Two cheers for a great February industrial production report! With a gigantic *

 

 - by New Deal democrat

Two cheers for the very good industrial production report for Feburary! Total production increased 0.7%, and manufacturing production increased 0.9% to the highest level in over two years:

But the biggest news is that the headline number wasn’t just the best since the pandemic: it was the highest number of all time since the index was first reported over 100 years ago:

The reason total production did even better than manufacturing was all about utilities, which in January and February were 5% higher than any previous month ever:

So why only two cheers and not three? 

Industrial production is a coincident indicator. It was the King of Coincident Indicators, but has faded in importance since the turn of the Millennium, as other sectors of the economy have become more important.

And the best leading indicator for industrial production is new orders. I track this weekly via the Regional Fed reports, which showed marked increases in the past several months, and even more significantly via the ISM manufacturing index, the most recent graph of which, showing new orders in gold, is below:

New orders spiked in December and January to the best levels in nearly three years, but then fell back into contraction in February.

And there is a very likely reason for that: front-running tariffs. If I expect input costs to increase sharply due to the imposition of tariffs, I am going to complete as much production ahead of time as I can.

We won’t know for at least another month whether that has indeed been the case. But it is a gigantic *, and a reason for withholding the final, third, cheer.

February housing construction rangebound, but at recessionary or near-recessionary levels

 

 - by New Deal democrat

As usual, the month’s important housing data starts out with construction. 

For a quick refresher, I follow this because housing typically leads the rest of the economy by a year or more. After the very leading, but very noisy and heavily revised new home sales (which will be reported next week), permits turn first. Single family permits in particular are the least noisy of all the housing indicators. Next, with a month or two delay, come starts, which are also much noisier and work best as a three month average. Among the most important data, next - with a significant delay - comes housing units under construction, which is the actual total economic activity. Finally - again frequently with another significant delay - comes employment in housing construction, which is part of the monthly employment report. And mortgage rates lead all of the above.

Last month I noted that new home sales have been rangebound for the last two years, which suggested that permits would likely continue to be the same. And in February they were, as permits declined a slight -17,000 on an annualized basis to 1.456 million. Single family permits (red, right scale below) declined only -2,000 to 992,000. Starts (light blue) on the other hand rose 156,000 annualized to 1.501 million. Their three month average rose to 1.459 million, the highest in exactly 12 months:

It is possible, as with so much other post-pandemic data, that there is some unresolved seasonality, because as you can see the December-February period was also the 12 month peak for starts one year ago. Most likely we are seeing a noisy recapitulation of the slight improvement in permits since last summer, but still as with single family home sales, within a narrow range.

In the past 18 months, I have paid much more attention to housing units under construction (gold in the graph below), the “real” measure of economic activity in housing. Throughout late 2022 and all of 2023, it levitated at near record levels. It then precipitously declined, and as of February it is down -14.8% from its peak:

For the last number of months, I have suggested that this number would likely stabilize soon. It took longer than I thought, and with a deeper decline than I thought, but it appears that since November it has indeed stabilized at a level of roughly -15% below peak.

This is significant for recession forecasting purposes. Below I show units under construction YoY (red, left scale) and in absolute terms (blue, right scale):

On the one hand, on a YoY basis units under construction have never declined so strongly without a recession occurring in the near future. But the typical decline from peak has been much more than -15%, often being -35% or more, although recessions have occurred with only a-10% or so decline.

The last shoe I would expect to drop significantly before a recession is the number of jobs in homebuilding (blue in the graph below):

Like units under construction before them, housing employment has levitated for the past year, even rising to ever new levels. As you can see from the last 1980s, sometimes this has persisted for several years before the downturn. I am not expecting it to persist for much longer now. A -10% or so downturn in this type of employment has typically been the recession onset marker.

Finally, here is an updated graph of the YoY change in mortgage rates (*10 for scale, inverted), together with the YoY% change in total (light red) and single family (dark red) permits:

Mortgage rates dipped from 7% to 6.2% in late summer, shown as the nearly 10% improvement (blue) in the graph above. So far permits are nearly unchanged YoY. Some at least temporary improvement in their YoY comparisons seems likely, but since mortgage rates have recently risen back close to 7%, permits and starts are more likely generally to continue in their recent range.

The conclusion: this month’s housing construction report showed more of the same rangebound data. That’s “good” in that it is not worsening, but we can expect employment to finally follow the rest of the data down sometime soon. That would confirm that the housing sector is recessionary.

Real retail sales show a significant downdraft, but still expansionary

 

 - by New Deal democrat

Let me start out this month with a historical review of why I have always paid so much attention to real retail sales.

Going back almost 80 years, real retail sales have always turned down in advance of a recession, and they have almost always turned negative YoY simultaneously or close thereto with the onset of recessions. Here’s their record prior to the Great Recession:

With the exceptions of 1952, 1966, and a couple of months in the 1980s, this was a flawless indicator.

Here is the record of the updated version from the 1990s until the pandemic. For reasons I’ll explore more fully below, I also include real sales using CPI ex-shelter (light blue, narrow) and real personal consumption of goods (red):

Again, a flawless record, with the exception of a few oddball one month outliers. Real sales ex-shelter and real spending on goods have similar records.

 

There is also a demonstrated 50+ year history that consumption leads jobs.  It is the change in demand revealed by sales which leads employers to add or subtract workers.  And real retail sales have traditionally been an excellent measure of consumption. Here’s the modern record from the 1990s until the pandemic:

So I have a nearly flawless short leading indicator of the economy, and also a noisy but leading indicator for employment as well.

With that into, this morning’s read for February retail sales was a disappointment. In nominal terms, sales increased 0.2%. But because consumer inflation also increased 0.2% in February, real sales were unchanged. Ordinarily this would not be a particularly negative result, but January’s measure, which already was a negative -0.9% nominally, was revised further downward to -1.2%. That’s a significant downdraft. But it is also at variance with the result using CPI ex-shelter and real spending on goods. In particular, when we exclude shelter prices from CPI, although there has been a genuine downturn in the past two months, real sales remain above all levels except for last autumn:

Note that I have started including ex-shelter real sales in the last few months because house prices have had such an outsized distorting effect on the shelter component of CPI since the pandemic.

Now let’s update the YoY figures since the pandemic.

Following the pandemic stimulus, real retail sales as an indicator completely failed, likely partly because consumers had stuffed themselves on goods purchases with their stimulus funds, and turned to spending on services instead. And partly because house prices as measured by “owners equivalent rent” distorted CPI to the upside, the unit of comparison was unusually unrepresentative. 

That being said, we are now three years out from the stimulus, and shelter prices are less distorting now than in the past several years, so I expect the historical relationship to gradually re-assert itself.

And YoY, even with the downdraft of the past two months, YoY real retail sales are higher by 0.3%. Excluding shelter, they are up 1.1%:

Neither measure is recessionary. 

Finally, because as I noted above, consumption leads employment, per the above paradigm, let’s plug in the latest real sales and consumption data and compare it with the latest jobs trend:

The two series have been coming much closer to being in sync. Keep in mind that based on the most recent QCEW updates, I anticipate further downward revisions to last year’s jobs data. Still, while the forecast remains for positive employment reports, the suggestion is that there is likely to be continued deceleration.