A Recession-and-State Forecasting Model

摘要

The state forecasting model (SFM) developed herein predicts the probability that a recession will begin or end within the next quarter. It is also important to forecast the state, as models prone to false alarms are not useful. The model combines parsimony, predictive accuracy, and ease of estimation using readily available data. No regressor is benchmark-revised, and only one is seasonally adjusted. The SFM correctly forecasts, out-of-sample, the binary state in 163 of 164 quarters for an overall proportion of correct forecasts of 0.994. The results demonstrate the discriminating ability, longevity and striking forecasting ability of the model. A benchmark model based on the newly revised Leading Economic Index of the Conference Board has zero marginal predictive power in the presence the SFM.