Which Interest Rate Should We Use In The Is Curve?

Do interest rates effect investment and the GDP? If so, which ones, and by how much? Research on this topic over 5 decades has produced conflicting results. Yet, this question is of critical importance to the viability of Keynesian macroeconomics. This paper attempts to explain why results have been conflicting. It also attempts to determine with some finality which rate(s), if any, are related to GDP through the standard Keynesian mechanism: the IS curve. The paper tests exhaustively (1) a variety of real and nominal rates, (2) different hypotheses about how businesses calculate “real” interest rates (3) how the number of lags used affects results, (4) whether small sample size inherent in annual time series data adversely affects results, and (5) whether lack of hetroskedasticity and autocorrelation controls in earlier studies influenced their findings. This paper concludes only the real prime or Federal funds rates, lagged two years and the nominal current mortgage rate are significantly related to variation in the GDP, and running the prime rate alone picks up most of the variation in both. The prime rate was found to be twice as important as the mortgage rate. It also finds relatively small size (40 observation) annual data sets do not lead to problems achieving statistical significance, at least in simple IS curve models. It also finds that post - 1980 White and Newey - West correction methods for hetroskedasticity make it far more likely that any of a wide variety of interest rates and lags will be found statistically significant than was the case in earlier studies, but that correcting for multicollinearity between rates again leaves only the real prime and Federal funds rate lagged two periods and perhaps the current nominal mortgage rate significant. The effect of changes in the prime rate and mortgage rates on the GDP, though systematic, appears to be small, implying the IS curve may be nearly vertical and the Fed’s interest rate policy of little significance unless rate changes are draconian. We estimate that even a five percentage - point change in the real Federal funds and prime rates changes GDP only 2.4%, and employment only 1.2% maximally (using Okun’s law). Other findings were that nominal interest rates deflated by adaptive expectations models of inflation using the past two year’s inflation seem to best describe how businesses calculate real rates. Rational expectations models were least successful. Other rates examined include the ten year treasury rate, the Aaa and Baa corporate rates. They were seldom found statistically significant, but the mortgage rate’s estimated marginal effect seems to also capture these rates’ effect on the economy.