Specifying “Demand”: Nick Rowe Meets Steve Keen on His Own Ground

You might well ask: “Whaddaya mean by ‘his,’ buster?”

Nick does a full-faith effort here (including the comments) to characterize Steve Keen’s position (aggregate demand = GDP + change in debt), using Nick’s preferred language and mental modeling. It’s a darned good effort, but I think it’s crippled (as is Steve’s construct) by a conceptual failing about the nature(s) of “demand.”

The problem is perhaps best revealed here:

Aggregate actual nominal income equals aggregate expected nominal income plus amount of new money created by the banking system minus increase in the stock of money demanded.

Nothing in the above violates any national income accounting identity.

The last statement is neither right nor wrong, because “demand” is not an accounting measure. You’ll never find “demand” anywhere in the national accounts, in balance sheets, income statements, or flows of funds. Demand is a (potentially) useful economic concept and construct.

In its general form, demand is conceived as a curve, not an amount. It describes what people, at a given moment, would spend over an ensuing period, at various price points. (You can’t include a curve in an accounting identity.)

But if you assume a price point — say, the price point that exists at that given moment — you can specify demand at that moment as a number, an amount, a point on the curve: how much people would spend over the ensuing period at that price point, if nothing changed and supply was unconstrained. This works, for instance, if you assume that that moment’s price point will pertain over the ensuing period — not crazy for short periods. You can say “this is how much people, at this moment and this price point, want to spend over the ensuing period.”

That numerical amount — demand at that moment — could say something useful about the state of the economy at that moment (especially in the context of other measures).

Demand in the textbook understanding is always demand at a moment. It’s an “instantaneous flow.” (Google that term to to see how flow-over-time measures for water, electricity, etc. that encompass or surround a moment can be used to estimate/derive such an instantaneous measure, and what formulas can be used to do so.)

Aside: Nick says in the comments that demand (or at least “”money demanded”) is a stock, and change in demand is a flow. I think he’s conceptualizing it wrong — the stock of demand?? — but he’s intuiting what I’m thinking: “demand” is like a stock measure because it describes a moment.

So here’s the question: “What was demand (for widgets, or aggregate demand) on July 31, 2011?” Gimme a number. How has that number changed over time? Graph it for me.

Have you ever seen a fever chart of aggregate demand over the decades? Would be darned interesting, no?

So I’m kind of amazed that economists aren’t, haven’t been, all over the problem of defining a formula to specify a measure of aggregate demand for an economy at a given moment. Steve Keen’s trying to do that. So is Ed Lambert.

You need a formula that draws on now-available, post-hoc accounting measures to derive an estimate of this economic measure for that moment. What accounting measures, and what formula combining those measures, deliver the most useful (accurate?) estimate of that moment’s “demand”? (The measure’s usefulness will ultimately depend on the larger model(s) in which it is employed, but we can begin by thinking in more general terms.)

Simple accounting measures don’t work. GDP over the ensuing period doesn’t do it, for instance. That’s “quantity actually supplied/bought/sold” during the period, not quantity demanded over the period, or demand at the beginning of the period. Supply constraints, price changes, etc. could (almost certainly do) mean that those numbers are quite different.

So what could work? There are an infinite number of possible formulas to estimate this measure, employing an infinite number of accounting measures. The most useful measures might be stock measures (describing the “demand moment” we’re examining), or flow measures (describing a period or periods preceding, succeeding, and/or encompassing that moment), or some combination of the two. It would be great if we could come up with a formula that relies on measures antecedent to the “demand moment” we’re estimating, because then we could estimate current “demand” in semi-real-time (subject to delays in measurement and reporting).

So what about Steve’s formula — GDP plus change in debt? I find it problematic because he seems to be specifying demand for a period, not a moment. Saying “demand for (the period) 2011 was GDP plus change in debt in 2011” is not very useful; it simply restates existing accounting measures for a period using a different word (“demand”). I want to know: what was demand on January 1, 2011? (If we’re using ensuing-period — say, 12-month — accounting measures to make the estimate, we may need to be precise in describing our measure: something like “ensuing-12-month-derived demand on January 1 was…”)

Also — assuming in my construct that Steve is deriving today’s “demand” from ensuing-twelve-month GDP and change in debt (I don’t think he’s actually doing that) — we don’t know what GDP and change in debt will be over the next twelve months. So the measure gives us no idea of what demand is today.

(It seems quite possibly or even likely to me, though, that  useful measures of “instantaneous demand”  will incorporate some debt/lending measures. Intuitively: when people borrow more they spend more, increasing the demand that producers face.)

This is all why I’m rather taken with Ed Lambert’s work. He’s given us a formula, based on most-recent accounting measures (Real GDP, Labor Share of Income, Capacity Utilization, and the Unemployment Rate), to calculate a measure he calls “effective” demand, at a given moment, i.e. now or any point in the reported past. And he graphs that measure over time relative to other measures.

Is it, will it, be a useful measure — allowing prediction or at least coherent understanding? That remains to be seen. But I’d sure like to see other economists developing competing measures of demand-at-a-given-moment, and accompanying models that make predictions based on those measures. It could result in some healthy Darwinian natural selection in the field.

Cross-posted at Asymptosis.