Pakistan's Air: 3.9 Years, 128,000 Deaths, and an Accounting Left Unfinished

The first page of the report carried a single number: 3.9 years. Beneath it sat the sentence that gets repeated in every press note — that this is Pakistan's "first harmonised, multi-sector emissions inventory." Beneath that, with no publication date, four city names: Lahore, Karachi, Islamabad-Rawalpindi, Peshawar.

I went looking for the date first, then the methodology, then the one sentence that would say what baseline the 3.9-year estimate is measured against and how wide its margin is. None of the three were there. What was there: roughly 128,000 premature deaths a year, 6.5 percent of GDP, and in Lahore a life-expectancy loss of "up to 7 years."

The file had been labelled as sport. Inside there was not a single player, match or contract. That mismatch is not new to me. But a wrong label and a wrong question are two different things. A city whose annual average PM2.5 sits at 110 micrograms per cubic metre has a sporting calendar that is also a public-health document, whether or not anyone reads it that way.

The geography testifies against Lahore. The city sits in a basin; winter builds a temperature inversion, a warm lid over cold trapped air that does not lift from November to February. In those months Punjab's fields are burned, brick kilns run at capacity, and the diesel fleet idles in still air. Seasonal, and therefore easy to defer — "it's winter, it will pass." The number in the report is not seasonal. 110 is an annual average, which means the winter months when readings run into the hundreds are being averaged against the cleaner months. Bad days are being hidden inside good ones.

The WHO annual guideline is 5 micrograms per cubic metre. Divide 110 by 5 and you get 22 times. That is the figure that travelled. It also raises the first question: which station, which year, how many days? Monitor count and monitor placement can move an average either way, and the report is silent on both.

The report says it combined ground measurement, satellite data, fuel consumption, vehicle fleet counts and industrial records. That is the right direction — fuel and kiln counts are ground truths that satellite-only inventories lack. But naming a method and showing a method are different acts. This report does the first.

It covers four cities and four pollutants: PM2.5, carbon monoxide, nitrogen oxides, sulphur dioxide. Four cities are a sample, not a country, and the sample's limits are written nowhere.

The 22x and its denominator. Any ratio lives on its denominator. Here it is the WHO annual guideline of 5, tightened from 10 in 2026. Against the older guideline the ratio is not 22 but 11 times. Both numbers are true. The question is whether the report chose the new guideline because it is scientifically correct or because it doubles the ratio. Both answers are possible, which is the problem: without a stated reason, the reader cannot adjudicate.

The second question is the kind of average. Annual and 24-hour guidelines are different instruments — 5 and 15 — measuring chronic and acute risk respectively. Lahore is being hit both ways and the report supplies only the annual figure.

The third is the station. A city's "average" is the average of its monitors, and monitors go where there is power, security and institutional willingness. A monitor on a traffic corridor and a monitor in a government colony garden are two different countries. Silent on placement, the number is not weakened but left unfinished.

The fourth is what 110 actually measures — PM2.5 alone, or all particulates including natural dust. Punjab and Sindh carry a substantial dust load, and dust cannot be legislated out of the air. That matters directly to the "up to 50 percent reduction" claim that travelled alongside the report.

Unverified is not the same as wrong. It means the paper that would let the number stand up in a hearing was not attached.

How 3.9 years is built. Life-expectancy loss is not measured; it is constructed. Concentration-response functions translate PM2.5 levels into mortality risk, then age structure, cause distribution and exposure duration are layered on. Three inputs are assumptions. First, the baseline: "3.9 fewer" compared to what — the counterfactual of WHO-compliant air, or the previous decade? If the former, the claim means a Pakistani would gain 3.9 years in 5-microgram air. That is valid and powerful. If readers are not told, they assume the loss happened recently. Second, exposure duration — forty years of 110 produces a far larger loss than ten. Third, the cause list. Which diseases are counted changes the answer, and the report does not say.

Without those three, 3.9 and 7 are valuable signals, not settled tools. The likeliest point of a range whose range is missing. And the number printed at the front is the one that turns out most wrong later.

On Lahore's 7 years: a national figure of 3.9 and a city figure of 7 tells you the country is an average and Lahore is a case. Correct in direction. It also means that removing Lahore pulls the national number well below 3.9, which the reader was not told. When a national average is dragged upward by one city, the city is the story, not the average.

128,000 deaths: a numerator without its base. Pakistan's population is roughly 240 million. Divide and you get about 53 deaths per 100,000 people per year — a figure that lands somewhere among road crashes, TB and malaria. Which matters, because a disease without a line item is a disease without a budget line. The harder question is attribution: how many of these are PM2.5, and how much is all pollutants combined. The report lists four pollutants but does not say which one the mortality figure rests on. Nor whether household air pollution from cooking fuel is included. If it is, outdoor policy alone will not move the number; if it is not, the true toll is larger. Definition was left out, so the number will be debated and the definition will not.

6.5 percent of GDP is the hardest figure of all, because it requires pricing a life. Human-capital valuation counts forgone earnings; value-of-a-statistical-life valuation is far more generous and produces a far larger number. Health expenditure, lost shifts, children pulled out of school, unpaid care — each inclusion changes the total. And the loss lands on households, not the treasury, which means the people living in the most polluted districts absorb the largest share of a cost that no report itemises. That is the fastest inequality pollution produces, and it needs no number changed.

The eighty percent question. This is the report's best work and it got the fewest headlines. Lahore source apportionment: transport 35 percent, industry 28 percent, brick kilns 17 percent. Karachi: industry 49 percent. Add the Lahore three and you get 80 percent. Where is the other twenty? Plausible candidates — natural dust, crop residue burning, domestic fuel, waste burning, transboundary inflow — are all things a city government cannot control. The report charted eighty percent and left unnamed the twenty that carries the least political will. And the percentages do not specify whether they cover primary particles only. Much of PM2.5 is secondary, formed in the atmosphere from NOx and SO2, both of which the report measures. If those sit inside the missing twenty, then transport and industry matter more than 35 and 28 suggest, because their emissions show up as PM2.5 somewhere else. That is not carelessness so much as an international habit: strip the uncertainty so the inventory fits on a slide.

"Up to 50 percent." Every measure has three parts — cost, enforcement machinery, timeline — and the available material does not give them with the clarity the target requires. Retrofitting kilns needs money; tighter vehicle standards need inspection capacity; stopping waste burning needs local politics. And the question the report should have asked: who measures whether it worked? If the measuring agency is the enforcing agency, the result is pre-decided. Pakistan has issued short winter orders — school closures, kiln shutdowns — with no public document an outsider can audit.

On the field. At rest an adult moves six to ten litres of air a minute. At football intensity that rises ten to twentyfold. Same air, same road, same evening — many times the dose. Which yields a demand absent from any government document: the sporting calendar should be set by air, not by broadcast. A 6 a.m. session and a 7 p.m. match on the same winter day are different exposures. Watching matches for years, I have suspected that some of what commentary calls fatigue is air nobody measured. I am not claiming results are decided by pollution; I am claiming that if you measure a player's capacity and not the air in his lungs, you have measured half of it. And a stadium holds players and spectators in the same air, with the player at the highest dose for ninety minutes. No spectator can prove what that evening did to them, which makes it easy to postpone — postponement is not what happens to the player. I apply the same logic I apply to signing-on fees: the real cost of a thing is not what is paid at the transfer but what is paid every year afterwards. Air pollution is a signing-on fee the country pays annually, with no audit.

The contrarian angle. The easiest attack is that an NGO report is partisan. Convenient, therefore suspect. Had the state published its own harmonised inventory, this report would not be standing in for it. The question is not who wrote it; it is why anyone else had to. Conversely, its defenders forget that hiding uncertainty marks a weak claim, not a strong one. A report that writes its limits earns more trust, because the reader knows how far to walk with it. Third, both camps skip the shared airshed: Punjab is on both sides, the air is one, and there is no wall at the border in winter. No domestic measure is complete, and no report contains the transboundary share. Finally — is the report cherry-picking the worst numbers? Possibly. But before concluding that, one should find a competing public figure for the same cities. I did not. Where there is no contradiction, "bias" is only an allegation.

What to demand. Not another report. One page — methodology, with a date, a confidence range, and a name for the missing twenty percent. Until it arrives, anyone writing on this should state plainly that these numbers are strong enough to set an agenda and too weak to target policy. Any government will find it convenient to say uncertainty means waiting; that wait is the most expensive decision of the next decade. What exists, publish with its limits. What does not exist, demand in writing, publicly, with a date. The real question is not 3.9 years. It is why the first document to count more than a hundred thousand premature deaths a year in Pakistan was assembled outside the state, and why it carries no date.