Breathing Space: Combating traffic congestion

Every urban Indian probably has a traffic horror story, one he/she initially thinks is so terrible it deserves an audience of sympathisers, only for subsequent experience or casual conversation to prompt a revaluation. Indian cities, those sprawling vortices of life, have oft been touted as the “engines” of future growth, going by the role that similar agglomerations like New York and London have played and continue to play for their and the global economy. 

Yet most developing countries witness awful traffic snarls on a daily basis that should be considered synonymous with the development process, but only highlight the progress to be made. The economic, environmental, even physical (think tension, road rage and accidents) costs of congestion are all too evident to a denizen of these afflicted cities, consequently becoming of interest to policymakers as well, who rely on various instruments to either directly or obliquely tackle traffic congestion and its concomitant problems, to varying effects.

A recent proposal motivated by this agenda was a 200% increase in tax on an additional (i.e. beyond the first) automobile owned by a family in Mumbai. It was felt that this would help congestion, rather bluntly, by reducing the potential number of vehicles plying the roads of the city. While the proposal received meagre press coverage, it and the abysmal state of roads in Mumbai today, excavated and reduced to strips, highlight the growing realisation that the present situation is untenable. Congestion is by no means an Indian, or even emerging or poor country phenomenon, and many developed countries have increasingly resorted to certain forms of incentivised traffic control that suggest the way forward. 

This article considers the Mumbai reform proposal, by articulating some results of the congestion pricing literature and describing other nations’ attempts to implement the prescribed policy measures, while also briefly remarking on some other features of our urban development.

Congestion pricing in theory

The earliest and simplest model dealing with congestion is static in nature, focusing on a single road, and is based on the realisation that traffic congestion is an externality imposed by motorists on their fellow travellers.

As in traditional economic analysis, one requires some conception of “travel demand”, which here is measured by vehicle flow, i.e. the quantity of trips demanded per unit of time. Travel is in some ways an intermediate good or service for a consumer, a means to an end (like work, shopping, recreation etc.). The cost of travelling depends on various things, like fuel costs, the opportunity cost of time spent in a queue as against working, costs of delayed/early arrival at destination etc. The dominant measure here is obtained from the speed-flow relationship. The fewer the number of vehicles, the higher the speed a vehicle can travel at. This speed declines with the number of vehicles, but if congestion is excessive, speeds may even increase, as more travellers opt away (termed “hypercongestion”). Travel costs then use the time cost interpretation (while also adding other costs like fuel and maintenance), and are derived from the speed-flow curve.

Traditional analysis equates the price of travel to private marginal cost (equal to the average social cost when there are many travellers), which neglects the externality effect. The social marginal cost includes this effect, and its incorporation into the analysis determines the efficient outcome, which can be arrived at through the imposition of a toll, the difference between these two marginal costs, called the Pigouvian tax.

By accounting for the “proper” cost of travel, the toll raises the price faced by travellers and thus reduces the number of trips demanded. Inevitably, this reduces consumer welfare as fewer people (those with a lower willingness to pay lose out, although they might resort to other modes of transport) travel, and those that do pay a higher price. 

On the other hand, the toll leads to an efficient outcome, resulting in gains from its implementation vis-à-vis the no-toll scenario. Finally, the proceeds from the toll (accruing to the government) could be channelled back to citizens, directly or indirectly (e.g. by reducing other tax burdens, or by funding infrastructure projects, preferably linked to the congested area itself). Social welfare increases as the efficiency terms dominate. However, the worsening of consumer welfare post-toll indicates why relatively few nations have actually implemented congestion pricing, another reminder of the importance of political considerations and acceptability for successful policy promulgation.

Things look more promising from an implementation perspective once one introduces a dynamic component, such as the dependence of travel demand on time or the evolution of travel flows over space and time. The most tractable approach involves considering congestion at a “bottleneck”, in recognition of the fact that they are quite pervasive and require a different sort of analysis. 

Now, individuals are assumed to care about both the cost of travel delays, analogous to the time cost above, and also a schedule delay cost incurred when one arrives at the destination at a time other than his/her preferred time. The bottleneck is assumed to have a fixed capacity (i.e. only a limited number of vehicles can traverse the stretch at a point in time). Consequently, the traffic inflows and outflows help determine the length of a queue (if any) and the duration of traffic delay and also schedule delay. Naturally, if inflows are below full capacity, there is no bottleneck queue and the rate of inflows equals that of outflows. In the other case, there is queuing and outflows equal the capacity of the bottleneck. It is assumed that travellers can decide their starting times (the time taken to reach the bottleneck from the origin, and to reach the destination from the bottleneck are assumed to be identical for all agents and often ignored).

In the no-toll case, the outcome is a Nash equilibrium with the choice variable being the departure time. An individual’s choice would depend on his relative evaluation of the costs of travel versus schedule delay, which are often correlated with the income level. The optimal outcome is one determined by a hypothetical social planner maximising citizen welfare. Both the decentralised outcome and the optimal outcome result in no travel delays as traffic flows are maintained at capacity for the duration of the period. However, at the optimum, departure times are chosen so as to minimise aggregate schedule delays.

The social optimum could be achieved with the aid of a time-varying congestion toll, which would depend on a variety of factors (including preferences over arrival times) and could be quite complex. However, the time delay cost would vanish, and schedule delays too might be lower, although travel costs would be higher under the toll. The toll would also tend to benefit those who valued travel time more, suggesting regressiveness. The fact that only departure times differ between the decentralised and optimal outcomes suggests a greater degree of “acceptability”, i.e. a higher chance of implementation.

This model has been extended to cover elastic travel demand, heterogeneity in travel costs, preferences over arrival time etc. The last extension in particular suggests more gains from the imposition of a toll, as arrival times could be tailored according to preferences, raising welfare. Another extension has been to the study of networks, and the pricing of various constituent links. Here, the optimal outcome is for each linked to be tolled efficiently, which is often infeasible. A related concept is cordon pricing, wherein access to and within an entire area is tolled.

These two approaches are simple, focusing on a single congested path (although several authors have used more realistic settings). Naïve adoption of an optimal congestion toll without consideration of other distortions in the system (e.g. other taxes) can reduce welfare, an accounting of which leads us to the theory of the second-best. Many studies have sought to move away from the ideal, first-best scenario highlighted above toward a constrained sort of analysis. Inevitably, welfare gains will be lower than what could be achieved in the ideal scenario.

One example is when not all links in a network can be tolled, for political or even logistical reasons. This includes an often seen situation where tolled and toll-free routes coexist. The existence of a (perhaps imperfect) substitute suggests travel demand elasticity is a factor. Indeed, the optimal second-best toll includes not only a marginal congestion cost for the tolled route, but also accounts for route diversion effects owing to toll spillovers on the toll-free route, which in turn depend on price elasticity of travel demand and route capacities. 

The relative efficiency with regard to the ideal depends on the degree of information available and the reach of the instrument at hand. Second-best tolls are also more complex, so even well intentioned programs could be damaging if potential spillovers are not accounted for. Allowing for dynamic aspects brings efficiency gains following the aforementioned channels.

It is difficult in practice to implement a time-varying toll such as the one suggested by bottleneck analyses. A compromise has been the introduction of differential pricing over time slots, e.g. during rush hour and off-peak hours, which compare favourably from an efficiency perspective with flat (invariant) tolls. The more the number of steps in such a “step” toll, the greater the efficiency.

Another issue might be the inability (due to informational or political problems) of differentiating between users, which could produce further distortions, like possible substitution between modes, besides direct efficiency losses. One should also consider the interactions with other distortions that motivate the second-best theory. Tolls raise the cost of living and thus lower the real wage rate, thereby discouraging work incentives. This could be mitigated through judicious use of toll revenues, like improving capacity, but simple redistribution could further dampen labour supply.

We would also do well to realise that individuals might not always be able to predict what traffic will be like, either due to general or idiosyncratic uncertainty. In such cases, real time information sharing systems could be useful (e.g. through radio, GPS and sites like Traffline.com)

Tolls might also affect incentives for capacity expansion, which has been the long-standing palliative for congestion problems. On the one hand, travel costs go down with an increase in capacity. On the other hand, the number of trips made goes up as well when the price of a trip declines. With underpricing of road use, the second effect is negative (using the argument in the simple model that marginal social benefits are less than corresponding costs), so the net effect is ambiguous. Choosing a toll lower than the first-best level could raise the marginal benefit from expansion, as the first positive effect would dominate. In the bottleneck model, expansion would alter departure times, which should be accounted for while setting tolls.

Two topics that bear some relevance in a second-best environment are parking fees and public transport, both of which are sometimes used as indirect decongesting mechanisms. Parking is a vital problem in most developing nations, with encroachments rampant and not enough facilities. Parking fees are a significant component of total travel costs, and so optimal pricing should follow the marginal cost approach described earlier. This is rarely practiced, however, distorting both incentives for travellers regarding modal choice, and the possible creation of facilities on account of an effective zero charge due to encroachment. A spatially differentiated parking toll could bring about a desired parking pattern.

If the ultimate aim of any policy is to bring the number of vehicles down (which also reduces emissions), one must consider alternative arrangements like public transport that, as seen above, would also affect second-best road tolls. Marginal cost pricing would suggest low public transit fares (when capacity is adequate), which neglects the impact of underpriced roads on fares that would tend to subsidise public transit (to reduce overall congestion). The subsidy would depend on scale economies in public transit, as well as on the degree of congestion reduction due to modal diversion (which would not matter if road travel were optimally priced).The first factor arises when marginal costs are below average costs, for example due to fixed costs or a decline in waiting time when frequency increases. The relative importance of these two factors is time, mode and location specific. In practice, public transit subsidies are often motivated by distributional concerns, and the related idea of acceptability.

From theory to practice

Actual congestion pricing schemes are hardly universal, with the incidence of higher tolls on travellers, regressiveness and consequent acceptability issues being prime factors. The complexity of optimal tolls obtained in more realistic settings is also not ideal. Nevertheless, the sweeping transformation that information technology has brought to our lives suggests that such complexity can be handled at an increasingly lower cost.

The classic application of congestion pricing is the Area License Scheme (ALS) OF Singapore, instituted in 1975 in the business district, which has been increasingly reliant on technology like smart cards. Spillover effects onto substitute routes have been a problem over the years, despite the gradual accounting for of such effects in toll setting. The tolls are time varying, depending on average speeds measured in the last quarter, and are frequently revised.

The London cordon-pricing scheme, applicable to vehicles traversing its business district, began in 2003 and relied on video identification using number plates. It has been successful in that traffic has declined in the area while road speeds have gone up, with the already extensive tube and bus systems proving very useful. The United States too employs time varying tolls in certain areas in an approach called “Value Pricing”, where some routes have to be left toll free, but results have been mixed with spillover effects often underestimated.

In the context of what we have learnt, a higher tax on additional vehicles would be a crude tool (although other proposals include revised parking charges and fuel taxes).

Automobile expansion is inevitable, being a facet of a higher standard of living and a tangible embodiment of social mobility. In the absence of capacity expansion plans, it is true that a congestion toll would reduce the number of vehicles plying a route. Yet, the ideal congestion toll brings about the socially efficient outcome, and more realistic tolls are designed to get close to that outcome. 

The proposed tax hike would distort consumption decisions, for example by inducing consumers to switch to cheaper modes of transport (although it is not difficult to imagine congestion itself as a factor in a vehicle purchasing decision). Further, route or even cordon pricing is applicable to specific areas or roads, while such a tax hike would dissuade consumers from overall travel. Given the obvious correlation between income and the number of vehicles owned, such a tax would be progressive, although the behavioural response to so exorbitant an increase (captured by elasticity) hints at limited revenue being actually generated. Considering also that the affluent would tend to have a greater willingness to pay high tolls, potential revenue could also be forgone. 

One could also question the impact such an increase would have on congestion, as multiple vehicle owners are unlikely to make up a significant proportion of the vehicle driving population. The muted response of automakers is surprising as well, and could be construed as scepticism regarding the plan’s outcome and even its implementation. Finally, issues of fairness in treating purchasing cohorts differently and penalising large families should be considered.

Interestingly, a committee under the aegis of the Mumbai High Court last March mooted the idea of congestion pricing through a sort of cordon tax imposed on vehicles in the Central Business District, while also recommending paid parking for residents and in older, less capacious areas (see here). It especially advocated the use of technology in gathering and disseminating information and setting tolls (recall that the use of instruments like GPS alleviate uncertainty discussed above). They too proposed a cap on new registrations alongside these measures.

A few final points are in order. First, vehicles have other externalities as well, the most insidious being accidents and pollution. The fuel tax and emission standards are the appropriate instruments with regard to the latter, and India too has been utilising them. Second, it is imperative that a decongesting plan employing whatsoever instrument should also keep in mind the availability of alternatives, and not just from a pricing viewpoint. While the aim of keeping road traffic manageable is laudable, a policymaker must realise that although he could certainly influence mode-specific demand, he should not try to distort overall travel decisions, i.e. going to work, making purchases etc., which he could do via ill-thought policies. 

Public transit is a natural outlet for decongestion, and urban planners and local, state and central governing agencies have focused increasingly on improving their quality, through schemes like the JNNURM, metropolitan area metro services (including monorail services) and Bus Rapid Transit schemes (BRTs). There are questions regarding the viability of the Metros, and the half-baked nature of implementation of BRTs, which often do not have separate lanes and therefore lose most of their raison d’être. Nevertheless, the successful implementation of BRTs in some cities like Ahmedabad bodes well.

References for the curious reader

The pioneer in the field of congestion pricing was the versatile economist William Vickrey, who proposed the simple bottleneck model. A comprehensive account of his work and that of his peers and followers can be found in the (technical) text, The Economics of Urban Transportation, 2007, Routledge, by Kenneth Small and Erik Verhoef, particularly chapters 2 and 3. 

Good and less technical reviews are ‘Traffic congestion and road pricing’ by Robin Lindsey and Verhoef in the Handbook of Transport Systems and Traffic Control, Pergamon, 2002; and ‘The Rationale for Road Pricing: Standard Theory and Latest Advances’ by Kenneth Button in Road Pricing: Theory and Evidence, Research in Transport Economics (Volume 9), 2004.

An account of the externalities imposed by automobiles with a decidedly American perspective is ‘Automobile Externalities and Policies’ by Ian Parry, Margaret Walls and Winston Harrington in the Journal of Economic Literature, June 2007.

There have been papers on urban poverty and transport using data from Mumbai by authors affiliated with the World Bank. Note also that transport costs (including those due to congestion) play a key role in theories of trade and economic geography. A good resource on road taxes is the study, ‘Road User Taxes in India’ by Mahesh Purohit and Vishnu Purohit, commissioned by the Planning Commission (available here).

An interesting experiment conducted by researchers affiliated with Stanford University and Infosys in Bangalore (called INSTANT) provided decongesting incentives to commuters, described here. 

Lalit Contractor has a MPhil in Economics from Oxford University and is curious about Economics and its interactions with politics and society.