"Microfounded" and Useful Models

To have fake micro foundations for your model is not a feature, but a bug.

Mark Thoma, last March:

Economist's View: "Microfounded and other Useful Models": More on today's apparent theme, at least for the moment, economic methodology. This is from Simon Wren-Lewis:

Microfounded and other Useful Models.... This title harks back to... Blanchard and Fischer’s Lectures on Macroeconomics. That textbook was largely in the mould of modern microfounded macroeconomics, but chapter 10 was not, and it was entitled ‘Some Useful Models’. One of their useful models is IS-LM…. [H]ow can something more ad hoc be more useful?… Can Krugman’s claim that they can be more useful than microfounded models ever be true?…

Let me pick on my friends here:

Down the hall David Romer, the latest edition of whose truly excellent Advanced Macroeconomics textbook contains this:

Do we really want to commit ourselves ex ante for reasons of pure theory to a model in which:

1. each 1% change in expected real GDP next year induces a corresponding 1% change in real GDP this year;

2. each 1% point change in the one-year-ahead real interest rate induces a corresponding change in the opposite direction equal to the inverse of the average and representative household's coefficient of relative risk aversion

3. the trend growth of output is equal to the average and representative household's pure rate of time discount divided by its coefficient of relative risk aversion;

and then use that model to try to analyze business cycles?

The answer is: NO!!

David Romer throws (3) over the side immediately: "we suppress the constant term".

(1) is embarrassing, in that it means that when (6.8) is iterated forward it forces aggregate demand to depend only on the real consol interest rate (with shorter-term interest rates affecting aggregate demand only to the extent that they affect the console rate).

(1) and (2) together impose a rigid proportionality--depending on the average and representative household's coefficient of relative risk aversion--between expected business-cycle movements in output and in the consol rate, which David Romer immediately throws over the side as well by bringing in (unmodeled) fluctuations in business investment and net exports.

The bottom line? We have a new Keynesian IS curve that is "in contrast to the traditional IS curve… derived from microeconomic foundations". But we don't believe what the NKIS curve's derivation tells us. We don't believe what it tells us about (a) which interest rate aggregate demand depends on, (b) the quantitative relationship between expected future output and current output, (c) the relationship between the responsiveness of aggregate demand and the average representative household's coefficient of risk aversion, or (d) the dependence of the trend growth rate of the economy on the average representative household's pure rate of time preference and coefficient of relative risk aversion.

What do we keep from the derivation? Two things. First, that it "implies an inverse relationship between" the real interest rate (not the real consol rate, mind you) and aggregate demand; and, second, that for each Old Keynesian conclusion there is a set of (almost surely wrong) microfoundations that gets you into the neighborhood of that conclusion

Why keep this implication of the microfoundations while rejecting the others? The book is silent.

Why is there an inverse relationship between the real interest rate and aggregate demand? I have asked a number of first-year graduate students taking Econ 202b here at Berkeley. The answers I get tend to be:

Because if the average representative household is following an optimal consumption plan over time, it must plan for its marginal utility of consumption to fall at a pace. equal to the difference between the real interest rate and the pure rate of time discount, which means it must plan for consumption to rise by an amount proportional to the difference between the real interest rate and the pure rate of time discount. Thus when the interest rate is high, the average representative household must be planning to raise its consumption rapidly, which means its spending must be low.

And then, if I am mean, I ask: "So if the interest rate goes up, people think that they are making too much utility this period and that their marginal utility of consumption is too low, and so they decide to make themselves poorer and less happy today by not working?". Then the students seem confused.

The right way to teach it, I think, is to say:

1. that production is demand determined,
2. that production is equal to income,
3. that households today plan to take their income and spend it on (a) consumption goods and (b) adding to their stock of financial assets,
4. that planned financial asset accumulations are a positive function of both the real interest rate r and the level of income Y,
5. that if the interest rate r goes up everybody tries to cut back on their consumption spending in order to boost their holdings of financial assets--and so production and incomes fall until Y is low enough that there is no desired net accumulation of extra financial assets which are not there to be accumulated.

It is, I think, powerful when presented as a disequilibrium process that drives the economy toward S-I=0. That creates a downward-sloping relationship between r and Y anywhere and anywhen that economy-wide net financial asset accumulation depends positively on Y and r.

It is, I think much less powerful when presented as the mysterious dictates of necessity: that Y and C "must" be equal and that the Euler equation "must" be satisfied.

But in what I regard as the relatively weak and confusing way to teach it, it has microfoundations--microfoundations that happen not to describe the world we live in terribly well, but microfoundations…