Environment

Driving Delhi: The impact of driving restrictions on driver behaviour

  • Blog Post Date 08 February, 2016
  • Perspectives
  • Print Page
Author Image

Gabriel Kreindler

Massachusetts Institute of Technology

gek@mit.edu

Author Image

Jasmine Shah

J-PAL South Asia at IFMR

jshah@povertyactionlab.org

In an attempt to address Delhi’s grave pollution problem, the state government experimented with a driving restrictions policy for a fortnight in January. Based on a phone survey of a sample of 614 drivers in the city, this column describes how the policy changed drivers’ behaviour in terms of labour supply, number of daily trips, travel modes, and satisfaction, between restricted and unrestricted days while the policy was in effect.

Many large cities in developing countries suffer from severe traffic congestion and high pollution levels. Together, these issues are likely to have considerable and varied negative consequences on human health and economic activity. Indian cities are particularly affected, with six out of the top 10 most polluted cities in the world being from India (World Health Organization (WHO), 2014).

While pollution has many causes, traffic congestion is a particularly salient contributor, and several cities worldwide have adopted driving restrictions as a policy to ration road space and bring down pollution levels. Prodded by the Delhi High Court, the Delhi government recently joined this trend, and implemented a similar policy on a trial basis. Between 1st and 15th January 2016, the Delhi odd-even policy prohibited cars with license plates ending with an odd digit from being on the road on even days of the month, and vice versa. The government issued several exemptions1 from the policy, notably for women drivers, vehicles running on compressed natural gas (CNG), and two-wheelers. The policy was in effect every day except Sundays, from 8 am to 8 pm.

Potential benefits and costs of driving restrictions

The potential benefits of driving restrictions have been studied and debated extensively, both for the Delhi trial period and for similar policies worldwide. The early consensus was that these policies did not lead to any change in aggregate pollution levels, and researchers argued this was due to drivers purchasing additional and potentially more polluting vehicles (Eskeland and Feyzioglu 1997, Davis 2008, de Grange and Troncoso 2011, Bonilla 2013, Gallego et al. 2013). However, more recent studies find that driving restrictions in Quito and Beijing reduced congestion and pollution (Carrillo et al. 2015, Viard and Fu 2015). A study of Delhi’s odd-even policy that compares data across time within Delhi and just outside Delhi also finds that the policy was effective in reducing peak pollution levels (Energy Policy Institute at the University of Chicago (EPIC India), 2016).

Meanwhile, the policy’s potential costs also deserve close scrutiny, and have received comparatively little attention. One natural concern is the inefficiency introduced by constantly disrupting commuters’ daily habits, in terms of forcing them to find alternate – and possibly unfamiliar – travel modes, which can lead to cancelled trips, reductions in labour supply, and a direct wellbeing cost2. There is surprisingly little evidence on these issues, as well as on the extent to which drivers comply with restrictions. These questions are addressed in a recently concluded study led by one of the authors (Kreindler) and conducted by the Abdul Latif Jameel Poverty Action Lab (J-PAL) South Asia.

Impact on drivers’ behaviour

The study enrolled 614 car drivers3 who came under the purview of the driving restrictions in Delhi, and tracked their behaviour over phone surveys throughout the 15-day trial period, including both days when they were allowed (unrestricted) and days when they were not allowed to drive (restricted). Drivers were recruited in petrol pumps from across Delhi, in the days leading up to the trial period. This strategy is likely to have reached a broadly representative sample of drivers in Delhi, even though sampling was not strictly random. For the 18% of drivers from households with at least two cars, we label the car used when they were first recruited into the sample as the “primary vehicle”; these drivers may still be affected by the policy, because the other cars may be used by other household members. Drivers whose primary vehicle had an odd license plate number were considered ‘restricted’ on even days, while those with an even number were considered restricted on odd days. Since odd and even license plate numbers are essentially randomly assigned, any systematic difference between the restricted and unrestricted groups of drivers will be due to the policy, instead of any other factor.

The results below describe how the policy changed drivers’ behaviour between restricted and unrestricted days, in terms of labour supply, number of daily trips, travel modes, and satisfaction. It is thus important to keep in mind that these results do not describe how drivers’ behaviour changes during the policy relative to before the policy; rather they capture the day-to-day variation in outcomes while the policy was in effect. This information complements and is complemented by existing studies of the aggregate impact of the policy.

The policy led to a substantial and significant decrease in the use of primary cars on restricted days, as intended and as expected. At the same time, there was also a small yet significant non-compliance with the policy. On a typical unrestricted day, that is, a day when the respondent is allowed to drive their primary car, 50% of respondents travelled using their primary car4. This simply means that even in our sample of regular drivers, some people do not make any trips on some days, while others use alternative travel modes, even if they are allowed to use their car. The fraction of respondents who travel using their primary car drops to 17% on restricted days, a two-thirds decrease. However, this means that 17% of respondents reported driving on days when their vehicle was restricted, that is, they most likely did not respect the rule, or travelled before 8 am when the policy enters into effect5.

It is unlikely that the non-compliance finding is spurious. Respondents who misreport are likely to underestimate non-compliance – an illicit behaviour – rather than overestimate it. The fraction of drivers travelling before 8 am – when the odd-even policy comes into effect – is 3.5% on restricted days, which means around 14% of drivers were actually non-compliant. Taking sampling uncertainty into account, non-compliance is above 10% with high confidence. Meanwhile, only three respondents in the entire study ever reported paying the Rs. 2,000 fine.

The 32% of drivers who switched away from their primary car on restricted days chose from a diverse range of alternatives, depicted in Figure 1. Indeed, all travel-mode categories other than primary four-wheelers saw increases, most of which were statistically significant. First, the large reduction in primary car usage was attenuated by an increase in other (unrestricted) household vehicles, from 2.6% to 6.4%, and an increase in two-wheelers, from 9.0% to 15.1%. Note that these are short-term changes, and in the medium and long term this difference may be larger – if drivers need time to learn to adjust to other modes of travel – or may be attenuated further if people purchase additional vehicles, as previous research suggests (Davis 2008). Secondly, on restricted days some drivers switched to taxis – from 1.4% to 2.5%, and to auto rickshaws – from 1.2% to 4.6%. Taken together, the use of any motorised vehicle, private or hired, fell from 65% to 46%, a decrease of less than one third.

Restricted days witnessed a significant shift to public transportation, with approximately 5% more drivers turning to metro, and 4% more to carpooling. The shift towards using the bus is also positive; however, it is smaller and not statistically significant. In the entire study sample, there were no reports of drivers using bicycles to commute, and walking was chosen by a very small fraction of respondents both on restricted and on unrestricted days.

Figure 1. Impact of the odd-even policy on daily travel modes

Figure 1 also reveals that some drivers simply find it too inconvenient to travel if their primary car is not available. Indeed, approximately 8% fewer drivers make any kind of trips on restricted days, compared to unrestricted days. Around half of these trips turn out to be work-related. These results show that the odd-even policy likely upended the work routine of a fraction of the affected drivers, although we cannot say with certainty whether these trips were permanently cancelled, or mostly rescheduled to unrestricted days.

Overwhelmingly, Delhi drivers report being satisfied with their daily commute. Indeed, 90% of drivers who made a trip on an unrestricted day say they were satisfied with the trip. The satisfaction rate is lower on restricted days, yet at 83% it is still very high. The results on inconveniences follow the same pattern: we detect an increase in reports of inconveniences, from 11% on unrestricted days to 15% on restricted days, which is mostly accounted for by increases in reports related to late arrivals and trip timing. However, we are not able to detect any changes in actual (self-reported) trip duration, departure time or distance. This may be related to imprecise answers from respondents, who very often ‘round-off’ these values to the nearest 15 minutes or 5 kilometres. Importantly, we do not find any evidence of reports of getting lost or stranded, being harassed during the commute, or feeling tired after the commute.

What this means

In the press, the odd-even policy was mostly judged a success in terms of public acceptability, while its impact on pollution levels is still being debated. The results described above support the view that, on average, Delhi drivers received the policy well, and many were successfully able to change their travel routines in response to the policy. However, the analysis also reveals a few different points, while leaving yet other questions unanswered.

A significant share of drivers – very likely at least 10% – did not comply with the policy. These drivers may have chosen to break the rule either because the inconvenience of switching to another transport mode was too high, or because they thought that enforcement was sufficiently unlikely. Both issues should be of concern for policymakers, yet the policy responses differ. Unfortunately, it is not possible to credibly distinguish between the two hypotheses with available data.

The results highlight the great diversity of alternatives available to drivers. There does not exist any main alternative to car driving; instead, drivers are remarkably evenly divided between using other household vehicles, taxis, public transport options, and travelling less. In particular, this implies that the effect of the policy on the use of any kind of private or hired motorised vehicle is significantly smaller than the effect on primary four-wheelers, and the switch to public transport is less than one might expect.

In many ways, the debate around the effects of the odd-even policy highlights the need for more rigorous and carefully designed studies, in addition to smart policy experiments and quality data, to uncover which policies work most effectively to contain congestion and pollution. This in turn requires policymakers to work in close collaboration with researchers – not after the policy has been designed and launched – but at the early stages of the policy design.

Notes:

  1. The full list of exemptions can be found here.
  2. For instance, Viard and Fu (2015) show evidence that the driving restrictions in Beijing increased leisure and reduced labour supply of workers with flexible hours.
  3. The sample was male drivers, owners and regular users of a petrol or diesel car.
  4. The travel mode numbers are slightly smaller than those quoted in the paper. For Figure 1, we categorise each respondent as using exactly one (main) travel mode, while in the survey respondents could select multiple options. In order of priority, the options were: if the respondent used a primary four-wheeler, other four-wheeler owned by household, two-wheeler, taxi, auto rickshaw, carpool, metro, bus, walking, and no trip.
  5. Respondents to the phone survey were asked about same-day trips, so no trips after 8 pm were recorded. None of the car trips on restricted days was for emergency purpose.

Further Reading

No comments yet
Join the conversation
Captcha Captcha Reload

Comments will be held for moderation. Your contact information will not be made public.

Related content

Sign up to our newsletter