Blog o’ Burns

King, Robert G. “Will the New Keynesian Macroeconomics Resurrect the IS-LM Model?” Journal of Economic Perspectives 7, no. 1 (Winter 1993): 67-82.

Part of the downfall of the original Keynesian model was its inability to explain the theoretical foundations for the existence of sticky prices. The past two decades has seen the emergence of new Keynesian economists using micro foundations to explain why both real and nominal rigidities might exist in the market. These exciting findings have led to the temptation to accept old Keynesian assumption as valid; however, King argues that new Keynesian work will never be able to reconcile the IS-LM model with rational expectations. As such, the IS-LM model may still be a useful way to present results, but it simplifies reality too greatly to be a reliable way to derive those results.

As all of us have learned in previous macro courses, the IS curve, which basically serves as the aggregate demand curve, is downward sloping under the assumption that savings is a positive function of income and investment is a negative function of interest rates. The LM curve is upward sloping since money demand depends positively on real income and negatively on nominal interest rates (which reflect the opportunity cost of holding money). A typical classroom analysis would go like this: an increase in the money supply (represented by shifting the LM curve out) leads to an increase in aggregate demand (which shifts the IS curve out) and leads to lower real and nominal interest rates. However, in the real world is it possible for expectations to shift the IS curve to such an extent that both real and nominal interest rates actually rise. This may happen because: 1) an increase in the money supply may signal higher future prices and lower the cost of investment at any nominal interest rate; and 2) investors recognize that an increase in the money supply will increase aggregate demand and will therefore invest more to capture the expected gains from additional consumption. These dynamic real world possibilities are difficult to account for using typical IS-LM analysis, which considers expectations as exogenous to the model.

To be fair, the IS-LM model was not initially designed to capture all of the nuances of real world behavior. It was primarily a short cut to demonstrate the impact of short run shocks in aggregate demand to help formulate policy responses. During the 1950s and 60s, intertemporal models of consumption and investment began emerging to reconcile short-run and long-run behavioral observations. These included Friedman’s permanent income hypothesis and Jorgenson neoclassical theory of investment, both of which placed importance on expectations of the future: future income, future production, and future costs.

It wasn’t until the Lucas critique that expectations began being thought of as endogenous to macro models. This critique directly assaulted two distinctions made by Keynes: the distinction between the short and long run and the separation of the saving/investment and monetary sectors. The incorporation of rational expectations blurred these distinctions since: 1) expectations about the future means that people’s behavior are based on intertemporal factors; and 2) expectations link conditions in the savings sector to conditions in the monetary sector. In general, rational expectations imply that all relevant variables useful for forecasting income and interest rates should be included in the consumption function– not just current income.

Apart from the theoretical weaknesses of the IS-LM model, it also does a poor job explaining variation in empirical studies. The model predicts that a positive IS shock will raise output and temporarily raise nominal interest rates. However, empirical studies have shown that such positive shocks tend to rates the rate of inflation for the first few quarters and permanently raises inflation across all time horizons. Thus, real interest rates decline. King argues that without any justification of being a reliable policy tool, the IS-LM model should be abandoned and only used to simplify the presentation of more complex findings.