On the Accuracy and Properties of Recent Macroeconomic Forecasts

Abstract

The aim of this study is to contribute to the measurement and analysis of errors in economists' predictions of changes in aggregate income, output, and. the prIce level.n11 sample studies of forecasts can be instructive, it their limitations must be recognized.Compilation of consistent forecast records extending over longer periods of tine is necessary to establish a reasonably reliable base for assessments of forecasting behavior and.performance.Thus the historical record of post-World War II forecasts assembled in the 1960's by the NB is here extended and updated.The end-of-year predictions of annual percentage changes in GNP earn good marks for overall accuracy when judged according to realistic rather than ideal standards.Moreover, they are found to have improved significantly in the period since the early l96O s compared with the previous years after World War II.The corresponding predictions for GNP in constant dollars (real growth) and the GNP implicit price index (inflation) are considerably poorer.The former suffer fran large turning point errors, the latter from large underestimation errors.Indeed, forecasts of inflation are not much better than projections of the most recently observed inflation rates, and they lag behind the actual rates much like such projections.But the errors in forecasts of real growth are negatively correlated with the errors in forecasts of inflation, which helped to make the nominal GNP predictions more accurate.Abstract -2 - Forecasts for the year as a whole can be satisfactory when based on a good record for the first two quarters; they tend.to be more accurate than forecasts with longer spans.An exaiination of the recent multiperiod predictions fr veU-kiown econometric models and.business outlook surveys shov that the errors for real growth and inflation cumulated rapid.lybeyond.the spans.of 2 to 1 cpiarters.Previous studies have shon the cumulation to be as a rule less than proportional to the increase in the span, but in the period of recession and recovery 1973-75 the build-up of errors was much greater.Again the nominal GNP forecasts benefitted from offsetting errors as the rise in prices was heavily underestimated and the downturn in real activity was missed.Forecasters were generally unprepared for the concurrence of accelerating inflation and slowing, then declining output rates: they optimistically (and.probably also from a lingering faith in a simple Phillips trade-off) kept anticipating less inflation and more growth.At the present time, the predictive value of detailed multiperiod forecasts reaching out further than a few quarters ahead must be rather heavily discounted.No doubt, in periods less turbulent than the recent past the longer forecasts can be considerably more accurate, but thiz fairweather argument is not very persuasive or helpful.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 34 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.