Residential building construction and spending sound a warning for goods-producing employment

 

 - by New Deal democrat

The final data point from last week that I wanted to catch up on was construction spending, and especially residential construction spending.

As reported last week, in June total nominal spending declined -0.3% in June, but is higher 6.2% YoY. The more leading residential sector also showed a -0.4% decline, and is higher by 7.3% YoY:

Since producer prices for construction materials declined another -0.1% in May and are down -2.5% YoY, “real” total construction declined -0.2%, while residential construction spending declined -0.3%:

Finally, the Inflation Reduction Act, which conferred favorable tax benefits for “restoring,” led to a sharp increase in manufacturing construction spending, which rose only 0.1% for the month - its slowest rate of increase in nine months - to another new record:

Which means it decelerated to a “mere” 19.1% increase YoY. Only one month’s data, of course, but it may mean that the surge is plateauing.

Finally, let me tie this in to the discussion of leading employment metrics from the Establishment jobs survey I wrote about on Wednesday. Recall that recessions typically don’t happen until after both manufacturing and residential construction jobs, and more generally goods-producing jobs, have turned down. 

As of July, manufacturing jobs are treading water, but construction jobs, and in particular residential construction jobs, were still increasing, as were goods-producing jobs as a whole.

But there is good reason to believe that residential construction jobs will also turn down at some point in the next 6 months.

Here is the historical look at all employees in residential construction (blue), residential construction spending (red), and total building units under construction (black) from the inception of the first two series:

Notice that for the limited history that we do have, building units under construction and nominal residential construction spending have tended to peak simultaneously or close thereto. By contrast, the number of employees in building construction has not peaked until 6-12 months later (presumably, it takes a few months for employers to understand that the Boom is over, and to start laying off their crews).

By the end of last year, residential construction spending and the number of building units under construction both had clearly peaked. The number of building units under construction is now down almost -9%. 

Unless there is a sudden turnaround caused by lower mortgage rates, which hasn’t happened on any significant basis as of the present, we should expect layoffs in the residential building trades to start happening within the next few months. That might also cause a peak in goods-producing jobs, which would put employment on a very short term pre-recession footing.

Yet another reason why the Fed has waited too long to start lowering interest rates, and why they should not hesitate to make a 50 basis point cut in September.

Initial jobless claims, ex Hurricane Beryl distortions, remain very positive

 

 - by New Deal democrat

Officially reported initial jobless claims declined -17,000 to 233,000 last week. The four week moving average rose 750 to 240,750. With the typical one week delay, continuing claims rose 6,000 to 1.875 million, the highest number since November 17, 2021:

On the more important YoY basis, initial claims were down -9.7%, while the 4 week average was nigher by 0.3%. Continuing claims were higher by 5.8%, up by more than their recent range but still well below their YoY readings from several months ago and last year:

Because this is the first week of August, I’ll dispense with my usual note about the unemployment rate and the “Sahm rule.”

Instead, there’s a more important point to discuss: the impact of Hurricane Beryl, which made landfall in Texas on July 8, on jobless claims in July.

Up until then, my “unresolved seasonality” hypothesis was working very well. But after that, claims were indeed higher YoY.

Well, Ernie Tedeschi had a good point last Friday in his discussion of the jobs report. While the BLS downplayed any impact by Hurricane Beryl, it looks like it did play a role in depressing hours worked:

 https://nitter.poast.org/ernietedeschi/status/1819374644862112238#m 

That got me wondering whether it had also shown up in jobless claims. Because first you employer says we don’t need you this week, because we don’t have power. Then it puts you on temporary layoff.

After Hurricane Sandy a decade ago, and several Nor’easters and wildfires since, I created a workaround by subtracting the State(s) affected by the weather event, and looked at jobless claims including all the other States. This data is not seasonally adjusted, so you then have to apply the weekly seasonal adjustment to get the important result.

And indeed when you subtract Texas, YoY jobless claims (red in the graph below) are still lower than they were last July:

In July 2023, initial claims in Texas, not seasonally adjusted, averaged about 16,000. This year, beginning with the week of July 13, they spiked as high as 32,000 during the week of July 20. The next week they fell to just over 25,000, and this week [not yet updated on the FRED graph] they declined again to a little over 20,000:

When we take the YoY increase in Texas out of the total, NSA initial claims were about 207,600 during the week of July 27, and 199,000 this week. The seasonal adjustment for both weeks is roughly +15%.  When we then apply that seasonal adjustment, we get 240,000 initial claims last week, and 228,000 this week.

What this means is that initial claims ex-Texas remained lower during all but one week in July, and were lower still this week. In other words, when we take out the Beryl-related distortions, initial jobless claims have remained lower YoY, in line with the unresolved post-pandemic seasonality hypothesis.

Last year initial claims declined sharply as August progressed, to under 220,000. So the ultimate test of my hypothesis for the increase we have been seeing this summer will take place over the next 4-5 weeks.

But for now, importantly, initial jobless claims are not forecasting a recession at all, but rather continued growth.

UPDATE: The same is true for continuing claims. These spiked in TX by 20,000 one week after initial claims did (note: most current state data is only through July 20):

Because these are not seasonally adjusted, here’s what the YoY numbers look like (blue in graph below is TX; red is all States ex-TX):

When we take out the TX spike, continued claims were only 1.891 million NSA for the week of July 20. Applying the seasonal adjustment of .966% gives us 1.805 million ex-hurricane distortions, only 1.7% higher than the same week last year - the best YoY reading all this year for continuing claims.

Why the leading elements of the Establishment Survey in the jobs report still forecast expansion

 

 - by New Deal democrat

Continuing my catching up this week, let’s take a look in some further detail about why I didn’t think Friday’s jobs report portended recession - at least, not yet.

As I always point out, the jobs report does contain some leading numbers. These are generally employment in more cyclical industries that, when they turn down, start the cascade into the broader economy. Generally speaking, these are all goods-producing industries, and specifically manufacturing and construction. Usually I also include temporary help, but that has clearly been undergoing some structural changes, so I am temporarily discounting that.

Anyway, here is a closer look.

First, here are goods producing jobs (gold) vs. manufacturing jobs (blue) back in the post-WW2 era:

During this time, manufacturing was a much bigger slice of the goods-producing pie, so it is not surprising that the two lines look very similar. Except for the recession caused by the oil embargo (1974) and the Fed-engineered recession of 1981-82, caused by the sudden reversal and sharp increase in rates by Paul Volcker, these always started declining before jobs generally started to decline.

Next, here is the era including when residential construction jobs (red) were specifically broken out:

Note that these have relatively speaking become a larger slice of the pie, as a larger population needs more residences built. They too have always turned down (except for the pandemic) before the overall jobs number have turned.

Finally, here is the post-pandemic close-up:

Manufacturing employment has stalled, but has not in any meaningful sense turned down. Residential construction employment and goods-producing employment generally have continued to rise.

Historically, even more leading than the number of manufacturing jobs (red) has been the average work week in manufacturing (blue). In fact, it is one of the 10 “official” leading indicators in that Index. The below two graphs show the leading relationship on a YoY basis:

Here is the post-pandemic close-up:

There were significant YoY declines in 2022 into early 2023, but the YoY comparison has stabilized this year. This suggests that manufacturing employment is not poised to decline significantly in the months ahead.

Here is the absolute number of weekly hours in manufacturing worked historically:

I include this because the situation changed after 1982, with employees typically working more overtime. It has usually taken not just a decline, but a decline all the way to 40.5 hours before a recesion has begun. Except for last winter, we have remained above that line.

For completeness’ sake, here is temporary employment:

This has been almost relentlessly declining for over two years. There is pretty clearly something structural rather than cyclical going on here.

Finally, short term unemployment (blue in the graph below) was historically one of the leading components of the index, before it was replace by initial claims (red):

You can see that they follow similar trajectories, although initial claims are considerably less noisy.

Here is the post-pandemic look:

both have increased in recent months, with the former starting before the latter. Again, I suspect this has to do with recent immigrants being unable to find jobs, and thus not applying for unemployment.

To conclude, not everything is positive, but the two negative elements both have special considerations. The bulk of the leading data from the Establishment Survey of the jobs report is still forecasting economic expansion in the next few months.

Credit conditions in Q2 improved, and are typical of an economy having come *out* of a recession, not going in to one

 

 - by New Deal democrat

The Senior Loan Officer Survey, the premier quarterly measure of the loose- or tight-ness of bank lending, was published yesterday for Q2. And since lending conditions are a long leading indicator for the economy, and several of the metrics contained in this release have a good and lengthy track record, let’s take a look.

And to cut to the chase, the news is positive.

The first measure that has a lengthy and accurate historical record is the percentage of banks tightening or loosening criteria for making commercial and industrial loans. This is one of those data series where a positive number is negative, as it means more banks tightening than loosening. A negative number means more banks are loosening conditions.

Although the data for lending to both large and small firms was negative, it was less negative than at any point in the last two years, with on net only 8% of banks tightening:

Going back 30+ years, this is typically what we see 2 to 4 quarters *after* a recession has ended, not going in to one.

The second measure is the percentage of banks reporting strong demand for commercial and industrial loans. This is divided into large and small banks, and large and small firms, giving us four series. And in this case, positive means positive for the economy as well. In the graph below, the two series for large firms are in light and dark rend, the two fo small firms are in light and dark blue:

Although the graph looks a little like spaghetti, the trend in the past 6 quarters is clear. The percentage of banks reporting weakening demand got smaller and smaller, until finally this past quarter on net the data was positive, with the percent of large banks reporting stronger demand was higher by 10%+, and the percent of smaller banks reporting weaker demand dwindled to under 10%, for an unweighted average of +4%. Across all four divisions, the numbers are the best since Q3 of 2022.

Again, this is very much what we see just after having come *out* of recessions rather than going into one.

By no means are all of the long leading indicators so sanguine; in particular real money supply and the yield curve are still negative. But credit conditions are improving.

Economically weighted ISM indexes show (and forecast) an economy still - barely - in expansion

 

 - by New Deal democrat

As I was traveling last week, I did not write about several data series that I normally update. I plan on taking care of that this week. There’s also a little excitement in the markets today. Typically when there has not been drastic *hard* news, the action is all about leveraged positions being unwound in disorderly fashion, setting up a “V”-shaped market correction. We’ll see.

In the meantime, this morning the ISM non-manufacturing index for July was reported. Since I have started paying more attention to this as part of an economically weighted short term forecasting tool along with the ISM manufacturing index, which is one of the items I wasn’t able to get to last week, let’s take a look now.

Last Thursday the manufacturing index deteriorated further, with the headline number declining from 48.8 to 46.8, and the more leading new orders subindex declining from 49.3 to 47.4. Since 50 is the dividing line between expansion and contraction, that puts both metrics further into contraction (although note that both were lower during 2022 and also during 2015-16, when there were no recessions):

This morning the non-manufacturing index bounced back from contractionary levels in June, with the headline index increasing from 48.8 to 51.4, and the new orders subindex increasing from 47.3 to 52.4. That puts both metrics back in expansion:

Since the turn of the Millennium, when we use an economically weighted average of the non-manufacturing index (75%) with the manufacturing index (25%), it has generated a much more reliable signal, when we use the 3 month average, requiring it to be below 50. 

So what does it tell us now? Including July, here are the last sis months of both the manufacturing (left column) and non-manufacturing index (center) numbers, and their monthly weighted average (right) :

FEB 47.8  52.6. 51.4

MAR 50.3. 51.4. 51.1

APR 49.2  49.4.  49.3 

MAY 48.9. 53.8. 52.5

JUN 48.5. 48.8. 48.7

JUL. 46.8. 51.4. 50.2

And here is the same data for the new orders components:

FEB 49.2  56.1. 54.4

MAR 51.4. 54.4. 53.6

APR 49.1. 52.2. 51.4

MAY 45.4. 54.1. 51.9

JUN. 49.3  47.3. 47.8 

JUL.  47.4. 52.4. 51.2

While the single month average for both the headline and new orders components showed contraction in June, it did not trigger a signal based on the three month average. For July, the current three month weighted average of the two for both the headline and new orders components is the same: 50.3.

Once again, the recession warning signal has not been triggered. And once again, as I wrote last month, “the signal for the combined weighted ISM indexes remains expansionary - but just barely - in its forecast for the next few months.”

Just like last Friday’s Establishment Survey portion of the jobs report, this is a vary weak expansionary economy. The Fed should have cut last week. This is more data that it should start cutting rates at its September meeting, if there are no panic-induced emergency cuts before then.

Weekly Indicators for July 29 - August 2 at Seeking Alpha

 

 - by New Deal democrat

My “Weekly Indicators” post is up at Seeking Alpha.

Unlike the jobs report, the high frequency data has only shown slight weakening in a few metrics in the past several months. One which did turn from positive to neutral, per my blog post on Thursday, was initial jobless claims.

To keep up to the virtual moment on the economic data, click over and read. It will also reward me a little bit for my efforts organizing and highlighting the metrics.