Four Census Retail Surveys, Four Clocks: Reading the U.S. Retail Data

The U.S. Census Bureau runs four retail trade programs: the Advance Monthly and Monthly Retail Trade Surveys (MARTS and MRTS), the Annual Retail Trade Survey (ARTS), and the Quarterly E-Commerce Report. Together, they produce data on retail economic activity in the United States, according to the Census Bureau's retail page.
The four programs run on different schedules. The two monthly surveys, paired under the MARTS and MRTS labels, are the series most often cited when people discuss consumer demand in near real time. The annual survey, ARTS, covers the same sector over a longer horizon. The Quarterly E-Commerce Report isolates online activity every quarter. Each program answers a slightly different question, so a single headline figure from any one of them shows only part of the picture.
Retail sales track closely with consumer spending, the largest single piece of U.S. economic output. A monthly figure that comes in above or below consensus, the average forecast among economists, can move Treasury yields (the interest rates the government pays on its debt, which anchor many other borrowing costs), shift expectations for the Federal Reserve's policy rate, and reprice consumer-facing stocks within minutes. A borrower's cash flow, a card issuer's loss projections and a retailer's inventory plans all depend on the same underlying demand. Many of the models built around those decisions draw on the Census series.
The practical caution concerns revisions and coverage. Monthly estimates are preliminary and get revised as more survey responses arrive, so the number that moved a market on release day may not match the figure in a later data vintage (a dated snapshot of the data as it stood at a particular time). The annual and quarterly programs are the ones to check when testing whether the monthly picture holds. Analysts building trend or seasonal models should pull the revision history for each series rather than relying on first prints, and should reconcile the e-commerce series against the total before drawing conclusions about shifts between online and store sales. Reading the series against one another, rather than in isolation, is where the useful signal tends to sit.
In my view, forecasts deserve the same scrutiny as any other projection. The consensus number is an expectation, and the released figure is a measurement. Treating the two as interchangeable is the most common error in reading retail data, and the one most worth avoiding.


