Early gift for the White House: an absolutely perfect PCE inflation print, with core at 2% (!) target. War is over, if you want it.
Many, including myself, were worried about divergence with CPI; and PCE is what the Fed actually watches. But that didn't happen.
Let's dig in. /1
To give a sense of the trend here, let's look at the 3-month core PCE rate. As Powell notes, there had no actual drop throughout the year, all the news cycle events just washing out.
Finally we see a drop. It's just two months, but it's encouraging, especially under the hood. /2
We can break PCE inflation into the three categories that Powell and other Fed officials are watching: goods in deflation, housing peaked, and the rest of services slowing. We see all; notably in services, which has a much different balance of items than CPI. /3
In general, core services ex housing - what Powell watches for wage pressures - runs hotter in PCE than CPI. This divergence jumped this month but had been elevated in 2022.
Digging into this further is an ongoing project (I got covid this week so couldn't prep this, alas). /4
Last, services aren't dragged down by imputed prices, a concern @jasonfurman and others have flagged in previous months as falling stock prices are factored into financial services.
Though this happened earlier in the year, the last three months have been in historical norms. /5
Core services still remains elevated versus pre-pandemic trend, and goods won't be in deflation forever. But the trends over the past 3 months are exactly what a "soft landing" would have predicted, and it is important for the Fed to let these positive developments play out. 6/6
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Quits and hires rate are now a notable step below 2019 levels.
Quits does better predicting labor market conditions and there's an increasing trend in job openings across the 21C. So, at this critical moment, worth weighing quits more when we are watching labor market. /1
If the concern is in (wage-inertial?) services, hires and quits levels are also below there, and have been falling consistently.
There has been a lot of shifting and upgrading in the labor market during the recovery, and that process has played itself out. Time to reassess. /2
If you are still focused on job openings because you think 'openings over vacancies' is a proxy for output deviations that gives a nonlinear Phillips Curve that fits this recovery, well, sorry bub, that's over. It fell off months ago.
A year ago these numbers were above 4 percent. I understand the yawns and the sense it's old news, but this is just a massive and wild achievement. Let's dig in and discuss the last mile and what just happened. /1
Last mile: weighted contribution to overall inflation by major categories over past 6 months.
We are at 2%. Core goods pull that down about -0.5%; easy to see how to replace that. For all the fears on non-housing inflation, it's only pulling up 0.2% compared to 2018-2019. /2
Here's a different version of approaching that chart. Non-housing services is volatile month-to-month, both now and before 2020.
We can debate where underlying is between 2 and 2.5 percent right now, but rates are set as if underlying is between 4 and 5 percent. Or: too high. /3
Just to go back to this for a minute, worth flagging that in the current New Keynesian grad textbook:
whether end of a transitory cost-push shock causes the price level to go back to previous trend (i.e. negative inflation) is a policy choice by the Fed; it’s not inevitable. /1
That model isn’t ideal - there’s no long-term costs to recessions - but it gives a sense. And given that we didn’t force a recession in response to reopening while long-term inflation expectations remained anchored, the choice looks like the right one. /2
Before we dig into an interesting upside CPI report, we now have the full 2023 data now.
And we can see, using the best proxy we have for core inflation going back to the 1940s, that 2023's disinflation looked like the post-WWII period, not the 1970-80s one. Remarkable story.
/1
Big story is leveling out of core CPI inflation over the past 5 months in the low 3 percent range. I'm curious how next month's retroactive seasonal adjustment impacts this, and throughout the past years we've seen these leveling out then down shifts during the disinflation. /2
We'll probably get a big round of discussion on is housing following this. There's ~4 months of housing clocking in a bit higher than the downward trajectory people had been expecting. After next month's seasonal readjustment, probably a time to dust off those models again. /3
And there it is: even though we're well into the recovery, and have had unemployment under 4 percent for 2 full years, 2023 had the 5th highest job growth in the 21st century. (6th as a percent of employment.) Note the contrast with the 2010s below.
Let's dig into December. /1
You can see job numbers having more variance over the past six months, especially coming off the big job growth the past two years. That means we should be more careful not to read too much into any one month.
And stepping back, this is the job gain rate of a soft landing. /2
Some weakness in the household survey; participation and (even prime) employment rates were down a bit. Should generally go with the employer survey, but with JOLTS weaker perhaps time to pay more attention?
Either way, participation rate is higher than prepandemic estimates. /3
Well #JOLTS was fun while it lasted. Here's quit rate vs inflation over the past ~20 years. November 2023 is hard to read, because it's so in the middle of the best values.
There's nothing to see here other than a solid labor market alongside big disinflation.
But but but... /1
But but but....job openings!
Whatever relationship is there, we've completely slid off of it. We all had a blast debating whether 'the Beveridge Curve is steep so declines in v/u could come mostly from v' it....was a nonstarter? We just got the disinflation without it.
Why? /2
One reason why is the openings series has a time trend. During a delirium dream I thought coding notebooks could be the new microblogs; this half-finished attempt shows time series trends in that data, including plotting the residuals of all 50 states. /3 github.com/mtkonczal/BLS-…