Economic forecasting at high‐frequency intervals

Abstract

Abstract Forecasting on the basis of the daily flow of monthly or more frequent statistical reports on the economy can enhance the predictive accuracy of quarterly structural models. The high degree of serial correlation in economic data can be used advantageously in quarterly forecasting for a horizon as long as 6 months—perhaps somewhat longer. The model used for high‐frequency (weekly) forecasting of the US economy has a national accounting structure and tries to follow the choice of indicators that are used in preparing early estimates of national income and product accounts (NIPA). Estimates are separately generated for the income side and the product side of NIPA. At the level of GDP and closely related aggregates a third prediction is also generated from estimates of the principal components of major monthly indicators. A simple average of three estimates of GDP, together with detail on NIPA components and scores of monthly indicators has been produced every weekend, summarizing the business week's flow of information. This procedure is followed not only for producing a steady stream of high‐frequency forecasts but also for providing adjustment factors that can be used for model recalibration, without judgemental input. The tracking of the US economy is illustrated for the period starting before the invasion of Kuwait until the end of the Gulf War.

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 — 28 citations exist upstream.

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