A new analysis from the Federal Reserve Bank of Chicago reveals that the University of Michigan's Index of Consumer Sentiment—one of the most closely watched economic indicators in the U.S.—is currently underestimating consumer confidence by 25 to 30 index points. The finding comes from a Chicago Fed Letter published in 2026 that introduces the Composite Consumer Sentiment Index (C-CSI), a daily measure designed to correct for biases in traditional sentiment surveys. The report shows that the link between consumer sentiment measures and actual consumer spending has sharply weakened in recent years, with the Michigan index now showing a near-zero correlation with annual real personal consumption expenditures growth after averaging 0.69 before 2020.

Before 2020, both the University of Michigan Index of Consumer Sentiment and the Conference Board's Consumer Confidence Index showed stable and sizable correlations with annual growth in real personal consumption expenditures—0.69 and 0.6 on average, respectively. Since 2020, those correlations have declined sharply, with the Michigan index dropping to near zero currently and the Conference Board index falling to its lowest level in 40 years. The C-CSI, by contrast, maintains a consistently higher correlation with consumer spending growth since 2018 and shows a substantially higher correlation over the past year compared to both traditional indexes. The new composite index is constructed using four daily sentiment measures—Morning Consult's Index of Consumer Sentiment, the Rasmussen Consumer Index, the Federal Reserve Bank of San Francisco's Daily News Sentiment Index, and the Goldman Sachs Social Media Economic Sentiment Index—along with the Conference Board's monthly index.

The report finds that measurement concerns with the Michigan index existed for some time but were roughly zero on average before 2024. According to the authors, the Michigan survey's transition to a completely web-based platform in April 2024 marked a turning point, after which estimated measurement errors "have grown consistently more negative." The analysis decomposes the measurement error into two sources: survey methodology issues, which contributed 0 to 15 index points of upward bias before the online transition and then 0 to 20 points of downward bias afterward, and what the report calls the "vibes" story, which consistently contributed 5 to 10 index points of negative bias throughout the sample period. The report notes that "it seems unlikely that the method of survey collection or the survey questions on their own are primarily to blame" since nearly all other sentiment indicators in the model also use online methods, pointing instead to "a more complex explanation" that's idiosyncratic to the Michigan index itself.

The breakdown in traditional sentiment measures has real economic consequences. The report's forecasting exercise shows that using the Michigan index predicts annual real personal consumption expenditures growth of 1.3% for October 2026, while a model using the C-CSI forecasts 2.0%—comparable to the current 12-month growth rate of 2.1% through April 2026. The authors explain that consumer sentiment surveys don't operate in a vacuum: "Responses are just as likely to be driven by prevailing economic and financial conditions as they are to potentially drive them." The report suggests the Michigan index may have ended up with "a suboptimal mix" of survey collection methods and topics covered after its online transition, creating an interaction effect that produces inconsistent measurements over time. While traditional sentiment measures have struggled in forecasting, with both the Michigan index and C-CSI models underperforming a benchmark model without sentiment indicators for much of the past few years, the C-CSI model has been trending upward and even outperformed the benchmark since late 2025.

The report concludes that the C-CSI "correlates more strongly with one-year-ahead annual PCE growth than traditional consumer sentiment measures and can be used to more accurately predict future consumer spending," though it cautions that "caution should be exercised when doing so given the inconsistency of the recent forecasting performance of consumer sentiment measures more generally." For economists and policymakers who've relied on the Michigan index for more than 40 years, the message is clear: traditional sentiment surveys need serious recalibration, and composite measures that blend multiple data sources may offer a more reliable window into what consumers are actually thinking—and spending.