The Output Gap Concept: On Inflation Forecasts and New Keynesian Measures

Abstract

Inflation forecasting is a continuous issue, not only in central bank business, but also in economics in general, combining the two major pillars in economic science, theory and empirics. The Great Recession or more precisely the behaviour of inflation following the severe contraction of output and the subsequent recovery process challenged the contemptuous forecasting models. Those models often rely on one of the most famous macroeconomic concepts, the Phillips curve. Within this framework, inflation is linked to a measure of economic activity, e.g. the widely used output gap concept, which is defined as the deviation of actual from the inflation-stable or potential output. Relying on this framework, inflation dropped during the crisis too less and increased afterwards too weak. Consequently, the question about the usefulness of output gaps as an instrument for inflation prediction raises and additionally the question if new measures retrieved from New Keynesian models can improve forecasts in a simple forecast framework. Current research indicates that for example nonlinear modifications of New Keynesian models could be sufficient to solve or model the puzzling behaviour of inflation in the last decade. Similar results are also indicated in non-DSGE Phillips curve forecasts. The New Keynesian models have the additional advantage that specific economic aspects like financial frictions can be incorporate, which could be particular important during the Great Recession. Therefore, this analysis not only replicated traditional measures (statistical and reduced-form methods) of the output gap, but also two DSGE measures, with and without financial frictions. Considering the last three decades, the mentioned replicates of the US output gap measures differ in several aspects: The DSGE measures are more volatile than the traditional measures, account within this sample for the largest outliers, but allow in contrast to the traditional measures for a deeper interpretation of fundamental drivers of past output gaps. The uncertainty about the level of output gaps, and sometimes even about the sign, is high, as previous literature already indicated. Using output gaps as a precise instrument is therefore at best questionable. Another observation that suggest a limitation of such measures for the purpose of inflation forecasts, is their time-inconsistency. Both, traditional and New Keynesian methods produce different estimates for final and real-time approaches. Whereas literature of the years before the Great Recession at least found a good in-sample fit and a worse out-of-sample fit of Phillips curve-styled models, this analysis results a poor in- and out-of-sample fit and supports therefore the suggested breakdown of the Phillips curve relationship. This indicates a poor predictive power of these output gap estimates on inflation nowadays. The intended usefulness of financial frictions in New Keynesian measures of output gap cannot be observed, since the extended DSGE measure produces qualitatively similarly estimates to the baseline model, i.e. in terms of correlation, standard deviation, and the behaviour within the forecast framework. Reformulations of the forecast equation in a nonlinear or New Keynesian (including of expectations) way don’t result in a restoring of the Phillips curve. Because of that, the suggested possible solution to the missing deflation and inflation puzzle in a DSGE framework does not apply in this analysis, which uses New Keynesian models solely to estimate the output gap and not for forecasting purposes. However, using the change of the output gap instead of the levels results in statistically significant coefficients for the traditional measures using the consumer price inflation and for all measures using an inflation measure derived from the GDP implicit price deflator. Although, these models are not able to beat the naïve benchmark without an output gap measure. The implications from these findings are unclear since they are at odds with the traditional understanding of the output gap, saying that inflationary or deflationary pressure is induced when actual output is above or below its potential counterpart. Plausible would be that the dynamics of output gap induce additional pressure, but not instead of the level. Therefore, additional research is necessary to investigate the implications of this finding and to state if these results apply in general or represent a model or sample specific feature.