On the usefulness of macroeconomic forecasts as inputs to forecasting models

Abstract

AbstractA forecasting model for yt based on its relationship to exogenous variables (e.g. x̌t) must use x̌t, the forecast of x̌t. An example is given where commercially available x̌t's are sufficiently inaccurate that a univariate model for yt appears preferable. For a variety of types of models inclusion of an exogenous variable x̌t is shown to worsen the yt forecasts whenever x̌t must itself be forecast by x̌t and MSE (x̌t) > Var (x̌t). Tests with forecasts from a variety of sources indicate that, with a few notable exceptions, MSE (x̌t) > Var (x̌t) is common for macroeconomic forecasts more than a quarter or two ahead.Thus, either: available medium range forecasts for many macroeconomic variables (e.g. the GNP growth rate) are not an improvement over the sample mean (so that such variables are not useful explanatory variables in forecasting models), and/or the suboptimization involved in directly replacing x̌t by x̌t is a luxury that we cannot afford.