I'm going to offer a partial defence of Michael Mandel against Scott Sumner. (Read Scott to see what Michael said and might have meant, but it doesn't really matter what he really meant, because it's an interesting question anyway). The question is this: would the invention of a lot of nifty new goods that people want to buy increase Aggregate Demand?
It seems to make a lot of sense. After all, if my firm invented a nifty new good, that would increase demand at my firm. But that doesn't mean that if all firms invented nifty new goods, that would increase demand at all firms. That would be a fallacy of composition. They might just be taking demand away from each other, leaving Aggregate Demand unchanged.
Start with the New Keynesian theory of the IS curve. People are trading off present vs future consumption. They choose that combination where:
MU(C1)/MU(C2) = (1+r)/(1+i)
MU(C1) means the marginal utility of consumption in period one, which is a decreasing function of consumption in period one. MU(C2) is the same for period two. i is the rate of time preference (degree of impatience), and r the rate of interest.
Let Nt be a measure of the niftiness of goods in period t. How would changes in N1 and expected N2 affect that equilibrium condition?
If gadgets became niftier, that would presumably increase the marginal utility of consuming an extra gadget, for any given number of gadgets. But if gadgets suddenly got niftier this year, you would expect gadgets to be at least as nifty next year too. If N1 and expected N2 increased by the same amount, this probably wouldn't affect the equilibrium condition at all, because both MU(C1) and MU(C2) would probably increase by the same percentage, leaving their ratio unchanged.
But if you expected gadgets to get much niftier next year than this year, you would postpone some consumption until next year. What matters is the expected growth rate of niftiness, n.
If U=log(C.N), where C is the number of gadgets and N their niftiness, then MU=N/C, and we can write the equilibrium condition as:
C1/C2 = (1+i)/[(1+r)(1+n)]
An increase in n, the expected growth rate of niftiness, for a given interest rate r, will reduce consumption demand today, as people save more because they are waiting for the niftier gadgets that will be in the stores next year. Another way to think of it is that an increase in n is like expected deflation, that also causes people to postpone consumption, because if the price per gadget stays the same, but the niftiness of gadgets increases, it's like a fall in the price of gadgets for a given degree of niftiness.
You could make exactly the same argument whether we are talking about nifty gadgets for consumption or nifty gadgets for investment.
Now let me make the best case possible for the argument that the invention of nifty new goods increases AD. I'm going to rig all the assumptions to make it work.
Suppose that all the nifty new goods hit the market in even-numbered years, and none hit the market in odd-numbered years. Just because.
Assume that nifty new goods don't take any more labour (or other inputs) to make than the old goods. It takes the same amount of labour to make one gadget; it's just a niftier gadget.
Assume prices are a constant mark-up over wage costs. That means monopolistically competitive firms maximise profits and the elasticity of demand for an individual firm's gadget is independent of niftiness, and is the same in odd and even years. (Dodgy assumption, that.)
Assume nominal wages are sticky and are the same in even and odd years.
These assumptions mean that actual and expected inflation are the same in even and odd years. (That's inflation unadjusted for niftiness; the rate of increase of price per gadget, which is not what Statistics Canada tries to measure, because it tries to adjust for niftiness.)
The IS curve will shift up in even years, when all the nifty new goods come out, and shift down in odd years, when they don't. (The natural rate of interest will be higher in even than in odd years).
If the central bank does something daft, like keeping the same nominal rate of interest in even and odd years, we will now get Aggregate Demand driven booms in even years and recessions in odd years. Because the expected growth rate of niftyness will be lower in even years (where it will be zero) than in odd years (where it will be positive).
But a sensible monetary policy wouldn't let this happen. And Scott would want a central bank that targets something sensible like NGDP, and not interest rates.
(Real Business Cycle theorists would also predict higher output and employment in even than in odd years. Workers will supply more labour in even years, making lots of nifty new hay while the sun shines and real interest rates are high so they consume less leisure due to intertemporal substitution effects. But RBC theorists don't like to talk about AD.)
[Mostly for fun. But any competent macroeconomist ought to be able to take a crack at questions like this, so it was a challenge. Plus, who knows, maybe there is something to it empirically.]