How to solve Beijing’s traffic congestion woes*

The difficulty with Beijing’s traffic is self-evident to anyone who drives or indeed who tries to walk around its streets as a pedestrian. Some stirring photographs are here.

There seem to be about 4 million cars owned in the city and this stock is experiencing explosive growth.  According to recent articles in China Daily the number of vehicles on Beijing’s roads grew by 800,000 in 2009. This suggests that forecasts of car ownership of 7 million by 2015 are, if anything, on the conservative side.  With current road use policies the levels of congestion on Beijing’s streets can be expected to markedly worsen.

But current policies are not being maintained.  Parking fees in central Beijing are being increased and the number of cars that may be on the road during the week is limited by the last number of vehicle license plates – odd or even numbers can only be used on alternative days. The subway system has recently been expanded and roads are being improved and expanded although this has occurred at high cost given the high opportunity values of land in this vast and densely populated city. If the historical experience of American and European cities is a guide then many of these policies will either fail to reduce congestion or have, at best, only very limited effectiveness.  

The difficulty is that these policies fail to address the core underlying issue – there are excessive demands for road use that stem from the fact that the price of road use is too low.

People take journeys in their cars when the value to them of the journey exceeds their private costs.  These private costs include the depreciation and maintenance costs of owning a car, pollution costs they individually suffer, traffic accident risks, insurance costs, petrol costs and the time costs they incur particularly when traffic is congested.  These cost calculations however do not account for the fact that, when an extra motorist joins a stream of traffic, they increase the travel costs of other drivers in that stream. 

Suppose the private costs a motorist (“Mr Li”) encounters on a particular journey are 40 RMB but, by making that journey, the travel costs of 10,000 other cars on the road are increased as a consequence by some small amount per vehicle say 0.001 RMB. This cost represents the value of extra pollution and congestion costs each other road traveller incurs as a consequence of Mr Li’s travel. Collectively this extra social cost is 10,000*0.01 or 100 RMB in total.  Mr Li will take his journey when his journey’s value exceeds his private costs of 40 RMB. But from society’s viewpoint he should account not only for his private costs but also the costs he inflicts on other motorists which are 100 RMB. Thus from society’s viewpoint Mr Li should only travel when the benefits he gets from his journey exceeds the much higher social cost of 140 RMB.

If motorists, such as Mr Li, do not account for the larger social costs they inflict they will take too many journeys from a social viewpoint and there will be excessive road congestion.  This is what economists call an externality – drivers are excessively using roads because such use is socially under-valued.

Congestion pricing

The solution here is to price travel by levying a charge on motorists for use of congested roads that captures the social costs caused by vehicle travel. The charge could be levied by setting up gantries around Beijing which charge motorists when they pass certain points by detecting their travel on an In Vehicle Unit or IVUs. They might for example be installed at entry and exit points onto Beijing’s major ring roads and arterials. This system has been successfully used to manage congestion in Singapore for 45 years.  The scheme is also used in Stockholm and London via area cordons that are set around the city and which price access to the city. Cordons are probably less appropriate for Beijing since the city has a vast size and does not have clearly identified high density CBD areas.

Alternatively, more recent approaches use GPS or other technologies to monitor car movement by satellite using GPS (or other related mobile) technologies. Then charges are attached to driving a car that depends on current congestion conditions. Electronic charging of all travel in the Netherlands is now being trialled using such technologies.

Ideally the price charged for making a journey will depend on the time of day the journey is made. Travel during the morning and afternoon peak periods is most intense so that at these times charges should be highest.  The idea is to provide motorists with incentives to switch their travel times away from these peak periods to ‘smooth the peaks’. Providing the highest charges at peak times means that only those with the most highly valued journeys will travel at these times.

The idea is to increase the charges on traffic use until excessive congestion at each time is eliminated.  This will involve experimentation since there are always complicated network wide effects – pricing one road will affect the demand for travel on other roads. Beijing has considerable natural advantages to operating such an electronically-based system given the pattern of ring roads around the city and the fact that most users of Beijing’s roads are Beijing residents.  In other world cities there are significant problems with occasional users of roads that add significantly to the complexity of pricing since visiting vehicles may not carry the necessary IVU.

It is also clear that the Beijing authorities have good information about real time congestion conditions in their city.  A map, available online, shows these congestion densities1.

Supply measures and parking policies

It is important to understand why supply-driven measures won’t work and why parking charge increases in Beijing – while probably a good idea – will only partially work to reduce congestion. 

People who drive their cars are well aware of the misery of congestion.  Many potential journeys in a city are not undertaken simply because congestion is so bad. This means there are high levels of latent demand for travel in a city like Beijing. These are people who would take their car to travel if only congestion was less severe. Supply measures such as providing more roads or providing increased availability of public transport do provide a temporary release from congestion but this is likely to be swamped both by the surge in demands that were previously latent and by growing levels of car ownership.  Measures such as allowing only certain number plates to use roads on particular days have similar effects. Apart from being costly policies – there is no guarantee that those with certain numbers who are excluded from travel have low-valued trips planned – those not using the road will create temporary congestion relief that again induces a surge in demands that were previously latent.  Congestion remains the same.   Anthony Down’s book, Stuck in Traffic, calls such effects of supply-oriented policies in releasing latent demands triple convergence.   Such phenomena lead to Down’s Law of Peak-Hour Traffic Congestion – on urban commuter expressways, peak-hour traffic congestion rises to meet maximum capacity. Duranton and Turner (2009) provide striking evidence supporting this law for US2 cities. Put simply, supply policies will fail to reduce congestion – it is essential to put a price on what really is in deficit supply, namely the use of road space.

Parking policies require careful discussion.   For parking markets to work effectively anyone should be able to park anywhere it is legal to do so by paying a sufficiently high parking charge.  This has the normal rationale of setting supply equal to demand in the parking market. If parking charges are set too low then motorists have to search the streets (to ‘cruise’) for a parking spot which adds to both congestion and pollution.  Donald Shoup in his The High Cost of Free Parking estimates that up to 74 per cent of traffic in downtown areas are searching for a parking spot.  To the extent that recent proposals to raise parking charges in Beijing to a maximum of 10 RMB per hour reduce such cruising then the move is a positive.

Unfortunately however the demand for parking is distorted by the fact that scarce road space is under-priced so it is probably questionable whether charging a fee of up to 10 RMB per hour is enough to eliminate excess parking demands and consequent cruising.

It is worth noting that the Beijing administration explicitly see hefty parking fees as a way of reducing congestion.  This will be true if all traffic heading into central foci in Beijing is terminating there and not travelling through these foci to reach points elsewhere in the city.  Such ‘through’ traffic will be unaffected by these parking charges. Indeed there might be increased incentives for motorists to make ‘through’ journeys if terminating traffic is reduced – this is again an instance of triple convergence – latent ‘through’ traffic demands are released because terminating traffic is restricted.

 The effectiveness of using parking charges as a substitute for proper congestion pricing therefore depends on the pattern of trips within Beijing and whether in fact those driving into highly congested areas are, in fact, terminating their trips there.  My casual observations as a foreigner who is only starting to observe what happens in Beijing is that many through journeys are in fact because the city itself has a rather diffuse set of foci. If my observation is sound then parking policies will have limited impacts on overall congestion.

 What to do with revenues raised

The idea of congestion pricing is traditionally not primarily related to raising revenues.   There might be specific arguments for reversing this view in the case of Beijing. Charges on motorists fall mainly though not exclusively on the relatively affluent residents of this city – those who can afford a vehicle – and hence congestion charges can be regarded as progressive in their impact. Parking charges have the same type of progressive impact and can also be seen as extracting revenues from resources (‘parking spots’) which are in somewhat (not entirely given the capacity to build above or below-ground car parks) restricted supply so standard economic theory suggests such parking taxes will efficiently provide revenues in the sense of implying low deadweight  losses.  So there might be a revenue gaining by-product (or ‘double dividend’) advantage of such parking fees.

 One use for revenues yielded by congestion and parking charges is to compensate those disadvantaged by the charges.  Revenues could, for example, be spent on improving public transportation in Beijing – the very good bus and subway systems in this city are often very over-crowded so it might make sense to provide better services particularly for those deterred from using their car by using the congestion and parking charges.  These funds could also be spent on improving environmental amenities in the city such as parks and other public goods that are also used by all.

Finally, if particular roads or particular parts of a road network yield very high revenues from charging efficient tolls this can be a signal to expand or improve such roads subject to the constraint that construction costs – included the costs of reclaimed land – are not prohibitively high.  This type of reasoning – due originally to Mohring and Harwitz in 19623 – is a more satisfactory approach to thinking about investing in roads than taking such investment decisions on the basis of forecast demands when congestion costs are unpriced.

 Final comments

 The most appropriate way to allocate a scarce resource is to use prices to allocate that resource to those users who value the resource most.  This means the total value people get from that resource is maximised. This argument is sound both for consumer and investment goods and is also true for the roads we drive on.

 Congestion causes unnecessary social costs because motorists are not constrained to pay for all of the social costs they impose on others.  Pricing congestion is the only feasible way of managing congestion.

 *  Harry Clarke is Professor of Economics at La Trobe University in Melbourne Australia.  He is currently visiting the School of Economics, Peking University.  Contact at h.clarke@latrobe.edu.au.

 1 See http://210.75.211.252/publish/portal1/tab185/e

 2 See G. Duranton & M.A. Turner, ‘The Fundamental Law of Road Congestion: Evidence from US Cities’, NBER Working Paper 15376, September 2009.

 3 H. Mohring & M. Harwitz, Highway Benefits: An Analytical Framework, Evanston, Illinois, Northwestern University Press, 1962, 70-87.