Consumer confidence is a survey-based measure of how households assess economic conditions and their own prospects. Quant researchers use it as a possible signal about future consumption and the business cycle. It records reported attitudes, not purchases: households can feel more optimistic while spending less because income, credit or prices constrain them.
There is no single universal consumer-confidence scale. Before using a reading, identify the publisher, country, questions, reference period and index construction.
What does the index actually combine?
In the United States, The Conference Board publishes a monthly Consumer Confidence Index. Its methodology separates two components:
| Component | What respondents assess |
|---|---|
| Present Situation Index | Current business and employment conditions |
| Expectations Index | Business conditions, employment and family income six months ahead |
The headline averages the five underlying question indexes; the present component uses two and the expectations component three. Consequently, an improving headline does not establish that expectations improved. Read the component releases. The Conference Board's technical note describes the construction; its release page identifies the series and survey cutoff.
For each question, the technical note uses seasonally adjusted positive and negative response proportions to form a positive share of their sum. That relative value is indexed to its average in 1985. A reading of 100 is a historical benchmark, not “everyone is neutral,” “100% confident,” or an economic growth rate.
Other publishers make different choices. The University of Michigan's US Index of Consumer Sentiment starts from favourable minus unfavourable response percentages, adds 100, then aggregates with its own 1966 base and an adjustment constant. Its expectations questions also have different horizons. It is a separate series, not a substitute observation on The Conference Board's scale. See Michigan's index calculations.
The OECD's standardised CCI is an amplitude-adjusted index with long-term average 100. That benchmark differs from a fixed base year's level. Equal readings from different constructions do not establish equal household optimism; rescaling alone cannot harmonise different questions or populations.
Eight points is not eight percent spending growth
Suppose a fictional monthly survey index, using an unchanged method and a consistent release vintage, rises from 80 to 88. These are index units, neither currency amounts nor annualised growth rates.
Its relative numerical change is:
Both calculations describe the index. Neither says consumption grew by 8% or 10%; 88 also does not mean 88% of respondents are confident. Report “up 8 index points” and name the survey. A relative change in an index can depend on its construction and origin, so avoid translating it into an amount of economic optimism.
Use the release that was available
A survey's reference month, interview window and publication timestamp are different pieces of information. A late-month response is not information a model could have used earlier that month. Keep the release timestamp and time zone, preliminary/final flag, index definition and vintage alongside each observation.
The Conference Board's technical note describes preliminary estimates followed by final estimates published with the next month's release. Michigan separately schedules preliminary and final releases, with different interview cutoffs; see its release calendar. Preserve both versions when available. Replacing a preliminary value with a later final value in a historical trading signal introduces look-ahead bias.
For an announcement study, define the surprise against a forecast recorded before publication, using the same named index and release. For a consumption forecast, test whether the survey adds information beyond already available income, employment and spending data. Correlation alone establishes neither causation nor a profitable trading rule. Price responses can depend on what the market expected and what else was announced.
Confidence can help describe a changing outlook. It does not mechanically date a recession, and a threshold cited for one publisher's expectations subindex cannot be applied to every headline confidence series.
For an applied example of release-vintage discipline, the Real-time macro forecasting capstone compares forecasts using information available at each origin. Membership required. Its target is US industrial production; it supports the timing workflow, rather than teaching confidence-survey construction.
Sources
- The Conference Board — Consumer Confidence Survey Technical Note, May 2021: sample coverage, question indexes, seasonal adjustment and base-period construction, pp. 1–2.
- The Conference Board — US Consumer Confidence: publisher's monthly release and component definitions.
- University of Michigan — Index Calculations: sentiment, current-conditions and expectations construction.
- University of Michigan — 2026 Release Dates: preliminary/final releases and interview cutoffs.
- OECD — Consumer confidence index: standardisation and the long-term-average benchmark.