Did shoppers really spend 3% more in January?

US retail sales jumped 3.0 per cent in January from December, ending talk of an imminent consumer slump and sending economists back to their growth models. Commentators praised the American shopper, who seemed untouched by the Federal Reserve's rate rises. Anyone who looks at how retail sales are put together knows that January figures are bent out of shape by seasonal adjustments built for a normal, pre-pandemic economy. The 3.0 per cent gain looks much more like a quirk of leftover seasonality than evidence of a boom.

Before adjustment, retail spending falls every January after the December splurge. The statistical agencies apply big seasonal factors to smooth out that predictable slump.

The size of that smoothing is easy to underestimate. Unadjusted retail sales typically drop by a double-digit percentage between December and January, as gift buying stops and households pay off holiday bills. The seasonal factor has to reverse almost all of that fall, so the published January number is a small residual left over after two very large figures have been netted against each other. A modest error in either one produces a large swing in the headline.

How the factors are built

The Census Bureau estimates those factors with a statistical programme, X-13ARIMA-SEATS, that infers each month's normal pattern from the history of recent years. Recent Januaries have been anything but normal. Relief cheques landed in early 2021, and lockdowns and reopenings rearranged household routines, so the model is extrapolating from a history that no longer describes how people shop. The problem has form: economists spent 2015 arguing that first-quarter GDP was understated by residual seasonality, and the Bureau of Economic Analysis changed its methods in response.

A mild January fooled the formula

In January 2023, though, mild weather and lingering post-holiday clearance sales meant that unadjusted spending fell by less than the model expected. Households that would normally sit out a cold snap went out and spent instead.

The adjustment formula then applied its usual large uplift to a smaller-than-normal drop, and produced a surge in the headline number that shoppers never delivered. Spending did not really expand. It simply fell less than it does in a typical frozen January.

The same distortion showed up elsewhere in the January data. Payrolls rose by 517,000 on first estimate and the unemployment rate fell to 3.4 per cent, its lowest since 1969, in another series where a small quirk in the raw figures is magnified by a large January factor. Two strong numbers from the same month and the same weather do not amount to two independent pieces of evidence.

The composition of the gain fits that reading. The biggest increases came at restaurants and car dealers, the categories most sensitive to weather and to the timing of promotions. Car sales also had a supply story behind them: dealers finally had more new vehicles to sell after two years in which the semiconductor shortage kept showrooms thin, so part of the jump reflected pent-up orders being filled rather than fresh demand.

Prices did part of the work

Retail sales are measured in dollars, not volumes. Consumer prices rose 0.5 per cent in January and were 6.4 per cent higher than a year earlier, so a slice of the headline gain reflects higher price tags rather than fuller baskets. Some of the categories that rose most, such as restaurant meals, are also among those where prices are climbing fastest, so their real gains are smaller still. Deflated even roughly, the real increase is smaller than the figure that set off the celebration.

Incomes tell another story

The apparent boom also bore no relation to household income. Real disposable income is growing slowly, and the personal saving rate is still near record lows. Shoppers did not gain 3 per cent more purchasing power. They spent leftover gift cards and leaned on credit cards. Average hourly earnings, adjusted for inflation, are still lower than a year ago, so the typical worker's pay buys less than it did when the Fed started raising rates.

The strongest objection is that January did bring a genuine income boost. Social Security benefits rose 8.7 per cent at the start of the year, the largest cost-of-living adjustment since 1981, and many employers hand out annual pay rises in the same month. That money is real, but it is a one-off step rather than a new trend, and it arrives in precisely the month the seasonal model already treats as exceptional.

Borrowing to keep spending

What households did spend increasingly came from borrowing. The New York Fed reported last week that credit-card balances ended 2022 at a record, having risen sharply through the year, and interest rates on those balances have climbed with the Fed's policy rate. Spending financed by revolving credit at rates near 20 per cent cannot run ahead of income for long, and the savings built up during the pandemic have been drawn down for more than a year.

That matters for who wins and who loses from the headline. Retailers and restaurant chains enjoyed a strong month and their shares rallied, but lenders with large credit-card books face rising delinquencies if incomes stay flat. Holders of short-dated Treasuries lost money as yields rose on the report, and any rebound in bond prices depends on the spring data exposing the January gain as a mirage.

The Fed has more reason than most to be careful here. It was misled late last year by consumer-price figures that the annual seasonal revisions later rewrote, and it will not want to react hard to a single month of spending data that may be revised in the same way. A central bank that has just been burned by soft numbers can equally be tricked by strong ones.

Bond traders have taken the headline at face value, pushing Treasury yields higher as they price further Fed increases. The first test comes on Friday with January's personal income and spending report, and the February retail figures follow in mid-March along with the usual revisions to January. Treating a seasonal quirk as proof of a sturdy consumer is a mistake. Once the distortion washes out, probably in the spring data, spending is likely to fall back into line with stagnant real incomes.