Guide

University of Michigan consumer sentiment explained

Harbor Retail ran discretionary promotions off the headline University of Michigan consumer sentiment print alone. When the preliminary March 2025 reading jumped 8 points on a gasoline-price dip, marketing launched a “spring confidence” sale across patio and outdoor SKUs — then watched same-store sales miss plan by 6% as the final release revised sentiment down 5 points and the Index of Consumer Expectations stayed flat. Buyers had felt better about today but not about six months ahead; the headline masked a split that retail sales confirmed two weeks later. The desk rebuilt around Michigan's component structure, not its top-line number, and cut false promo triggers from 17 to 10 in twelve months.

The Surveys of Consumers, conducted by the University of Michigan since 1946, is one of two major U.S. household confidence gauges (alongside the Conference Board's consumer confidence index). It publishes twice each month, separates current conditions from forward expectations, and uniquely reports median one-year and five-year inflation expectations that Fed officials cite in FOMC minutes. This guide explains how the Michigan index is built, what preliminary vs final means, which subcomponents lead spending, how inflation expectations interact with PCE and wage growth, the Harbor Retail refactor, a technique decision table, pitfalls, and a production checklist.

What the Michigan survey measures

Michigan polls roughly 600 U.S. households by telephone each month on personal finances, business conditions, and buying plans. Unlike hard spending data from the Census Bureau or BEA, sentiment is subjective and leading: households answer how they feel today and what they expect over the next year before those attitudes show up in GDP consumption. The survey is a probability sample designed to represent the continental U.S. adult population; results are seasonally adjusted and indexed to 1966 Q1 = 100.

A headline reading of 78 means sentiment is 22% below the 1966 baseline quarter, not that 78% of Americans are pessimistic. Month-over-month moves of 2–4 index points are common noise; sustained 6+ point trends across both preliminary and final releases usually precede shifts in durable-goods and big-ticket categories by one to three months.

Two headline indexes

Index of Consumer Sentiment — the full survey composite covering both current conditions and expectations.
Index of Consumer Expectations — forward-looking questions only; historically more correlated with recessions and the Sahm rule than the headline alone.

Michigan also publishes an Index of Current Economic Conditions (sometimes called Present Situation) built from questions about personal finances and whether now is a good time to buy major household goods. When current conditions rise but expectations fall, consumers often describe a “last good month” before they pull back — exactly the pattern Harbor misread.

Index construction and subcomponents

Each survey question receives a relative score: the percent giving favorable replies minus the percent giving unfavorable replies, plus 100. Component scores are weighted and averaged into the headline indexes. Michigan publishes several subcomponents that matter more for forecasting than the top line:

Personal finances — whether households are better or worse off than a year ago and expected finances over the next year; ties to personal income and real wage growth.
Business conditions — expected unemployment and general business conditions over the next year; leads labor-market turning points.
Buying conditions for durables — whether now is a good time to buy cars, homes, and household durables; most directly linked to auto and appliance sales.
Inflation expectations — median expected change in prices over the next year and next five to ten years; Fed officials watch for de-anchoring above the 2% target band.

The labor market differential (percent saying jobs are plentiful minus percent saying jobs are hard to get) often turns before the headline sentiment index and correlates with payrolls momentum with a short lag. Buying-conditions indexes for vehicles and houses lead hard data on auto sales and housing starts by roughly one to two months when moves are sustained across both preliminary and final releases.

Preliminary vs final release

Michigan publishes twice per month:

Preliminary (mid-month) — typically around the 10th–15th, based on roughly the first half of monthly interviews. Markets react immediately because it is often the first household mood read after payrolls.
Final (end-month) — typically the last Friday of the month, incorporating the full sample. Revisions between preliminary and final can be large when late-month news (gas prices, earnings reports, geopolitical shocks) shifts respondent answers.

Trading desks treat preliminary as a flash estimate and final as the durable print. Harbor Retail now logs both but acts on neither alone: promotional rules require the final expectations index to confirm the preliminary direction, or they wait for the next month's preliminary if the revision is larger than 4 index points.

Add both dates to your economic calendar. The preliminary often lands the same week as CPI; the final often lands near PCE and income data, which helps separate sentiment-driven narratives from confirmed spending flows.

Inflation expectations: why the Fed watches Michigan

Michigan is the only major monthly U.S. household survey that publishes median one-year and five-year inflation expectations in the same release as sentiment. Conference Board confidence does not include comparable inflation series; the New York Fed Survey of Consumer Expectations is separate and less historically continuous.

Fed policymakers reference Michigan five-year expectations when debating whether inflation psychology is anchored near 2%. A sustained rise in one-year expectations above 3.5% with flat five-year readings often reflects energy or food shocks rather than de-anchoring; when both horizons move together, bond markets usually reprice rate-cut timing and real yields adjust before CPI confirms the shift.

Compare Michigan inflation expectations to realized PCE and wage growth from average hourly earnings. When households expect inflation above wage growth for several months, buying conditions for durables typically soften even if the headline sentiment index holds up on strong current-condition answers.

Michigan vs Conference Board consumer confidence

Both surveys ask households about jobs and buying plans, but methodology and timing differ enough that they diverge for months at a time:

Sample and method — Michigan uses a rotating panel with telephone interviews; Conference Board uses a new mail and online sample each month. Panel effects make Michigan smoother; Conference Board can spike on one-off news.
Baseline year — Michigan indexes to 1966 Q1; Conference Board to 1985. Levels are not comparable across surveys, only direction and relative extremes.
Inflation expectations — Michigan only. If your macro model needs household inflation psychology, Michigan is the primary source.
Release cadence — Michigan preliminary is mid-month; Conference Board is typically the last Tuesday. Michigan final and Conference Board often land the same week.

Harbor Retail now requires both surveys' expectations components to improve before national discretionary promos, not just Michigan headline or Conference Board alone. Divergence between the two is logged as “mixed household read” and triggers category-level tests instead of chain-wide sales.

Harbor Retail sentiment desk refactor

The legacy rule fired promotions when Michigan headline sentiment rose month-over-month, regardless of component split or preliminary vs final revision risk. That produced three costly errors in eighteen months: patio goods on a gasoline-driven preliminary spike, electronics on a current-conditions bounce while expectations fell, and home goods during a month when five-year inflation expectations jumped 0.4 percentage points. The refactor:

Expectations gate — national promos require Index of Consumer Expectations up month-over-month on the final release, not preliminary alone.
Buying-conditions filter — durables promos require the vehicle or household-goods buying-conditions subindex to confirm, mapped to Harbor SKU categories.
Inflation ceiling — no broad discretionary promos when one-year inflation expectations rise more than 0.3 pp month-over-month unless five-year expectations are unchanged or lower.
Hard spending cross-check — control-group retail sales three-month trend must not be negative before expanding promo depth.
Conference Board confirmation — Expectations subindex from both surveys must agree in direction for chain-wide campaigns; single-survey moves get A/B tests only.

False promo triggers fell from 17 to 10 per year; same-store sales hit rate on sentiment-driven campaigns improved from 58% to 74%. The desk still publishes Michigan headline in the morning note but operational rules run on expectations, buying conditions, and inflation expectations.

Technique decision table

Question	Best source	Weaker substitute
Household mood and recession lead	Michigan Index of Consumer Expectations (final)	Headline sentiment alone
Household inflation psychology	Michigan 1-year and 5-year median expectations	Conference Board confidence
Big-ticket purchase intent	Michigan buying conditions for durables	Headline sentiment
Confirmed consumer spending	Retail sales control group	Any sentiment index
Labor market household view	Michigan labor market differential	Conference Board jobs-plentiful spread
Flash household read mid-month	Michigan preliminary sentiment	Conference Board (end-month only)
Durable vs Conference Board comparison	Both expectations indexes, same month	Either headline alone
Real purchasing power context	Personal income and PCE deflator	Nominal sentiment level

Common pitfalls

Trading preliminary as final — revisions of 4+ points are common; operational rules should wait for final or hedge preliminary exposure.
Ignoring expectations vs current conditions split — rising present situation with falling expectations often precedes spending slowdowns.
Comparing Michigan levels to Conference Board levels — different baselines (1966 vs 1985); compare direction and extremes, not absolute numbers.
Equating sentiment with spending — confidence can rise while real consumption flatlines if income or credit constraints bind.
Overweighting one-month inflation-expectation spikes — energy shocks move one-year expectations faster than five-year; read both horizons.
Skipping buying-conditions subindexes — headline sentiment can improve while vehicle-buying conditions deteriorate.
Missing the panel effect — Michigan's rotating panel smooths month-to-month noise but can lag sudden shocks that Conference Board captures faster.
Using sentiment for services-heavy models without ISM Services — Michigan durables buying conditions miss services spending; pair with ISM Services PMI for breadth.

Production checklist

Log Michigan preliminary and final headline, expectations, and current conditions each month.
Record median one-year and five-year inflation expectations with month-over-month deltas.
Chart labor market differential and durables buying conditions alongside headline.
Flag preliminary-to-final revisions larger than 3 index points in the release note.
Compare Michigan expectations direction to Conference Board expectations same month.
Cross-check control-group retail sales before acting on sentiment alone.
Pair sentiment moves with personal income and real wage trends from BEA.
Map category promos to relevant Michigan buying-conditions subindex, not headline.
Archive University of Michigan Surveys of Consumers press release PDF each release.
Add preliminary and final dates to the shared economic calendar with revision notes.

Key takeaways

Michigan consumer sentiment indexes household mood to 1966 Q1 = 100, with separate current conditions and expectations composites.
Preliminary mid-month and final end-month releases can diverge materially; treat preliminary as flash, final as operational.
Median one-year and five-year inflation expectations are Michigan's unique Fed-watched contribution among major confidence surveys.
Harbor Retail cut false promo triggers 41% by gating on final expectations, buying conditions, and inflation expectations instead of headline sentiment.
Pair Michigan with Conference Board expectations, retail sales, and income data — sentiment leads but does not replace hard spending measures.