Air pollution and traffic-related mobility in Pakistani cities

Urban cities in Pakistan are a hotspot for local air pollution. Around 40% of the country’s 220 million people live in cities where air pollution levels exceed the World Health Organisation’s recommended limit. This air pollution is not only limited to a certain season or source but has a perpetual bond with Pakistanis.  Particulate matter (referred to as PM from here on) is one of the constituents of air pollutants. This PM, along with other aerosols produced from industries, vehicles, domestic heating or cooking, agricultural activities, and power stations can travel across boundaries. This means that a pollutant associated with a power station 100 km away from a city is able to transcend its local boundaries if the prevailing wind allows.

However, the conversation around air pollution abatement strategies are seasonal. Localised, limited efforts and conversations are concentrated at pointing at the crop residue burning season (one of which happens around October/November).  This leads to worsened air quality with the onset of winters due to a combination of weather conditions and the fact that our efforts assume a uni-causal explanation for smog (crop burning) rather than a multi-causal one.

Recently, the Punjab government in Pakistan announced a smart lockdown in 10 districts of the province, which they claim would reduce traffic and therefore improve air quality. The lockdown means that educational institutions and offices will stay closed. Moreover, markets and business centres would only be allowed to open before 3 pm on Friday and Saturday. Policies such as restricting traffic in car-centric urban and sub-urban Punjab on weekends in ad-hoc and would not mitigate the hazard on most days when people have to go to work or kids to school.

During 2020-22, Google released traffic mobility data in assessing the impact of lockdown measures and correlating mobility with pandemic levels. In Pakistan, the lockdown was lifted by 2021 and people were mostly commuting without any such restrictions. Similar to COVID-19 assessment, we have used mobility data from November 2021 to Oct 2022 to find any relationship of PM2.5 concentration levels with mobility.

The purpose of this piece is to highlight the relation between hazardous air and mobility through a linear regression model (OLS). The results show that mobility is not naturally constrained by hazardous levels of pollution. These results paint the Punjab Government’s “smart lockdown” policy in a problematic light, as they highlight problems that would render its effectiveness limited, and question its thoughtfulness.

Measurements of Air Pollution

For the purposes of this paper, we look at PM2.5 concentration in major urban centres of Pakistan. The data for this measurement was taken from AirNow (https://www.airnow.gov/), an open source database for monitoring air pollution, and PAQI. In Figure 1, we plot the weekly average PM2.5  concentration in each major city between November 2021 and October 2022.

From Figure 1, we can see that high air pollution levels are generally experienced from late Autumn to early Spring, though even outside those seasons we see undesirable levels. The reason being a combination of weather conditions and the spatial dependence of three of the four cities (barring Karachi) being located in airsheds. Other reasons may exist, but we feel as though the above two explanations are plausible conclusions rooted in fact.

Furthermore, we can see that while Islamabad and Karachi have high amounts of air pollution, Lahore and Peshawar see PM2.5  concentration often well above 300-400. Keep in mind that these are weekly averages, and on a daily level, these figures may be even higher. This is inline with what we observe, smog and visible air pollution is far more prevalent in Punjab, around Lahore and its peripheries.

Figure 1. Weekly average for PM2.5 concentration (μg/m3) from Nov-21 to Oct-22 for Peshawar, Islamabad, Lahore, and Karachi.

Model

To attempt to estimate the effect of air pollution on mobility, we ran a regression which tested the effect of varying levels of air pollution on different types of mobility. We limited this test to Karachi, Islamabad, Peshawar and Lahore. This was partly due to data limitations, but also because these cities, in particular Lahore and Peshawar, are noted for having strong and lasting episodes of smog.

We ran four individual regressions with a single dependent variable identifying mobility for a particular type of activity. These dependant variables were:

 Retail and Recreation

• Grocery and Pharmacy
• Parks
• Transit Stations

The logarithm of these variables was taken to get a percentage change, rather than an absolute figure, as Google’s units of measurement were arbitrary (percentage change from a base in 2020).

Our independent variable, whose effect we want to discern on the above observables, was PM2.5 concentration levels. We divided the concentration levels into “bands”. These were 

• Band 0 (PM2.5 concentration <50 µg/m3)
• Band 1 (PM2.5 concentration 50-150 µg/m3)
• Band 2 (PM2.5 concentration 150-250 µg/m3)
• Band 3 (PM2.5 concentration 250+ µg/m3)

These were treated as categorical variables with the coefficient on each in our regression being interpreted as the percentage difference in that particular type of mobility as compared to a baseline – Band 0 where we have “good” air quality. A positive coefficient would mean that we see more mobility when air pollution is high. Vice-versa, a negative coefficient would mean we see less mobility when air pollution is high.

The year of observation was taken as a control variable to deal with yearly trends e.g. COVID hesitancy in late 2021 in the immediate aftermath of the pandemic. Thus, our coefficients can be interpreted without that particular bias in play.

Results

The results are summarised below in Table 1.

For Retail:

• The coefficient for PM2.5 bands is positive and statistically significant at a 99% confidence level in all but the highest band (which is statistically significant at a 90% confidence level). This suggests that an increased PM2.5 concentration is associated with an increase in people going to retail and recreational activities.

For Grocery and Pharmacy:

• Similar to Retail, the coefficient for PM2.5 bands is positive and statistically significant at a 99% confidence level for all but the highest band (which is statistically significant at a 95% confidence level). This indicates that an increase in PM2.5 bands is associated with an increase in people going to grocery and pharmacy shops.

For Parks:

• The coefficient for PM2.5 bands is statistically significant at a 90% confidence level in the first band, and 99% confidence level in the highest band. In fact, mobility to parks almost halves when pollution levels are at their highest. Thus, the model suggests that there is a significant reduction in people going to parks when pollution levels are very high.

For Transit:

• The coefficient for PM2.5 bands is positive and statistically significant at a 99% confidence level for just the first band. This implies that an increase in PM2.5 bands is associated with an increase in transit access for certain levels of pollution, and for higher levels, there is no discernible effect.

 Discussion

The most rudimentary takeaway from this exercise is that air pollution isn’t a significant factor in people’s mobility and activity. This could be for a variety of reasons. One explanation is that many of these tasks are probably essential (such as buying food, paying bills, going to offices etc), and thus are downwards sticky so regardless of circumstances people must do them (even when the air is hazardous). Another, is that smog season overlaps with times of the year where it is relatively more pleasant to engage in outdoor activities, compared to the blistering summer. Given that mobility to parks is negatively affected by air quality, and that is a non-essential activity, we believe the first explanation is more compelling.

The importance of these results comes in when we look at some of the solutions proposed by the Government. In the context of these findings, the “smart lockdown” is at best irrelevant, or at worst cruel. If the lockdown is actually implemented, it would be curtailing ordinary people from essential tasks that they clearly feel compelled to do even at the expense of their health. More likely, it will be ineffective and collapse once the Government realises that these tasks are necessary and enough outroar erupts. Neither outcome is a good one.

Clean air is a basic right for our health. Ultimately, gimmicks and simple-minded policies aren’t going to put a dent in the problem, rather they will probably create more problems than they solve. This is a multi-causal problem, and requires a multi-prong approach. This includes, but is not limited to, industrial regulations, congestion and emission charges (particularly on the most polluting vehicles), more public transit, fewer highways, underpasses and bridges and a movement away from car-centric cities and reckless industrial practices. Some of these solutions may be long-term, but many can be achieved in the short-term, and are likely to be far more effective than the gimmicks this Government has proposed.