Federal withholding steals from every paycheck and dumps it back as a single spring refund.
The Commerce Department’s July retail-sales release — a 0.6 percent month-over-month decline in retail and food-services sales, the steepest monthly drop since May 2025 — is the bill for that mechanism. The June release the prior coverage described as a 0.2 percent gain sits between the spring bump and the July drop on the same release schedule.
Three numbers carry the report. First, the 0.6 percent headline drop. Second, the 0.2 percent decline excluding gas stations and auto dealers, the two most volatile categories; the exclusion is doing real work, with AAA reporting the national average gas price at $4.08 a gallon on Friday, up from $3.85 a month earlier — a 6 percent monthly rise on a category that runs roughly 4 percent of the Consumer Price Index basket and a much larger share of lower-income household budgets. The arithmetic is straightforward: twenty-three cents a gallon, on roughly five hundred gallons a year for an average driver, comes to on the order of a hundred dollars a year before the second-round effects on goods that move by truck. Third, the year-over-year figure: retail and food-services sales stood roughly 5 percent above year-earlier levels on the Federal Reserve Economic Data (FRED) series as of Friday. The 5 percent is the baseline; the monthly volatility is what sits on top of it.
Here is what the Commerce Department’s own release says about the April-May bump it carried. Households dipped into the refunds; the spending showed up in the retail-sales series; by July the refunds were spent. The April-May bump and the July pullback are not two stories. They are one story with two halves. The transfer was the spending. There was no recovery behind it.
The 5 percent year-over-year figure conceals more than it reveals. The Monthly Retail Trade Report is a nominal series; it is not deflated. Five percent nominal over twelve months, against the inflation households actually faced, is not five percent real growth in household consumption. The nominal number flatters the months it crosses, particularly when price increases have been concentrated in the categories the report covers. The Producer Price Index released this week shows the wholesale-side pattern feeding the same nominal pressure — producer prices are cooling, but the nominal retail series still carries the price levels households are paying now.
The composition of the decline tells the story the headline masks. Restaurants rose 0.5 percent while goods retailers fell. Consumer electronics and online retailers led the goods pullback. The pattern is consistent with households holding routine service commitments — restaurants sit close to social and recurring spending that is harder to cancel month-to-month — and cutting discretionary goods purchases first. Electronics and online are the clearest discretionary categories in the goods side of the report. If the April-May refund money landed anywhere in the goods side, it landed there, and the absence of refund money is now visible there.
The Joint Committee on Taxation’s distributional analyses document that tax refunds are concentrated in lower- and middle-income households.
The Congressional Budget Office’s literature on fiscal multipliers documents the second half of the pattern: transfer-recipient households have higher consumption propensities — meaning every additional dollar of refund translates more fully into current spending — than higher-income households. The April-May bump and the July pullback are the consumption profile those publications predict.
The structural choice is documented. Federal tax-withholding architecture imposes a lump-sum liquidity event on most households each spring. Withholdings throughout the year exceed actual liability for the median refund-receiving household; the refund is the return of the over-withheld portion, arriving as a single payment between February and April. The 2020 Form W-4 redesign — the most significant overhaul since 1987 — modernized the withholding calculation but preserved the underlying refund architecture. The volatility the July numbers document was not on that redesign’s scope. The mechanical smoothing fix — revised withholding tables so that refund recipients receive the over-withholding as additional take-home pay throughout the year rather than as a single spring lump — has been on the table at Treasury for decades and has been deferred each time because the lump-sum refund is a popular political signal of “money back” from the federal government. The July data documents what the popularity costs the data signal.
The July release is not a recovery indicator — the April-May bump was a fiscal-timing artifact; July is the structural read. It is not a recession indicator either. A 0.6 percent monthly decline after a refund-driven bump is what the underlying position looks like when the timing operation runs out; it is not what the underlying position is breaking.
The score is the score. The author of the tax code does not get to grade the demand.