When should we care about consumer sentiment? Evidence from linear and Markov-switching models

Using monthly data over 1978:01-1996:12 we examine the usefulness of the Index of Consumer Sentiment (ICS) in predicting aggregate consumption expenditure and its components. Bivariate causality tests in a linear model show the indicator value of ICS in predicting durables consumption. We propose Markov-switching models that differentiate between regimes of high and low consumption volatility where the former regime is characterized by a strong and markedly different relationship between consumption and ICS. We conclude that the benefits from including consumer sentiment in models of consumption are largest in periods when conflicting economic and sociopolitical news costs high overall uncertainty and wide swings in near term expectations of personal income. These high volatility periods are not always related to recessions. Interestingly, ICS loses its predictive power in the presence of forward-looking financial variables such as interest rate spreads and stock returns in linear as well as Markov-switching models.