January 31st, 2012

Not sure if anybody will read to the end of this, but here goes.

Brad deLong recently (my italics):

It turns out in economics to be remarkably hard for lots of people to distinguish between:

  • behavioral relationships–things that tell you how people will change their behavior to respond to changes in the economic environment and economic policy;
  • equilibrium conditions–things that tell you what configurations of the economic environment are consistent and are not rapidly-changing out-of-equilibrium phenomena seen for an eyeblink of time, if that long;
  • and accounting identities–things true by the metaphysical necessity of the definitions that are devoid of interesting substantive implications.

Hum. The doctrine of instantaneous equilibrium looks suspiciously like an analytical convenience that has got metaphysical notions above its station. It seems remote from real life. Getting information and taking decisions take time and effort. We run out of time before we optimise and then things change.

Just for fun, let’s build a toy qualitative model that reflects these facts, and see where it takes us. (Probably unoriginal, but I’m far too effete and lazy to plagiarise.)

Léon Walras created general equilibrium as a metaphor for capitalism with his bucolic dream of a village operating on pure barter.
(Never mind that the popular economists’ story of money being adopted as a convenience in barter market economies has been shown to be quite wrong. Money was invented as a unit of account by urban temple bureaucracies and long-distance traders, not as a means of exchange by imaginary village barterers.)

Walras’ image is of a peaceful Swiss Bronze Age mountain village, a picture postcard without the chocolate. On market day all the farmers and shepherds bring their produce in kind: sacks of grain, baskets of parsnips, jugs of milk, goats and chickens. Each seller wants to go home with a different bundle of goods than he (it’s presumably a patriarchy) arrives with. The relative prices are established by multilateral haggling, in the famous process of tâtonnement – groping. Eventually everybody is Pareto-happy and the prices and quantities are frozen. Bingo, general equilibrium. In the absence of money and money prices, it’s general equilibrium or chaos. Also, it’s unclear whether the individuals are subject to a hard budget constraint in their initial haggling. If they are, how does the tâtonnement get started? But it applies fully at the end.

We’ll fast-forward 3000 years from Walras-la-Vallée to Keyneston, 1935. Keyneston is an autarkic city-state, with factories run by industrialists, shops, wage-earning workers, a government, bonds, and money. The week is divided into three periods. Monday morning is market time, when contracts are negotiated to fix all the activities carried out in the second period from Monday noon to Saturday. On Sunday everybody reviews how things went in the week, taking account of information about others, and draws up plans for the Monday trading session. These plans are internally consistent, but not as a rule compatible.

Now here’s the crucial tweak. There is imperfect information and high search costs for everything, and a strict time limit to trading. The initial offers of participants are not random but good guesses based on previous experience – what happens is strongly path-dependent. The bell rings at noon, trading stops, and all participants must abide by their last offers, whether to buy or sell. Since there is money, the hard budget constraint on each is the income they expect to receive under the last offer for their labour or goods for sale, plus their holding of money. Money acts as a short-term cushion against trading uncertainty.

In this universe, everybody starts the week with a portfolio of contracts to buy and sell, and an invisible portfolio of regrets about what went wrong. Some goods lie unsold on the shelves; some workers are unemployed; some investment projects don’t get carried out; and everybody has a money balance rather different from what they had expected – in the case of of the newly unemployed or employed, a lot different. It’s a thoroughly disequilibrium world, for ever, as it never catches up with its shocks.

It’s also very dynamic. The portfolios of regrets expand during the week as the players learn more about the deals struck by others. On Sunday they review the week and draw up revised strategies and opening bids for Monday. Some of the fluctuations cancel out, and are just noise around an unchanging mean, for example on relative prices. Others reflect collective net shifts in sentiment. In this way we can reintroduce the standard behavioural hypotheses of textbook equilibrium macroeconomics: Keynes’ consumption function, propensity to invest, and liquidity trap; the quantity theory of money; Prescott’s technological shocks.

As an example, suppose an exogenous increase in the propensity to consume. In period 1, consumers don’t succeed in buying everything they want, but they do increase their spending and run down their money balances. Shopkeepers and industrialists have more sales than expected, excess money balances, and forced disinvestment in stocks. These enter the period 2 trading with plans to hire more workers and restore stocks to normal, plans which again will be partly fulfilled. In this way the shift in the consumption function drives an increase in economic activity, just as in the equilibrium Keynesian model.

I make no claims that this world, more realistic than the equilibrium models, offers advantages for analysis. You can make trade cycles work within it, but it’s hard to discuss the baseline level of output. But here’s the thing, In a world when no individual is ever in subjective equilibrium except by a fluke, still less entire markets or the whole economy, the accounting identities still hold in full as long as there any transactions at all: including Say’s Identity (total supply = total demand), and Keynes’ Lemma (savings = investment).

The impact of a bond-financed government stimulus cannot possibly be inferred from these identities. For this, you have to look at the behavioural responses embedded in the regret portfolios and revised bidding strategies. It’s very plausible, via Keynes’ propensity to consume, that there will be a positive multiplier in succeeding cycles. It’s plausible too that this will be counteracted to some extent by crowding-out of private investment as interest rates rise; though incompletely so, unless the crowding-out works so much faster than the consumption effect that the latter never gets a purchase on behaviour. (The speed and resiliency of bebavioural responses have become important variables.) These are empirical hypotheses that can be verified or disproved. SFIK, the consumption effect seems to be quite strong, and the crowding-out negligible. Please correct me.

The doctrine of Ricardian equivalence is another empirical hypothesis to add to these. It’s a shame that the author of towering, revolutionary texts on rent and trade should have his name only immortalised by his successors through this footling speculation: I’d rather call it “the Lucas-Trichet fallacy”, but I suppose the great man is stuck with it.

I’m not sure if I get this right, but the idea seems to be that when the government expands its purchases financed with bonds, all taxpayers, without exception (1), immediately learn of this sneaky plan (2), accurately calculate their increased tax liability to pay the bond interest to the indefinite future (3), and reduce their consumption by an amount equal to the discounted present value of this tax liability (4), using the same discount rate as the bond interest (5).

Stated this way, the full equivalence doctrine has as many holes as it has assumptions. None of them are remotely plausible. Take 2: recall candidate Obama’s exchange with Republican hero Joe the Plumber. It turned out that Joe did not grasp the distinction between his marginal and his average tax rate, so he would not be capable of the fancy long-range calculations of tax liability.

Some of course can: let’s take some more sophisticated players and ask how likely is behaviour (4). Taxpayers Krugman and deLong are on record as denying Ricardian equivalence – as Keynesians they expect a stimulus to lead at a permanent increase in output, so it will be partly self-financing. Inflation hawks in central banks predict that the accompanying monetary expansion will erode the real value of nominal wealth. If you expect the real rate of interest on bonds to turn negative, why sacrifice present consumption to build up your holdings of them? Either group or both may be wrong, but their behaviour for sure won’t be “Ricardian” today. So even if taxpayers Lucas and Fama are cutting their spending in line with their beliefs, the effect must be partial.

The consumption of the very rich Romney household is not public and he doesn’t blog his theory, but we do know that the young Mitt financed his studies by selling inherited stock. That is, he sensibly used wealth as a cushion to maintain a satisfactory level of consumption; not, as full-throttle Ricardian equivalence requires, maintaining capital at the expense of consumption.

Come to think of it, this alleged absolute priority for maintaining capital makes no sense. What’s the point of being rich if you are never prepared to spend it in the face of shocks? A classier argument can be drawn from the writings of Saint Milton: to spend is to tax. Under progressive taxation, the permanent income of the better off is lowered by the stimulus anyway, however the tax is spread in time, or disguised as inflation. So their equilibrium wealth is down too. They won’t therefore try to restore their starting level of wealth in full, let alone immediately.

Complete Ricardian equivalence is IMHO a non-starter: you have to believe too many highly improbable things before breakfast. At most there is a possible Ricardian effect that reduces the consumption multiplier from a government deficit-financed stimulus. The hypothesis competes with many others; let’s not forget Laffer-type shirking by high taxpayers in the face of higher marginal rates, and contrariwise a shot to animal spirits raising expectations of future income and hence current consumption. Let the econometricians fight this one out. Pending hard data in its favour, I propose to shelve Ricardian equivalence as “not proven”.