DafT’s deeply regressive fantasy formula

Flicking through Google Reader, catching up, something caught my eye in George Monbiot’s latest:

Cost-benefit analysis is systematically rigged in favour of business. Take, for example, the decision-making process for transport infrastructure. The last government developed an appraisal method which almost guaranteed that new roads, railways and runways would be built, regardless of the damage they might do or the paltry benefits they might deliver(8). The method costs people’s time according to how much they earn, and uses this cost to create a value for the development. So, for example, it says the market price of an hour spent travelling in a taxi is £45, but the price of an hour spent travelling by bicycle is just £17, because cyclists tend to be poorer than taxi passengers(9).

I was vaguely aware that the government had complicated infrastructure cost-benefit formulae that included attempts to put value on people’s time, but I wasn’t aware that they had gone so far as to value the time of cyclists versus the time of taxi passengers.  So I followed the reference #9.  I’ve read some absurd documents in my time, but I wasn’t quite prepared for this.

When deciding whether a transport project — a road “improvement”, a high-speed railway, a bicycle path — is value for money, the Department for Transport consider the value of the time that users of the new infrastructure will save.  When deciding how much time to give each phase of the traffic lights, or whether a bicycle lane gets priority over road, transport agencies will consider the value of the time of the competing users.  In particular, DafT are interested in the value of the time that employers will save, because your time doesn’t matter, but the time you waste at work does.

How does DafT determine the value of an hour that an employee spends cycling for work, compared to one that the employee spends on a train or behind a steering wheel?  All it needs to know is the average hourly wage of employees using each mode.

It uses the 1999-2001 National Travel Survey.  Specifically, there is a dataset that counts all journeys made by mode and by income band, so we know whether the rich and the poor are more or less likely to drive, cycle, ride the bus, etc.  If a mode is over-represented amongst the rich and under-represented amongst the poor then DafT consider the time of users of that mode to be more valuable.

No, really.

If you can’t spot what’s wrong with this, and why I am cringing, I don’t know where to start.

Perhaps with the fact that we’re trying to derive the value of time “wasted” on travel-for-work from a dataset of travel modes that the rich and poor use outside of work?  A dataset, indeed, that includes students and the unemployed.

Or the fact that we’re specifying the average value of an hour of time to two decimal places, despite the vast range of values being averaged.

Or with the idea of specifying the value of one Great British hour, despite the massive variation in income and modal share between cities and regions?

Or the assumption that time spent travelling is always time wasted (my favourite office is a good long off-peak and under-crowded train ride with a netbook and an android), and that time “saved” by faster journeys is converted into economically productive time?  (Rather than, e.g., additional journeys.)

Maybe we should go back to the silly assumption that the work of the highest paid is the most economically important?

Or perhaps point out the critical fact that the demographics of users of a transport mode prior to investment do not necessarily reflect the demographics of users after investment, and that investment which enables modal-shift can “save” time too.

Or question what DafT are doing using inflation-adjusted figures derived from a decade-old version of the National Travel Survey when there is not one but eight more recent datasets?  In 1999-2001, the railways, the buses, and cycling were all at their very lowest ebb.  Were DafT to use the latest numbers, the value of an hour on a bicycle would be considerably higher.¹

The whole thing is absurd.  According to DafT, if I take a taxi to meet my client, there is more value in their giving me a faster taxi ride than in enabling me to get the meeting even quicker by bicycle.

This is cargo cult mathematics and cargo cult economics.  These numbers — given on the DafT website down to the exact pennies-per-hour — are pure fantasy.  I am actually embarrassed for the department that they are not only using these numbers, but are proudly publicising the pioneering way that they have been derived.  Were it not for the fact that transport policy and funding is such an unsexy topic, the press would be in gales of laughter at this nonsense.

And I would be gales in laughter were it not for the fact that, as I understand it, this crap is the foundation of a deeply regressive and damaging political programme.  When modelling the impact and benefits of investment in a transport intervention, the Department factors in this hypothetical “value of time” of users.

That is, because a taxi passenger is more likely to be very highly paid than a cyclist, when all other variables are equal the department should invest in schemes that favour taxis ahead of schemes that favour cyclists.  Aberdeen Cars is right: traffic light timings remind the economically-inactive cyclist that she does not matter.  Stabiliser can find some answers in this policy.

It becomes a positive feedback loop.  Invest in trains ahead of buses and the rich will use the trains while the poor ride the bus.

Imagine this happening at the Department for Health.²  Should we invest in cutting the waiting lists for lung cancer surgery or prostate cancer surgery?  Well, we value the time of the average prostate cancer patient more…

This is more than absurd.  This is a fraud.  This is a crude imitation of science and statistics being employed to disguise political decisions — to invest in transport for the rich and not for the poor — as pure objective economics.

1. Here  is the 2009 NTS.  The dataset we want is NTS0705.  Do the maths properly if you like, but even a glance at the journeys-by-mode-by-income table and it’s instantly obvious that cycling has flipped from being highly under-represented in the richer categories in 2001 to being very slightly over-represented in 2009.

2. I mean, imagine it happening this blatantly.  I would be surprised if there were not many many ways that state healthcare is subtly weighted in favour of the rich, whether designed and deliberate or not.

As usual, this is hastily bashed out heat-of-the-moment blog post, not a careful scholarly article.  The thesis I am certain of, but the details are always open to amusing malapropisms and embarrassing subtle errors in calculations.  If you are distracted by them, point them out and I will fix it.

24 thoughts on “DafT’s deeply regressive fantasy formula”

  1. I wrote a piece last week with one picture showing why I felt the London Conservative party didn’t ‘get’ cycling. It shows the hundreds of bikes at the office of one well-known City company. Thing is, this Dft stuff is comlpete rubbish in London at least. Have a look at this, Dft: http://cyclelondoncity.blogspot.com/2011/06/picture-that-shows-why-tories-dont.html

    It’s actually wealthy people that cycle too. So, even ignoring whether you should be taken to task about the blatant bias in your modelling, the fact is that your bias is based on completely the wrong assumptions

    1. This is all true, of course, and the decade old NTS clearly doesn’t reflect the cycling population of today, with the rise of the environmentally-aware middle-classes and the so-called MAMIL.

      But the numbers would remain skewed in favour of the car and train for the simple reason that rich people make way more journeys than poor people — 1.3 times as many journeys, and massively larger distances (slide 11 in Harry Rutter’s street talk, itself from the 2009 NTS, illustrates the latter nicely). So while the richest people cycle as much as the poorest people, they make twice as many car journeys and almost four times as many rail journeys.

  2. Seems like this would argue in favor of more and larger congestion charges. Did they somehow manage to dance around that conclusion without stepping in it?

    1. Oh, it came in nearly a decade ago — exactly when the then government were pursuing the idea of road use charging. Road charging is itself regressive — indeed, Ken Livingstone never misses an opportunity to remind us that he got the CCharge idea from Milton Friedman. Livingstone defended that on the grounds that the massive investment in tube and buses that came with it meant that the poor came out of the deal much better off anyway.

  3. As well as congestion charging, you could almost argue for Zil lanes from this standpoint. It’s like a drunken back of a beer mat calculation writ large. The assumptions alone tell you every you need to know about our transport priorities. And what about including all of the long, happy and productive QALYs those young people cycling and walking might be giving up for the benefit side of this equation?
    This particularly struck me: “Any impact on pedestrians, cyclists or others on changes in the amenity of the journey should be addressed under the journey ambience sub-objective.” It is true a serious traffic accident can really put a downer on the ambience.

  4. I find this staggering, the sheer utter incompetence of it! Yes, I know that economics is the dismal science for a reason, but this is farcical.

    The assumption that only poor people use bicycle to travel is totally unfounded. The research that has been done on cycle commuters shows that they are more likely to come from the higher economic groups, than the lower.

    Then we have the time of a car driver is worth more that a car passenger. What!?! This means that we should prioritise single occupancy cars over multi-occupancy to get the greatest economic benefit? Utter nonsense!!

    As for the highest value being placed on the Taxi/Minicab passenger…

    1. Well, presumably the formulae that these numbers are plugged into use the sum of the value of time of occupants rather than the average — so a car with a passenger is worth more than a car without (but two single-occupant cars is worth more than the same two people sharing one car), and a full bus is worth a handful of single-passenger taxis…

      Though, I wouldn’t put it past them…

  5. The biggest horror here seems to be the anti-democratic character of the model. It would appear, as it has been explained, that the time of a child walking or cycling to school is valued at zero, the time of the pensioner is valued, on average, low, and, of course, the time of women and racial minorities will also be valued below that of white men as well. Well, the child cannot in principle vote on this, but the others can, and I’m sure that if this large majority of the population knew about it, they would not vote for a politician who thought this was reasonable.

    Thanks for educating us.

    Vole O”Speed

  6. Minor point: we (I say we, we do it for a job because the DfT make us, not because I think it’s a good idea) do value your time if you’re not working, it’s just if it’s working time then it’s your time plus your employers time. Though you are entirely right to say that this misses the point with rail and this is one of the issues with the HS2 case as part of that assumes that the time you will save would have been a waste otherwise. Take that out and HS2 falls apart.

    My main point though, it’s worse than perhaps you think or have said. It isn’t just the value of time that’s the issue here as if cycling is faster (enough) then that comes out on top of cars. It’s the vehicle operating cost. Distance saving for bikes = worthless, as VOC for bike is 0. Save a km in a car and that’s a benefit and that’s going to get you some nice DfT money. Or would have when they had some.

    Other stupidity: If I get you out of your car and onto a bike , that’s a disbenefit due to the loss in fuel tax revenue. Environmental improvement or the improvement to your health, worthless.

    Those are off the top of my head, there’s more and it’s got progressively worse since they introduced WebTAG a few years ago. The only ray of light I can offer, you’ve got a new genation of engineers who also think it is unintelligble garbage and aren’t wedded to their cars and predict-and-provide road building. NATA (the predecessor to WebTAG) is now broadly discredited, it will just take the DfT a while to catch up and (one can but hope) kill WebTAG too!

  7. Actually there’s a bigger flaw than just valuing people by their journey options: it’s valuing any commute time as anything other leisure time. If you choose to live somewhere out of down and drive in to work, that’s as valid a decision as living 30s walk away from your office -but the time you spend in your car isn’t a cost to your company -it’s your decision. Why should we care about people who choose to live in the sticks and then drive in? Costing time of people using different transport options differently is not just “institutionalised motorism”, it’s daft economics to do for commutes -as they take place outside working hours, by their very nature

    1. I think you have hit on something there — if someone chooses to live so far from where they work, clearly, they do not place a high value on their time. This looks like a fundamental mistake in their methodology.

      The other, more serious, mistake, is that I think the sort of analysis they have done assumes independence of choices — my decision to ride my bike, is not affected one bit by all the other people choosing to commute in the largest possible vehicles they can afford. This is clearly Not True, and dependent choices can converge on suboptimal sticky points (locally utility-maximizing, not globally utility-maximizing).

      And, like other economic analyses, it assumes things like rationality, and perfect information. If, for example, all those people driving cars knew how much it was hurting their health to not get exercise (and also did not mistakenly believe that their health, like their driving skills, was above average), they would spend more time exercising. This can lead you pretty obviously to ride a bike to work, since you get spend a little time exercising every morning, and then arrive at work instantly after finishing your morning aerobics. There is no car that will do that for you. For a range of commute distances (3-12 miles), this delivers a given amount of exercise and transportation in less time than any possible alternative — and the longer life that results, is WORTH MORE to the higher income earners.

  8. Zil lanes – no way!

    We should instead have special lanes for Bromptons as these perform a valuable economic service of conveying the rider to higher value tube & rail services hence their rider value is £35 not £17. Why Brompton owners (who by definition have not winced at forking out £700) should have to slum it with the BSO underclass is a national scandal. As a first move ASLs should be exclusively reserved for Bromptons.

    You know it makes economic sense ;-)

  9. :-0 unbeleivable! I despair.
    Dr2chase, if we could get this message across (commuting+workout all in one) I reckon more people would go for it but as you say many over estimate their health aswell as their driving.

    1. I better jump in here before somebody else gets here and starts shouting “it’s the environment, stupid!” ;)

      The message that cycling is healthy and sedentary lifestyles are not healthy has been gotten across. People know these things, but it affects very few people’s behaviour. “Promoting” cycling and “encouraging” cycling, by whatever means, is not very effective, not very efficient, and affects a very limited demographic. We’ve tried it more than enough times to be confident of that.

      People know that cycling is cheap and simple and quick and healthy and clean and green. It is not a lack of reasons to cycle that is stopping them.

      The conditions on our streets are *the* overwhelming reason that people do not cycle.

      1. I think my argument is that if you choose to do an economic analysis, you need to review the economic setting. In order for market behavior to be welfare-maximizing, you need good (perfect!) information, rational actors, and independent utility functions. We don’t have independent utility functions — the driving population’s choice to drive, makes cycling unpleasant for others. This is a consequence of (ahem) “sharing the road”. Do what the Dutch do, and suddenly, your choice to ride a bike is unaffected by others’ choice to drive a car.

        So, the right place to apply DaFT’s analysis, is the Netherlands. Otherwise, DaFT could easily be reinforcing the stickiness of a local, suboptimal, welfare maximum. I realize that this is a fancy way of saying “the conditions on our streets are deplorable”, but economists are fond of thinking that they have Proved Their Results With Mathematics; as a one-time mathematician, I think it is perfectly worthwhile to point out that their castle of proof is built on foundations of sand.

        I also don’t think that the health message has necessarily sunk in. Actions speak louder than words, though it is possible that (1) people discount their future life; (2) sharing the road with cars is really that awful; (3) the driving status quo is heavily marketed; (4) it is really easy to get sold the wrong bicycle, at least here in the US, which reduces the convenience and/or safety of cycling.

      2. This basically goes back to what Ben Goldacre says: on the list of variables that affect people’s everyday decisions and lifestyle choices, distant theoretical health implications are near the bottom. It’s difficult to tell people to make healthy choices, you need to make the healthy options easy so that they become the natural choice.

  10. @dr2chase – that is actually a rather silly remark. People are priced out of city centres so move further way to get a house they can afford. The choice is then between a higher commuter fare or a higher mortgage service charge, insurance, property taxes, etc. It says nothing about how much they value their own time.

    Actually, my choice was to live on the edge of a large stretch of National Trust common land, which implied a long distance from London – I guess that means I placed a fairly low (economic) value on my travel time. If however I had chosen to live in Woking or Camberley instead of in Putney or Wimbledon that, I think, would not imply anything whatever about my travel time values.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: