Priced to the paisa, on a 1996 guess

The paise are the interesting part. They are the residue of a calculation, and a calculation has inputs.

The largest input is what the driver needs to live on. In Maharashtra's fare formula that is a single figure, set at 15,470 rupees a month in 2017. It was not worked out in 2017. It was carried forward from 1996 and adjusted for inflation.

So what happened in 1996?

The unions asked for 4,250 rupees a month and itemised the request. This is the only line-by-line breakdown of a Mumbai driver's living costs that anyone has put on paper, before or since.

Quoted in the Khatua Committee report, 2017, which records the award as "about 88% of the demand".

The committee's reasoning survives in one sentence:

That is the whole method. There is no family size in it. No food basket, no wage norm, no statement of whether the driver is the household's only earner. The figure was not derived. It was landed on.

The 1996 committee was chaired by P.M.A. Hakeem, then the state's Principal Secretary for Transport. Sixteen years later Maharashtra appointed a one-man committee to build a proper fare formula, and appointed Hakeem. He took his own 1996 number and indexed it.

In 2017 a fresh committee under B.C. Khatua reviewed the formula and declined to reopen the figure. It said why. The 1996 number had been "accepted by both the taxi drivers as well as consumer organizations". The case for keeping it was that nobody had objected, which is a claim about peace rather than accuracy.

Smaller things survive the same way. Until at least October 2012, the Mumbai minimum fare covered 1.6 kilometres, which is a mile. Hakeem's own terms of reference asked him to replace it with a metric kilometre. The tariff card issued that October still reads 1.60. By 2021 it reads 1.50. No order anyone can find made the change.

And the report that sets the fare is published without the chapter that sets the fare. The Khatua report runs to 224 pages. The copy on the transport department's website stops at page 148, mid-table. Chapter 7, which contains the auto rickshaw fare structure, is missing, along with the annexure comparing the new recommendations to the old.

So we rebuilt the model.

The method can be verified, because the taxi half of the report survives intact. Running the committee's own inputs through its own procedure reproduces its published answer exactly, to the paisa. So the reconstruction works, and the same machine can be run on the auto rickshaw inputs whose own pages are missing.

All inputs are the committee's own except repairs and maintenance, which is summed from its printed line items because the row carrying the total sits on a missing page. Fixed cost works out at 10.62 a kilometre, fuel adds 1.70 at 2017 CNG prices, giving a 2017 fare of 12.32. The minimum fare is set at 1.5 times that, which is 18.48. The minimum actually in force until 2021 was 18.

Run the same machine on current vehicle and fuel prices and it gives 18.19 rupees a kilometre against a notified 17.14.

That result is worth almost nothing, and it matters to say so plainly. We used the committee's structure, its kilometres, its working days, its shift weighting and its cost of living. The notified fare was produced the same way. We copied the state's method and found that the state agrees with itself.

Step outside the method and there is nothing to stand on.

Start with the household. The formula pays for a driver's family. It has never said how many people are in it. Not in 1996, not in 2012, not in 2017. Three people or five is the difference between a fare that covers a household and one that does not, and no document addresses it.

Then the kilometres. The formula assumes an auto earns money over 68 kilometres a shift. That figure came from a choice between two surveys. The state's own motor vehicles department checked 60 vehicles and reported 67. An independent survey of 500 drivers across eleven busy suburbs reported 94.8. The committee moved a fraction and stopped beside its own department.

Rebuilt on the committee's method with current vehicle and fuel prices, which are the least well sourced inputs here. The point is the gap, not either number.

Note which way this one cuts. Assume a driver covers fewer paying kilometres and the fare has to rise to recover the same annual costs. On this input the committee chose the figure that favours drivers. It remains a choice nobody has verified, and the answer moves a quarter with it.

An auditor asked whether 17.14 is the right price would have to say the question cannot be settled. Not because the sums are hard. Because two of the inputs have never been established.

One component can be checked against data the committee never used.

Take the shares from the 1996 request, since it is the only breakdown there is. Housing was one rupee in eight. Applied to the 2017 cost of living figure, the formula sets aside about 1,820 rupees a month for a driver's rent.

The following year the national sample survey measured what urban households in Maharashtra actually pay. Those that rent paid 4,569 rupees a month. Close to a third of the state's urban households rent rather than own.

So the formula's housing line runs at about 40 per cent of the measured figure. And that measurement covers all of urban Maharashtra, where most cities are cheaper than Mumbai.

The reason it fails here and not elsewhere is mechanical. Indexing a basket assumes its parts move together. Food, clothing and fuel roughly track a general price index, because that is largely what the index is built from. Mumbai housing has tracked nothing for thirty years. Apply one index to the whole basket and the component that moves fastest quietly falls behind the rest.

A fair objection at this point: if the total lands in roughly the right place, who cares how it divides?

Because the formula does not treat every cost the same way. Some costs have their own line and get re-priced at every revision. Others are buried inside the indexed lump and are never re-priced at all. Fuel is the first kind. Housing is the second. And over the same period they moved by almost exactly the same amount.

A driver was paid 1.66 rupees a kilometre for the fuel he burns and was not paid 1.62 rupees a kilometre for the room he sleeps in. Nothing decided that. No committee weighed housing against fuel. It follows entirely from which cost happened to get its own line in 1996 and which one did not.

This is also why it went unseen for thirty years. The gap does not arrive; it accumulates. Its size each year is roughly the share of a driver's costs that goes on rent, multiplied by the amount rent outruns general inflation. At a fifth of costs and three points of excess, that is about six-tenths of one per cent a year.

Nobody strikes over six-tenths of a per cent. Compounded across thirty years it is a fifth of a driver's income, and by then the number that caused it has been reaffirmed twice and is defended on the grounds that everybody accepted it.

A housing shock would not move a Mumbai auto fare by a paisa. A fuel shock moves it at the next revision. That asymmetry is the whole finding, and it is not really about autos.

Indexing works on one condition: that the shape of the basket stays put. A price index is an average of many prices. Multiplying a basket by that average keeps the basket honest only if its components move together. The moment one component runs away from the average, the index quietly underpays anyone whose basket is heavy in it, and overpays anyone whose basket is light.

In urban India the component that runs away is housing. So the question of who loses has an answer, and it is not random. A driver who inherited a room pays little rent and the formula treats him well. A driver who came to Mumbai for the work and rents by the month is the one the index has been shortchanging, by about six-tenths of a per cent a year, for thirty years. Indexation looks neutral. Applied to a basket fixed in 1996, it transfers income from the migrant to the settled, and it does so without anyone choosing it.

This is the same machinery that sets dearness allowance for government staff, the variable component of statutory minimum wages, and most Indian pensions. All of them price a basket once and index it thereafter. All of them inherit whatever composition happened to be true on the day the basket was written, along with whoever that composition happens to suit.

The fix is not complicated and nobody wants it. Take the basket apart every few years and price each line against what it actually costs, the way the formula already does for fuel and vehicles and tyres. Every one of those lines gets re-priced at each revision without difficulty. Only the driver's own cost of living is exempt, and it is exempt because it is the one line that would require somebody to state, in public, what they think a driver deserves.

There is one more thing the formula does not do, and it explains the strikes better than any of the arithmetic.

The formula does not pay a driver. It prices a vehicle.

The cost of living enters the calculation as 15,470 rupees, multiplied by twelve, and then by 1.7. The 1.7 is there because roughly 70 per cent of Mumbai autos run two shifts. So the fare is built to fund about one and seven-tenths people per auto per year.

How that money divides between those people is not in the formula. It is settled by a private arrangement between the person holding the permit and the person driving. Many Mumbai drivers do not own the auto they drive. They rent it by the shift, at a price the tariff card does not mention and no committee has examined. The last published figure for that rent is about 200 rupees a shift, and it is from 2012.

How many drivers rent rather than own is itself unknown. The number usually quoted is three in four. Follow it back and it comes from a consumer body's chairman, interviewed for a student research paper in 2012. Nobody has counted.

A fare can therefore be adequate on paper and reach nobody in particular. The formula funds a vehicle and its 1.7 notional drivers. The rental market decides who receives it. Those are two separate questions and only one of them has ever been asked.

Which is why the drivers who strike are not really disputing a calculation. There is no calculation to dispute. There is a figure from 1996, indexed for thirty years, applied to a household nobody has counted, over a distance nobody has measured, divided by a contract nobody regulates.

The last time anyone asked what a Mumbai auto driver needs to live on, the answer was 3,750 rupees a month, and the reasoning was that the committee considered it adequate.