On the future rate of unemployment in the US

The current status and near future of inflation and unemployment in the U.S. has ignited fierce debates on the reasons behind both macroeconomic variables. There are quite a few academic economic models with the Phillips curve in the centre of many. Any academic knows that unemployment is likely among the principal driving force behind inflation. To be able for this mission the change in the rate of unemployment must be contemporaneous or lead the induced changes in the rate of inflation. This is a dogma despite observations contradict all such models.
Six years ago, we presented a model which links labour force to unemployment and inflation. In other words both variables are driven by the only force – the change in labour force. Year by year, this model has been providing accurate predictions in all biggest developed countries. Econometrically, the change in labour force is cointegrated with the rate of inflation and unemployment, as the Engle-Granger and Johansen tests have shown. Since both variables have the same root, we introduced a generalized model. This model links all three variables together. It also allows resolving some metrological and methodological problems and improving prediction. Unfortunately for the mainstream models, measurements show that unemployment lags behind inflation and labour force by 2.5 and 5 years respectively. In that sense, we called our model the anti-Phillips curve, i.e. the curve with reversed causality. The original model covered the whole period between 1963 and 2005 by one empirical relationship. It provided a much better prediction of inflation and unemployment than any other model (by a factor of 2 lower RMFSE at a two year horizon). Among other predictions, we foresaw the possibility of a deflationary period from 2012 with very low inflation after 2010. The original model is now enhanced by the introduction of structural breaks related to the major changes in monetary policy. Such structural breaks are manifested in the change of coefficients in the linear links between inflation, unemployment and labour force. The linearity is retained, however. For the U.S. we have obtained the following model:

ut = 1.2lt-5 + 0.21pt-2 +2.03; 1984>t≥1956
ut = 1.9lt-5 + 0.425pt-2 +1.53; 2008>t≥1984
ut = 3.4lt-5 + 0.84pt-2 + 2.00; t≥2008

where ut is the rate of unemployment at time t; lt-5 the relative change rate of labour force (= dlnLF/dt) five years before , t-5; pt-2 is the rate of (CPI) inflation two years before. There are two breaks in 1984 and 2008. Both are obtained by a free search minimizing the RMS model error between 1960 and 2010. One can guess that the change in monetary policy increases the sensitivity of unemployment to the change in labour force and inflation approximately by a factor of 2.

Figures 1 through 3 present the measured and predicted rate of unemployment and the model error. Considering the fact that the rate of unemployment is predicted at a two and a half year horizon the agreement is excellent. Of special importance is the current peak in unemployment which follows from the change in monetary policy in 2008.

Figure 1. Monthly observed and predicted rate of unemployment in the U.S. One can expect the rate to fall below 8% by the end of 2011 or in early 2012.

Figure 2. Annual observed and predicted rate of unemployment in the U.S. One can expect the rate to fall below 8% by the end of 2011 or in early 2012.

Figure 3. The annual model error. The model predicts at a two year horizon. We also present cumulative annual error between 1965 and 2010.