Nowcasting the Surveys of Consumers

This memo is a summary of the full report Nowcasting the Surveys of Consumers: Using Morning Consult’s Daily Economic Intelligence to Track and Predict the University of Michigan’s Consumer Sentiment Index.

In our connected, ever-changing and fast-paced world, a monthly measure of consumer confidence often isn’t enough for companies to make timely decisions about their business operations. To fill this gap in coverage, we at Morning Consult ask on a daily basis the same five questions that underpin the University of Michigan’s Consumer Sentiment Index. Furthermore, we do so at an unprecedented scale, conducting online surveys of roughly 6,000 U.S. adults daily, nearly 10 times the amount contacted by phone each month for the University of Michigan’s Surveys of Consumers.

Though strongly correlated, the two indexes aren’t identical

As the figure below demonstrates, Morning Consult’s daily read on consumer confidence, our Index of Consumer Sentiment, closely tracks the University of Michigan’s Consumer Sentiment Index on a monthly basis (correlation coefficient = 0.97), but the two are not identical. For this reason, and to aid users of the University of Michigan index, we built a statistical model that translates their index to ours daily, with the former able to predict the latter.

We came up with a new daily tracker for the University of Michigan index …

The statistical model uses the history of the Surveys of Consumers preliminary and final releases, Morning Consult’s daily ICS, the monthly IBD/TIPP Personal Finances Index, the daily S&P 500 index and AAA retail gas prices to generate the daily estimate of the University of Michigan index shown in the figure below. When averaged over the appropriate monthly survey windows, the daily estimate exactly matches the published data from the University of Michigan. And when that data is not yet available, our model can be used to predict it.

… and provide a monthly nowcast for it on the Bloomberg terminal at ECOS

We use the model to “nowcast” the Surveys of Consumers and make these findings available the day before the consumer sentiment index’s release on the Bloomberg terminal at ECOS. Our own out-of-sample backtesting, shown below, suggests that these nowcasts would have outperformed consensus forecasts for the preliminary releases of the University of Michigan index from March 2020 to August 2022 by a statistically significant margin, as well as held their own relative to the final releases of the same consensus forecasts.

Conclusion

The model described here combines Morning Consult’s ICS with other high-frequency data sources to produce a daily tracker and monthly nowcast for the University of Michigan’s Consumer Sentiment Index. The nowcast consistently improves on consensus forecasts that are often relied upon as an alternative. For further information, download the full report.