This morning’s retail trade report for December is a bit tricky to assess because retail control beat meaningfully and was revised up marginally for past months, but restaurant sales, which are not part of control, were down sharply. Normally, this is not an issue and so having expertise in how to treat non-control components of retail trade does not matter much. But this month makes for an exception. I will be curious how the bean counters at the Atlanta Fed treat this, because their quarterly bean count should be roughly just-the-facts (with no imposition of their chronically upbeat priors) when dealing with how to treat data related to the last month of the quarter.
But the following should be close. Real PCE growth for December looks to be tracking up somewhere in the range of 20 to 25 basis points, with some very marginal tendency to revise up past months, which I ignore here as noise. That leaves the quarterly growth rate tracking at 3.1% (ar), although I could be off by 10 to 20 bps there. In contrast, the Bloomberg polling has the consensus for the quarter at 2.6% (ar), which does seem reliably too low – again.
Quarterly bean counts can be arbitrary because of base or “launch” effects and because of the noise introduced by auto sales volatility. To control for those issue, I like to chart out the level of real PCE ex-autos and then try to eyeball a trend an objective trend line to indicate the underlying growth rate. This exercise suggests that underlying growth remains about 3%, where it has been, although there is a hint of maybe needing to revise that down slightly, rather than up slightly. Anyhow, because the underlying growth rate is near the quarterly growth rate, the so-called statistical launch into Q1 looks about neutral. For example, if real PCE were to grow sequentially at an annualized rate of 2.5% in each of January, February and March (from its estimated December value), then the growth rate for Q1 would be 2.6% (ar).
Let me conclude here with an aside about how this mapping from retail sales and the auto SAAR to real PCE estimates works. Much of it is just rote in the way you might expect. But there is also a big component that involves extrapolating the recent rate of growth of components of PCE that are not informed by hard data ahead of the income and consumption report. These are mostly services. For much of the past few years, the momentum there has been quite strong, which is why estimates of broader real PCE growth have tended to be strong for any given results on retail sales or the auto SAAR. But growth there has moderated recently to a pace of about 2% (ar), which is reflected in my estimate for December, and I assume also in what the consensus estimate will be. But we actually learn about how this stuff behaved only when the income and consumption report is released.
Separately, regarding the high print in the cost of foreign travel component in the import price index, I wouldn’t worry too much about that. Americans aren’t going to want to be found outside the USA after January 20 anyway. So, no impact on my macro view. Joking aside, I am leaving my core PCE bean count and its slicing exactly unchanged.
Source: BEA, FH estimates
Data are actual to December.
