Wednesday, 31 August 2011

The lies we tell

Former Canterbury Professor, David Giles, now at Victoria University in Canada, recently posted on his blog about the pitfalls about teaching something you know to be wrong at a lower level requiring the students to unlearn it at higher levels. He then presented an example from econometrics courses regarding the necessity of the assumption of normally distributed errors for certain results in the standard linear regression model.
Being a microeconomist not an econometrician, this set me to wondering what are the common lies we tell in lower level micro. By this, I don’t mean statements about the world of which someone might contest the empirical validity (e.g. minimum wages cause unemployment), or even the assumptions we make or, more often, are accused of making but never do (e.g. “people are rational”). Instead, I am referring to things that are unambiguously conceptually wrong.

I can think of three common ECON 100 microeconomic lies:
  1. Diminishing marginal utility: We have known since Pareto and Hicks that this concept is meaningless, and neither necessary nor sufficient to show either downward-sloping demand or even diminishing marginal rates of substitution, but it seems to be
  2. Consumer surplus: The answer to no known economic question, this is typically taught at first year not as a useful approximation but the actual value of the difference between the willingness of consumers to pay and the amount they actually had to pay.
  3. Deadweight losses of taxes constructed by comparing total surplus with and without a tax: The deadweight loss triangle corresponds quite nicely to the idea of Pareto inefficiency (at least in contexts where partial equilibrium analysis is reasonable), but the ECON 100 derivation that comes from adding up total consumer plus producer plus government surplus rests on no welfare foundation that I know of. It is not Kaldor-Hicks, since government revenue is considered as a separate entity to the surpluses of individuals.
Any others that anyone can think of? How about the lies we tell in macro?

Note: Please don’t get me wrong; what I am calling lies here are not attempts to pull the wool over students’ eyes or sell a political viewpoint. I don’t think there is ever any justification for diminishing marginal utility, but consumer surplus and deadweight losses are useful concepts to teach that give appropriate conclusions if used in the right contexts. It certainly is better for students to have to unlearn these concepts slightly, than to be hit with duality theory and general equilibrium in their first year.



  1. I am surprised about your distaste for diminishing marginal utility.

    Personally, I think its an easier concept for people for wrap their heads around than trying to get them to go the whole hog and think about income and substitution at the very beginning.

    On top of that it is a relatively "realistic" assumption, which does help to reinforce one of the principle "laws" of economics.

    As a result, I was wondering if you could fill me in regarding why you dislike it to such a degree?

  2. My textbook when doing economics was written by Susan St John, I submit everything after the first chapter on supply and demand graphs should be in this category.

  3. Seamus:
    You sure that Consumer's surplus is the answer to no known economic question? "How much would you pay me not to ban you from buying apples?"

    I've never thoroughly read Mike Burns on this question:

    And what's wrong with separating out the costs of taxation from the benefits of government spending, then adding them together afterwards?

  4. @Matt:

    Diminishing marginal utility is easier than doing income and substitution effects (which I would never do at 100 level), but for what gain? The concept itself has no operational meaning that can be described in terms of observable behaviour--it has to be reasoned by introspection. You can use it to derive (said with sarcastic inverted commas) downward-sloping demand, but if you do, you are assuming separable utility, and ruling out all kinds of reasonable behaviour, such as that the MRS between gas and electricity can't depend on the number of electrical appliances owned! If you want to get to downward-sloping demand without income and substitution effects, why not just ask students to reason from introspection directly, rather than asking them to reason diminishing marginal utility from introspection and plus some clearly wrong assumptions.

    I see no benefit here, but a large cost. Namely, while it is easy to unlearn consumer surplus and replace it with compensating or equivalent variations in higher courses, it is far less easy to throw away the implicit utilitarianism of diminishing marginal utility and face up to the fact that one's views on equity and redistribution have to be reasoned from well-articulated value judgements rather than derived from some careless notion of cardinal utility.

  5. @ Nick

    "How much would you pay me not to ban you from buying apples?" is a well-defined economic question with a well-defined answer. But the answer is called "equivalent variation" not "consumer surplus", and it is the area above the price to the left of the Hicksian demand curve that cuts the Marhsallian demand curve at the current price.

    Of course, we can always call the EV "Hicksion consumer surplus" as some authors do, and then that "consumer surplus" is well defined, but that is not the consumer surplus we teach our students to shade.

  6. @ Nick

    And finally, "what's wrong with separating out the costs of taxation from the benefits of government spending, then adding them together afterwards?"

    There is nothing wrong with this in principle, but who is to say that the government revenue from a tax equates to the dollar value of the benefits of the expenditure resulting from that tax. Imagine that you put on a sales tax that raised $1m to purchase $1m of public goods, and that consumers had a collective willingness to pay for the extra public goods was $2m. Adding up costs and benefits in this way would suggest a negative deadweight loss.

    But the deadweight loss does represent something real, the Pareto inefficiency that comes from counting backwards from the point where S=D to put a monetary value on the mutually beneficial trades that don't occur as a result of the tax distortion. As I said, there is nothing wrong with the concept of a deadweight loss--it is just the method of getting there by summing total surpluses that I take issue with.

  7. "Diminishing marginal utility: We have known since Pareto and Hicks that this concept is meaningless"

    And we have known since Von Neumann and Morgenstern that it is not meaningless, since Von Neumann utility is cardinal. Diminishing marginal utility is a (usually true) claim about what gambles people prefer.

  8. @ David

    I would put it this way. With the usual ordering axioms, we can represent choices over lotteries with an ordinal utility function, and the implied preference ordering is invariant to any monotonic transformation. With the addition of the independence axiom we can *represent* these preferences with a von-Neumann-Morgenstern utility function that as the expected utility property. With that representation, diminishing marginal utility has implications for observable behaviour (namely, risk aversion). But the observation of risk aversion in behaviour tells us nothing about how we should represent preferences over certain bundles, and certainly doesn't present a useful way of explaining demand curves.

  9. @ Seamus

    Very interesting. I just never felt that the concept really "hurt" people at the first year level - as they are still just getting their heads around the idea that trade creates value. As a result, the concepts are viciously partial eqm.

    In one sense I always felt that the learning process in economics was one of peeling back assumptions - so moving from diminishing marginal utility to actually moving on and thinking about more than one good, the related preference relations, and income and substitution effects, felt like a natural progression. I guess the fact that DMU isn't sufficient violates this view though - and I hadn't really thought of that.


  10. Seamus: OK. EV and CV. But I think those are the least of our worries. The "Hicksian" demand curve is holding U constant, IIRC? As opposed to the Marshallian holding Y constant? Neither U nor Y would be held constant if we messed with P. Only the Clowerian demand curve makes sense!

  11. Nick: Yes, the Hicksian demand curve holds U constant, but the area below it and above price still gives the answer to your question: How much would you pay to stop me banning apples.

    As for the Clower demand curve, this exposes my lack of knowledge of macro (please don't rat on me to my former colleagues at the Bank of Canada). I take it that it is a GE demand curve when some exogenous variable shakes an entire system leading p to change?

  12. Seamus; Actually, I just made that up, sort of. But it's a demand function where quantity of apples demanded depends on the price vector, but takes into account the quantity constraints agents will face in buying and/or selling other goods, if the price vector is not at market-clearing.

    For example: start at the competitive equilibrium price vector. What happens to the quantity of apples demanded if we arbitrarily raise the price of apples, holding all other prices constant? The normal answer: it falls, because people will want to substitute away from apples towards pears.

    But that answer is total rubbish. Because if we hold the price of pears constant (by assumption), and if the quantity of pears supplied stays the same (suppose), apple consumers who *want* to buy fewer apples and more pears won't actually be able to buy any more pears.

    So they are just as likely to say "Sod it! I wanted to buy fewer apples and more pears, but since I can't actually buy more pears (there's a shortage) I'm just gonna have to keep on buying apples."

    That's the big lie we tell in all of micro, both partial and general equilibrium. The difference between EV and CV is peanuts in comparison.

  13. Nick: I am struggling to see how this is a lie we tell in micro. In a world with quantity rationing rather than price rationing in other markets, there is still a well-defined demand curve for a particular market, and it will typically still be downward-sloping due to income effects. We can still do our partial equilibrium analysis with the demand and supply curve in the remaining market, including putting in quantity rationing with legislated prices and the like.

    Or are you worried about the fact that if you have quantity rationing in all but one market (apples), the partial-equilibrium approximation that one can ignore the feedback effects on consumer income from changes that affect price. That is not a lie we tell. In fact, it is the very reason we say that we can't stop at S&D and have to do GE analysis--either GE with price rationing (advanced micro), or GE with quantity rationing (new Keynesian macro).