Construction spending in January declined, manufacturing construction tanked; but the AI data center Boom continued

 

 - by New Deal democrat

This morning the construction spending report for January was released (note that this is still about 3 weeks later than usual, so last autumn’s government shutdown continues to reverberate in the data). In the past I have used it to help track the long leading sector of housing, but in the wake of the Inflation Reduction Act, plus the chaos now in Washington, it has also been useful to track manufacturing. And finally, via private construction of water supply, as a proxy for construction of AI data centers.

And with the exception of that last item, the numbers in January, even nominally, were all negative. Total construction spending declined -0.3%, residential spending down -0.8%, and manufacturing spending down -2.0%. Non residential spending as a whole declined -0.4%. But spending on water supplies increased a sharp 3.3%. Below are all of the above, normed to 100 as of one year before (January 2025). I also show the PPI for construction materials to show how much spending there was for each in “real” terms:

Since the cost of construction materials (red) increased 6.6% during the 12 month period, only spending on (likely AI data center related) water supply increased in real terms.

Here is the same data measured YoY:

Note that almost every sector of construction has either slowed down or turned negative in the past year, and in particular manufacturing construction has declined sharply. Only residential construction has rebounded slightly on a nominal basis, and in real terms bottomed last spring. In contrast, the Boom in water supply construction is apparent. Notably, as shown in the graph below, even spending on water supply construction in real terms has declined since last September:

Most importantly, this paints a picture of spending in the two leading sectors - housing and manufacturing - declining through 2025. 

This is yet more evidence that the only sector which has been keeping the economy out of recession recently has been that releated to AI data centers.

American political (and military) support for Israel appears to be on borrowed time - maybe 10 years

 

 - by New Deal democrat

Occasionally on Sundays I have posted on things other than the economy. Some dramatic poll results I saw this past week called out for such a post.

A popular poster over at Bluesky named Micah posted what comes pretty close to my overall view of the 2024 Presidential campaign:

My theory of the Harris campaign, which makes exactly nobody happy: 1. she was put in an incredibly deep and arguably insurmountable hole by Biden 2. she made significant errors, largely in not separating herself from him enough 3. her results were still in the upper range of plausible outcome

Every incumbent party on the planet got walloped and Harris/Dems did better than almost all of them She overperformed in swing states where she campaigned heavily compared to the rest of the country She absolutely could have run a better campaign (Gaza!) I don’t know if it would have been enough

Given what she was handed, I think Harris ran as close to a perfect campaign as could have been reasonably expected.

But what I wanted to focus on in this post is his mention of her position on Gaza, which was to embrace Biden’s position of unwavering support for Israel.

I’m not sure if supporting the civilian population of Gaza would have on net gained her votes, because there is a significant bloc in the Democratic Party which would refuse to vote for anything short of complete support for Israel.

But whether or not that was true in 2024, the below poll results from Gallup published several weeks ago strongly indicates that US support for Israel is on borrowed time.

The headline result was that, for the first time, more respondents sympathized with Palestinians than supported Israelis, by 41%-38%:

But that isn’t the biggest result. Rather, it is the breakdown of support by *age* that is most breathtaking.

Americans over 55, most of whom remember the 1974 Yom Kippur War, and many who remember the 1967 War, continue to support Israel by a large margin, 49%-31%:

But middle-aged Americans, ages 35-54, now decisively sympathize with Palestinians, by 46%-28%:

And the most shocking result of all, Americans adults younger than 35 sympathize with Palestinians by a decisive 53%-23%:

Unsurprisingly, a similar survey from a year ago found that the deterioration in sympathy towards Israel was across the political spectrum, but most dramatic among Democrats. Negative views towards Israel among GOP supporters rose from 27% to 37%, but among Democrats it rose from a slim majority of 53% to a decisive one of 69%. In both cases younger adults (below age 50) ended with majority antipathy towards Israel, 50% among GOP leaners and 71% among Democrats:

Usually, basic political attitudes are formed while a person is very young, and are maintained throughout adulthood. Not only do the above polls show that attitudes have been shifting against Israel even among older age groups, but that a solid majority of younger Americans no longer support it politically. Under the theory that “progress will be made, one funeral at a time,” it seems likely that support for Israel will become toxic across all the political spectrum except for some hard core GOPers within 10 years. The only mitigating factor, as for as Israel is concerned, is that power in Washington is concentrated towards a Gerontacracy, but even so, unless there is a complete reversal of trend, how much longer can Israel hold on to American political - and even more importantly, military and diplomatic - support? And without such support, what is its future then?

Weekly Indictors for March 16 - 20 at Seeking Alpha

 

 - by New Deal democrat

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

Unsurprisingly, the big news this week is how the damage from the oil shock caused by the idiotic Iran war is spreading into more sectors of the economy. There was also a very unusual development in the bond market.

As usual, clicking over and reading will bring you up to the virtual moment as to the state of the economy, and reward me a little bit for gathering and organizing it all for you.

Prepare for a major shock in March and April CPI

 

 - by New Deal democrat

Prepare for the CPI to increase over 1.5%, and perhaps even 2.5%, just by the end of April. That’s the mesage conveyed by the huge spike in gas prices so far this month due to the war with Iran.

Let me start with the blockbuster graph. This is the monthly % change in gas prices at the pump since the onset of the data, measured weekly by the E.I.A:

Just through March 16, gas prices had risen from $2.94/gallon at the end of February to $3.72, an increase of 26.7%, the largest such monthly increase on record. Only two other months approached 20%: May 2009 at 18.9%, and March 2002 at 20.3%. And in both of those cases prices were rebounding from very low levels near the end of recessions. And remember, we still have over a week to go in March, and daily prices as I type this are already at $3.92/gallon according to GasBuddy.

My past rule of thumb has been that to correlate gas prices with CPI, divide them by 14, and add 0.15% for underlying inflation elsewhere. It’s definitely not perfect, but it is usually in the ballpark. For purposes of this post, however, I am being more conservative, dividing the change in gas prices (blue in the graphs below) by 16 and only adding 0.10% to arrive at the CPI forecast (red). Here’s what that looks like, first from 1998 through 2012:

And here is 2013 to the present:

You are reading that correctly. Even with a conservative measurement, a CPI increase of 1.8% is forecast. If you examine the graphs carefully, it is easy to see that the increase most often doesn’t show up in just one month. More often the increase is only 1/3 to 1/2 the forecast number in previous cases when there have been big increases in gas prices in a month. But more often than not, when we include the subsequent month as well as the month of the gas price increase itself, the outcome is very close to the monthly forecast. For example, after plummeting in 2015, in January 2016 the forecast due to the jump in the price of gas was 1.1%. The CPI in January and February of that year increased 0.7%. And more recently, in March and May 2022, the model forecast CPI increases of 1.2% and 0.8% respectively. In March of that year the CPI increased 1.1%, in April 0.3%, in May 0.9%, and in June 1.3%, for a total of 3.7%, due to a 48.6% peak increase in gas prices during that period. Jut through the 16th of this month, the increase in the price of gas was over half of that. At today’s average gas prices, that would be a 32.6% increase. 

In other words, based on past history and using conservative assumptions, the model forecasts a 1.8% increase in CPI between March and April. Using normal assumptions it would forecast a 2.1% increase in these two months. And if I were to plug in today’s $3.92/gallon average vs. $3.72, the model would forecast a 2.5% increase in consumer prices by the end of April. 

Three days ago I wrote about how real aggregate nonsupervisory payrolls were continuing to increase through February, negativing an imminent recession. I used the graph below to show the decomposition of that metric:

Real aggregate nonsupervisory payrolls have held up because wage increases of about 3.8% YoY have consistently been above the inflation rate of 2.5%-3.0%. But just the 1.8% increase in CPI forecast using conservative methods would mean a 4.0% YoY inflation rate. A 2.5% increase in inflation by the end of April would mean a YoY% increase of 4.7%, swamping wage increases. 

In the past, with one outlier (1979), real aggregate nonsupervisory payrolls have declined by between -0.3% and -1.4% from their peak until the onset of recessions, with a median of -0.8%. If we get a 2.5% increase in consumer inflation by April, that will likely more than meet that threshold.

A detailed look at interpreting expansionary and pre-recession layoff and unemployment signals

 

 - by New Deal democrat

This week, in addition to my usual look at jobless claims, especially in view of my post earlier this week breaking down the components of aggregate nonsupervisory payrolls,  I want to compare them with several other indicators of increased joblessness in terms of their expansion and pre-recession dynamics. 

First, let’s look at this week’s update. It is simply very hard for me to conceive of any recession being imminent as long as people aren’t getting laid off. The historical look at the 4 week moving average of claims shows that they have *always* trended higher before a recession begins, with the very least warning being 2.5 months in 1981:

Further, the monthly average of initial claims has almost always been higher by 10% or more YoY before the onset of recessions, with the exception of the oil shock of 1974 and the 2008 Great Recession:

Which means that this week’s data is a strong negative of any imminent recession, regardless of the copious amounts of other soft data. As I pointed out yesterday, it has been one of the three pillars holding up this expansion.

Initial claims declined -8,000 to 205,000, one of the lowest numbers in the past 50 years. The four week moving average declined -750 to 210,750. With the typical one week lag, continuing claims rose 10,000 to 1.857 million:

As usual, it is the YoY% change which is more important for forecasting purposes. So measured, initial claims were down -9.1%, the four week average down -7.9%, and continuing claims down -1.3%:

These are simply very positive numbers, totally inconsistent with any imminent recession.

Let’s do our typical look at how these compare with the unemployment rate. To back up a bit, here is the historical look at the 4 week average of claims compared with the unemployment rate, measured YoY:

Simply put, going back 60 years, initial claims have always started rising before the unemployment rate and always peaked before the unemployment rate, with the closest point being only 2 weeks during the 1980 recession.

Now here is the post-pandemic look:

Since last July, initial claims have been forecasting that the YoY comparisons in the unemployment rate would improve. Here is the same data in absolute terms:

Although the unemployment rate ticked up 0.1% last month, both initial and continuing claims still indicate downside pressure on that rate.

 Now let me expand this discussion to three other metrics of employment softness: layoffs and discharges from JOLTS, the number of newly unemployed by less than 5 weeks, and aggregate hours of nonsupervisory workers, both of which are from the monthly payrolls report. I included this last metric because, as I pointed out several days ago, it typically is the first component of aggregate payrolls to decline, with average wages holding up usually into the onset of recessions.

First, here is the historical look at each, broken down into two time periods because layoffs and discharges were not reported until 2001:

Now here is the post-pandemic view:

The short term unemployment less than 5 weeks metric is very noisy, so to extrude signal, can only be looked at on a 3 month moving average basis. Comparing the metrics, one 3 occasions short term unemployment turned up first; but on 3 other occasions initial claims turned up first. One time they turned up simultaneously. Meanwhile, the (inverted) number of hours worked always turned up last, frequently not until the onset of the recession.

Now let’s look at the same data YoY, first historically:

And now the post-pandemic view:

On a YoY% change basis, initial claims *always* turned up first, with the exception of 2001 when they turned up simultaneously with the three month average of short term unemployment.

So the first lesson is that initial claims are a much less noisy metric than short term unemployment, and are the more leading of the two.

Secondly, notice that while (inverted) aggregate nonsupervisory hours have always turned up last, they are the least noisy of any of the four metrics, and they have also turned up well before the onset of recessions.

In other words, when we compare all four metrics, the best thing to do is to look at initial claims, especially YoY, and then look for confirmation by aggregate nonsupervisory hours. Only when both signal is there the most reliable indication of a recession being close at hand. 

And at present neither are signaling. This pillar of the post-pandemic expansion remains firmly intact.