Winner of the New Statesman SPERI Prize in Political Economy 2016

Thursday, 17 April 2014

Abusing economic analysis: UK Treasury edition

For US readers, this is about the misuse of dynamic scoring in analysing tax changes

Ask most people if they think a particular tax - like fuel duty - should be reduced, and they will say yes. If you ask people do you think income taxes should be raised to pay for a cut in fuel duty, you will get a rather different response. So just asking people if they would like one particular tax to be cut without saying how it will be paid for is pretty meaningless. Unless of course your aim is to provide ‘evidence’ that taxes are too high, and you are not too worried about the nature of that evidence.

There is a slightly more sophisticated version of this trick, and the UK Treasury have just played it. Each individual tax potentially distorts the pattern of economic activity. If that pattern without any taxes is near some ideal, then we can call taxes ‘distortionary’. If we taxed apples and used this money to subsidise the production of pears, people would eat too many pears and not enough apples. However there is one tax that is not distortionary, because it does not influence incentives and therefore this pattern of economic activity. It is a poll tax - a tax levied on each individual independent of their income, wealth or what they spend their money on. Economists call this a lump sum tax. So cutting any distortionary tax, and paying for this by raising a poll tax, is bound to produce beneficial results in terms of reducing distortions.

There is only one problem with paying for a particular tax cut by raising a lump sum tax - in the UK we do not have a poll tax. We did very briefly - it was not very popular, because people care about fairness as well as the distortionary impact of taxes. For this reason, you should not expect to find a government department like the Treasury modelling the benefits of cutting fuel duty by assuming it was paid for by raising a poll tax. Unfortunately, that is exactly what has been done in a Treasury/HMRC report released this week

George Osborne is not planning to reintroduce a poll tax - veneration of a past Conservative Prime Minister would not go that far. I think the argument the Treasury would use to justify what they have done is simplicity. If you pay for a cut in fuel duty by, say, raising income taxes, you have to model the impact of two taxes on economic behaviour rather than just one. I don’t think that is a very good excuse, but even if we think it has some validity it has a direct implication: an individual study of this kind is meaningless on its own. It can only be used in conjunction with other studies that look at the impact of raising other taxes. Will Treasury officials therefore stop their masters using the numbers from this exercise to justify cuts in fuel duty? No prizes for guessing the answer. (They might if they could but they don’t have that degree of influence.)

So what could have been the beginning of an intelligent discussion of the costs and benefits of particular taxes (as in the Mirrlees review, for example) has been turned into a simple propaganda exercise.

Unfortunately it gets worse. Fuel duty is particularly ‘distortionary’ because its rate is high (see Chart 2.1 of the Treasury paper). There might be a good reason for that. The tax could be high because it is trying to offset damage that is not prevented by the market: road congestion, pollution and of course climate change. In terms of the language of economics it is (at least in part) a Pigouvian tax designed to offset externalities. In that case the tax is not distortionary at all: a world without fuel tax would not be ideal, and imposing a fuel tax gets us nearer that ideal. As Chart 3.1 from the paper indicates, these beneficial impacts of fuel duty are not modelled by the Treasury’s CGE model. (This is why, as John McDermott notes, this kind of partial dynamic modelling tends to be attractive to right wing outfits. Is it significant that the paper does not actually include the words ‘climate change’, and just uses the vaguer term environmental damage?)

So what the Treasury have done is modelled all the benefits of cutting the tax, but ignored all the costs. If this was but one stage in a process that would subsequently look at the cost of these externalities, and would realistically model how these tax cuts were paid for, fine. As a stand alone exercise, I’m afraid the Treasury study is worthless.

As Chris Giles notes in an excellent report, this is really part of a political exercise to build the case for tax cuts. It has two unfortunate side effects. First, it just encourages the suspicion among many that anything coming out of the UK Treasury at the moment is worthless propaganda. Second, it encourages those on the left who think that mainstream economics is inherently biased. But if you saw an opinion poll that asked people if they thought a particular tax was too high, without also asking what tax they would increase to balance the books, you would not say that this shows opinion polls are inherently biased. Instead you would just conclude that the person commissioning the poll had a political agenda. You might also ask whether the polling company should have accepted the commission.


  1. Second, it encourages those on the left who think that mainstream economics is inherently biased.

    As a former mainstream economist in a government/institutional setting, I am coming more and more to that conclusion. Whether it's the resolute refusal to consider the distributive, as opposed to aggregate, impacts of economic actions or the use of micro-foundations to (implicitly) promote an individualistic, even atomised, view of societal economics, I am wondering if there is much hope.

    I am reminded of a series of articles in The Economist, I think in the 90's, which reported variousl studies showing that students of economics, especially graduate students, took systematically more self-oriented and less altruistic positions on a series of moral dilemmas. I denied any implications at the time, but now I can't help wondering.

    1. Here is something to encourage you not to give up hope:

    2. Tim Harford writes

      Economists may appear ethically impoverished on the question of co-operating in the prisoner’s dilemma but they seem to have a far more favourable attitude to immigration from poorer countries. To an economist, foreigners are people too.

      Very patronising, particularly to the people who are really feeling the consequences of globalisation - not journalists or tenured professors.

      I think people on the left have a problem with unfettered globalisation - including labour flows. Why? Because it breaks down communities and leads to a game of the survival of the fittest. Who benefits from globalisation? Large multinational firms certainly. Large flows of labour lead to rising congestion, high rents, low wages. Liberalisation of trade can push up prices of vital home produced crops in poorer countries. Small poor countries do not have the resources to defend themselves in the WTO. The opening of markets encourages automation further dampening wages. However, globalisation is good, and necessary, it just must not be implemented too quickly. It is the pace of the change and the dislocation such rapid change can bring. That is what most people I think believe, it is not because of a dislike of foreigners.

      Economists are not necessarily terribly wise in these matters. They need to talk to historians and anthropologists.

    3. Paul Krugman blog September 12, 2013 'Insurance Company With An Army Blogging', "As we need to remember now and then, the federal government is basically an insurance company with an army."

  2. Two further comments on the Treasury study.

    1. There are references to the Diamond-Mirrlees proposition that producer taxes are inefficient. But the fuel tax is a tax on final consumption as well as a tax on intermediate use, so Diamond-Mirrlees cannot justify the proposition that the fuel tax should be cut.

    2. As far as I can see without working through the details of the model, there's nothing in the nature of the dynamic CGE model that actually addresses tax inefficiency. The tax distortion is an assumed input into the model not a conclusion from it.

  3. «As a stand alone exercise, I’m afraid the Treasury study is worthless.»
    «The tax distortion is an assumed input into the model not a conclusion from it.»

    Conversely it is very useful to its sponsors as you write and it is one of the standard techniques used by clever academics in general and their sponsors.

    The technique is simple: if the conclusion is a given, the skill is in finding a celebrity and a set of assumptions under which it can be proven, and then announce with great emphasis that the conclusion has been proven by the celebrity. It won't matter that the assumptions under which it can be proven may be exceptionally narrow, or absurd, as those are technicalities. What matters is that a celebrity can state as a true fact that the conclusion has been proven, and indeed it is a true fact, even if worthless.

    I see that this technique has been at the core of Economics for a very long time, because the central truthiness of Economics, that the distribution of income determined by the markets is unique and optimal and solely based on marginal productivity, can be proven only starting from extremely narrow assumptions that have been designed for that purpose (and are also contradictory among themselves).

    Thanks to Arrow, Debreu an many others crafting carefully the right assumptions, many Economics celebrities can state with honesty that it is a proven fact that the optimal distribution of income depends solely on personal productivity.

    I think that finding assumptions under which the conclusions desired by wealthy sponsors can be proven, and used as such by celebrity endorsers, is a very valuable skill, that can seriously enrich an academic.

    I also think that finding and corroborating generally applicable, relevant, reliable conclusions can be even harder work and lead to difficult, poorly paid careers.

  4. «the skill is in finding a celebrity and a set of assumptions under which it can be proven,»

    One of my examples is Modigliani's equivalence theorem: I reckon that it should be more properly described as "Proof that someone clever can find a set of peculiar assumptions under which debt and equity funding are in some narrow sense equivalent".

    I think that for aspiring Economists on a path to success there are three levels of success in designing the assumptions that lead to conclusions desired by a sponsor:

    #1 Find assumptions that imply or are logically equivalent to the desired conclusion, but worded so that it is not obvious. Begging the question is however not that challenging, even if ski;fully done.

    #2 Find assumptions that are internally inconsistent, but don't look so without careful analysis. Requires higher skill than merely begging the question.

    #3 Find assumptions that are inconsistent *with the desired conclusion* but so subtly that it takes great skill to figure that out. This is the golden prize.

    Usually #2 and #3 require strong application of "ceteris paribus" lubricant, but that needs to be disguised, because some cynical spoilsports take the presence of "ceteris paribus" to be a sign of "artistic licence".

  5. Does anyone know who peer reviewed it?

  6. I don't see how poll taxes are non-distortionary and not liable to alter behaviour.
    If you replaced fuel duty with a flat rate levy on every citizen, surely this would encourage individuals to use their cars more and perhaps even buy a car when they currently didn't have one since they believe they 'need to get their monies worth'? What am I missing?

    1. Its the reduction in fuel tax that is doing this, not the poll tax. Suppose you increased income taxes instead. Then you would still get the effect you mention, but in addition you might discourage people from working because they paid higher tax.


Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time.