UMCSENT: University of Michigan Consumer Sentiment Index Monthly Since 1952

The UMCSENT series tracks the University of Michigan Consumer Sentiment Index — the longest continuous survey of American household confidence, running since 1952. UMCSENT captures how consumers perceive present financial conditions and one-year-ahead expectations across spending, income, employment, and prices; its turning points have historically led aggregate consumption shifts by one to two quarters. The series is produced by the University of Michigan and distributed through FRED under a University of Michigan copyright (citation required); Eco3min provides the chart and the analytical layer, and points to FRED as the distribution channel for the data itself.

Dataset: US Consumer Sentiment — University of Michigan

Source: Federal Reserve Bank of St. Louis (FRED). Chart generated and served by FRED.

Macro Takeaway

Dataset Overview

Licensing note: UMCSENT is © Surveys of Consumers, University of Michigan, and is flagged “Copyrighted: Citation Required” on FRED. Eco3min therefore does not redistribute the file; the series remains freely accessible from FRED via the direct link below. At the request of the source, the FRED data is delayed by one month.

Dataset Variables

The FRED CSV contains the following columns. Each row represents one reference month.

Missing values appear as “.” in the FRED CSV.

FRED Direct CSV Access

The monthly series since January 1978 is available from FRED under series code UMCSENT:

https://fred.stlouisfed.org/graph/fredgraph.csv?id=UMCSENT

The quarterly 1952–1977 history is available separately as UMCSENT1:

https://fred.stlouisfed.org/graph/fredgraph.csv?id=UMCSENT1

Both URLs return CSV files usable directly in pandas, R, curl, or any data tool — no download or API key required.

Using the Dataset in Python

import pandas as pd

url = "https://fred.stlouisfed.org/graph/fredgraph.csv?id=UMCSENT"
df = pd.read_csv(url, parse_dates=["observation_date"], na_values=".")

print(f"Latest reading: {df['UMCSENT'].dropna().iloc[-1]:.1f}")
print(f"Historical average: {df['UMCSENT'].mean():.1f}")
print(f"Record low: {df['UMCSENT'].min():.1f} on {df.loc[df['UMCSENT'].idxmin(), 'observation_date'].date()}")

Using the Dataset in R

library(readr)

url <- "https://fred.stlouisfed.org/graph/fredgraph.csv?id=UMCSENT"
df <- read_csv(url, na = ".")

head(df)
summary(df$UMCSENT)

Both examples load the data directly from FRED — no download or API key required.

Methodology

The University of Michigan Consumer Sentiment Index is constructed from a monthly telephone survey of approximately 500 US households, conducted by the Survey Research Center at the University of Michigan. Respondents answer five core questions covering personal financial situation (current and expected), business conditions (one-year and five-year horizons), and durable goods purchasing conditions. Each answer is scored on a relative scale (better, worse, same), and a diffusion measure is computed per question. The composite index is the unweighted average of the five sub-indices, normalized to 100 in the first quarter of 1964.

A preliminary estimate is released mid-month (around the 14th–15th) based on the first ~250 interviews; the final value, incorporating the full sample, is published at the end of the month. Both readings are seasonally adjusted. UMCSENT is not revised after publication, though the underlying micro-data is occasionally re-weighted as Census household composition estimates are updated.

The chart above is generated and served by FRED and always reflects the latest observation FRED distributes.

Data Quality & Provider Notes

UMCSENT is published by the University of Michigan and mirrored on FRED as the public reference. The survey has run continuously since 1952, with quarterly readings through 1977 (series UMCSENT1) and monthly observations from January 1978 onward (series UMCSENT).

• Licensing (important). The series is © Surveys of Consumers, University of Michigan, and carries a “Copyrighted: Citation Required” flag on FRED — the FRED copy is reprinted with the university’s permission. Redistribution of the file itself requires the owner’s permission, which is why Eco3min links to FRED rather than mirroring the data.
• Release latency. The preliminary UMCSENT reading is released around the 14th–15th of each month at 10:00 a.m. ET; the final value follows around the 26th–28th. At the request of the source, the free FRED version is delayed by one month; earlier access requires the University of Michigan’s subscription products.
• Revisions policy. The headline UMCSENT value is not revised once the final monthly reading is published. Survey methodology has remained stable, with periodic adjustments to question wording (notably 1980 and 2005) documented in the University of Michigan release notes.
• Alternative sources. Bloomberg (CONSSENT), Refinitiv/LSEG, and Haver Analytics carry the same series with earlier access under commercial license. The University of Michigan subscription product provides the full micro-data; the delayed FRED version is sufficient for time-series analysis.
• Known gaps. Quarterly data only from 1952 through 1977 (UMCSENT1); monthly observations from January 1978 onward (UMCSENT). The pre-1978 quarterly readings are not directly comparable to monthly values without resampling. Methodology shifted to telephone-only collection in 1978.

When working with UMCSENT in a research workflow, distinguish between preliminary and final readings — backtests run on real-time data should use preliminary values for the first two weeks of each month, then switch to the final reading after release.

Common Pitfalls When Using UMCSENT

UMCSENT is one of the most-cited sentiment measures, but several recurring misuses distort the signal.

1. Preliminary versus final readings. UMCSENT publishes a preliminary estimate mid-month and a final value at month-end. The two often differ by 1–2 points, occasionally more during volatile periods. A common error is comparing a preliminary reading to a prior month’s final value, generating spurious month-over-month moves.
2. Levels are not comparable across decades without context. A UMCSENT reading of 75 in 1985 carried different macro implications than 75 in 2025 — labor force participation, household debt composition, and the share of wealth held in financial assets have shifted materially. Comparing absolute UMCSENT levels across regimes without these adjustments tends to mislead.
3. Confidence versus consumption. UMCSENT measures stated expectations, not realized behavior. The 2022–2023 episode demonstrated that sentiment and spending can decouple for extended periods when households finance consumption out of accumulated savings or credit. Treating UMCSENT as a near-term consumption forecast in isolation produced poor calls during this period.
4. UMCSENT versus Conference Board Consumer Confidence. The two measures often diverge by 5–10 points and weight different questions — UMCSENT leans toward financial conditions and expectations, while the Conference Board weights labor-market perceptions more heavily. Citing them interchangeably as “consumer confidence” is a common analytical shortcut that loses signal.

Historical Regimes

1952–1972 — Survey establishment and postwar expansion. The survey was quarterly through 1977 and averaged around 90, reflecting steady household optimism in a low-unemployment, low-inflation environment. The first major dip occurred in 1970 (74), tracking the Nixon-era recession.

1973–1982 — Stagflation collapse. UMCSENT fell to 51.7 in May 1980 — among the lowest readings ever recorded at the time — as the combination of oil shocks, double-digit inflation, and the Volcker tightening crushed expectations. The 1974–75 recession had already pushed UMCSENT below 60.

1983–1999 — Reagan-Clinton expansion. UMCSENT recovered to a sustained range of 85–110, peaking at 112 in early 2000 during the tech boom. The 1991 recession produced a moderate dip to roughly 67, well above the 1980 trough.

2000–2007 — Erosion despite expansion. UMCSENT trended down from the 2000 peak even as GDP grew, reflecting rising household debt and concerns about real purchasing power. Hurricane Katrina (2005) and Iraq War sentiment effects compounded the trend.

2008–2009 — Global Financial Crisis. UMCSENT collapsed to 55.3 in November 2008, the lowest reading in nearly 30 years at the time. Recovery was slower and shallower than in prior cycles, plateauing in the high 70s for most of 2010–2013.

2014–2019 — Late-cycle high. UMCSENT averaged near 95, reaching 101.4 in March 2018. The 2018–2019 trade-war episode produced a moderate dip without recession.

2020–2026 — Pandemic, inflation shock, and the all-time low. UMCSENT fell to 71.8 in April 2020 (pandemic shock), partly recovered, then plunged to a series-record 50.0 in June 2022 as headline inflation peaked above 9%. The 2022 trough was a regime change — for the first time in the series’ history, inflation rather than unemployment drove the low. The slow recovery through 2024–2025 has tracked disinflation more closely than labor-market conditions, with the trajectory of US federal debt to GDP serving as a longer-run anchor for household expectations.

Related Macroeconomic Datasets

UMCSENT captures the demand-side, behavioral layer of macroeconomic monitoring. Pairing it with output, activity, and fiscal indicators provides a multi-channel view of cycle dynamics.

Macroeconomic Dataset Hub

This dataset is part of the Eco3min macro-financial data repository.

Explore the Eco3min Dataset Hub

Sources

• University of Michigan, Survey Research Center — Surveys of Consumers (primary source)
• Federal Reserve Bank of St. Louis — FRED series UMCSENT / UMCSENT1

Dataset Reference

Eco3min Research. (2026). US Consumer Sentiment — University of Michigan UMCSENT.<br> https://eco3min.fr/en/us-consumer-sentiment-dataset/