Consumer spending is the biggest piece of GDP, but GDP arrives once a quarter and months late. Nobody waits for it. Two reports land every month claiming to track the same thing. One counts dollars that changed hands. The other asks people how they feel.
Why One Monthly Report Gets So Much Attention
Household spending is roughly two thirds of US GDP. Nothing else comes close. If consumers pull back a few percent, the whole number bends with them.
The retail sales report is the fastest look. The Census Bureau publishes it about nine or ten business days after a month ends, from a survey of roughly 4,800 companies called the Advance Monthly Retail Trade Survey. By mid-August you have a picture of July.
It does not cover everything you buy. The survey counts retail stores, online sellers, restaurants and bars. It leaves out almost all other services: rent, doctor visits, tuition, insurance, plane tickets. Those are most of what households spend, and a separate report on personal consumption measures them. Retail sales covers roughly a third of the total.
Core Sales and the Control Group
The headline moves for reasons that say nothing about the average household, so analysts publish stripped-down versions. Core retail sales usually means sales excluding motor vehicles and parts, and sometimes gas stations come out too. Cars are expensive and lumpy. One strong month at auto dealers can swing the headline while nothing else changed.
The control group drops four categories:
Auto dealers, for the reason above.
Gas stations, because those sales track the price per gallon, not how much anyone drove.
Building material stores, because that spending is mostly construction, which GDP counts as housing investment.
Food services, because a restaurant meal is a service and GDP picks those up elsewhere.
What survives is close to the goods spending that feeds the GDP calculation, which is why traders read the control group first.
The Headline Counts Dollars, Not Items
Retail sales are reported in current dollars. The Census Bureau adjusts for seasons, holidays and trading days, but not for prices at all.
Picture a month where gas stations sell the same gallons as the month before and the pump price rises 5 percent. Gas station sales rise 5 percent in the report. Nobody bought an extra gallon. Run it backwards when fuel gets cheaper and the report shows a decline that is really a discount.
So a rising headline can mean people bought more, or that the same cart cost more. Telling them apart means setting the sales change against CPI for the same goods.
The First Print Gets Revised
The advance report is built from a subsample. A month later the fuller Monthly Retail Trade Survey covers more firms and the number changes. Once a year the series is benchmarked against a larger annual survey.
The Census Bureau puts the average revision to that advance monthly percent change, ignoring direction, at about two tenths of a percentage point. Set that against a typical monthly change of a few tenths. A report that beats the forecast by a tenth of a point is inside the measurement error.
What the Confidence Surveys Ask
The Conference Board publishes its Consumer Confidence Index on the last Tuesday of each month, drawn from an online panel. The scale sets 1985 equal to 100, so a level only means something against its own history. It splits in two: a Present Situation Index for how people rate business and job conditions now, and an Expectations Index for what they expect six months out. Its questions lean hard on employment, including whether jobs seem plentiful or hard to get.
The University of Michigan publishes the Index of Consumer Sentiment, a preliminary reading mid-month and a final one at month end. Its base period is the first quarter of 1966. Five core questions build it, covering personal finances, business conditions, and whether now is a good time to buy big household items, and they split the same way. Michigan also reports what households expect inflation to be a year out and five to ten years out, figures the Federal Reserve watches closely.
Current conditions track what people can see, like prices and whether their employer is hiring. Expectations move on news and politics, and they swing much harder.
Why Sentiment and Spending Drift Apart
Here is the honest part. Confidence surveys have often failed to predict spending.
Since 2021, Americans have told surveys the economy was bad while retail sales kept climbing. People remember the price level, not the inflation rate, so groceries that stopped rising still feel expensive against 2019. Answers have also grown more partisan, moving with which party holds the White House rather than with anyone's budget. And Michigan moved from telephone to web interviewing in 2024, where people answer more negatively.
The reason is simple. A survey answer costs nothing. A purchase costs money, and it shows up in the data whether or not the buyer felt good about it. That is why a soft survey is a weaker signal than soft spending.
Analysts still watch it for good reasons. The Conference Board's Expectations Index is one component of its Leading Economic Index, and Michigan's inflation expectations feed how the Fed judges whether high prices are getting locked in. Treat a weak confidence reading as a question to go check, not as an answer.
How to Read a Release Without Getting Fooled
Compare to the expectation, not to zero. Markets have already priced in a forecast, so a 0.4 percent gain against a 0.6 percent forecast disappoints even though sales rose. Every economic calendar publishes the consensus beforehand.
Read the revision to the prior month. Each release carries two numbers: the new month, and a correction to the last one. A strong new month paired with a big downward revision can leave total spending lower than you thought. Headlines rarely mention it.
Prefer the trend to the print. One month of retail sales carries weather, holiday timing, a single product launch and sampling noise all at once. A three-month average of the control group tells you more about where consumers really are, and it is the reading that still looks right a year later.
