Residents of many large cities endure oppressive levels of traffic congestion, and ample evidence suggests that the rise of Uber, Lyft, and other transportation network services (TNS) has exacerbated the problem. Policymakers can use congestion pricing to combat heavy traffic, where drivers pay a fee to enter a busy area, usually during rush hour. Although these fees have traditionally been levied on privately owned vehicles, cities are beginning to impose congestion charges on TNS rides. But how high should these charges be set? While it might seem obvious that TNS and private rides should face the same congestion charge, in a new job market paper, AREC PhD student Julian Gomez-Gelvez shows that when one or two TNS companies dominate the market, the optimal charge could be much lower.

Using Congestion Pricing to Correct Market Failures

The rationale behind congestion charges is that when people decide whether to take a trip to a center city during rush hour, they don’t consider the costs they impose on other drivers by slowing them down. In other words, there’s an externality. The congestion charge causes drivers to internalize those costs, and as a result, people will make fewer trips and congestion will be lower. The standard advice from economists is that the congestion charge should equal the marginal external congestion costs.

So if the goal of these policies is to reduce congestion, and an Uber ride causes just as much congestion as any other ride, the TNS ride should face the same congestion charge—right? That was basically the logic that London policymakers used recently when they decided to impose the same congestion charge on (previously exempted) TNS rides as on all other rides.

However, Julian points out that with TNS rides, another force is in play: market power. In many cities, either Uber or Lyft dominates the market for TNS rides; other companies match their prices if they want to compete. This market power allows Uber and Lyft to charge prices above the company’s cost of providing the ride, where the price is the amount the customer pays the TNS company and the cost is the amount that the company pays the driver. The more market power the TNS company has, the higher its profits. If we imagine two hypothetical cities that are identical except that in the first city Uber is a monopolist but in the second city Uber and Lyft compete against one another, we’d expect the competition in the second city to drive down prices and profits.

In general, when firms have market power, consumers lose because they pay the higher prices. But in this case, the market power and resulting prices already cause the consumer to internalize some of the external congestion costs caused by the ride. Consequently, the congestion charge on TNS rides doesn’t need to be as high as the congestion charge on other rides.

Evidence from Bogotá

Having made this theoretical point, Julian uses a model to compute the optimal congestion charge for TNS rides in Bogotá, Colombia. Bogotá makes an interesting case study as a large city (7.5 million inhabitants) with some of the worst traffic congestion in the world, despite having an extensive public transportation system. And in Bogotá, Uber exerts considerable market power, accounting for about 70% of all ride-hailing trips in 2019. In Julian’s model of transportation in Bogotá, Uber, as a monopolist, sets the price of each ride to maximize its own profits, and consumers decide whether to take a TNS ride or an alternative such as public transportation. The consumer makes this decision based on the expected travel time of the TNS ride and the value of the alternative travel mode. The monopolist sets the price accounting for the effect of the price on travel time (that is, setting a higher price reduces ridership and congestion, increasing demand for the ride).

His main conclusion is that the monopolist’s markup—the difference between the price the consumer pays and the price the driver gets—is about 70 percent of the congestion cost. In other words, when consumers decide whether to take Uber in Bogotá, the price already includes most of the congestion cost. Therefore, the optimal congestion charge on a TNS ride is less than the congestion cost of the trip.

These results have clear policy implications for the numerous cities that already have or are considering congestion charges, and in which TNS companies have a substantial presence. Although the optimal congestion charge for TNS rides varies from city to city depending on consumer preferences, availability of alternative travel modes, and other factors, if TNS companies have some market power, the optimal charge is less for TNS rides than for other rides. The idea of differential congestion charges may face political and technical hurdles, but the case could be made on the basis of a more progressive approach because low-income earners are heavy users of TNS providers.