It must have been a good feeling when Scott Sumner saw Karl Smith’s blog post last Thursday announcing that he had proved that Scott was right in asserting that Simon Wren-Lewis had committed a logical blunder in his demonstration that Robert Lucas and John Cochrane made a logical blunder in denying, on the basis of Ricardian equivalence, that government spending to build a bridge would be stimulative. I don’t begrudge Scott such innocent pleasures, and I feel slightly guilty for depriving him of that good feeling, but, you know the old saying: a blogger’s gotta do what a blogger’s gotta do. For any new readers who haven’t been following this twisted tale of claim and counterclaim, charge and countercharge, response and rejoinder, see my three previous posts (here, here, and here, and the far from comprehensive array of links in them to other posts on the topic).
My main problem with Scott’s argument against Wren-Lewis was that, at a crucial stage in his argument, he relied on the national income accounts identity that savings equals investment. Now in the General Theory, Keynes himself also asserted that savings and investment were identically equal and made a rather strange argument that the identity between savings and investment had a deep economic significance because there had to be an economic mechanism operating to ensure the ultimate satisfaction of the identity. That was a nonsense statement by Keynes, as pointed out by Robertson, Haberler, Hawtrey, Lutz and others, because if two magnitudes are identically equal, there is no possible state of the world in which the two magnitudes would not be equal, so there obviously is no mechanism required (or possible) to ensure equality between the magnitudes. The equality is simply a consequence of how we have defined the terms we are using, not a statement about what can or cannot happen in the world. The nonsense statement by Keynes did not invalidate his theory, it merely meant that Keynes was confused about how to interpret his theory.
I cannot resist observing that this is just one example of many showing that the notion that the original intent of the Framers of the Constitution has any special authority in Constitutional interpretation and adjudication is totally wrong, based on the misconception that the original inventor, discoverer, or articulator of a concept has any power to control its meaning and interpretation. Keynes, let us posit, invented the income-expenditure theory. But his understanding of the savings-equals-investment equilibrium condition of the theory was obviously wrong and defective. The Framers of the Constitution may have invented or may have first articulated any number of concepts mentioned in the Constitution, e.g., the prohibition against cruel and unusual punishment, due process of law, the right to bear arms, equal protection. That they invented or articulated those terms first does not give the Framers ownership over the meaning of those terms in the sense that their understanding of the meaning of those terms cannot establish an immutable understanding of what the terms mean any more than Keynes could impose the notion that savings is identically equal to investment simply because he provided the first articulation of a model that hinged on the equality of savings and investment. Sorry for that digression, but I just couldn’t help myself.
Now back to Scott. Based on the presumed identity between savings and investment, Scott asserted that the reduction in savings by which households would seek to smooth their consumption in response to a temporary increase in taxes would necessarily imply a reduction in spending on capital goods (i.e., a reduction in investment). But savings and investment are not identical; their equality is a condition of equilibrium. If savings fall, there has to be an economic mechanism (perhaps, but not necessarily, the one posited by the Keynesian model) that restores equality between saving and investment. The equality cannot be established by invoking an identity between savings and investment that is purely conventional and is the result of a special definition that ensures the equality of savings and investment in every conceivable state of the world, a definition that drains the identity of any and all empirical content.
Here’s what Karl Smith had to say on the subject on his blog:
In a perfect world I’d lay out a concise logical proof that Simon Wren-Lewis and Paul Krugman are wrong. And number each point. They’d respond saying which of my points were wrong, and why. Then I’d reply. . . .
Perhaps I can help.
DY = DC + DS + DT = DC + DS + DG Λ DG > 0 Λ -DC < DT ==> DY > 0
Karl’s notation is a bit cryptic. This is how I understand it:
DY = change in Y (income)
DC = change in C (consumption)
DS = change in S (saving)
DT= change in T (taxes)
DG = change in G (government spending)
The first equation says that a change in income can be decomposed into a change in consumption plus a change in savings plus a change in tax payments. This is derived from the definition of income in the income-expenditure model, namely that income is disposed of either by spending it on consumption, paying taxes or saving it. There is nothing else (in the model) that one can do with his income.
The next equation simply makes the substitution of G for T, which in the example under consideration were assumed to change by equal amounts.
The symbol “Λ” means something like “and furthermore,” so that we are supposed to assume that DG > 0, i.e., that government spending has increased. Then we are given another assumption, -DC < DT, which means that, because of consumption smoothing, the temporary increase in taxes is not financed entirely by a reduction in consumption, but partly by a reduction in consumption and partly by a reduction in savings, so that the reduction in consumption is less than the increase in taxes. This is Karl’s rendition of Simon Wren-Lewis’s argument that a temporary increase in taxes to finance the construction of a bridge would imply an increase in Y because G will increase by more than C falls. Karl continues:
Which is false.
Proof by example:
Let DG = DT = 2, DC = -1, and DS = –1
Here Karl is saying let us assume that G and T both increase by 2. That part is fine. The problem is what comes next. He assumes that to finance the increase in taxes, consumption goes down by 1 and savings goes down by 1. Why is that a problem? Because he is reasoning in terms of an accounting identity rather than in terms of an economic model. Wren-Lewis was making an argument in terms of the implications of the income-expenditure model which consists of (yes!) definitions, causal or empirical functions (consumption, investment, etc.) and an equilibrium condition. The change in income cannot be derived from a simple definition, it is derived from the solution of the model. The model has a solution. You can solve for Y by taking the initial conditions and the empirical functions and applying the equilibrium condition. You can also express the equilibrium value of Y in a single equation as a reduced form in terms of all the parameters and initial conditions. If you want to solve for DY in terms of a change in one of the other initial conditions, like G and T or consumption function, you have to do so in terms of the reduced-form equation for Y, not in terms of the definition of Y. Doing that leads to the nonsense result that, I am sorry to say, Karl arrives at below.
Then both inequalities are satisfied and by the first equation.
DY = –1 –1 + 2 = 0
Which is what we were required to show.
It’s a nonsense result, because his solution does not correspond to the equilibrium condition of the model, which is either savings equals investment or expenditure equals income. In Karl’s nonsense result, savings is not equal to investment (because investment has not changed while savings has fallen by 1) and expenditure is not equal to income (because DC + DG + DI > DC + DS + DT). This is just the ABCs of comparative-statics analysis.
Now in a subsequent post, Karl seems to have retracted his “proof,” admitting:
S = –1 is not allowed [because investment has not changed].
Karl actually has interesting things to say about how to think about the effects of an increase in government spending and taxes in terms of a neo-classical analysis which is worth reading and thinking about. But the point is that to make any statement about the consequences of a change in the initial conditions or parameters of a model, one must reason in terms of the equilibrium solution of the model, not in terms of the definitions within the model, and certainly not in accounting identities that are completely separate from the model.
Finally, just one comment about Lucas and Cochrane. As Karl points out in his more recent post, Lucas and Cochrane offered reasons for rejecting the stimulative effect of building a bridge that were themselves couched in the very terms of the Keynesian income-expenditure model that they were criticizing. Thus, Lucas offered as his explanation for why building the bridge would have no stimulative effect that the increase in spending associated with building the bridge would be offset by a reduction in consumption associated with the taxes needed to finance the bridge as if that were an obvious internal contradiction within the model. Karl suggests a better response that Lucas and Cochrane might have given, but their response was simply an attempt to show that there was some gap in the logic of the model. That is why they invited such a brutal counter-attack from the Keynesians.
PPS Just to be clear, as Scott notes in a comment below, Noah did not mention Scott in his post.