How to forecast for recessions and recoveries

Automatic forecasting for production and inventory can lead to large errors when the economy goes into decline, and when it returns to growth. Armstrong and Green describe how forecast errors and their exacerbation of business cycles can be reduced by using the simple rule of not extrapolating a trend when causal forces are not acting to support it.

Widely used methods of automated forecasting for production and inventory control tend to contribute to the severity of recessions. This is because of the assumption that causal forces support the historical trend is violated when recessions occur. We describe an approach to forecasting that should ameliorate the damage caused by business cycles.

In the early 1990s, Fred Collopy and Scott Armstrong published a series of papers in which they showed that nearly all inventory models are fallacious because they assume that the causal forces will support the historical trends. In fact, they found no practical time series where an assumption of supporting causal forces was warranted.

While the assumption of supporting causal forces is unfounded, the world often looks as though trends are supported by causal forces. Consequently, in normal times, standard extrapolation models perform adequately. However, when the historical trends are contrary to the expected causal forces-referred to as "contrary series"-the forecast errors from standard extrapolation models become very large.

To deal with this and related problems, Collopy and Armstrong developed and published a set of 99 rules as the basis of "rule-based forecasting." Although a number of RBF programs have been developed privately, there is no commercial package. While applying RBF without software is onerous, a simple and inexpensive rule can achieve much of the benefit of RBF by reducing errors when forecasting contrary series.

Here is the rule: When a time series is identified as "contrary," do not extrapolate a trend.

Read more…