Uneasy Money

There’s a lot of talk about the Phillips Curve these days; people wonder why, with the unemployment rate reaching historically low levels, nominal and real wages have increased minimally with inflation remaining securely between 1.5 and 2%. The Phillips Curve, for those untutored in basic macroeconomics, depicts a relationship between inflation and unemployment. The original empirical Philips Curve relationship showed that high rates of unemployment were associated with low or negative rates of wage inflation while low rates of unemployment were associated with high rates of wage inflation. This empirical relationship suggested a causal theory that the rate of wage increase tends to rise when unemployment is low and tends to fall when unemployment is high, a causal theory that seems to follow from a simple supply-demand model in which wages rise when there is an excess demand for labor (unemployment is low) and wages fall when there is an excess supply of labor (unemployment is high).

Viewed in this light, low unemployment, signifying a tight labor market, signals that inflation is likely to rise, providing a rationale for monetary policy to be tightened to prevent inflation from rising at it normally does when unemployment is low. Seeming to accept that rationale, the Fed has gradually raised interest rates for the past two years or so. But the increase in interest rates has now slowed the expansion of employment and decline in unemployment to historic lows. Nor has the improving employment situation resulted in any increase in price inflation and at most a minimal increase in the rate of increase in wages.

In a couple of previous posts about sticky wages (here and here), I’ve questioned whether the simple supply-demand model of the labor market motivating the standard interpretation of the Phillips Curve is a useful way to think about wage adjustment and inflation-employment dynamics. I’ve offered a few reasons why the supply-demand model, though applicable in some situations, is not useful for understanding how wages adjust.

The particular reason that I want to focus on here is Keynes’s argument in chapter 19 of the General Theory (though I express it in terms different from his) that supply-demand analysis can’t explain how wages and employment are determined. The upshot of his argument I believe is that supply demand-analysis only works in a partial-equilibrium setting in which feedback effects from the price changes in the market under consideration don’t affect equilibrium prices in other markets, so that the position of the supply and demand curves in the market of interest can be assumed stable even as price and quantity in that market adjust from one equilibrium to another (the comparative-statics method).

Because the labor market, affecting almost every other market, is not a small part of the economy, partial-equilibrium analysis is unsuitable for understanding that market, the normal stability assumption being untenable if we attempt to trace the adjustment from one labor-market equilibrium to another after an exogenous disturbance. In the supply-demand paradigm, unemployment is a measure of the disequilibrium in the labor market, a disequilibrium that could – at least in principle — be eliminated by a wage reduction sufficient to equate the quantity of labor services supplied with the amount demanded. Viewed from this supply-demand perspective, the failure of the wage to fall to a supposed equilibrium level is attributable to some sort of endogenous stickiness or some external impediment (minimum wage legislation or union intransigence) in wage adjustment that prevents the normal equilibrating free-market adjustment mechanism. But the habitual resort to supply-demand analysis by economists, reinforced and rewarded by years of training and professionalization, is actually misleading when applied in an inappropriate context.

So Keynes was right to challenge this view of a potentially equilibrating market mechanism that is somehow stymied from behaving in the manner described in the textbook version of supply-demand analysis. Instead, Keynes argued that the level of employment is determined by the level of spending and income at an exogenously given wage level, an approach that seems to be deeply at odds with idea that price adjustments are an essential part of the process whereby a complex economic system arrives at, or at least tends to move toward, an equilibrium.

One of the main motivations for a search for microfoundations in the decades after the General Theory was published was to be able to articulate a convincing microeconomic rationale for persistent unemployment that was not eliminated by the usual tendency of market prices to adjust to eliminate excess supplies of any commodity or service. But Keynes was right to question whether there is any automatic market mechanism that adjusts nominal or real wages in a manner even remotely analogous to the adjustment of prices in organized commodity or stock exchanges – the sort of markets that serve as exemplars of automatic price adjustments in response to excess demands or supplies.

Keynes was also correct to argue that, even if there was a mechanism causing automatic wage adjustments in response to unemployment, the labor market, accounting for roughly 60 percent of total income, is so large that any change in wages necessarily affects all other markets, causing system-wide repercussions that might well offset any employment-increasing tendency of the prior wage adjustment.

But what I want to suggest in this post is that Keynes’s criticism of the supply-demand paradigm is relevant to any general-equilibrium system in the following sense: if a general-equilibrium system is considered from an initial non-equilibrium position, does the system have any tendency to move toward equilibrium? And to make the analysis relatively tractable, assume that the system is such that a unique equilibrium exists. Before proceeding, I also want to note that I am not arguing that traditional supply-demand analysis is necessarily flawed; I am just emphasizing that traditional supply-demand analysis is predicated on a macroeconomic foundation: that all markets but the one under consideration are in, or are in the neighborhood of, equilibrium. It is only because the system as a whole is in the neighborhood of equilibrium, that the microeconomic forces on which traditional supply-demand analysis relies appear to be so powerful and so stabilizing.

However, if our focus is a general-equilibrium system, microeconomic supply-demand analysis of a single market in isolation provides no basis on which to argue that the system as a whole has a self-correcting tendency toward equilibrium. To make such an argument is to commit a fallacy of composition. The tendency of any single market toward equilibrium is premised on an assumption that all markets but the one under analysis are already at, or in the neighborhood of, equilibrium. But when the system as a whole is in a disequilibrium state, the method of partial equilibrium analysis is misplaced; partial-equilibrium analysis provides no ground – no micro-foundation — for an argument that the adjustment of market prices in response to excess demands and excess supplies will ever – much less rapidly — guide the entire system back to an equilibrium state.

The lack of automatic market forces that return a system not in the neighborhood — for purposes of this discussion “neighborhood” is left undefined – of equilibrium back to equilibrium is implied by the Sonnenschein-Mantel-Debreu Theorem, which shows that, even if a unique general equilibrium exists, there may be no rule or algorithm for increasing (decreasing) prices in markets with excess demands (supplies) by which the general-equilibrium price vector would be discovered in a finite number of steps.

The theorem holds even under a Walrasian tatonnement mechanism in which no trading at disequilibrium prices is allowed. The reason is that the interactions between individual markets may be so complicated that a price-adjustment rule will not eliminate all excess demands, because even if a price adjustment reduces excess demand in one market, that price adjustment may cause offsetting disturbances in one or more other markets. So, unless the equilibrium price vector is somehow hit upon by accident, no rule or algorithm for price adjustment based on the excess demand in each market will necessarily lead to discovery of the equilibrium price vector.

The Sonnenschein Mantel Debreu Theorem reinforces the insight of Kenneth Arrow in an important 1959 paper “Toward a Theory of Price Adjustment,” which posed the question: how does the theory of perfect competition account for the determination of the equilibrium price at which all agents can buy or sell as much as they want to at the equilibrium (“market-clearing”) price? As Arrow observed, “there exists a logical gap in the usual formulations of the theory of perfectly competitive economy, namely, that there is no place for a rational decision with respect to prices as there is with respect to quantities.”

Prices in perfect competition are taken as parameters by all agents in the model, and optimization by agents consists in choosing optimal quantities. The equilibrium solution allows the mutually consistent optimization by all agents at the equilibrium price vector. This is true for the general-equilibrium system as a whole, and for partial equilibrium in every market. Not only is there no positive theory of price adjustment within the competitive general-equilibrium model, as pointed out by Arrow, but the Sonnenschein-Mantel-Debreu Theorem shows that there’s no guarantee that even the notional tatonnement method of price adjustment can ensure that a unique equilibrium price vector will be discovered.

While acknowledging his inability to fill the gap, Arrow suggested that, because perfect competition and price taking are properties of general equilibrium, there are inevitably pockets of market power, in non-equilibrium states, so that some transactors in non-equilibrium states, are price searchers rather than price takers who therefore choose both an optimal quantity and an optimal price. I have no problem with Arrow’s insight as far as it goes, but it still doesn’t really solve his problem, because he couldn’t explain, even intuitively, how a disequilibrium system with some agents possessing market power (either as sellers or buyers) transitions into an equilibrium system in which all agents are price-takers who can execute their planned optimal purchases and sales at the parametric prices.

One of the few helpful, but, as far as I can tell, totally overlooked, contributions of the rational-expectations revolution was to solve (in a very narrow sense) the problem that Arrow identified and puzzled over, although Hayek, Lindahl and Myrdal, in their original independent formulations of the concept of intertemporal equilibrium, had already provided the key to the solution. Hayek, Lindahl, and Myrdal showed that an intertemporal equilibrium is possible only insofar as agents form expectations of future prices that are so similar to each other that, if future prices turn out as expected, the agents would be able to execute their planned sales and purchases as expected.

But if agents have different expectations about the future price(s) of some commodity(ies), and if their plans for future purchases and sales are conditioned on those expectations, then when the expectations of at least some agents are inevitably disappointed, those agents will necessarily have to abandon (or revise) the plans that their previously formulated plans.

What led to Arrow’s confusion about how equilibrium prices are arrived at was the habit of thinking that market prices are determined by way of a Walrasian tatonnement process (supposedly mimicking the haggling over price by traders). So the notion that a mythical market auctioneer, who first calls out prices at random (prix cries au hasard), and then, based on the tallied market excess demands and supplies, adjusts those prices until all markets “clear,” is untenable, because continual trading at disequilibrium prices keeps changing the solution of the general-equilibrium system. An actual system with trading at non-equilibrium prices may therefore be moving away from, rather converging on, an equilibrium state.

Here is where the rational-expectations hypothesis comes in. The rational-expectations assumption posits that revisions of previously formulated plans are never necessary, because all agents actually do correctly anticipate the equilibrium price vector in advance. That is indeed a remarkable assumption to make; it is an assumption that all agents in the model have the capacity to anticipate, insofar as their future plans to buy and sell require them to anticipate, the equilibrium prices that will prevail for the products and services that they plan to purchase or sell. Of course, in a general-equilibrium system, all prices being determined simultaneously, the equilibrium prices for some future prices cannot generally be forecast in isolation from the equilibrium prices for all other products. So, in effect, the rational-expectations hypothesis supposes that each agent in the model is an omniscient central planner able to solve an entire general-equilibrium system for all future prices!

But let us not be overly nitpicky about details. So forget about false trading, and forget about the Sonnenschein-Mantel-Debreu theorem. Instead, just assume that, at time t, agents form rational expectations of the future equilibrium price vector in period (t+1). If agents at time t form rational expectations of the equilibrium price vector in period (t+1), then they may well assume that the equilibrium price vector in period t is equal to the expected price vector in period (t+1).

Now, the expected price vector in period (t+1) may or may not be an equilibrium price vector in period t. If it is an equilibrium price vector in period t as well as in period (t+1), then all is right with the world, and everyone will succeed in buying and selling as much of each commodity as he or she desires. If not, prices may or may not adjust in response to that disequilibrium, and expectations may or may not change accordingly.

Thus, instead of positing a mythical auctioneer in a contrived tatonnement process as the mechanism whereby prices are determined for currently executed transactions, the rational-expectations hypothesis posits expected future prices as the basis for the prices at which current transactions are executed, providing a straightforward solution to Arrow’s problem. The prices at which agents are willing to purchase or sell correspond to their expectations of prices in the future. If they find trading partners with similar expectations of future prices, they will reach agreement and execute transactions at those prices. If they don’t find traders with similar expectations, they will either be unable to transact, or will revise their price expectations, or they will assume that current market conditions are abnormal and then decide whether to transact at prices different from those they had expected.

When current prices are more favorable than expected, agents will want to buy or sell more than they would have if current prices were equal to their expectations for the future. If current prices are less favorable than they expect future prices to be, they will not transact at all or will seek to buy or sell less than they would have bought or sold if current prices had equaled expected future prices. The dichotomy between observed current prices, dictated by current demands and supplies, and expected future prices is unrealistic; all current transactions are made with an eye to expected future prices and to their opportunities to postpone current transactions until the future, or to advance future transactions into the present.

If current prices for similar commodities are not uniform in all current transactions, a circumstance that Arrow attributes to the existence of varying degrees of market power across imperfectly competitive suppliers, price dispersion may actually be caused, not by market power, but by dispersion in the expectations of future prices held by agents. Sellers expecting future prices to rise will be less willing to sell at relatively low prices now than are suppliers with pessimistic expectations about future prices. Equilibrium occurs when all transactors share the same expectations of future prices and expected future prices correspond to equilibrium prices in the current period.

Of course, that isn’t the only possible equilibrium situation. There may be situations in which a future event that will change a subset of prices can be anticipated. If the anticipation of the future event affects not only expected future prices, it must also and necessarily affect current prices insofar as current supplies can be carried into the future from the present or current purchases can be postponed until the future or future consumption shifted into the present.

The practical upshot of these somewhat disjointed reflections is, I think,primarily to reinforce skepticism that the traditional Phillips Curve supposition that low and falling unemployment necessarily presages an increase in inflation. Wages are not primarily governed by the current state of the labor market, whatever the labor market might even mean in macroeconomic context.

Expectations rule! And the rational-expectations revolution to the contrary notwithstanding, we have no good theory of how expectations are actually formed and there is certainly no reason to assume that, as a general matter, all agents share the same set of expectations.

The current fairly benign state of the economy reflects the absence of any serious disappointment of price expectations. If an economy is operating not very far from an equilibrium, although expectations are not the same, they likely are not very different. They will only be very different after the unexpected strikes. When that happens, borrowers and traders who had taken positions based on overly optimistic expectations find themselves unable to meet their obligations. It is only then that we will see whether the economy is really as strong and resilient as it now seems.

Expecting the unexpected is hard to do, but you can be sure that, sooner or later, the unexpected is going to happen.