The consensus view is that the Bureau of Labor Statistics (BLS) tracks consumer food spending reasonably well. The CPI food spending weights tell a rather different story, one in which the official basket has drifted meaningfully away from where Americans are actually putting their money.
The starting point is retail data. Since April 2022, consumers have spent more on food service than on food and beverage stores. The aggregate figures are unambiguous: food service tallied $103,555 against food and beverage stores at $85,526, giving food service a 54.77% share of that combined total. In other words, dining out has become the majority habit, not the minority one.
Where the CPI Food Spending Weights Still Point
The BLS weights its Consumer Price Index using the Consumer Expenditure Survey (CE), a multi-year trailing average. That process means the basket updates slowly relative to how quickly spending patterns can shift. The BLS currently weights food at home at 8.231% of the CPI and food away from home at just 5.290%. Run those figures against the retail-sales split and the divergence becomes clear: the CPI food spending weights assign the larger share to grocery spending at the precise moment real-world behaviour has moved the other way.
The practical consequence shows up in the year-on-year CPI readings. Food at home is running at 2.7% annually, food and beverage at 2.9%, and food away from home at 3.4%. Because the BLS still overweights the cheaper category in its basket, the composite food inflation figure it publishes is pulled down relative to what a basket matched to actual spending would produce. The argument, advanced by MishTalk, is that the BLS has been underreporting food inflation since April 2022 for precisely this structural reason.
BLS Consumer Expenditure Data Complicates the Picture Further
The enrichment here comes from the BLS’s own consumer expenditure research, which makes the drift look worse, not better. According to the U.S. Bureau of Labor Statistics consumer expenditures report for 2023, food away from home expenditures rose 8.1% that year, adding an average of $294 per consumer unit compared with 2022. Food at home expenditures also rose, but by a more modest 6.1%. The gap between the two growth rates is directly at odds with a CPI basket that still weights grocery spending more heavily than restaurant and takeaway spending.
The same report contains a detail that cuts against any simple “consumers are trading up” narrative: spending on produce, specifically fresh fruits and vegetables, fell 10.1% in 2023. Rising overall food-at-home expenditure combined with a double-digit drop in fresh produce suggests consumers are not simply buying more groceries; they are shifting toward less perishable, often more processed, categories within the grocery aisle. That kind of within-category substitution is precisely the sort of compositional shift a slow-updating index basket struggles to capture.
The CPI food spending weights problem does not exist in isolation. The same structural critique applies to shelter costs, where the BLS assigns tenants’ and household insurance a weight of just 0.414% within a shelter component that makes up 35.473% of the total index. The BLS treats home prices, property taxes, and homeowners’ insurance as capital expenses rather than consumer expenses, which removes three large and fast-rising cost items from the inflation measure that households feel most acutely.
Kevin Warsh has called for more timely market-based measures in the CPI. The retail-sales split and the 2023 consumer expenditure data suggest the problem is less about timeliness and more about whether the basket’s structural weights bear any resemblance to actual spending. More frequent updates of a mismeasured basket would still be a mismeasured basket, just with fresher timestamps. The BLS has the survey machinery to do better; the question is whether the corrected weights would produce inflation readings that anyone in Washington would find politically comfortable.
The 2023 consumer expenditure data referenced above is already published and available. The gap between those figures and the current CPI basket weights is not a forecast: it is an arithmetic fact waiting to be closed.
