What to look for if housing construction does forecast a recession

 

 - by New Deal democrat

No data today, but since it is mainly a housing week, let me pick up on a topic I discussed at the end of yesterday’s post; namely, if housing does indeed forecast an oncoming recession, what should we expect next in that sector?

To cut to the chase, ultimately we need to look to construction employment.

Briefly for background, I won’t bother reposting the graphs, but the most leading aspect of housing data are mortgage interest rates. After that the most leading data are new home sales (which are very noisy) and permits, with single family permits being the least noisy. Permits, starts, and sales all lead prices.

And, reposting from yesterday, permits substantially lead housing units under construction:

Permits also lead residential construction spending adjusted for headline inflation:

Now let’s compare housing units under construction with inflation adjusted residential construction spending. In the graph below I measure each YoY, and in the case of construction spending, subtract YoY headline inflation so that what is shown is the % by which YoY residential construction spending exceeds or trails overall inflation. Finally, I also include the YoY% change in employment in residential construction:

Inflation adjusted residential construction spending has typically led housing units under construction, and both have led residential construction employment.

For completeness’ sake, let’s compare house prices as measured by the Case Shiller repeat sales index with adjusted residential construction spending:

Construction spending has typically led house prices in the past 20 years.

In fact house prices adjusted for inflation have even lagged residential construction employment, and did not even turn down in the 2001 recession:

Now let’s bring the rest of the goods producing sector (mainly manufacturing but also notably non-residential construction employment) into the mix.

As we already know, manufacturing as measured by the ISM index has been contracting since 2022:

Again, although I won’t repost the graph, because manufacturing is only about 1/4 of the US economy, for recession forecasting purposes I have begun economically weighting it with the ISM services index.

So in the following graphs I compare the YoY changes in employment in residential construction, construction generally, manufacturing, and the entirety of goods production employment. 

First, here is the historical record from 1950 through 2002 (note that the subcategory of residential construction employment was only added in 1988):

Next, here is the period from 2003 until just before the pandemic:

Now here is our post-pandemic period:

Here’s the upshot of these three graphs: Focusing on manufacturing or residential construction employment alone is not enough. If one turns negative but not the other (e.g., 1966, 1984, 1994, 2002) a recession typically does *not* happen. It is only when there is a more broad-based downturn across multiple goods-producing sectors that a recession typically occurs. 

As you can see from the final graph, that YoY downturn has already manifested in manufacturing. It has not manifested in either residential or non-residential construction, nor in goods production generally.

Indeed, on an absolute basis, but residential and total construction employment are still increasing:

And total goods-producing employment only turned down in the past month (and that may be reflective of hurricane impacts):

To sum up: with permits, starts, and housing units under construction all down from their peaks, at levels at least close to consistent with an oncoming recession, the big item to look for is employment in residential construction, and construction generally. If manufacturing employment remains negative, and construction employment turns down, that would strongly indicate that more likely than not a recession is approaching.

Leading housing construction data stabilizes, while units under construction continue free-fall. With a continued hurricane asterisk

 

 - by New Deal democrat

Housing data for October, like that for September, has to be viewed with an asterisk, due to Hurricanes Helene and Milton. Since the effects of both were in the South Census Region, where relevant in the analysis below I will also discuss the numbers excluding that region.

The most leading metric, after revisions housing permits (gold in the graph below), declined -9,000 to 1.416 million. Excluding the South Census Region, permits increased 5,000. Single family permits (red), which are just as leading and have very little noise, rose 5,000 to 968,000, with revisions the highest since April. Excluding the South, they would have been unchanged. Housing starts (blue), which tend to lag permits by a month or two, and are much more noisy, declined -42,000 to 1.311 million. Excluding the South, they *rose* 20,000:

Accounting for the regional issues, this is a slight rebound from lows earlier this year, and it is what I have been expecting, for the simple reason that for 60+ years, mortgage rates have always led housing permits, and over the summer mortgage rates declined:

Given the upturn in mortgage rates in the past two months, this mild upturn is likely to be short-lived - although I am not expecting any new lows.

The bad news is that units under construction, the measure of real economic activity in this sector, declined another -29,000 (-19,000 excluding the South), or another 1.7% from its October 2022 peak, and is now down -14.4% from that peak. 

This is very important, because in the past it has declined on average -15.1% and by a median of 13.4% before the onset of recessions:

Two months ago I hoisted a yellow flag “recession watch” for the construction sector, based on this metric being down more than -10%. As of last month, and certainly continuing this month, it is at the level consistent on average in the past with the onset of a recession. The only reason I am not hoisting a red flag now is that, excluding the South, units under construction are only down -12.4%:

There are several other silver linings which suggest that we should not put too much weight on this decline.  The first is that in the lead-up to past recessions, the downturn was led by single family units under construction, the number of which continued to decline into recessions. This year the number of single family units under construction has remained stable for the past four months (far right of graph below):

Additionally, as shown in the long historical view of housing permits, with the exception of the tech producer-centered recession of 2001, they continued to decline sharply into the recession, and typically well after recessions had begun:

In our present situation, since it appears permits have bottomed, then housing units under construction will not decline too much further before bottoming as well. 

For those reasons, at this point there is only a housing sector ‘recession watch,’ meaning a heightened possibility, and not a ‘warning,’ meaning one is more likely than not. Once the hurricanes-induced asterisks are resolved, the issue will become whether mortgage rates continue to head back higher - in which case recession risks increase - or are at the top of their range going forward.

Whither housing? A look at interest rate and inflationary considerations

 

 - by New Deal democrat

Starting tomorrow we get to the time of month when the data on the important long leading sector of housing begins to be reported. So let me update a few important points about where this sector is likely going and its effects on the economy.

As we probably all know by now, housing costs in the form of Owners’ Equivalent Rent and Rent of Primary Rresidence are about 1/3rd of the entire CPI, and about 40% of core inflation. Where are they going?

Let’s start with rents. The Philadelphia Fed has an experimental “New Tenant Rents Index” and “All Tenants Regressed Rents Index” that are culled from the same data that gives rise to the CPI numbers, and are designed to lead the Rents portion of that index.

They were recently reported for Q3, and while unfortunately they are not presented in graph form, Ben Casselman of The Economist has provided such an update:

The graph suggests that the CPI measure of rents lags the “All Tenants” index by 2 quarters on average. Past day, most notably from 2009-10 and 2020-21 suggests that CPI for rent will continue to decelerate to approximately the same level as the All Tenants index, unless the New Tenants index rises significantly in the meantime. In Q3 the All Tenants YoY% increase was 3.8%, and the New Tenants YoY increase was 1.0%. On average the CPI rents YoY% change in Q3 was 4.9%. *If* the New Tenants index remains somnolent, this suggests the YoY% change in CPI for rents should decline to under 4% YoY in the next 4-6 months. In the past 6 months, the CPI for rent has risen at a 4.4% rate, so some further deceleration must occur if this forecast is to come true.

Similarly, the continuing slow decline in the house price indexes (FHFA is shown below) suggests that the CPI for Owners Equivalent Rent should also continue to decelerate:

There’s no magic 1:1 monthly correspondence to the lag in CPI vs. house prices. Above I show both the quarterly (dark red through Q2) and monthly (light red through August) changes in the FHFA index. Over the last two quarters, the YoY increase in the FHFA index has averaged roughly 5.5%. This suggests that the OER is currently “aiming” for roughly a 2.2% YoY increase in the next 12-18 months, depending upon the course of house prices in the near future.

But if shelter inflation can be expected to continue to moderate, the news is not so good about new construction.

As I always say, mortgage rates lead new construction. Here’s the latest updated graph of rates (blue, left scale) vs. permits (red, right scale):

More or less mirror images, with a slight lag between rates and permits. TNiehter the summer decline in rates, nor the subsequent increase, have yet been reflected in permits. Generally this suggests that permits will stay in the same range as they are presently.

Let me elaborate on this a little bit, and make a long term statement as well.

On the left below is a snapshot of the Treasury yield curve, showing its steep inversion in summer just before the Fed made its first rate cut (light red), and its current generally flat status (dark red):

As with the graph of mortgage rates above, what is most interesting here is that longer rates beginning at about the 7 year maturity have risen even as the Fed has cut rates twice. This suggests that the bond market, at the moment, thinks this will lead to a little more inflation. If mortgage rates do not meaningfully decline in the near future, the housing market is going to remain moribund.

And here is where the longer term secular issue comes in. It seems very likely that the next Administration is likely to want to pursue a high-deficit, easy money policy. If so, it almost certainly will ultimately get its wish with both the Congress and the Fed (after new appointments and possibly firings as well).

Such a policy is going to be inflationary, and is likely going to mean that long term interest rates like for mortgages will not return anywhere close to their post-pandemic lows.

Interest rates move in very long cycles. There was a declining rate long-term secular trend that started in 1981 and almost certainly ended in 2021:

Both the increases in the Fed funds rate and long term Treasury rates definitively broke those long term trend lines in the three years since.

And if there is going to be inflationary fiscal policy, and accommodating monetary policy, there is every reason to believe that we have started a new secular era that will be much like the era from the 1950s through 1981, which featured both increasing Fed interest rates and increasing inflation:

It is hard to imagine housing making any kind of strong positive contribution to the economy with those headwinds, unless by some miracle middle and working class incomes increase faster than inflation on a sustained basis.

Weekly Indicators for November 11 - 15 at Seeking Alpha

 

 - by New Deal democrat

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

More the recent ‘same’ this week: short term and coincident indicators are a little noisy, but continue to say the economic OK. Meanwhile the longer leading indicators are mixed and weighing on future growth.

As usual, clicking over and reading will bring you up to the virtual moment on the economic situation, and reward me with a little contribution towards my lunch money for collating and organizing it for you.

Production turns more negative

 

 - by New Deal democrat

Industrial and manufacturing production slid further in October, by -0.3% and -0.5% respectively. They are also down respectively -1.2% and -1.8% from their late 2022 highs:

It’s a good thing I suppose that manufacturing is no longer such an important part of the American economy, because as the below graph of the past 50 years shows, before 2001 and especially before the Great Recession, and decline YoY *always* coincided with or at least immediately heralded a recession:

But since the accession of China to regular trading status, even downturns of about -5% or more, as in 2015-16 and 2018-19 have not necessarily meant recession.

And as this close-up of the post pandemic record show, on a YoY basis production is only down about -1%:

 So, to reiterate what I wrote earlier this morning, it’s a good thing that consumption particularly of services continues to be positive.

Real retail sales jump nicely, but we’re not out of the woods on consumption just yet

 

 - by New Deal democrat

Let me start with my usual reminder that real retail sales is one of my favorite economic indicators, because it tells us so much about the state of the consumer, and since consumption leads employment, it is a short leading indicator for that as well.

In October retail sales rose 0.4% on a nominal basis. After adjusting for inflation, they rose 0.2%. There were also important positive revisions to September, which increased from a 0.4% gain to a 0.8% gain. The below graph norms both real retail sales (dark blue) and the similar measure of real personal consumption of goods (light blue) to 100 as of just before the pandemic:

Since the end of the pandemic stimulus in spring 2022, real retail sales had been trending generally flat to slightly declining, while real personal consumption expenditures on goods continued to increase. In the last four months, however, real retail sales appear to have broken out to the upside, in accord with the positive trend in real spending on goods, which is very positive.

An important reason why this breakout is so positive is that, on a YoY basis, real retail sales had been negative almost all this year. And over the past 75 years, a negative YoY comparison has usually meant a recession followed shortly (although that wasn’t the case in 2022 and 2023). But with this month’s data, YoY real sales are up 0.3%:

I had been increasingly concerned about the YoY trend in sales over the previous four months. For example, three months I concluded by saying that “the longer real retail sales go without posting a positive YoY number, the more concerned I will be.” The October number plus the September revision takes some of the pressure off.

Finally, real sales are a good if noisy short leading indicator for employment, as shown in the historical pre-pandemic graph below:

Despite the recent uptrend in real sales, since consumption leads employment and a 0.3% YoY gain is still very weak, this continues to forecast that on a YoY basis, job gains are likely to continue to decelerate. As you probably recall, the very poor October jobs report was distorted by hurricane impacts. Most of these will probably be reversed in the November report. But the three month trend of reports in the range of 75,000 to 175,000 appears likely. Here is the post-pandemic close-up:

Obviously this was a very good report. But we’re not completely out of the woods yet.