Table of contents

Keynes and The Keynesians


Keynesianism as an ideology oftentimes represents not the economic philosophy of John Maynard Keynes but rather “the policies of the Keynesians.” The General Theory of Employment, Interest, and Money and its interpretations have been debated since it was published in 1936. Of these, Hick’s neoclassical synthesis model, the IS-LM, has been argued almost as much as The General Theory itself.

Keynes’s attitudes toward the IS-LM apparatus are of little consequence to those who consider themselves Keynesians or to those who vehemently oppose the supposed Keynesians. IS-LM has become a symbol, one that has been attacked and subsequently defended on numerous occasions. To understand the merits and shortcomings of these criticisms, the core purpose of the IS-LM apparatus must be understood. The IS-LM model is, of course, wrong, but when was it meant to be useful?

Solow offers a simple explanation, “IS-LM survived because it proved to be a marvelously simple and useful way to organize and process some of the main macroeconomic facts.” The simplification of many of Keynes’s ideas into the IS-LM model has provided a useful framework for applying intuition to monetary and fiscal actions. Beyond intuition, Solow views IS-LM as an evolutionary step towards more comprehensive models. The need for this stepping-stone was particularly acute during the great depression. Solow asserts that “the General Theory of Employment is the Economics of Depression.” By extension, this makes IS-LM the theory of depression. Indeed, many of the questions Keynes considers are a careful analysis of economies below full employment. In contrast, criticisms of Keynesianism, especially from the monetarist school, were especially poignant during the inflationary crises of the 70s and 80s. Intense inflationary environments provide the most intense test of the IS-LM apparatus and bring to the forefront its shortcomings.

I will first provide a very brief overview of the critical assumptions of the IS-LM model before considering two of its most essential failures. The first is the assumption of fixed nominal prices and wages. The second is its omission of expectations.

The IS-LM Model

The IS-LM begins on a foundation of a short-run horizon with homogeneous labor. Hicks later relaxes the assumptions of homogeneous labor but assumes a fixed rate of wages, $w$. The classical model assumes a static marginal-efficiency-of-capital ratio, whereas the IS-LM lets the money supply be a function of income and interest; this is where the LM curve is derived. The IS-LM model can be modeled as

$$M = L(I,i), I_x=C(I,i), I_x=S(I,i)$$

Here all investment, consumption, and money demand depend on income and the interest rate. The issues lie not in the formation of the IS-LM model but in its core assumptions. The fixed nature of wages in the short-run and the lack of incorporating future expectations into the functions of the model.

Price and Wage Dynamics

IS-LM is built upon the assumptions of nominal wage and price rigidity. Solow goes as far as to say that “there is ample evidence that real wages have no strong endogenous pattern in modern industrial economies.” These are fair assumptions for fair times but fail during crises when macroeconomics is most needed. In times of exceptionally low or exceptionally high employment, endogenous wage movements may impact the analysis of the IS curve. First, I weigh the argument against the IS-LM.

Axel Leijonhufvud offers a compelling argument against the Keynesian contributions to our understanding of unemployment. The price-level, $p$, typically enters the LM curve as an exogenous variable to normalize inflationary or deflationary effects on the money demanded. The classical interpretation is the value theory of labor. Firms have a labor demand function, $N^d(w/p)$, laborers have a labor supply function, $N^s(w/p)$, and the labor market clears when $N^S = N^d$. Keynes takes a different approach where wages are fixed at a level $w = \bar w$ and set above the classical solution $\bar w > \hat w$. The labor supply function tells us how much unemployment a system will have, given the wage and the price level. Labor market equilibrium as a concept does not exist in the Keynesian model. In either case, if wages are free to move, then the $p$ adjusts, and the story does not change. The classical case fully assumes the neutrality of money; only the real values are affected.

In the Keynesian case, wages are nominally rigid. The case of low employment, i.e., the depression economy that Keynes was primarily concerned with, presents a deflationary environment. Because $\bar w > \hat w$, that is, sticky wages are higher than the equilibrium level of wages given the price-level, unemployment is high. Small endogenous wage movements at this level can drastically shift the IS curve and create significant effects in either direction.

Taking the opposite case, an economy at full employment or higher, the inflationary environment is likely to be hot. Internal movements of wages at this level may have substantial effects on the price-level, creating self-fulfilling circles of inflation that the IS curve does not account for. These exceptionally tight labor markets cannot survive for long without intense inflation.

There is the third case, an economy with low employment and also high inflation. In this case, wage movements, or lack thereof, are indeed maligned with the IS-LM analysis. The fall in real wages during stagflation leads “to a fall in the general price level, worsens the real burden of debt and has a contractionary effect on the economy.”

These criticisms extend beyond the wage rate and also apply to endogenous movements of the price level. Price level may shift during inflationary periods, and is not independent of money supply, money demanded, consumption, investment, or government spending. Treating it as such is particularly disastrous during periods of high inflation. Solow readily admits that “when the macroeconomic problem is dominated by partial adjustment to inflation and its aggregative and distributional consequences, IS-LM is not the right model, or is only a part of the right model.” Solow reconciles the use of IS-LM despite its fundamental disregard for endogenous movements of real price levels and thus wage level, by proclaiming that the IS should be viewed as “a locus along which the wage-employment bargain is an equilibrium given a goods-market outcome that is an equilibrium given the wage-employment bargain.” In other words, the dynamics of labor markets and goods markets reconcile themselves within a period under normal circumstances along the IS curve.

If the IS-LM is so fundamentally flawed in inflationary environments, why study it at all? First, I return to the training of intuition. IS-LM does not stand the test of minute econometrics and lacks micro-foundations. What Leijonhufvud sees as a failure, is actually a prominent feature of the IS-LM apparatus. Keynes himself was adamantly against turning the General Theory into a precise mathematical model. This Keynesian spirit has been the double-edged sword of the wider economic world. One on hand, it reminds generations of economists to consider ideas and implications and avoid falling into the trap of the econometricians. On the other, it is the very reason we lack a definitive model of Keynesianism, and allows for these debates. Something Paul Krugman believes is responsible for the “dark age of macroeconomics.” To borrow from Hicks himself, “The General Theory of Employment is a useful book; but is neither the beginning nor the end of Dynamic Economics.”

Furthermore, the intuition of the IS-LM stands the test of time, and Paul Krugman illustrates the liquidity trap of the 2008 crisis. Despite massive increases in money supply due to expansionary fiscal and monetary policy, inflation remained steady in the 2010s. It is in the present inflationary crises that the IS-LM becomes an inexact approximation because of the potential change in norms around wage and price setting. The lack of incorporation of norms is a much larger macroeconomic issue than IS-LM could hope to tackle. Endogenizing norms is a much bigger endeavor that the field must take, but that does not detract from the utility of the IS-LM apparatus as a whole.


The IS-LM apparatus has little to say on the matter of expectations. Solow views a minor problem and a major problem arising from this. The minor problem is the investment component of effective demand. Of course, investment decisions are based on discounting rates and forward-looking decision-making. Solow offers adding a separate variable known as expected output and computing effective demand as a function of expected output, current output, and interest rate. National income is now modeled as a function of the interest rate and the discounted value of all future flows of national income. The pre-existing $Y-r$ space does not become a $(Y+Y_e)-r$ space; rather, three dimensions are needed to model a $Y-Y_e-r$ space. This, of course, says nothing about the endogenous effects of monetary and fiscal policy on expectations, and the further effects of expectations on monetary and fiscal policy. In short, the IS curves can be adjusted to account for expectations partially through existing variables and partially through exogenous intervention. Unfortunately, it seems that IS-LM is resigned to a partial medium-run equilibrium, but for Solow, this is no matter because there is no model that can fully incorporate the future into the present. Nevertheless, he encourages “an Icarus waiting to try.”

The more pressing issue, Solow’s major problem, enters through the LM curve. IS-LM is a flow model, but the LM curve is defined as the set of points that clears the money market, $M_s = M_d$. Here lies the rub; money supply is a stock variable. Thus, the LM curve represents a stock relation, and a stock equilibrium must exist throughout the period.

Solow states that “the very existence of a demand for money as a liquid asset presupposes that expectations may be unfulfilled” . Money demand is viewed as the sum of money demanded for speculation and transaction $M_D = M_t + M_{sp}$. Conceptually speculation demand will only happen if there are expectations for future states that are misaligned with the current state of the economy or with the current state of expectations. If future information were perfect, there would be no speculation, no expectations, and therefore no money demand. It is hard to imagine money demand without considering speculation, but we may set it to zero as an exercise. In the real world, this would surely equate to a liquidity trap, but the role of expectations becomes clear when viewing $M_t$ as a function of national wealth. Differing expectations from households and firms will fundamentally change consumption and investment patterns, most importantly, willingness to take on debt. Once again, the problem of expectations rears its ugly head.

Hicks attempted to settle the issue by allowing for stochastic liquidity preference, but Solow points out three issues with this approach. 1) The inherent difficulty in integrating stocks and flows, 2) The difficulty of fitting money into equilibrium models, and 3) probabilistic equilibria do not make a sufficient concession to true uncertainty, i.e., the model functions as usual with a small nod to expected value theory without truly describing the range of possible outcomes. Solow claims that “none of these difficulties is peculiar to IS-LM.” The problem of stocks and flows is a more general one, not a pure indictment of the model.

Incorporating expectations into modeling remains a difficult task for predicting long-run equilibria, but attempting to predict a long-run equilibrium is a fool’s errand, to begin with. It is fundamentally impossible to forecast structural change or catastrophe, much less to prescribe appropriate fiscal and monetary policies in the face of this uncertainty. Nevertheless, IS-LM need not clear an impossible bar to be a useful starting point for our intuition. Nor does it need to be perfect for us to recognize its place in the development of economic thought. It would be quite a travesty for my career if Hicks and Keynes solved macroeconomics in 1936.

Keynes, The Classics, and Beyond

Robert Solow’s primary defense of the IS-LM relies on an analogy of biological evolution. Keynesianism and the neo-classical synthesis are an iteration of “the Classics,” and they have, since their inception, been replaced by a proliferation of different economic schools. It is worth assessing Solow’s claim by examining the development of classical economics, Keynesianism, the neoclassical synthesis, and today’s schools.

Keynes’s primary purpose in the first part of The General Theory is a refutation of Say’s Law of markets. Say’s fundamental claim lies in the accounting identity that “products are paid for by products”. In Keynes’s view, Say’s law claims that supply creates its own demand; the past economists, including Ricardo, Malthus, and Marshall, implicitly assume this accounting identity in their classical theories of money, interest, and employment.

Keynes iterated on classical theory with his notions that schedules of aggregate supply and demand determine economic activity. Aggregate demand arises from investment, consumption, and potential inducement to invest from the government. Simple Keynesian crosses show that an increase in overall spending increases national income and, depending on the macroeconomic environment, may increase inflation. Especially unique to Keynes is his emphasis on the role of expectations in investment schedules. For example, if investors believe that their marginal capital efficiency is greater than the current interest rate, they will invest and potentially borrow to invest. Furthermore, Keynes’s views on unemployment show that economies never reach full employment; thus, the General Theory concerns itself with the economics of depression. While the IS-LM apparatus gives intuition to some of these facts, much of Keynes’s economics is lost in Hick’s neoclassical synthesis.

IS-LM and the neoclassical synthesis bridge to more complicated labor closed and open economies models. Unfortunately, the powerful simplicity of the IS-LM is also one of its most significant weaknesses. Some would say it perfectly captures Keynes’s core ideas, but others would disagree for the abovementioned reasons, namely the role of price levels and expectations. Krugman describes these as differences between part 1ers and chapter 12ers. Still, IS-LM fills its role nicely as intuition for economists.

Blanchard describes the state of modern economics as being fundamentally influenced by Keynes. Contributions of aggregate demand and expectations were reincorporated into new-Keynesian, new-Classical, and new-growth models. IS-LM has led to the development of the next phase of macroeconomics and continues to provide a helpful starting point for an entire generation of economics.