Predicting the directional change in consumer sentiment

Abstract

The Consumer Sentiment Index (CSI) is a widely monitored economic indicator. The index measures consumer expectations, which contain information about potential future changes in consumer spending. Thus, any change in the dynamics of the sentiment index should contain important signals about future consumer behaviour, as well as changes in the real economy and stock returns. In this paper we use Cox’s proportional hazards model to estimate dynamic transition probabilities for a reversal in the CSI. We predict the probability of both positive and negative runs in sentiment surviving in the same state. Lagged changes in consumer sentiment and the frequency of recent reversals of sentiment are found to be significant predictors of a change in state. The model has substantial predictive ability, as measured by both categorical accuracy and a probability scoring rule, the Brier Score (Brier, 1950) out of sample. The predictive accuracy of the model is robust to the impact of the global financial crisis.