Monday, March 18, 2024

Manufacturing and construction vs. the still-inverted yield curve

 

 - by New Deal democrat


Prof. Menzie Chinn at Econbrowser makes the point that the yield curve is still inverted, and has not yet eclipsed the longest previous time between onset of such an inversion and a recession. So he believes the threat of recession is still on the table.


And he’s correct about the yield curve, although it is getting very long in the tooth. In the past half century, the shortest time between a 10 minus 2 year inversion (blue in the graph below) to recession has been 10 months (1980) and the longest 22 months (2007). For the 10 year minus 3 month inversion (red), the shortest time has been 8 months (1980 and 2001) and the longest has been 17 months (2007):



At present the former yield curve has been inverted for 20.5 months, and the latter for 16.5 months. So if there is no recession by May 1, we’re in uncharted territory as far as the yield curve indicator is concerned.

My view for the past half year or so has been much more cautious. While there has been nearly unprecedented Fed tightening (only the 1980-81 tightening was more severe), on the other hand there was massive pandemic stimulus, and what I described on some occasions as a “hurricane force tailwind” of supply chain unkinking. If the two positive forces have abated, does the negative force of the Fed tightening, which is still in place, now take precedence? Or because interest rates have plateaued in the past year, is it too something of a spent force? Since I confess not to know, because the situation is unprecedented in the modern era for which most data is available, I have highlighted turning to the short leading metrics. Do they remain steady or improve? Or do they deteriorate as they have before prior recessions?

First of all, let me show the NY Fed’s Global Supply Chain Index, which attempts to disaggregate supply sided information from demand side information. A positive value shows relative tightening, a negative loosening:



You can see the huge pandemic tightening in 2020 into 2022, followed by a similarly large loosening through 2023. For the past few months, the Index has been close to neutral, or shown very slight tightness.

Typically in the past Fed tightenings have operated through two main channels: housing and manufacturing, especially durable goods manufacturing.

Let’s take the two in reverse order.

Manufacturing has at very least stalled, and by some measures turned down to recessionary levels.  Last week I discussed industrial production (not shown), which peaked in late 2022 and has continued to trend sideways to slightly negative right through February.

A very good harbinger with a record going back 75 years has been the ISM manufacturing index. Here’s its historical record through about 10 years ago (when FRED discontinued publishing it):



And here is its record for the past several years:



This index was frankly recessionary for almost all of last year. It is still negative, although not so much as before.

Two other metrics with lengthy records are the average hourly workweek in manufacturing (blue, right scale), which is one of the 10 “official” leading indicators, as well as real spending on durable goods (red, measured YoY for ease of comparison, left scale):



As a general rule, if real spending on durable goods turns negative YoY for more than an isolated month, a recession has started (with the peak in absolute terms coming before). Also, since employers generally cut hours before cutting jobs, a decline of about 0.8% of an hour in the average manufacturing workweek has typically preceded a recession - with the caveat in modern times that it must fall to at least roughly 40.5 hours:



The average manufacturing workweek has met the former criteria for the last 9 months, and the latter since November. By contrast, real spending on durable goods was up 0.7% YoY as of the last report for January, and in December had made an all-time record high.

But if some of the manufacturing data has met the historical criteria for a recession warning, it is important to note that manufacturing is less of US GDP than before the year 2000, and had been down more in 2015-16 without a recession occurring.

Further, housing construction has not meaningfully constricted at all. The below graph shows the leading metric of housing permits (another “official” component of the LEI, right scale), together with housing units under construction (gold, *1.2 for scale, right scale), and also real GDP q/q (red, left scale):



Housing permits declined -30% after the Fed began tightening, which has normally been enough to trigger a recession. *BUT* the actual measure of economic activity, housing units under construction, has barely turned down at all. In comparison to past downturns, where typically it had fallen at least 10%, and more often 20%, before a recession had begun, as of last month it was only 2% off peak!

The only other two occasions where housing permits declined comparably with no recession ensuing - 1966 and 1986 - real gross domestic product increased robustly. This was similarly the case in 2023.

An important reason is the other historical reason proppin up expansions: stimulative government spending. Here’s the historical record comparing fiscal surpluses vs. deficits:



Note the abrupt end of stimulative spending in 1937, normally thought to have been the prime driver of the steep 1938 recession. Note also the big “Great Society” stimulative spending in 1966-68, when a downturn was averted (indeed, although not shown in the first graph above, there was an inverted yield curve then as well). Needless to say, there as been a great deal of stimulative fiscal spending since 2020 as well.

Fed tightening typically works by constricting demand. Both government stimulus and the unkinking of supply chains work to stimulate supply. 

All of which leads to the conclusion that, while manufacturing has reacted to the tightening, the *real* measure of construction activity has not, or not sufficiently to be recessionary.

Tomorrow housing permits, starts, and units under construction will all be updated. Unless there is a sharp decline in units under construction, there is no short term recession signal at all.

Saturday, March 16, 2024

Weekly Indicators for March 11 - 15 at Seeking Alpha

 

 - by New Deal democrat


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

Not a lot of movement in any indicator this week. The long leading background, even after all this time, still tilts negative even though “less bad.” And the shorter term data continues mixed, as some sectors are very positive, and a few are very negative.

As usual, clicking over and reading will bring you up to the virtual moment on the economy, and bring me a little lunch money for my efforts.

Friday, March 15, 2024

Industrial and manufacturing production improve for the month, but 16+ month fading trend continues

 

 - by New Deal democrat


Industrial production is an indicator that has faded somewhat in importance in the modern era since China’s accession to normal trading status in 2000. Before that, a downturn in production was an excellent coincident indicator for a general downturn in the economy. Since then there have been several downturns, most importantly during 2015-16, when the broader economy, most notably housing and the consumer, did not follow. That was again the case of the downturn in 2023 - which has not resolved yet.

This morning’s report was another case of good news and bad news. The good news is that industrial production rose 0.1% for the month, and manufacturing production rose 0.8% from downwardly revised January numbers:



The bad news is that both remain down YoY, by -0.3% and -0.2% respectively (which, note, is not as bad a YoY comparison as either 2015-16 or 6 to 12 months ago), and both remain down from their respective peaks of September and October 2022. Here’s the YoY view:



In fact, the general trend over the past half year appears to be a further slight fade from a secondary peak in the summer of 2023.

This is, needless to say, particularly unwelcome in view of yesterday’s poor real retail sales report. If both manufacturing and the consumer are stallling out, that is not good. It will heighten the importance of this month’s report on personal spending, to see how well the broader measure of real spending on goods in particular is holding up.

Thursday, March 14, 2024

Good news and bad news Thursday: the bad news is real retail sales

 

 - by New Deal democrat


The bad economic news this morning was that after taking into account inflation, retail sales, which rose 0.6% nominally, were only up 0.2%, and last month’s number, which I described as making a “face-plant,” was revised down a further -0.3% to -1.1%.

In other words, the net result was that real retails sales were -0.1% worse than last month’s poor result as initially reported.

Which is bad enough. But it means that the last two months are the worst post-pandemic numbers in almost three years. Below I show them in comparison with real personal consumption on goods, the similar metric from the personal income and spending report, normed to 100 as of just before the pandemic:



I included the second number above because real retail sales and real personal spending on goods tend to track one another fairly closely over time, and both (/2) tend to forecast the trend in nonfarm payrolls. What has been compellling over the past half year is the marked divergence between the two spending measures, as retail sales have declined, while real personal spending has continued to increase.

 Here’s the record of both compared with jobs going back 15 years measured YoY:



On that same YoY basis now, real retail sales (blue) are down -1.6%, after a revised -2.0% in January, meaning a (noisy!) trend forecast of a YoY decline in jobs of over -0.5%, vs. the real personal spending forecast of roughly a 1% gain: 



Usually in the past (as, going back almost 75 years) such a decline in real retail sales has meant recession - but not in the last 18 months. I continue to expect the unusual large divergence between the two spending measures to resolve, hopefully in the direction of real personal spending.

Good news and bad news Thursday: the good news is jobless claims . . .

 

 - by New Deal democrat


This morning brought us both good and bad economic news.


The good news was that initial jobless claims continue very low, at 209,000, down -1,000 from last week, while the four week average declined -500 to 208,000. Even better, after major downward revisions, continuing claims rose 17,000 to 1.811 million:



Recall that continuing claims had been reported over 1.900 million, so as I said above, this was major!

On the more important YoY basis for forecasting, initial claims are down -14.3%, the four week average down -7.2%, and continuing claims, which before revisions had been running at about 10% higher YoY, are now only up 2.2%:



This is all very positive for continued good employment numbers in the months ahead (but see my next post today!).

Last week the unemployment number very much did NOT do what I expected, which was to remain steady or decline. Instead it rose to a new 2+ year high of 3.9%. I wondered whether, because unemployment includes both new and existing job losses, it followed continuing claims more than initial claims (although initial claims lead both).

Here’s the long term pre-pandemic trend (divided into two parts for easier viewing)(continuing claims /8 for scale):




Historically, as I’ve always pointed out, initial claims lead both continuing claims and the unemployment rate. The above graph shows that continuing claims also lead the unemployment rate, although with much less of a lead time.

So here is the post-pandemic record:



The divergence between initial and continuing claims beginning this past autumn looks like it indeed has passed through into the unemployment rate. Since the historical record remains that initial claims lead continuing claims, and in the past three weeks (post revisions!) continuing claims have declined sharply, we’ll see how this shakes out after the full month of March.

Wednesday, March 13, 2024

The most potent labor market indicator of all is still strongly positive

 

 - by New Deal democrat


On Monday I examined some series from last Friday’s Household survey in the jobs report, highlighting that they more frequently than not indicated a recession was near or underway. But I concluded by noting that this survey has historically been noisy, and I thought it would be resolved away this time. Specifically, there was strong contrary data from the Establishment survey, backed up by yesterday’s inflation report, to the contrary. Today I’ll examine that, looking at two other series.


Historically, as economic expansions progress and the unemployment rate goes down, average hourly wages for nonsupervisory workers improve at an increasing rate (blue in the graph below). But eventually, inflation (red) picks up and overtakes that wage growth, and a recession occurs shortly thereafter. Not always, as we’ll see in the graph below, but usually:



As you can see, there have been a number of exceptions to the rule, chiefly where inflation outstripped wage growth, but no recession happened anyway. Typically this has occurred because of the entry of so many more people (like women in the 1980s and early 1990s) into the labor force.

And we certainly see that inflation outstripped wages in 2022, not coincidentally when there were several negative quarters of real GDP. But with the decline in gas prices, in 2023 inflation subsided much more sharply than wage growth, and the economy improved more substantially. That has remained the case in the first two months of 2024.

But an even more potent indicator is one I have come to rely on even more: real aggregate payrolls for nonsupervisory workers. Here’s its historical record up until the pandemic:



There’s not a single false positive, nor a single false negative. If YoY aggregate payroll growth is stronger than YoY inflation, you’re in an expansion. If it’s weaker, you’re in a recession. Period.

And here is its record since the pandemic:



Real aggregate nonsurpervisory payrolls are positive, and they got more positive in 2023 compared with 2022. Currently they are 2.6% higher YoY than inflation.

In addition to the YoY comparison, real aggregate nonsupervisory payrolls have always declined, at least slightly, from their expansion peaks before every single recession in the past 50 years except for when the pandemic suddenly shut down the economy:



Not every slight decline means a recession is coming. But if real aggregate payrolls are at a new high, you’re not in a recession, and one isn’t likely to occur in the next 6 months, either.

And in case it isn’t clear from that long term graph, here’s the short term graph of the same thing:



Real aggregate nonsupervisory payrolls made a new all-time high in February. Despite the negative metrics in the Household survey, this is *very* potent evidence that not only are we not in a recession, but one isn’t likely in the immediate future either.