication of the foregoing is that planning models should have as their aim moving public policy debate to a more rational plane. They should be designed to reduce guesswork and supposition in public discourse and decision making. In order for this to be accomplished, it is necessary that the framework for planning models be: (1) easily comprehensible to policy makers, interest groups, and voters; (2) fast and easy and to implement; (3) neutral with respect to special interests; and (4) resistant to manipulation to generate preconceived results.

Transportation poses a major challenge in achieving these objectives. Public dialogue concerning transportation is characterized by a stridency and intensity of debate that reflects deep and difficult to reconcile divisions. Each faction, whether highway advocates, transit enthusiasts, partisans of public goods pricing, or land use revisionists, are convinced of the righteousness of their particular "pet" transportation alternative and often disdainful of oppositional viewpoints. For this reason, there is some reluctance by these groups to relinquish their freedom of action to a neutral agency or analytic tool.

In the view of the authors, it is precisely for this reason that the development of an objective and credible analytic planning tool -- a tool that makes assumptions transparent -- is essential. In its absence, transportation decisions are likely to be governed by a process that is largely political -- the faction with the most powerful "political guns" is likely to prevail, whether or not its views are in the long-term public interest.

At the same time, the public process is likely to be, and indeed must be, iterative in nature. Proposals issued from within the public arena should be subjected to analysis using an appropriate framework, models, and tools. The results of these analyses must then be assimilated by policy makers and potential voters, who make appropriate changes in their original preconceptions. Modified proposals are then again subjected to analysis until the prevailing political wisdom accommodates the analytic results. We have outlined such an iterative process in a separate paper (8).


The challenge is to design an appropriate regional congestion reduction strategy that integrates investment in infrastructure, demand management, regulatory changes, system management, and price and tax innovations. In an "optimal" strategy, it is likely that all types of elements will be present to some degree and that no one "fix" will dominate to the exclusion of others. Ideally, it would be wonderful if a "miracle tool" existed to generate such an optimal strategy that could command the full confidence both of the public and of specialists and could set forth just those elements comprising a perfect strategic mix. In practice, no such tool exists. As a result, it is necessary to assess a number of approaches that have been put forth and to review their advantages and disadvantages. Included among the paradigms most often referred to are: (1) benefit-cost analysis; (2) least-cost planning; (3) multi-attribute or multi-criterion analysis; and (4) dynamic optimization.

Least-Cost Planning

Least-cost planning (LCP) incorporates the following elements and


The process maximizes the number and range of transportation alternatives under consideration.

Subjectivity that would favor one alternative over another is minimized or at least made explicit.

All alternatives are constrained to achieve a standard of performance, such as a required level of regional accessibility, over a specified planning period.

If a given standard poses problems of measurement, other surrogate standards can be developed.

An efficient search among alternative strategies is conducted to determine that strategy which achieves the standard and minimizes net social cost (alternatively which maximizes net social benefit).

The process takes into account significant elements of uncertainty and risk.

Benefit-Cost Analysis, and Its Comparison to Least-Cost Planning

Benefit-cost analysis (BCA) refers to an analytic process that has been employed historically to select among a set of alternative projects. The selection criterion is as follows: projects are selected if their net social benefit contribution is positive. In contrast to LCP, which minimizes costs relative to a pre specified performance criterion, BCA quantifies benefits and costs using independent analytic approaches for costs and benefits. In particular, BCA has developed a highly evolved methodology for quantifying benefits based on the economic concept of "consumers' surplus." Consumers' surplus refers to the benefit which consumers experience when they are willing to purchase a commodity at a given price and the commodity in fact trades at a still lower price. To the extent that investments alter the prevailing price structure, and permit reductions in price, they generate consumers' surplus.

However, despite its greater scope and seeming analytic advantages, the application of BCA to system level regional transportation planning involves serious practical difficulties.

BCA developed in a project planning context (and its analytical features are attuned to this context), while LCP developed in a systems planning context. In congestion mitigation, the task at hand is to develop an overall regional, subarea, or corridor level transportation system strategy, with all manner of alternatives in the hopper. While there is ample precedence for this in LCP, a whole new analytic apparatus would have to be developed were BCA to be employed for this purpose. To implement BCA in real life instances and to use it for systems planning, especially in transportation where the "system" is immensely complex, would require the development of extensive new machinery and the undertaking of exercises that involve presumption and speculation (e.g. simulating market responses). This is a formidable, expensive task, which may yield no definitive practical outcome.

To apply BCA it is necessary to calculate consumers' surplus through the estimation of actual or virtual demand functions. In transportation, this would involve the assumption that transportation goods are priced at varying levels (possibly by time of day) coupled with an assessment of the public response to these various prices. In practice, any attempt to go beyond a purely heuristic estimation procedure and to initiate actual travel-related user fees is likely to elicit vigorous opposition; the notion of pricing public highways is extremely sensitive politically. The very mention of toll roads and congestion pricing has typically evoked hostility from travelers potentially affected by the imposition of fees and tolls. It is incorrect, in the authors_ view, to employ an analytic device for regional planning based on a presumption that is a political "no go." This is not to suggest that congestion pricing or tolling should be deliberately taken off the table. Rather, it expresses a concern about the use of a method whose implementation presumes the existence of actual or virtual markets. This necessarily highlights price responsiveness above other considerations in measuring benefits. Groups with a philosophical objection to congestion pricing might prefer a more neutral approach.

Aside from this political objection (which is in itself overwhelming), the data required to estimate demand curves and to derive publicly credible estimates of consumers' surplus for purposes of system planning appear to us so great as to preclude its implementation in practice. Transportation is sufficiently complicated; why select an investment analysis methodology that magnifies the complexity and expense?. Cameron's calculations of consumers' surplus for the Southern California transportation system highlight the empirical difficulties underlying (BCA, 9). His demand curves, for instance omit estimates of demand at operating costs per mile driven of more than $.30. He comments as follows: "...because...of the impracticality of estimating a full demand curve for auto travel...(the) estimate of the gross value of mobility, and ultimately of net transportation benefits should not be relied on too heavily". This concern might be compared with the similar concern expressed about environmental externalities. Estimates of non-market parameters (e.g. air pollution cost, cost of delay) that are then employed across all options and packages is difficult but represents a well-defined and developing "frontier" area of research. Several significant studies have appeared in the last two years and several conferences have been convened (and more are planned) on this subject. Estimating market functions under a myriad of institutional circumstances -- which change based on the set of transportation options in force -- represents a task which is orders of magnitude more difficult. The ultimate concern is that transportation planners in the face of these difficulties will throw up their hands and resort to subjective and politically determined criteria. The whole point of planning exercises is to undertake an analysis that is independent of subjectivity and political bias and then use it as a benchmark and reference point when the inevitable phenomenon of political give and take enters the decision process.

A further empirical objection is the following: since for most travel modes, the user does not pay the full cost of travel and often only a fraction of the total cost, empirical estimates of demand curves even using survey methods or heuristic devices calculate demand as a function of perceived private cost. If a transportation alternative, for instance, raises private cost, but at the same time lowers social cost by a greater amount, a benefit calculus based on empirically derived consumers' surplus would actually show decreased benefits.

BCA dampens rather than stimulates consideration of a large number and variety of approaches to transportation. The tendency in implementing this method will be to restrict alternatives and approaches to allow the analysis to proceed.

BCA, substantially more than LCP, depends on a subsequent choice of parameters and assumptions. If these are subjective, they compromise the objectivity, neutrality and freedom from manipulation that we view as central to a credible analytic tool.

Multi-Attribute Analysis

Multi-attribute analysis acknowledges implicitly two potential difficulties with BCA: (1) the calculation of net social benefit and, in particular, that the quantification of benefits may not be feasible; (2) there may be political resistance to the notion of relying on a single quantitative criterion. To implement a multi-attribute analysis a series of independent criteria are specified including cost effectiveness 10. These criteria are then weighted based on weights chosen by the user. Alternative investments are then ranked based on the sum of weighted measurements.

Dynamic Optimization

Dynamic optimization goes beyond the other approaches in searching among alternative time paths over which transportation investments, alternative policies, or innovative transportation measures are implemented. More so than other methods, it acknowledges that premature commitments to particular investment options may foreclose other options down the line. For example, if a region commits to a major rapid rail investment program and completes 80 percent of construction within ten years at which time a wholly new and significantly more cost effective transportation technology (say, personal rapid transit) becomes feasible, this latter possibility is foreclosed by the earlier decision. Dynamic optimization would account explicitly for these types of situations and would guard against premature commitments. While the logic of this methodology is well understood and widely practiced, its disadvantage, however, is that it requires hard numbers for futuristic transportation alternatives that are presently in the R&D phase. The analysis must, therefore, allow a significant range in cost estimates to account for future uncertainty.


Throughout the development of our LCP model, we have dealt with various criticisms of the least cost approach, both direct and implied. In developing our modeling framework, we have attempted to pursue a course that took account of these criticisms. As a result, we believe we have evolved a "sketch planning" tool that would provide useful information to transportation planners and concerned citizens.

Some of these issues and criticisms merit further elaboration, since they serve to highlight the differences between the available methods and, we believe, the advantages of LCP.

Criticism: The Need for a Network Element

The need for a network element in a least-cost model is often cited as a defect. We view this as a red herring. It argues for an unnecessarily elaborate, time consuming, and expensive modeling effort. First, network models exist and are resident within MPO's demand models. A LCP methodology need not "reinvent the wheel." Our model is designed to assimilate the output of a four-step network model and to compute costs and savings based on these inputs. Without this information, we would argue that the utility of information at this level of detail is limited. The costs most impacted by network considerations are congestion costs. However, congestion occurs principally in certain well-defined screenlines and volumes at these screenlines are roughly proportional to total vehicle miles of travel. VMT is a major output of the sketch-planning tool we are developing.

Criticism: That Benefits are Assumed Constant

The claim has been made that benefits are constant in our methodology. This is literally untrue and misleading in its implication. We specify net social benefit as the cost savings engendered by a given transportation strategy relative to the no-action base. Both the no-action and aggressive strategies are designed to meet a moving target -- a growing need for accessibility and mobility within the region. Presumed within these accessibility and mobility targets are implicit benefits. Least cost planning takes these as input without attempting to undertake an independent measurement of these benefits. While BCA highlights the individual consumer benefits (in the form of consumers' surplus) as a quantity which it is necessary to measure, the immense complexity of performing this measurement in the context of system planning is so daunting, difficult, and non-empirical in the absence of data, that it remains a theoretical option whose practicality is dubious.

It seems that we have two choices: hold off on our analysis until some way is found to practically calculate these benefits, or "wing it" (use cumulative discounted real cost savings as a surrogate for social benefits) until an algorithm is in place to credibly measure consumers' surplus. Under the latter option, won't we improve our planning?

Criticism: Does Not Account for Latent Demand

A failure to account for latent demand has also been cited as a decisive problem with our approach (and with LCP more generally). The essence of this is that LCP assumes an independence of demand and supply. A performance target in the form of a demand estimate (e.g., a load forecast) is specified, estimated, and simulated independent of how the target is met. While this poses few problems in energy, its application in transportation is problematic because new capacity in transportation may lead to new (latent) demand that would be excluded from the original demand estimate.

In our view, this objection has force to an extremely limited degree. What is commonly referred to as "latent demand" includes two distinct elements: (1) travelers who are meeting their accessibility needs in one way, take advantage of new capacity and shift their traveling behavior; (2) the new capacity stimulates an increase in end-use demand for accessibility. The first element is accounted for fully by a least-cost approach, since a least-cost model, in principle, is designed to optimize relative to a specified performance target that does not change because of any new capacity. For example, a work trip before the opening of a new bridge may have been subject to congestion cost, but it was accounted for. Now the commuter may incorporate the bridge in his/her more efficient traveling plans. The end-use requirement, i.e., one work trip, does not change as a result of the new bridge.

On the other hand, to the extent that supply increments actually stimulate "end use" demand, so that "stay at homes" suddenly are induced to travel (usually for social/recreational purposes) under the impetus of a new highway or bridge, the performance targets actually move, and the independence assumption is breached. But we view this component of latent demand as de minimis in comparison to the former component. Rather than throwing out the baby with the bath water, we would suggest that empirical data generated from regional experience be used to make estimates of latent demand on this second account and to adjust demand accordingly under the assumption of new supply. This can be implemented directly within a least cost framework.


While the political environment governing transportation is highly complex and that no jurisdiction will allow itself to abide mechanically by the results of a single-criterion analysis, that is all the more reason, the authors believe, to persist in advancing a method whose results are more unitary, definitive and neutral. Analysis should exist as a counterpoint to politics. If an unambiguous cost-effectiveness criterion is superseded by other considerations, this should be clear and acknowledged. To submerge estimates of net benefit within a broader and less rigorous notion of cost effectiveness and then to further submerge cost effectiveness within a melange of multiple criteria, hard and soft, would effectively suppress information that is useful, available, and in the public interest.

Of course, any "hard" estimate of net social benefit must be qualified considering uncertainties and defects in methodology. But it is crucial that we highlight separately the results of our analytical machine on the one hand and the political constraints that may force us to deviate from the prescriptions on the other. It is wrong, though, to disregard the output of a credible analytical model. The object of planning is, after all, to reach the best decisions possible under admitted methodological, informational, and political constraints.

Because of its tendency to confound analysis and politics, the authors strongly urge that multi-criterion methods be rejected in favor of single-criterion methods. Within the cohort of single-criterion methods, LCP should be favored over BCA. Put briefly, BCA is a proven technique for single project evaluation and has merit as a means of evaluating a small number of alternative projects. We are dubious, though, about its aptness or practicality in a systems context. Devising a regional transportation strategy is a classic systems planning problem which BCA simply has not been designed to deal with.

We take strong exception to the thesis advanced in COMSIS that LCP and BCA are approximately equivalent. We view LCP as a methodology that has evolved in the context of system evaluation, as opposed to BCA that has evolved in project evaluation. The proposed principle that all projects with positive net benefits be pursued is difficult to implement in an environment where any one transportation alternative affects the parameters governing all others. The classic economist's assumption "all other things being equal" doesn't work in regards the highly intricate and interdependent system of regional transportation.

We see other practical implementation problems in estimating benefits based on the classical BCA methodology. For instance, the authors of COMSIS note that under a consumers' surplus criterion, it is possible that "the encouragement of sprawl results in net external benefits on balance" (p. 3-32), though the precise measurement of these benefits is in all likelihood impractical.

On the other hand, it might be argued that LCP oversimplifies in that it abstracts from an explicit measurement of benefits, whether through the consumers' surplus or some other approach. We would suggest, however, that in practice these benefits are inputs to the analysis and are specified as target levels of electrical load in energy or mobility and/or accessibility in transportation. The region determines what is an appropriate level of transportation services (by means of, say, a four-step forecasting model implemented by an MPO); LCP is then designed to meet the targets specified by the planning agency.


The overarching aim of our least cost planning work has been to

develop a planning tool with the following characteristics:

The model could be made operational in a matter of 4-6 months of full-time effort; its use would therefore be timely to assist in making the critical transportation choices that face a metropolitan region.

Its data requirements would be within the scope of what is currently available including national statistics and available region-specific data.

Its outcomes would be concrete, unambiguous, and unbiased. An understanding of the last point is crucial. We are not advocating that all decisions be subordinated to the dictates of an "objective" model. Rather, that the model allow the chips to fall as they may, that the information it provides be folded into a public process where other considerations are accounted for (as they will be inevitably). What we don't wish to witness is a pseudo-analytical "model" that is either so wishy-washy in its recommendations that it is useless as an unbiased guide or else is trained to move in the direction of the prevailing political winds.

A truly neutral (and hence useful) model is one that provides untainted, tangible intelligence. This information can then be accorded whatever weight the decision makers and voters deem appropriate. A process in which all reasonable efforts are made to overcome the effects of bias and subjectivity (LCP) is very different from one in which subjective choice is an intrinsic aspect of the procedure (BCA).

In reading COMSIS, we reacted with major concerns. The proposed BCA modeling concept and process would: (1) require a significant amount of time to become operational; (2) would require data (especially demand data) that is not currently available and would have to be developed; and (3) would incorporate crucial subjective elements and would be susceptible to political manipulation.


The authors extend their appreciation to Professor Jerry Schneider and John Niles whose constructive comments and steady encouragement have helped shape this work and the development of the least-cost planning methodology.


1. Nelson, D., and D. Shakow. Least Cost Planning: A Tool for Metropolitan Transportation Decision Making, In Transportation Research Record, 1499, TRB, National Research Council, Washington, D.C., 1995, pp.19-27.

2. DeLucchi, M. A., Total Cost of Motor-Vehicle Use, Access, University of California Transportation Center, Berkeley, California, No. 8, Spring 1996, pp. 7-13.

3. COMSIS Corporation, Least-Cost Planning: Principles, Applications and Issues, Report prepared for the Office of Environment and Planning, U.S. Department of Transportation, Washington, DC, July 1995.

4. Hansen, M., Do New Highways Generate Traffic?, ACCESS, University of California Transportation Center, Berkeley, California, No. 7, Fall 1995, pp.16-22.

5. Moore, J. E. II, Ridership and Cost on the Long Beach-Los Angeles Blue Line Train, Transportation Research, Vol. 27A, No. 2, 1993, pp.139-152.

6. Central Puget Sound Regional Transit Authority, Phase I Study Options Results Report, September 9, 1994, p. 66.

7. Giuliano, G., The Weakening Transportation Land-Use Connection, ACCESS, University of California Transportation Center, Berkeley, California, No. 6, Spring 1995, pp. 3-11.

8. Nelson, D., and D. Shakow, Sustainable Transportation Through an Integrated Planning Process, Submitted for Publication in the Proceedings of the OECD Conference: Toward Sustainable Transportation, Vancouver, B.C., Canada, March 24-27, 1996.

9. Cameron, M. W., Efficiency and Fairness on the Road: Strategies for Unsnarling Traffic in Southern California, Environmental Defense Fund, 1994.

10. Giuliano, G., A Multicriteria Method for Transportation Investment Planning, Transportation Research, Vol. 19A, No. 1, 1985, pp. 29-41.