Using data from the Michigan Consumer Survey, I explore alternatives for constructing the Index of Consumer Sentiment (ICS) to improve its forecasting power regarding consumption and its components. Questions which seemed to matter in the past are no longer good predictors. For more recent sample periods, expectations of automobile purchases, unemployment, and current economic situations are more important than categories selected previously. An alternative index is constructed accordingly. Applying different techniques suggested in the literature, the new index significantly outperforms the ICS in both in-sample and out-of-sample tests. Furthermore, the new index also produces more accurate results when forecasting recessions.