There are inherent limitations in the IS-LM framework because thinking in terms of aggregate non-dynamic relationships will always, at least implicitly, ignore the way individuals react in accordance with rational, inter-temporal expectations. When Keynes developed his general theory, it was never in anyway general because it failed to incorporate expectations, and the time dimension beyond a single period. The IS LM framework developed by Keynes, however, cannot be made to incorporate rational expectations because the assumptions and foundations are inherently incompatible. The author explains that old economists were asking the wrong questions: it should not be what explains cycles, but rather, what explains differences in cycles—in assuming that cycles are not uniform in time, this latter form of the questions gets at the time dimension of cyclical behavior, and therefore opens up more questions about what are the determinants of this time dimension—indeed, this article deals with the journey to formulating this as a question, and the answers to this question.

Looking closely into the Keynesian system we see that he was generally short run oriented. We need to take his time period into account before we judge him too harshly; he was oriented mostly towards policy prescriptions and short run corrections for shocks for which he could not provide an explanation. By neatly dividing up all the interaction in the economy into three major relationships, Keynes was then able to develop his general theory with a few characteristics that King identifies as trouble spots. 1) liquidity preference responding to nominal prices as largely predetermined 2) expectations were largely exogenous, and onlyn effect to be reckoned with by the model, and not from within the model 3) behavioral rules were essentially fixed, and so public policy had the responsibility of seeing the variability in the exogenous shock; therefore, public policy should be developed on a case by case basis.

Milton Friedman criticized the Phillips curve which is consistent with Keynesian theory, for showing changes between the inflation and unemployment rates, rather than a series of events. The relationship was therefore oversimplified into a tradeoff that appeared static rather than dynamic. Even though Friedman criticized the static nature of the Phillips curve, his own formulation of the curve that incorporates expectations seems to be a static theory in other respects. It implicitly assumed that real wage increases would encourage the same increase of labor supply in the short and long run. The new-classical framework assumes, rather that labor supply is exogenous (population growth rate), so that in the long run labor is fixed and much more inelastic with respect to real wage changes. Also, by incorporating micro-foundations and the permanent income hypothesis, the neoclassical approach was able to account for the dynamic relationship between inflation and unemployment, by looking into the individual consumption function.

Indeed, behind this critique however there is a long literature of economists who have built upon the ISLM framework, curve by curve. On the income side of the IS curve, Friedman and his life cycle theory, Modigliani, and Brumberg, all imply the sensitive relationship between the patterns of income, savings and consumption over the lifetime of a rational agent that implicitly hold a stock of wealth. These variables, assumed to be largely independent of each other in ISLM analysis, are intimately related from the perspective of the individual in time. “The marginal propensity to consume depends in an essential way on the nature of the income profile under consideration: it is low for changes in income that are temporary; high for those that are permanent; and depends in a delicate way on the details of the stochastic process for situations in between.”(pg 76) Therefore, as the neoclassical research and theory continued, it became clear that econometric models should also include some variables that control for expectations about the future conditions of income, production, and cost levels.

The demand for money in the LM portion of the graph, is also complicated by expectations. It was deficient because it boiled down to static models—expectations of future inflation, could only influence cash holdings because inflation effects the nominal interest rate in the future. The choice is not simply between money and bonds, but also between different time-maturity bonds, which requires some inclusion of expectations. For these reasons, we see that Friedman implicitly includes wealth in his permanent income theory, and other economist investigate the influences of inflation directly on money demand by looking into “partial adjustment” phenomena in the money demand market.

These models confirm and build upon the rational expectations augmented neoclassical models that came into the picture around the same time. Rational expectations link the short and long run; money and savings-investment markets that were assumed in the IS LM framework to be mutually exclusive, now became essentially linked by individual agents. The permanent income hypothesis is also derived from the rational expectations framework, and gives a very sensitive account of consumer behavior—empirically its predictions are sounds, and theoretically, it demonstrates that IS curve should be dependent on everything—this is just evidence of how much as Keynes model overlooked. In the end, the author wants to change the direction teaching, policy analysis, and econometrics. Because this framework never even addresses, or recognizes the existence of the long run, asking questions and formulation questions from within this framework is could only tell half the story; which never makes for a good story.

Will the New Keynesian Macroeconomics Resurrect the IS-LM Model?
Robert G. King The Journal of Economic Perspectives, Vol. 7, No. 1. (Winter, 1993), pp. 67-82.