Retail Sales Firmed in August. The Signal Is Breadth, Not the Bounce.

Key Takeaways

• Breadth is the signal, not the bounce. August’s gain reached nearly the entire report — online, restaurants, electronics, health, and general merchandise all rose — which matters more than the size of the one-month jump.
• The headline overstates the trend. Much of August simply reversed July; smooth the two months together and nominal sales are growing near 4% annualized, right where the three-month trend sits.
• Inflation and gasoline flatter the top line. The 6.0% year-over-year rise falls to about 2.6% in our CPI-deflated proxy, and higher gas prices padded the month’s dollars.
• Goods demand is lopsided. Real durable-goods spending runs about 5.2% above a year ago while real nondurables sit slightly below, so one firm number blends two very different experiences.
• This is not the whole consumer. Retail covers goods plus restaurants, not the services that make up roughly two-thirds of consumption — so read it as resilient goods demand, not an overheating consumer.

The headline was hot. The trend was not.

Americans spent $773.9 billion at retailers and restaurants in August, up 1.24% from July and 6.0% from a year ago. Census rounds those changes to 1.2% and 6.0%.

But the first thing to notice is how much of August simply reversed July.

Sales fell 0.54% in July and rose 1.24% in August. Put the two months together and the average pace is about 0.35% per month, or roughly 4.3% annualized. That is almost exactly where the broader trend sits: nominal retail sales increased at a 4.1% annualized rate over the past three months, while our CPI-deflated measure ran at 3.9%.

So the 15.9% annualized rate implied by August alone is not the number I would carry forward. August more than erased July’s decline — sales finished the month about 0.7% above their June level — but smoothing through the volatility leaves something closer to a moderate 4% pace.

The control group tells a similar story. Its monthly annualized rate swung from roughly −5% in July to +17.6% in August, while its three-month nominal pace finished August at 5.0%. The trend strengthened, but nowhere near as dramatically as the one-month bar suggests.

That is why the bounce itself is less interesting than what was underneath it.

Breadth is the better signal

The August gain was spread across nearly the entire retail report. Online retailers rose 2.6%, restaurants and bars 1.2%, electronics 1.6%, health and personal care 0.9%, and general merchandise 0.7%. Building materials was the only major category in our tracking set to decline.

Breadth does not eliminate sampling error, and Census cautions that comparisons across individual industries have not been tested for statistical significance. But it does reduce the risk that the headline is being driven entirely by one volatile category.

The control group is especially important here. Excluding autos, gasoline, building materials and restaurants, sales rose 1.36% in August. Our CPI-deflated proxy rose 0.96% on the month, was 2.2% above a year earlier and increased at a 4.8% annualized rate over the past three months.

That is a firmer core reading than the headline bounce alone would suggest.

Prices still flatter the nominal numbers

Retail sales are up 6.0% from a year ago. Using headline CPI as a rough deflator, however, our inflation-adjusted proxy is up only 2.6%.

So more than half of the 6.0% increase in retail dollars disappears after a rough adjustment for inflation. The real increase is meaningful, but much smaller than the headline suggests.

Gasoline makes the distinction especially clear this month.

Gas-station sales jumped 3.06% in August, the largest increase among the major retail categories, and were 21% higher than a year ago. Yet gasoline prices themselves rose 3.9% in August and 27.4% from a year earlier.

Those aren’t perfectly matched baskets — gas stations sell more than gasoline — but the direction is clear. Higher prices explain much of the increase in dollars spent at the pump.

Strip gasoline stations out and retail sales still rose 1.08%, so August’s improvement was not simply an energy story. But gasoline did make the headline look stronger than the underlying volume story.

The bigger divide is inside goods spending

The pressure on necessities is still visible.

Grocery-store sales were only 0.5% above last August, while food-at-home prices were 2.2% higher. That implies a simple inflation-adjusted grocery-sales proxy running roughly 1½% below a year ago.

Our retail-to-PCE model finds the same divergence at a broader level. It estimates that real consumer goods spending rose about 0.6% in August and 1.8% from a year ago, but that average hides a very different story across goods categories. Real durable-goods spending is running about 5.2% above last year, while real nondurable spending is slightly below its year-ago level.

That does not mean one group of households is booming while another is collapsing; retail data cannot tell us that. It does tell us that the strength in goods demand is lopsided.

Falling or relatively soft prices for many durable goods are supporting real purchases, while faster price growth across nondurables is absorbing more of the increase in household spending. The aggregate headline blends those two very different experiences into one firm-looking number.

What August actually says about the consumer

This is where I would be careful with the word consumer.

Retail captures goods plus restaurants. It does not capture most services, which account for roughly two-thirds of household consumption. One firm retail report therefore cannot establish that the overall consumer is strong.

What August does tell us is narrower, but still important. Goods demand held up. The July decline was largely reversed, the improvement was broad, the control group was firm, and our nowcast points to real goods spending rebounding in August and remaining above its year-ago level.

At the same time, the report is less impressive than the 1.24% monthly gain or 6.0% year-over-year headline suggests. The two-month pace is much closer to 4% annualized, inflation cuts the year-over-year gain by more than half in our CPI-deflated proxy, gasoline prices boosted August spending, and nondurable goods remain weak in real terms.

For the Fed, that makes the retail report evidence against a sudden collapse in goods demand rather than proof of an overheating consumer. It adds another firm data point to the policy discussion, but retail alone cannot answer the rate question.

The bottom line

August was better than July, but not as hot as the headline makes it look.

A meaningful part of the increase was a rebound from July. Smooth the two months together and nominal sales are growing at roughly a 4% annualized pace, almost exactly where the three-month trend sits. Adjust roughly for inflation and the pace is similar.

What survives that adjustment is the important part. The improvement was broad, the control group strengthened, and real goods spending appears to be growing again. But that strength is uneven, with durables doing much more of the work while nondurables remain under pressure.

So I would not call this a strong consumer. I would call it resilient goods demand growing at a moderate real pace, with more pressure underneath than the headline suggests.

A note on method and sources

Retail figures are seasonally adjusted and come from the Census Bureau’s Advance Monthly Retail Trade Survey. One-month changes are reported directly; the three-month trend is annualized as (level_t / level_t−3)^4 − 1.

The “real retail” figures are CPI-deflated proxies, not official real retail-sales estimates. The total and control-group proxies use seasonally adjusted CPI-U. October 2025 is missing from the CPI series because BLS could not collect the survey data during the lapse in appropriations.

The PCE-goods nowcast uses rolling five-year regressions that map matched retail baskets into nominal BEA goods categories. The inflation-adjusted estimates use CPI commodity, durable and nondurable indexes as provisional deflators, so they should be interpreted as CPI-deflated PCE proxies rather than forecasts of BEA’s eventual chain-weighted real PCE series.

In a 48-month pseudo-out-of-sample exercise using revised data, the same rolling five-year bridge used in the current nowcast reduced RMSE relative to a last-month-repeats benchmark by roughly 75% for total goods, 64% for durables and 61% for nondurables. Because the exercise uses revised rather than real-time vintages, live forecast errors should be expected to be somewhat larger.