Academic journal article Economic Review (Kansas City, MO)

How Reliable Are Recession Prediction Models?

Academic journal article Economic Review (Kansas City, MO)

How Reliable Are Recession Prediction Models?

Article excerpt

The U.S. economy continues to advance briskly, defying forecasts of more moderate growth. Beginning in March 1991, the current expansion has become the longest peace-time expansion on record and is less than a year away from becoming the longest in U.S. history. To the surprise of some observers, economic growth has been particularly robust late in the expansion. In fact, over the last three years growth has averaged 4 percent annually, and indicators of growth for the first half of 1999 show no signs of significant slowing.

Despite these positive signs, few analysts believe the expansion can go on forever. As the expansion continues to age, economists will increasingly be called on to predict the next recession. Recession prediction models may help them gauge the likelihood of imminent recession.

This article examines the reliability of five popular recession prediction models. The first section reviews each model's theoretical strengths and weaknesses in predicting recessions. The second section evaluates how well these models have given advance warning of past recessions. Performance is measured both with recently released data as well as the data originally available to analysts. The article concludes that these models have demonstrated some ability in the past to predict recessions. When judiciously interpreted, the models can help resolve uncertainty about the possibility of future recession.

I. FIVE RECESSION PREDICTION MODELS

While a recession is commonly understood to be a widespread and prolonged decline in economic activity, using a model to predict recessions requires a more precise definition. One popular definition of recession is a consecutive 2-quarter decline in GDP.(1) The appeal of this definition stems from the fact that GDP is one of the broadest measures of economic activity. It is hard to imagine a widespread decline in economic activity without a decline in GDP.

Another definition of recession comes from the Business Cycle Dating Committee of the National Bureau of Economic Research (NBER), which officially dates the beginnings and ends of U.S. recessions. The Dating Committee defines a recession as a broad decline in aggregate economic activity (which is measured as a common movement in output, income, employment, and trade), usually lasting from six months to a year, and marked by widespread contractions in many sectors of the economy. The NBER explicitly shuns the GDP definition because it considers GDP to be too narrow a measure of economic activity to reliably date recessions.(2) Nevertheless, declines in GDP are closely correlated to recession periods as denoted by the NBER (Chart 1).

[CHART 1 OMITTED]

Many analysts use specialized models to predict the onset of recession. These models are useful primarily because the behavior of the economy during periods of transition between expansion and recession is fundamentally different than when recession is not imminent (Hymans). In other words, special times such as turning points in the business cycle call for special models. Moreover, empirical evidence lends support to this view. During expansions when recessions are not a threat, forecasters tend to rely on large-scale econometric models to chart the future of the economy. It is well known, however, that around taming points in economic activity, these models can produce large forecasting errors (Braun and Zarnowitz). As a consequence, when an expansion might be nearing an end, small-scale forecasting models can play a critical, if not a dominant, role in predicting future economic activity (Diebold and Rudebusch 1989).(3) This section describes five of these popular business cycle models: simple rules of thumb using the Conference Board's composite index of leading indicators (CLI), Neftci's probability model of imminent recession using the CLI, a regression-based model of the probability of recession called a Probit model, a GDP forecasting model, and a recession prediction model recently proposed by Stock and Watson. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.