“If every car in San Francisco was Ubered, there'd be no traffic.”This was the brash claim of Travis Kalanick, erstwhile CEO of Uber, the pioneer in ride-hailing services. Radically reducing both car ownership and traffic congestion were part of Uber’s founding vision. To what extent has this promise held up? Can ride-hailing services provide part of the solution to traffic congestion globally? Or are they part of the problem?
In a recent paper, we argue that the net effect of ride-hailing services on traffic congestion hinges on precisely the question of whether commuters see them as alternatives to cars or to public transport. With the coming of ride-hailing services, does the average commuter shift away from low occupancy car trips? Or does she substitute Uber for trips for which she would otherwise take a bus or local rail? The news is not good for Uber. Our findings offer some of the clearest evidence yet that ride-hailing may exacerbate, not ameliorate, congestion issues.
The evidence from American studies so far has been mixed. Studies using data from a variety of American cities suggest that ride-hailing services have the potential to cut down private car usage and taxi fleet numbers while complementing public transport. On the other hand, compelling recent work increasingly shows the opposite effect. In a research report based on surveys from seven major metropolitan areas, Regina Clewlow and Gouri Shankar Mishra at the Institute of Transportation Studies in the University of California, Davis note that commuters appeared to view ride-hailing as a substitute for certain kinds of public transit. Another study published in the journal Science Advances found that ride-hailing networks are the largest contributor to traffic congestion in San Francisco.
Why is the evidence so inconclusive? First, the city context matters. Cities can be vastly different in terms of transport infrastructures and cultures. In turn, that determines the impact of ride-hailing services on road-traffic scenarios. Second, early studies tended to be exploratory in nature and failed to establish causation. A related problem is that reliance on self-reported user data may obfuscate actual behaviour and usage patterns. The gold standard here continues to be actual traffic data. Last but not least, it is very likely that the factors that increase numbers of Uber or Lyft cabs on urban roads are related to traffic congestion. Thus causally differentiating the effect of ride-hailing on traffic congestion presents a chicken and egg problem.
In our study of Indian cities, we are able to deploy a variable that affected the numbers of ride-hailing vehicles on the roads but was not determined by levels of traffic on the roads. Over 2017 and 2018, drivers of Uber and Ola cabs, an Indian provider of ride-hailing services, went on strike in three Indian cities – New Delhi, Bangalore and Mumbai. These strikes were organized in protest against declining earnings, as a result of falling passenger fare rates as well as the driver bonuses these companies offered. We compare the levels of congestion during these periods with levels in these cities in previous years or with similar control cities. Our data comes from real-time traffic metrics available from Google Maps at 20 minute and half-hour intervals.
Effects on Traffic
With this clean design, our results offer strong confirmation for the thesis that ride-hailing services in fact increase levels of traffic congestion. This finding is evidenced by shorter traffic delays when Ola and Uber are off the roads during the strikes. We compare traffic levels in Delhi the same dates a year before the strike. In Bangalore, we compare the strike period to the same dates a year after the strike. For Mumbai, we compare the strike period to the same dates in Pune, a neighbouring city that did not witness strike-related disruptions in ride-hailing services. Overall, we find that morning and evening peak hour traffic delays reduced by 5.8-6.6% in Mumbai, 4.6% in Delhi and 7.3% in Bangalore during the respective strikes. These effects might appear small but they are highly statistically significant. Moreover, the effects are not negligible. In Delhi, the delay reduction is about 40% that of a public holiday. In Mumbai, it may be as much as half the delay reduction caused by holidays.
We go further and demonstrate that ride-hailing services substitute for, rather than complement, public transit. In the context of Delhi, we have data on ridership levels on the busy Metro Rail. We find that the strike increased ridership on the Delhi metro by a small but significant amount.
Given the setting of our study, do these results have relevance for cities internationally? The answer is an unqualified yes. India happens to be the second-largest market for Uber outside the United States. There have been relatively few rigorous studies of developing country settings before ours. This gap exists despite the fact that emerging market cities top traffic congestion rankings. An analysis by the Boston Consulting Group in India found that shifting from private cars to ride-hailing services on a large scale would reduce both private cars and traffic congestion levels. A study by the firm AlphaBeta in Indonesia argued that “shared mobility” on ride-hailing or public transport would reduce congestion levels. Both studies were commissioned or done in collaboration with Uber.
In contrast, our study proffers more rigorous but much less sanguine results. Transport policymakers should undertake more careful evaluations of the costs and benefits of ride-hailing services before deciding regulatory responses.
The full research can be read here.
Views are personal.
Deepa Mani is associate professor of Information Systems at the Indian School of Business, Saharsh Agarwal, is a doctoral candidate at the Heinz College at Carnegie Mellon University in Pittsburg, U.S., and Rahul Telang is professor of Information Systems and Management at CMU