Retail Sales is the U.S. Census Bureau's monthly estimate of the dollar value of goods sold (plus food-service receipts) by retail establishments — the single most timely read on the American consumer, who drives roughly two-thirds of U.S. GDP. Its core tension: it is the fastest, broadest consumer-spending gauge available, yet it is nominal (not inflation-adjusted), goods-heavy, survey-based, and noisy month-to-month — so a number that looks strong can dissolve once you strip out price effects, volatile categories, and revisions.

How it's calculated / formed

The headline figure comes from the Advance Monthly Retail Trade Survey (MARTS), released about 9–10 business days after the reference month ends (e.g. mid-June for May), typically at 8:30 a.m. ET (Census). MARTS samples roughly 4,800 firms drawn from the larger ~13,000-firm Monthly Retail Trade Survey (MRTS) and estimates current-month sales via a link-relative method — a current-to-prior-month sales ratio applied to the prior month's level — rather than imputing most non-responders (Census methodology). The advance estimate is revised the following month by the fuller MRTS, then again later, so early prints are provisional.

Figures are seasonally adjusted and adjusted for holiday/trading-day differences, but not adjusted for price changes (Census). The standard layers analysts watch:

• Headline: total retail and food services sales, MoM %.
• Ex-autos: removes motor-vehicle/parts dealers (auto sales are large and lumpy).
• Ex-autos and gas: also removes gasoline-station receipts, which swing with fuel prices, not real demand.
• "Control group" (a.k.a. core control): excludes auto dealers, gas stations, building-materials retailers, and food-services, plus a few minor categories (CME Group; Advisor Perspectives). This is the smoothest series and the one that feeds the goods-consumption component of GDP and the BEA's Personal Consumption Expenditures (PCE) goods estimate.

How it's used in practice

The market reaction is driven by the surprise versus consensus, not the level. Traders watch all four layers, but the control group carries the most analytical weight because it filters out the noisiest pieces and is what economists plug into GDP nowcasts (e.g. Atlanta Fed GDPNow) and PCE projections (CME Group). A common scenario: a strong headline alongside a soft control group (because gas-price inflation flattered the top line) gets faded once analysts read the internals.

Transmission to assets runs through the growth-vs-Fed-policy lens. A hot report signals consumer resilience: typically supportive for cyclical equities and the dollar, but if it stokes inflation/"higher-for-longer" rate fears it can pressure long-duration assets and lift Treasury yields. A weak report cuts the other way — recession-fear selling, or "bad news is good news" rate-cut hopes — with the reaction depending entirely on the prevailing macro regime (inflation-fighting vs. growth-scare). Sector desks also read the category mix for read-throughs into retailer and discretionary-stock earnings (LSEG/Lipper).

Adoption, debate & evidence

Retail Sales is universally tracked and routinely described as one of the most market-moving monthly releases (CME Group). Its adoption is not contested — but its interpretation is, on several well-documented grounds:

• Nominal, not real. Because the data isn't price-adjusted, rising sales can reflect inflation rather than more buying. The post-2021 period is the textbook case: pent-up demand and stimulus boosted volumes, but higher prices also inflated the dollar figures (MCA Merchandising). FRED publishes a deflated companion series, Real Retail Sales (RRSFS), precisely to address this (FRED).
• Goods-heavy. The only services category captured is food services (restaurants/bars); the rest of the ~70%-of-spending services economy is excluded (Federal Reserve FEDS Note). It therefore measures part of the consumer, not all of it.
• Revision risk. The advance estimate rests on a subsample and is regularly revised, sometimes materially, so single-print conclusions are fragile (FRED RSAFS notes).
• No demographic decomposition. Because firms report sales (not consumers reporting spending), the data can't be split by household income — a Fed FEDS Note built an alternative dataset specifically to overcome this (Federal Reserve, 2024).

The honest summary: as a directional, real-time consumer read it is genuinely valuable and a legitimate GDP/PCE input; as a precise statement of consumer health it is frequently over-read, especially when commentators cite the headline without adjusting for prices or stripping volatile categories.

Strengths & limitations

Works best as a timely, surprise-driven macro catalyst and as a GDP-tracking input via the control group. It is hard data (actual sales), not a survey of sentiment, which gives it credibility over soft indicators. Fails / misleads when read at the headline level in isolation: gas-price swings, auto lumpiness, inflation, and revisions can all flip the story. The #1 misuse is treating a strong nominal headline as proof of a strong real consumer without checking the control group and a deflated series.

Sources

• U.S. Census Bureau — About the Advance Monthly Retail Trade Survey (MARTS)
• U.S. Census Bureau — How the Data Are Collected (link-relative method, sample size)
• CME Group — About the Retail Sales Report (control group definition, market significance)
• Advisor Perspectives / dshort — control group composition and volatility vs. headline
• Federal Reserve FEDS Note (2024) — limitations: goods-heavy, no demographic split
• FRED — Advance Real Retail and Food Services Sales (RRSFS), the inflation-adjusted series
• MCA Merchandising — nominal vs. inflation-adjusted caveat

Dispute flags: exact control-group exclusion list varies slightly by source (minor categories like office supplies/mobile homes/tobacco are sometimes listed, sometimes not); the "two-thirds of GDP" consumer share is a commonly cited round figure, not a fixed constant.