One of Scott Sumner’s many contributions as a blogger has been to show over and over and over again how easy it is to lapse into fallacious economic reasoning by positing a price change and then trying to draw inferences about the results of the price change. The problem is that a price change doesn’t just happen; it is the result of some other change. There being two basic categories of changes (demand and supply) that can affect price, there are always at least two possible causes for a given price change. So, until you have specified the antecedent change responsible for the price change under consideration, you can’t work out the consequences of the price change.
In this post, I want to extend Scott’s insight in a slightly different direction, and explain how every economic analysis has to begin with a statement about the initial conditions from which the analysis starts. In particular, you need to be clear about the equilibrium position corresponding to the initial conditions from which you are starting. If you posit some change in the system, but your starting point isn’t an equilibrium, you have no way of separating out the adjustment to the change that you are imposing on the system from the change the system would be undergoing simply to reach the equilibrium toward which it is already moving, or, even worse, from the change the system would be undergoing if its movement is not toward equilibrium.
Every theoretical analysis in economics properly imposes a ceteris paribus condition. Unfortunately, the ubiquitous ceteris paribus condition comes dangerously close to rendering economic theory irrefutable, except perhaps in a statistical sense, because empirical refutations of the theory can always be attributed to changes, abstracted from only in the theory, but not in the real world of our experience. An empirical model with a sufficient number of data points may be able to control for the changes in conditions that the theory holds constant, but the underlying theory is a comparison of equilibrium states (comparative statics), and it is quite a stretch to assume that the effects of perpetual disequilibrium can be treated as nothing but white noise. Austrians are right to be skeptical of econometric analysis; so was Keynes, for that matter. But skepticism need not imply nihilism.
Let me try to illustrate this principle by applying it to the Keynesian analysis of involuntary unemployment. In the General Theory Keynes argued that if adequate demand is deficient, the likely result is an equilibrium with involuntary unemployment. The “classical” argument that Keynes disputed was that, in principle at least, involuntary unemployment could not persist, because unemployed workers, if only they would accept reduced money wages, would eventually find employment. Keynes denied that involuntary unemployment could not persist, arguing that if workers did accept reduced money wages, the wage reductions would not get translated into reduced real wages. Instead, falling nominal wages would induce employers to cut prices by roughly the same percentage as the reduction in nominal wages, leaving real wages more or less unchanged, thereby nullifying the effectiveness of nominal-wage cuts, and, instead, fueling a vicious downward spiral of prices and wages.
In making this argument, Keynes didn’t dispute the neoclassical proposition that, with a given capital stock, the marginal product of labor declines as employment increases, implying that real wages have to fall for employment to be increased. His argument was about the nature of the labor-supply curve, labor supply, in Keynes’s view, being a function of both the real and the nominal wage, not, as in the neoclassical theory, only the real wage. Under Keynes’s “neoclassical” analysis, the problem with nominal-wage cuts is that they don’t do the job, because they lead to corresponding price cuts. The only way to reduce unemployment, Keynes insisted, is to raise the price level. With nominal wages constant, an increased price level would achieve the real-wage cut necessary for employment to be increased. And this is precisely how Keynes defined involuntary unemployment: the willingness of workers to increase the amount of labor actually supplied in response to a price level increase that reduces their real wage.
Interestingly, in trying to explain why nominal-wage cuts would fail to increase employment, Keynes suggested that the redistribution of income from workers to entrepreneurs associated with reduced nominal wages would tend to reduce consumption, thereby reducing, not increasing, employment. But if that is so, how is it that a reduced real wage, achieved via inflation, would increase employment? Why would the distributional effect of a reduced nominal, but unchanged real, wage be more adverse to employment han a reduced real wage, achieved, with a fixed nominal wage, by way of a price-level increase?
Keynes’s explanation for all this is confused. In chapter 19, where he makes the argument that money-wage cuts can’t eliminate involuntary unemployment, he presents a variety of reasons why nominal-wage cuts are ineffective, and it is usually not clear at what level of theoretical abstraction he is operating, and whether he is arguing that nominal-wage cuts would not work even in principle, or that, although nominal-wage cuts might succeed in theory, they would inevitably fail in practice. Even more puzzling, It is not clear whether he thinks that real wages have to fall to achieve full employment or that full employment could be restored by an increase in aggregate demand with no reduction in real wages. In particular, because Keynes doesn’t start his analysis from a full-employment equilibrium, and doesn’t specify the shock that moves the economy off its equilibrium position, we can only guess whether Keynes is talking about a shock that had reduced labor productivity or (more likely) a shock to entrepreneurial expectations (animal spirits) that has no direct effect on labor productivity.
There was a rhetorical payoff for Keynes in maintaining that ambiguity, because he wanted to present a “general theory” in which full employment is a special case. Keynes therefore emphasized that the labor market is not self-equilibrating by way of nominal-wage adjustments. That was a perfectly fine and useful insight: when the entire system is out of kilter; there is no guarantee that just letting the free market set prices will bring everything back into place. The theory of price adjustment is fundamentally a partial-equilibrium theory that isolates the disequiibrium of a single market, with all other markets in (approximate) equilibrium. There is no necessary connection between the adjustment process in a partial-equilibrium setting and the adjustment process in a full-equilibrium setting. The stability of a single market in disequilibrium does not imply the stability of the entire system of markets in disequilibrium. Keynes might have presented his “general theory” as a theory of disequilibrium, but he preferred (perhaps because he had no other tools to work with) to spell out his theory in terms of familiar equilibrium concepts: savings equaling investment and income equaling expenditure, leaving it ambiguous whether the failure to reach a full-employment equilibrium is caused by a real wage that is too high or an interest rate that is too high. Axel Leijonhufvud highlights the distinction between a disequilibrium in the real wage and a disequilibrium in the interest rate in an important essay “The Wicksell Connection” included in his book Information and Coordination.
Because Keynes did not commit himself on whether a reduction in the real wage is necessary for equilibrium to be restored, it is hard to assess his argument about whether, by accepting reduced money wages, workers could in fact reduce their real wages sufficiently to bring about full employment. Keynes’s argument that money-wage cuts accepted by workers would be undone by corresponding price cuts reflecting reduced production costs is hardly compelling. If the current level of money wages is too high for firms to produce profitably, it is not obvious why the reduced money wages paid by entrepreneurs would be entirely dissipated by price reductions, with none of the cost decline being reflected in increased profit margins. If wage cuts do increase profit margins, that would encourage entrepreneurs to increase output, potentially triggering an expansionary multiplier process. In other words, if the source of disequilibrium is that the real wage is too high, the real wage depending on both the nominal wage and price level, what is the basis for concluding that a reduction in the nominal wage would cause a change in the price level sufficient to keep the real wage at a disequilibrium level? Is it not more likely that the price level would fall no more than required to bring the real wage back to the equilibrium level consistent with full employment? The question is not meant as an expression of policy preference; it is a question about the logic of Keynes’s analysis.
Interestingly, present-day opponents of monetary stimulus (for whom “Keynesian” is a term of extreme derision) like to make a sort of Keynesian argument. Monetary stimulus, by raising the price level, reduces the real wage. That means that monetary stimulus is bad, as it is harmful to workers, whose interests, we all know, is the highest priority – except perhaps the interests of rentiers living off the interest generated by their bond portfolios — of many opponents of monetary stimulus. Once again, the logic is less than compelling. Keynes believed that an increase in the price level could reduce the real wage, a reduction that, at least potentially, might be necessary for the restoration of full employment.
But here is my question: why would an increase in the price level reduce the real wage rather than raise money wages along with the price level. To answer that question, you need to have some idea of whether the current level of real wages is above or below the equilibrium level. If unemployment is high, there is at least some reason to think that the equilibrium real wage is less than the current level, which is why an increase in the price level would be expected to cause the real wage to fall, i.e., to move the actual real wage in the direction of equilibrium. But if the current real wage is about equal to, or even below, the equilibrium level, then why would one think that an increase in the price level would not also cause money wages to rise correspondingly? It seems more plausible that, in the absence of a good reason to think otherwise, that inflation would cause real wages to fall only if real wages are above their equilibrium level.