Nominal spending across the world’s largest economy fell more than expected last month, even as a key measure of underlying demand firmed.

As discussed in this week’s macro preview, retail sales were set to show a decline in Tuesday’s release attributable to lower car sales and gas station receipts, but the closely-watched “control group” — which informs GDP tracking — was expected to rebound. And that’s precisely what happened.

Headline retail sales fell 0.9% in May from April, larger than the expected 0.6% decline. Auto sales dropped 3.5%, while the gas station line reflected a 2% drop.

The control group, by contrast, notched a 0.4% advance, better than the 0.3% uptick expected by economists. The prior month’s control group print was revised slightly better to show a 0.1% decline from -0.2% as initially reported.

The headline drop was among the largest in years, but anyone who knows what they’re talking about (admittedly a dwindling share of the investing and, more importantly, the voting, public) will ignore the headline print in favor of the control group beat.

That’s not to say this was an unequivocally solid readout. It wasn’t. Seven of 13 categories showed a decline including a meaningful drop on the restaurants and bars line, the only services sector category in the report. Notably, the building materials and garden supplies line registered a near 3% MoM decline.

Also, April’s headline was revised to show a small decline from a small gain, which means headline nominal spending was flat or down every month this year besides March.

All in all, this was a mixed read. The ex-autos & gas print showed a minor drop (-0.1%) versus expectations for a gain, so it’s not accurate to write the miss off entirely to cars and pump sales. But, the rolling, three-month annualized pace of control group spending looks quite healthy.