January retail sales: once or so a year, it lays an egg. This was one

 

 - by New Deal democrat

It’s that time of month again for my favorite indicator for the consumption side of the economy: retail sales have been tracked for over 75 years. When they are lower YoY, that has historically been a good (not perfect) indicator that a recession is near. That’s because that same 75 year history empirically demonstrates that consumption leads jobs. In other words, it is the change in sales that causes employers to add or lay off employees (not the other way around, as I have sometimes seen claimed).

They are somewhat noisy, especially around the Holiday season, and they do get significantly revised, both of which were apparent in January’s report. In nominal terms, retail sales declined a sharp -0.9%, but December was revised higher by +0.3%. Since CPI tagged a strong 0.5% in January, that means real retail sales declined -1.3% for the month. Because I had the genius thought several months ago that because shelter prices (i.e., house prices and rent) were distorting CPI, maybe they were distorting the signal from real retail sales as well. So they are included below in light blue as well. Note the graph is normed to 100 as of just before the 2021 pandemic stimulus:

When we take out shelter, real retail sales show a solid uptrend since July 2022 (when gas prices were $5/gallon). Note that the December/January numbers have shown the sharpest m/m changes for each of the last three years. I am thus inclined to treat this month’s big decline as unresolved seasonal noise unless there is confirmation next month.

The best recession vs. expansion signal is the YoY% change in sales, which - pretty much until the pandemic - almost always forecast an imminent recession when it turned negative. Again, when we take out the distortions caused by shelter, the false recession signal almost completely goes away:

Even with January’s downturn, the forecast is for continued expansion.

Finally, because consumption leads employment, per the above paradigm, here is the update on that:

With the downward benchmark revisions in employment numbers for the past year, the two series are coming much closer to being in sync. While the forecast remains for positive employment reports, the suggestion from real retail sales is that there is likely to be continued deceleration in the YoY comparisons. In early 2024, the average monthly gain in employment averaged 180,000, so per this model I am expecting the next few months of job gains to average less than that.

Jobless claims: more of “steady as she goes”

 

 - by New Deal democrat

Now that we are well past the Holidays, seasonality has settled down and so have the comparisons for jobless claims.

Initial claims declined -7,000 to 213,000 last week, and the four week average declined -1,000 to 216,000. With the usual one week delay, continued claims declined -36,000 to 1.850 million:

On the more important YoY basis for forecasting purposes, initial claims were higher by a mere 0.9%, and for the first time in five months the four week average was *lower,* by -0.7%. Continuing claims were higher by 2.6%, about their average for the past year:

It’s too early in the month to talk about what this might mean for next month’s jobs report, but initial claims, along with the YoY% change in the S&P 500, constitute my “quick and dirty” forecasting model. Basically, if claims are higher YoY, and the stock market is lower, the economy is almost always in trouble. Here’s the update:

With jobless claims essentially unchanged from one year ago, and the stock market higher by about 20%, the verdict is: more of “steady as she goes.”

January CPI: a new paradigm, with the reappearance of some old suspects

 

 - by New Deal democrat

Needless to say, January’s increase of 0.5% in consumer prices was not welcome. And there was a changing of the guard somewhat, as several of the old suspects (food and energy) made new appearances.

Some of the spike probably has to do with unresolved seasonality. A lot of producers increase their prices at the beginning of each year, which shows up in the graph below of CPI since mid year 2022:

The highest bar on the graph is January 2023. January 2024 is only lower than February of last year. And this month’s change was the highest since then. So a lot of what we need to look for are changes in the YoY%, which is how almost all of the graphs below are presented.

The increased contribution of the old suspects is apparent in the graph below of core (red) vs. total inflation:

Core inflation has been holding nearly steady for a full year, with changes between 2.9% and 3.2% (3.1% this month). It is inflation in food and energy that has driven the change in total inflation. And although I won’t show specific graphs, food inflation has a lot to do with the price of eggs (bird flu resulting in the detruction of lots of chickens) and the deflation in gas prices halting at roughly $3/gallon.

On the plus side, inflation in medical care services was looking like a concern, but that has at least leveled off:

And what about shelter? Well, for the 20th month in a row YoY inflation ex-shelter came in at less than 2.5%, although it did hit an eight month high at 2.2%:

Both rent of primary residence and owners equivalent rent continued their slow descent, hitting their lowest YoY levels in almost three years, at 4.2% and 4.6% respectively:

Although it did not decline this month, the downtrend in transportation services inflation (motor vehicle insurance and repairs, in blue) also appears to be intact:

And the deflation in used car prices has definitely ended, although new car prices are flat:

To sum up: while the very gradual deflation in shelter and to a lesser extent in transportation services is helpful, the definitive end in the deflation of gas and used car prices, plus the spike in food prices due mainly to bird flu has re-ignited consumer prices. 

Finally, here’s an update of one of my favorite measures of average American well-being: real aggregate payrolls for nonsupervisory workers:

This declined in January, and is no better than it was in September. On a YoY basis, consumers still have over 2% more than they had to spend a year ago:

And a mult-month stall like this is not uncommon. But needless to say, if it continues several more months it will begin to flash an important recession watch signal.

An examination of the levitation in residential building employment

 

 - by New Deal democrat

While we are waiting for new economic data tomorrow, let me pick up on an issue I closed with yesterday: while manufacturing has turned down, goods production in the US economy is being held up by construction, and in particular residential construction. Given the severe hike in mortgage rates as well as house prices, I described it as “levitation.” So today let’s look at that levitation.

As per usual, I always start out with the fact that mortgage rates lead sales. The below graph includes mortgage rates (blue, left scale) compared with housing permits (red, right scale) and the even more leading, but very noisy new single family home sales (gray, right scale). The latter two are normed to their peak at 100:

And as mortgage rates increased from 3% to 7%, sales and permits declined about 20%, +/-5%. This is a serious decline, which has often but not always meant a recession has followed.

For completeness’ sake, here is real private residential fixed spending from the GDP reports compared with single family permits, which have been the least noisy leading housing metric of all. Again, both are normed to 100 as of their peaks, and permits are averaged quarterly for better comparison:

Note that real housing spending in the GDP has declined about 15%.

Next, the noisier housing starts (light orange) follow permits by a month or two. And housing units under construction (which vary by starts minus completions), a which are the “real” aggregate economic activity, follow with a significant delay, in this case by almost 1 year. Note again all three are normed to 100 as of the month of their peaks:

The first act of levitation after the pandemic was how units under construction stayed almost perfectly flat for about 18 months after they approached their peak. Historically this was a very long delay.

But at long last housing units under construction turned down, and did so with a vengeance, now down over 15%, which in the past has been consistent with a recession, although not always. Which leads us to the second act of levitation, which is the number of employees involved in housing construction (gold, right scale):

Employment in residential construction has continued to increase, despite the downturn in every other housing sector metric!

When might we expect this last shoe to drop? Here’s the longer term historical look:

In the case of the 2000s housing bubble, construction employment turned down with only a very slight delay. The 2001 recession was partly a tech bubble, partly the China manufacturing employment shock, and partly the September 11 terrorist attacks, so housing construction did not turn down meaningfully. Residential construction employment peaked 15 months after housing construction turned flat.

Perhaps the closest analogue is the 1980s construction boom. Here is a close-up on that era:

In the 1980s, residential construction employment increased for 22 months after actual construction turned down. 

Where does that leave us now? We are currently over two years after the actual peak in construction, and roughly 18 months since it turned down significantly. As the above historical graph shows, the last three organic recessions (i.e., not including the pandemic) started once residential construction employment was down 10% (and the trends in other sectors typically leading recessions also were in place).

I have been looking for residential construciton employment to turn down for at least half of year. My best guess is that it will finally turn down at some point in the next six months, if the remaining economic pieces (especially mortgage rates!) remain in place. Once they do turn down, they may follow a similarly sharp decline as construction already has, which would suggest a recession 12-18 months later on average. Please note this last sentence is *not* a forecast, only an average. There are many other moving parts to consider.

The Big Convergence: scenes from the January employment report

 

 - by New Deal democrat

There’s no new significant news until Wednesday, so let’s catch up graphically with a few important items from Friday’s employment report for January.

As I wrote then, probably the most important developments weren’t in the monthly numbers, but rather the annual revisions to both the Establishment and Household Surveys. 

For background, recall that the monthly report is a survey. But once every quarter, with unfortunately about a six month lag, an actual census of almost 100% of all employers is published, based on their tax reporting. This is called the QCEW. And through Q2 of last year, it indicated that the Establishment Survey had been too optimistic by over half a million jobs. That was resolved on Friday with a downward revision of 610,000 jobs over the past 12+ months.

On the other hand, all last year I was writing that the Household Survey was “frankly recessionary,” showing almost no growth YoY. But it appeared via ata from the Congressional Budget Office, that the survey had missed millions immigrants in the years since tha pandemic. That too was largely resolved on Friday, with the addition of over 2,200,000 to the number of people employed.

The downward revisions in the Employment Survey, and the upward revision in the Household Survey, completely resolved the discrepancy between the two:

There were a few other trends of note, all from the Establishment Survey.

I’ve noted in the past several months how goods-producing employment has stalled. The are two separate and contradictory trends playing out here. The first is that manufacturing employment has fallen by about -140,000 in the past two years. Up until the 1980s, this would absolutely be recessionary. But no more, as manufacturing employment is too small a share of the total. And in the past two years, it has been completely balanced by growth of over 370,000 in construction jobs:

What has been particularly surprising is the continued growth in residential construction jobs, despite the downturn in virtually every other metric in that sector:

If the goods producing sector as a whole has turned flat, that means that on net all the growth is coming from service providing jobs. But not all such jobs. In particular, as I have been highlighting for over a year, professional and business employment, which tends to be higher paying jobs, has seen a decline of roughly -250,000 jobs in the past 18 months. But the remainder of the services sector has added almost 4,000,000 jobs:

A fair amount of that growth has come from the provision of health care, which has seen employment grow at a steady 3% YoY clip:

The net effect of all these cross currents - at the moment - is slow but steady employment growth. I continue to focus particularly on the construction sector, where I just do not see residential construction employment continuing to levitate when the number of houses under construction is down over -15% from its recent peak.