Weekly Indicators for November 28 - December 2 at Seeking Alpha

 

 - by New Deal democrat

My Weekly Indicators post is up at Seeking Alpha.

First the long leading indicators turned. Then the short leading indicators turned. Now the coincident indicators are weakening to new expansion lows almost every week.

The silver lining is: the first long leading indicator may already have hit its worst levels.

As usual, clicking over and reading will bring you up to the virtual moment as to the economy, and reward me a little bit for my efforts.

November jobs report: decidedly mixed signals, most consistent with a still-growing but decidedly weakening economy

 

 - by New Deal democrat



Since early this year I have expected employment to follow the halt in consumption growth, decelerating over time to a stall. This has only intensified given the major decline in growth in payroll withholding tax payments, which are near recessionary. 

This expectation was partially met today in that the three month average in employment gains since February, which had decelerated from over 500,000 to 289,000 through October, decelerated further to 272,000 through November.

Aggregate payroll growth also declined YoY from 9.1% to 8.7%, but is still probably above the inflation rate.

Here’s my in depth synopsis.

HEADLINES:

• 263,000 jobs added. Private sector jobs increased 221,000. Government jobs increased by 42,000. 
• The alternate, and more volatile measure in the household report *delined* for the second month in a row, by -133,000 jobs. In the past 8 months, according to this report, only 12,000 jobs have been added! The above household number factors into the unemployment and underemployment rates below.
• U3 unemployment rate was unchanged at 3.7%.
• U6 underemployment rate fell -0.1% to 6.7%.
• Those not in the labor force at all, but who want a job now, rose 167,000 to 5.550 million, compared with 4.996 million in February 2020.
• Those on temporary layoff declined -44,000 to 803,000.
• Permanent job losers rose 127,000 to 1,368,000.
• September was revised downward by -46,000, while October was revised upward by 23,000, for a net decrease of 23,000 jobs compared with previous reports.

Leading employment indicators of a slowdown or recession

These are leading sectors for the economy overall, and will help us gauge whether the strong rebound from the pandemic will continue.  These tilted to the negative:

• the average manufacturing workweek, one of the 10 components of the Index of Leading Indicators, was declined -0.1 hour to 40.9, and is down -0.7 hours from its recent February peak of 41.6 hours. This is consistent with the onset of a recession.

• Manufacturing jobs increased 14,000, and are at a level higher than before the pandemic.
• Construction jobs increased 20,000, also at a level higher than before the pandemic. 
• Residential construction jobs, which are even more leading, declined by -2,600.

• Temporary jobs, which had been rising sharply, declined  by 17,200. Since the beginning of the pandemic, about 300,000 such jobs have been gained.
• the number of people unemployed for 5 weeks or less increased by 32,000 to 2,243,000, about 125,000 above its pre-pandemic level.

Wages of non-managerial workers

• Average Hourly Earnings for Production and Nonsupervisory Personnel rose $0.19 to $28.10, which is a 5.8% YoY gain, an increase of 0.3% from last month, vs. its 6.7% peak at the beginning of this year.

Aggregate hours and wages: 

• the index of aggregate hours worked for non-managerial workers declined -0.2% which is still above its level just before the pandemic.

•  the index of aggregate payrolls for non-managerial workers rose by 0.6%, and is up 8.7% YoY. This metric has been decelerating nominally almost consistently for the past 16 months.  Compared with inflation through October, it is up only 0.9% YoY (recessions typically start when it crosses zero).

Other significant data:

• Leisure and hospitality jobs, which were the most hard-hit during the pandemic, rose 88,000, but are still about -6% below their pre-pandemic peak.
• Within the leisure and hospitality sector, food and drink establishments added 62,100 jobs, but are still about -4% below their pre-pandemic peak. 
• Professional and business employment increased by only 1,000, over 1,000,000 above its pre-pandemic peak.
• Full time jobs increased 92,,000 in the household report.
• Part time jobs decreased -302,000 in the household report.
• The number of job holders who were part time for economic reasons rose 25,000 to 3,685,000.
• The Labor Force Participation Rate declined -0.1% to 62.1%, vs. 63.4% in February 2020.

SUMMARY

This report was mixed, with both strong and decidedly weak points. On the strong side, wages increased sharply, improving the YoY gains as well. The underemployment rate fell. Construction and manufacturing jobs increased. YoY aggregate payrolls decelerated further, but probably improved vs. YoY inflation. While the 3 month average gain in payrolls continued to decline, in absolute terms it remains good.

On the negative side, the manufacturing workweek has declined enough to be consistent with the onset of recession. Temporary and residential construction jobs declined. The number of short-term newly unemployed rose. All of these are leading indicators, and suggest further weakness to come. Aggregate hours worked declined, and are only up 1.3% in the past 9 months. Perhaps worse, employment as measured by the household report is only up 12,000 in total for the entire last 8 months, which is very much in line with the weak tax withholding data. In other words, I increasingly expect significant downward revisions to the recent establishment reports when they are revised next year.

This mixed report, taken all together, is most consistent with a still-growing economy, but one which is weakening, the theme for the entire last 6 months or more.

November manufacturing and October construction both decline, the former almost at recessionary levels

 

 - by New Deal democrat

As usual, we begin the new month’s data with the ISM manufacturing index. This index has a very long and reliable history. Going back almost 75 years, the new orders index has always fallen below 50 within 6 months before a recession. Recessions have typically started once the overall index falls below 50, and usually below 48.

Which means that today’s report for November comes very close to meeting all of the above criteria. The overall index declined below 50 for the first time since May 2020, at 49.0. The new orders subindex declined -2 to 47.2, the 5th time in the past 6 months that it has been below 50, and only 0.1 above its September low of 47.1:

This is a reading on the very cusp of recession.

Construction spending for October was also reported, and also showed declines of -0.3% for both total and residential spending:

Total construction spending is -1.3% below its July peak, while the more leading residential construction spending is -7.1% below its early May peak.

Since construction spending is a series with only a 30 year history, there is no reliable recession marker, except that residential spending turns down in advance.

Finally, below is a comparison of total residential construction spending vs. housing units under construction from the permits and starts report:

While there is a general concordance, there is no set pattern as to which turns first or simultaneously. I continue to expect housing units under construction to follow permits and starts and turn down shortly.

In sum, our two final reports of the day show declines in both leading sectors of manufacturing and construction, the former more seriously than the latter.

Strong personal income and spending contrast with near record low in saving

 

 - by New Deal democrat

Like retail sales earlier in November, personal income and spending both rose smartly, as shown in the below graph of real retail sales compared with real personal spending:

Real personal income was up 0.4%, and real personal spending increased 0.5%:

Nominally each increased 0.3% more; i.e., the PCE deflator was 0.3%. Each metric only had one better reading in the entire past year.

Since the spring 2021 stimulus ended, personal spending is up 3.9%. Personal income remains down -1.9%, although it has improved in the past few months (thank you, big decline in gas prices!):

With real income still down, you may wonder where all the fuel for increased spending has been coming from. The answer is a continuing decline in the saving rate, which at 2.3% for October, was the lowest in the entire 60+ year history of the series, except of one month (July 2005), as shown in the below graph which subtracts -2.3% so that the current level shows as zero:

In short, both sales and spending - two sides of the same coin - were strong in October, assisted in part by the continuing decline in gas prices. But with near-record low savings, consumers are more vulnerable to a negative shock than they have ever been.

Initial jobless claims get closer to signaling recession

 

 - by New Deal democrat

Today is one of those data-palooza days, so I’ll put up separate posts on personal income and spending, and the ISM manufacturing report and construction spending reports later.

But let’s start with weekly jobless claims, and the news here is OK for the week, but the trend is troublesome.

Initial claims declined -16,000 from last week’s 3 month high to 225,000. But the 4 week average climbed 1,750 to 228,750, the highest level since January 22 of this year. Continuing claims rose 57,000 to 1.608 million, the highest since February 26:

In the absolute sense, very few people are getting laid off, and those who are still are able to find new jobs pretty quickly. But the trend is deteriorating. At its recent pace, the 4 week average of initial claims is likely to go negative YoY by the end of this month.

In the past, if the 4 week average is more than 5% higher YoY for any significant period of time, and less reliably, if the slightly lagging continuing claims are higher YoY, a recession is almost always close at hand:

The first marker could be met by January 1. The second marker could be met by February.

Since initial claims is one of the last positive short leading indicators, this is a bad sign.

October JOLTS report shows continued deceleration in jobs market, with continuing gap in job openings filled

 

 - by New Deal democrat

For the past year, I have likened the jobs market to a game of reverse musical chairs, where there are more chairs than players. Some chairs are always left empty. The chairs are jobs, and the players are job seekers. Since the lowest paying jobs have always been left empty, there is tremendous pressure to raise wages. And as job-jumping is rewarded with greater pay increases, there is *continued* inflationary pressure.

The October JOLTS report showed both deceleration, but also that the game remains intact.

Let’s begin with the long term graph of openings, hires, quits, and total separations since the beginning of the series:

A year ago hires, quits, and separations were all at their highest levels since the series began, equivalent to 2006 and 2019. Openings were sky high, 50% higher than even their best levels before the pandemic. This was simply tremendous pressure on employers to raise wages and increase benefits in order to attract job seekers - and many slots went unfilled.

Now here is a close-up of the last 2 years in the above data:

All 4 series hit their best levels in the first few months of this year, and all have deteriorated since. Hires declined 84,000 in October to their lowest level since January 2021. Quits declined 34,000 to their lowest level since May 2021. Total separations did rise 18,000 for the month, but otherwise also were at the lowest level since May 2021. 

Finally, openings declined 353,000 for the month, but remained 64,000 above August’s level. Whether the last several months indicate a pause in the downward trajectory of that metric remains to be seen, but aside from August, October was the lowest level since June 2021.

Layoffs and discharges make bottoms during expansions and increase leading into recessions. They did increase by 58,000 in October, but remained generally in line with their level for the past 18 months:

By way of contrast, during the previous two expansions, quits averaged between 1.6M-2.0M per month. During the Great Recession, they rose as high as 2.650M.

In summary, the October JOLTS report shows that while the game of musical chairs remains intact, the overall theme of deceleration has continued. Some employers may be keeping job openings up formally, but choosing not to fill them due to wage pressures or a decline in overall demand.

House price indexes continue to show the top is in

 

 - by New Deal democrat

The FHFA and Case Shiller house price indexes were reported this morning, with both continuing to show that the peak in house prices took place during the summer.

For the month, the seasonally adjusted FHFA index rose +0.1%, vs. a +0.9% increase one year ago, and following two months of -0.6% and -0.7%, respectively. The Case Shiller national index declined -0.5%, following -0.5% and -0.9% declines in the previous two months.

Here are the absolute seasonally adjusted values. The FHFA index is down -1.2%, and the Case Shiller national index is down -2.6%, respectively from their June peaks:

And here are the YoY% changes, with the FHFA up 11.0% and the Case Shiller national index up 11.2% (Note: FRED has not yet updated the Case Shiller data):

The question now is, how far down do house prices go? In the 2007-11 bust, house prices fell a little over -20%. But in the smaller 1990-91 downturn, prices only declined about -3%. 

I have seen guesstimates of a -5% to -10% decline in house prices in this downturn, and that is a reasonable first dart-throw, although my guesstimate would be at the -10% end of that range. That’s because the Fed seems hell-bent on causing a sharp recession, and that recession will bring lots of joblessness, which in turn will mean more people unable to make mortgage payments, and so suffering foreclosure.

Speaking of the Fed, here is my updated graph of the YoY% change in the FHFA index (/2) vs. Owner’s Equivalent Rent in the CPI:

For literally over a year I have been writing that the house price indexes forecast OER rising all during this year, ultimately to record YoY% increases, and peaking at perhaps 9%; dragging core consumer inflation higher with it. Three of the four forecasts have com to pass. One is pending, with the most recent OER reading up 6.9%.

The continued YoY deceleration in the house price indexes give me confidence that, after continuing to rise for a few more months, probably beginning next spring or so, monthly OER increases will begin to decline as well. 

Will the Fed take heed?

New home sales adjusted for cancellations: still signs that a bottoming process might be taking place

 

 - by New Deal democrat

I had a correspondent question me about whether new home sales might actually be in the process of bottoming, due to the big increase in the percentage of cancellations, as shown below (via Bill McBride):

This is something I’ve been aware of, and commented on one month ago in the context of housing that was permitted but not started.

There are two responding points to be made.

The first is that this is not the first time new home sales have turned down. There is no reason to believe that there weren’t similar increases in cancellation rates in any of the other downturns caused by increases in mortgage rates, so the pattern in new home sale was probably similar in those downturns as well:

But let’s apply the data to current new home sales. Here’s the graph of the last two years I ran last Friday:

Below I show the raw data for new home sales (annualized, in thousands) in the first column, followed by the cancellation rate for that month, and finally the net sales after adjusting for cancellations:

 Apr 619 8.0% 569

May 569 10.5% 509

Jun 571 15.2% 484

Jul 543 18.4% 443 (LOW)

Aug 566 18.3% 540

Sep 583 20.4% 468

Oct 632 25.6% 470

As indicated above, the adjusted low to date was in July. October’s rate was in line with September’s and June’s.

Of course, the series could certainly go lower in coming months.

But as I’ve indicated a number of times recently, now that an oncoming recession is virtually certain, I am beginning to look for signs in the long leading indicators of how long that recession might be. And new home sales, for all of its noise and heavy revisions, is one good place to start looking.

 

Have new home sales made a bottom?

 

 - by New Deal democrat

Hopefully you are recovering from your turkey coma today. Here’s a little late commentary on Wednesday’s new home sales report.

New home sales are noisy, and heavily revised, which is why I prefer housing permits, and especially single family housing permits, as a source of information.

But . . . on the other hand, new home sales tend to be the very first housing metric that turns. In fact, during expansions they often peak so early that they are more of a mid-cycle indicator than a long leading indicator. Here’s the comparison of new home sales with single family permits from the 1970s through the Great Recession:

Sometimes the two series peak and trough simultaneously, but not that often (1977, 1985, 1998, 2005) new home sales hit their peak a few months before permits.

Why is that of note? Here is the current expansion:

As per my comment above, new home sales peaked first. But also look at the far right side of the graph. New home sales bottomed in July and are currently at their level from last spring, while permits are still heading south.

Beware revisions, but new home sales may be signaling a bottom for the housing market, which might be correct if long term interest rates are in the process of peaking now, despite the Fed continuing to raise short term rates.

Median new home prices are not seasonally adjusted, so the only good way to look at them is YoY. Remembering as always that sales lead prices, here is the YoY% change in each:

Again, very noisy, but the overall YoY trend in prices is down. The 3 month average gain over the past year has declined from over 22% to under 12%, suggesting that new home prices are on the cusp of turning down in absolute terms.

On a more general note, notice that, having made a recession warning, I am now on the lookout for the indicators that will tell me how long and deep the recession might be, and when it might bottom out. Watching for a peak in long term interest rates is an important part of that process.

Jobless claims have a poor week, rising to multi-month highs

 

 - by New Deal democrat

Initial claims for jobless benefits rose 17,000 this week to 240,000, a 3 month high. The 4 week average also rose by 5,500 to 226,750. Continuing claims one week ago rose 48,000 to 1,551,000, the highest number since March:

While one week like this shouldn’t set off any alarm bells, initial claims has been one of the increasingly few positive short leading indicators. Recently it has been more neutral, but it would only take about another 25,000 increase in the 4 week average to over 250,000 for this to turn fully negative, consistent with the “recession warning” I have written about several times in the past week.

At this point all 3 of my primary indicator systems - the long and short leading indicators tracked by the Conference Board and ECRI (pace Prof. Geoffrey Moore), the high frequency weekly indicators, and the “consumer nowcast” - are all signaling recession ahead.

On the bright side, manufacturers new orders rose 1.0% in October, and “core” capital goods rose 0.7%. This is one of the two components in the Conference Board’s Index of Leading Indicators that is still positive.

Also, new home sales will be released later this morning. I’ll update this post with a line or two later.

More on deteriorating tax withholding receipts and jobs reports

 

 - by New Deal democrat

I have a new post up at Seeking Alpha, in which I lay out all of the short leading indicators, and conclude that the conditions have now been met for a recession to begin at any point in the next 6 months.

There’s one graph I intended to use which didn’t make it through to the final published piece. Here it is:

Typically recessions have only begun when 8 of the 10 components of the Index of Leading Indicators are down compared with their levels 6 months previously. And so, I go through the list . . . .

In the piece, I note that the strong jobs reports have been the biggest reason why no recession has occurred yet. But in the past several weeks I’ve been pounding the table about the implications of the steep deceleration in tax withholding receipts since mid-year. Here’s the YoY% change in total tax withholding receipts since then:

July +7.8%*

Aug +10.2%

Sep +1.2%*

Oct +12.2%

Nov +3.7% (to date)*

*= less than YoY% change in CPI

I’ll get back to this chart further below.

In the meantime, consider that the monthly household report is prepared from a sample of 50,000. The monthly establishment report is prepared from a sample over 100,000+. But tax withholding is a full and complete report of what every taxpayer/employer in the US remits daily to the Department of the Treasury.

As a result, total tax withholding receipts should come fairly close to mirroring aggregate payrolls, especially for non-supervisory workers, in the household jobs report. The one important difference is that even bosses pay withholding taxes, up to $147,000 of salary, on Social Security, and also on Medicare without any income limit. With that in mind, here is a graph of the YoY% change in aggregate payrolls for both non-supervisory workers and all workers for the past year:

Note that beginning in April, the YoY trend starts to decelerate markedly.

Back in August, Investors Business Daily highlighted the below graph of the monthly 10 week YoY% change in tax withholding to date:

Note that, just like aggregate payrolls, with a one month delay it peaked and had declined ever since.

IBD hasn’t updated since, but the values are easy enough to calculate. As of the end of September, the 10 week YoY% change was 4.2%. As of the last daily report last week, the increase was only 3.7%.  

Note that beginning in July in the IBD graph, and the updated information for the 10 week totals, and the monthly totals from July and September in the monthly chart above, *all* of the YoY% increases in tax withholding are less than the rate of inflation.

In other words, since July *even in the aggregate* non-supervisory workers at least are making less, adjusted for inflation, than they were at the same time in 2021.

Further, per my rule of thumb for non-seasonally adjusted YoY data, when the YoY% change declines by more than half, which in the IBD metric it did by September, the data has likely made its absolute peak and started to turn down in absolute terms (i.e., if we could seasonally adjust it).

Over the weekend I took a look at how tax withholding performed in the year leading up to and during the 2001, 2008, and 2020 recessions. In each case, I found that when tax withholding declined on a YoY% basis to a level of 1% or less above CPI, that coincided with a variance of one month with either weak jobs reports of less than 100,000 jobs gains, or even outright job losses, in either the household or establishment aspects of the jobs report.

Now here are the m/m gains/losses in the household and establishment jobs reports for the last 12 months:

What tax withholding data is strongly suggesting is that the actual job losses in the household reports in June and October were signal, not noise, and that there is a strong likelihood that the establishment numbers are going to be revised downward in the future as more complete data is updated.

Coronavirus dashboard for Thanksgiving week 2022

 

 - by New Deal democrat

As we start Thanksgiving week, let’s take a look at the current state of COVID.

The Alphabet Soup of variants (most of which are direct descendants of BA.5), primarily BQ.1&1.1, has largely displaced their parent, which is down to 24% of all cases:

Typically new waves have peaked when the displaced variant is down to 10% or so of all cases, which should be the case with BA.5 in two or three weeks.

This is noteworthy, because as we will see below, the Alphabet Soup variants have not yet caused any real wave at all.

In the below graphs, I’m going to show the entire 2.5 year history of COVID as to each for comparison purposes.

Here’s BIobot’s waste particles data (dark line) vs. confirmed cases (light line):

Since the end of last year, many people have relied on home tests and not bothered with confirmation, so the “real” number of cases is roughly equal to the peaks of the first 3 waves of the pandemic. This makes sense since each new variant has been more immune-evasive than the previous variants.

Regionally, we do see the likely beginning of a winter wave in the West, and also perhaps in the Midwest (but not the Northeast at all!):

Despite this, hospitalizations are not elevated at all:

And deaths (thick line), when compared with cases (thin line), are near their all-time lows:

This has not gone unnoticed. Dr. Eric Topol, in his substack, calls the BQ.1.x variants the first displacing variants not to cause a new wave, pointing out that:

“in New York State, which has the highest level of BQ.1.1 in the US, there continues to be no sign of hospital admissions increasing. If anything, that rate is decreasing.”

He concludes:

“It would be tempting to interpret the lack of impact of BQ.1.1, relative to its immune evasion properties, as we’re out of the woods. A population-level immunity wall has been built up over 3 years, with all the infections and vaccinations. Further, our T-cell immunity from these exposures, which isn’t assessed with these neutralization antibody assays, is helping us defend against variants. The optimistic viewpoint is that there’s little more that the Omicron family can throw at us which will be much worse than what we’ve already seen. That we’re done, going endemic, that the acute phase of the pandemic with big waves is over.

“Not so fast. As Daniele Focosi reminded us this week, the SARS-CoV-2 mutation rate has increased by 30% in the past year. There still could be room within Omicron, and especially the XBB recombinants, to pose a significant threat. Moreover, there’s the dismal prospect of a whole new family of variants to emerge (e.g. Sigma) that are distinct from the mutation cascade we’ve seen from Omicron for over a year.”

For now, I’m gong to go with the more optimistic scenario. As to XBB, that was responsible for a wave in Singapore, that quickly came and went all in the month of October:

And no new lineage has managed to displace Omicron for the past full year.

In the meantime, I would still mask up in all public indoor spaces, not just for COVID, but to avoid this year’s bad flu outbreak as well.

Weekly Indicators for November 14 - 18 at Seeking Alpha

 

 - by New Deal democrat

My Weekly Indicators post is up at Seeking Alpha.

It had to happen sooner or later. Earlier this year, based on the long leading indicators, I went on “Recession Watch.” Now, for the first time in a very long time, I have escalated to “Recession Warning.” I believe there is much more than a 50/50 chance of a recession beginning in the next 6 months.

For all the gory details, click on over and read, which will bring you up to the virtual economic moment. As usual, it will also reward me a little bit for the efforts I made.

Existing home sales decline to recessionary levels; prices have clearly turned down; low inventory still a problem

 

 - by New Deal democrat

As I wrote earlier this morning, my primary interest in existing home sales at this point is prices. [Note: graphs below for sales and prices does not include October]

For the record, existing home sales fell to a new 2.5 year low (i.e., since the teeth of the pandemic lockdowns) of 4.430 million annualized:

Before the pandemic, the last time the number was this low was in 2012. Further, this is down -19.5% YoY, -26.4% from their recent secondary February high, and -35.3% below their October 2020 expansion high. This is the kind of number I would expect at the cusp of a recession.

More importantly, median prices declined seasonally by -1.5% for the month to $379,100. This is “only” 6.6% higher than one year ago, and is the slowest YoY% increase in over 2 years:

The highest YoY% change in the past 12 months was +17.6% in January.  This is the third month in a row that the rate of change has declined by over 50%, my rule of thumb for when a seasonally-adjusted data set would turn down. 

In short, I think we can safely say that existing home prices, were we able to seasonally adjust, have turned down from a peak during summer.

Inventory is also not seasonally adjusted. This turned down m/m, but YoY is -0.8% lower:

We are nowhere near solving the low inventory problem that we have had since even before the pandemic hit.

Core inflation using house prices rather than imputed rents

 

 - by New Deal democrat

Later this morning existing home sales will be reported for October, which will mainly be of interest to me only for what happened with prices, and secondarily whether the problem of low inventory which has existed for 3 years is moving in the direction of resolution.

In the meantime, yesterday Jason Furman got some traction, and amplification by Paul Krugman, of the below graph which measures core inflation using new rent indices (e.g., Zillow) rather than owners’ equivalent rent:

The implication is - one embraced by Krugman - that inflation is already not a problem. 

I don’t think this is really the case, because we are still using an imputation of hypothetical rents to measure house prices.

I already posted a graph of what core inflation ex-shelter would look like:

But let’s go one step further and measure what core inflation, *including* shelter looks like, using actual house prices as measured by the FHFA Index rather than owners equivalent rent, thus banishing the entire problem. Here it is:

Most importantly, as of August (the last month for which the house price index has been reported), YoY core inflation including house prices was 8.6%. At its peak 6 months earlier in February, it was 12.4%. If it were to continue to decline at that rate, in October core inflation using actual house prices would be 7.3% - declining fast, but still above the official 6.3% reading of core inflation using imputed rents. It would take until about next April for core inflation using house prices to get to the Fed’s comfort zone of 3% or less.

Housing permits and starts continue to fall, but housing under construction continues to (slowly) rise

 

 - by New Deal democrat



The monthly numbers for housing permits, starts, and single family permits all declined this month. Permits (red in the graph below) declined -38,000 annualized to 1.526 million annualized, and starts (blue) declined -62,000 annualized to 1.425 million, both the lowest since summer 2020. Single family permits (gold, right scale), which have the most signal and least noise, declined -31,000 annualized to 831,000, the lowest since May 2020 and before that, the lowest since April 2019:

Perhaps more importantly, here’s a variation on a graph I have run many times over the past 10 years, comparing the YoY change in interest rates, in this case mortgage rates (inverted, *10 for scale) with the YoY% change in total housing permits (red) and single family permits (gold):

These are well within the ranges of declines that have previously been consistent with recessions, with the exception of 1966, although frequently (not shown) the actual recession hasn’t started until there has been a -40% decline.

As I always point out, interest rates lead housing permits roughly by 3 to 6 months. As shown above, for example, this year mortgage rates turned negative (i.e., higher) YoY in April. Housing permits in total and for single family units followed in August. As of now, YoY interest rates have risen almost 4%, an increase only exceeded by 5% and 6% increases in 1980 and 1982 respectively, which coincided with -50% declines in housing permits (not shown). Thus, we should expect housing permits to have declined by about -40% by about January or February, to levels of 1.100-1.150 million and .700-.740 million units annualized, putting a likely recession start date in the 1st quarter of 2023.

BUT, this time it really is slightly different. In the past, housing permitted and not yet started, and housing units under construction, have typically turned down well in advance of the onset of any recession:

But because of a shortage of building materials, there have been record numbers of housing units that have been permitted, but have not yet been started (blue in the graph below), which appears to have peaked in July,  and consequently a big lag in housing under construction (red), which rose slightly again this month, although it has only risen 3.2% in the past 6 months:

Because housing under construction is the actual economic activity, this suggests that  residential housing has continued to contribute ever so slightly to GDP growth. 

This is reinforced by the generally coincident nature of employment in residential construction (gray in the graph below) compared with housing units under construction:

With the sole exception of the 2020 pandemic lockdown recession, construction, which is an even smoother metric than single family permits, has always peaked at least 6 months before the onset of recession, with a median time of 18 months, and as much as 47 months; and has declined at least 6.5%, and as much as 34%, with a median decline of 20% from peak:

In other words, there is a significant element of “it’s different this time,” in that construction has in the past typically followed a decline in permits by 0 to 11 months, with a median of 5.5 months. Further, a recession has typically followed a decline in construction by between 6 and 47 months, with a median time of 18 months; and has declined at least 6.5%, and as much as 34%, with a median decline of 20% from peak.

We are now 10 months out from the peak in permits, and construction has not yet rolled over. Nor has residential construction employment, which has also typically turned down months in advance of any recession. Thus past history would suggest no recession begins until at least 6 months from now, and possibly much later - depending on how quickly construction rolls over and how abruptly it declines.