We consider the revision properties of Federal Reserve Board staff estimates of the output gap after the mid-1990s and examine the usefulness of these estimates for inflation forecasting. Over this period, we find that the Federal Reserve's output gap is more reliably estimated in real time than previous studies have documented for earlier periods and alternative estimation techniques. In contrast to previous work, we also find no deterioration in forecast performance when inflation projections are conditioned on real-time rather than on final estimates of the output gap.

The canonical inflation specification in sticky-price rational expectations models (the new-Keynesian Phillips curve) is often criticized for failing to account for the dependence of inflation on its own lags. In response, many studies employ a “hybrid” specification in which inflation depends on its lagged and expected future values, together with a driving variable such as the output gap. We consider some simple tests of the hybrid model that are derived from its closed form. We find that the hybrid model describes inflation dynamics poorly, and find little empirical evidence for the type of rational, forward-looking behavior that the model implies.

We survey the evidence bearing on measurement error in the CPI and provide our best estimate of the magnitude of CPI bias. We also identify a "weighting" bias in the CPI that has not been previously discussed in the literature. In total, we estimate that the CPI overstates the change in the cost of living by about 0.6 percentage point per year, with a confidence interval that ranges from 0.1 to 1.2 percentage points. Roughly half of this bias is accounted for by the CPI's inability to fully capture the welfare improvement from quality change and the introduction of new items. Our bias estimate is smaller than that found in several earlier studies, in part because the BLS has recently made a variety of improvements to its procedures; our study highlights several potential areas for further improvement.

We survey the evidence bearing on measurement error in the CPI and provide our best estimate of the magnitude of CPI bias. We also identify a weighting bias in the CPI that has not been previously discussed in the literature. In total, we estimate that the CPI overstates the change in the cost of living by about 0.6 percentage point per year, with a confidence interval that ranges from 0.1 to 1.2 percentage points. Roughly half of this bias is accounted for by the CPI's inability to fully capture the welfare improvement from quality change and the introduction of new items. Our bias estimate is smaller than that found in several earlier studies, in part because the BLS has recently made a variety of improvements to its procedures; our study highlights several potential areas for further improvement.