Weekly Indicators for November 9 - 13 at Seeking Alpha

 

 - by New Deal democrat

My Weekly Indicators post is up at Seeking Alpha.

With Joe Biden assuming the Presidency on January 20, and a announcement of a successful vaccine by Pfizer that should be available by next spring or summer, the short and long leading indicators should once again be giving their usual signals for next year. And we have a pretty good idea what that means....

As usual, clicking over and reading should bring you virtually up to the moment about the economy, and also reward me a little bit for my efforts.

Lack of inflation in September consistent with weak demand; real wages increase, but will the pandemic derail the gains?

 

 - by New Deal democrat

Consumer prices were unchanged in October, both on a seasonally adjusted and unadjusted basis:

But while the lack of inflation is good news in isolation, the last two months can also be viewed as a sign of economic weakness - lack of demand - from a recession.

Digging a little deeper, for the past 40 years, recessions had typically happened when CPI less energy costs (red) had risen to close to or over 3%/year. We are nowhere near that now (last 15 years shown in graph):

Again,  note that the YoY% change in inflation has decelerated since the outset of the pandemic, potentially another sign of weakness.

On the bright side, because wages are “stickier” than prices, typically as recessions beat down prices (or at least price increases), in real terms wages rise. That has been the case for the coronavirus recession as well:

It is the “real” buying power of wages among those still securely employed during a recession that is one of the engines that usually restarts growth.

Also as a result, as I’ve noted for the past several months, real hourly wages for non-supervisory workers have finally exceeded their previous 1973 peak, although part of that has been the asymmetric loss of jobs among some of the lower paid occupations:

Finally, one of the most telling metrics of the overall health of the middle/working class is that of real aggregate wages. After declining -13.8% from February through April, they have now recovered to a point -3.5% below their peak, approximately at the same level as they were in autumn 2018:

If the rate of gains over the past 4 months were to continue - a *very* open question - aggregate real wages would exceed their February level in about 6 months.

Of course, this data like almost all other economic data, remains at the mercy of the course of the pandemic - which is basically out of control in much of the US. It is also very much subject to the public policy that has been stalled in Washington for the past half a year, and is going to continue to be stalled until at least January 20.

New and continued jobless claims: best week of the pandemic; can it survive a new emergency?

 

 - by New Deal democrat

This week’s new and continued jobless claims, both seasonally adjusted and unadjusted, declined to new pandemic lows - but at levels roughly equivalent to their worst readings during the Great Recession.

On a unadjusted basis, new jobless claims declined by 20,799 to 723,105. Seasonally adjusted claims declined by 48,000 to 709,000. The 4 week moving average also declined by 33,250 to 755,250. Here is the close up since the end of July (for comparison, remember that these numbers were in the range of 5 to 7 million at their worst in early April): 


Unadjusted continuing claims (which lag initial claims typically by a few weeks to several months) declined by 402,298 to 6,486,000. With seasonal adjustment they declined by 436,000 to 6,786,000, both also new pandemic lows:


New jobless claims have declined almost 90% from their March and April pandemic highs! But the seasonally adjusted numbers are still about 50,000 to 100,000 higher than their worst readings of the Great Recession:


Meanwhile, continued claims are about 72.5% below their May pandemic highs:


But these are also 150,000 (adjusted, weekly) and 35,000  (adjusted, 4 week average) higher than their worst levels of the Great Recession.

Directionally this week was very good news, but on an absolute level the level of layoffs continues to be very bad - just nowhere near like it was 7 months ago. This is a part of the slow continued improvement in most of the “weekly indicators” I update each Saturday.  

As the pandemic is once again out of control in the majority of the country, and near emergency levels in parts of the upper Midwest and Mountain States, with no signs of new infections abating at this point, I have to think this is going to reverse, and reverse badly. But it hasn’t yet.

Semptember JOLTS report shows jobs recovery has been real - but will it continue?

 

 - by New Deal democrat

Yesterday’s JOLTS report for September showed a jobs market that continues to be, ever so slowly, on the mend. Openings and quits were up (good), and layoffs and discharges were down (also good). The only negative was that hires actually declined, although slightly.

We are far enough  past the worst of the pandemic jobs losses that it is worthwhile to compare the state of the various JOLTS components with the 2 previous recoveries from recession bottoms in the series’ histories (this because the JOLTS data only dates from 2001. 

In the two past recoveries:

• first, layoffs declined
• second, hiring rose
• third, job openings rose and voluntary quits increased, close to simultaneously

Let’s examine each of those in turn. In each case, I break out 2001-19 in a first graph and then this year in a second.

What appears below is that, although there has been some variation, the past several months have recapitulated the pattern from the last two early recoveries: the first two data series to turn - layoffs and hires - have indeed turned, while the last two - job openings and voluntary quits - have appeared to bottom but have had a much less dramatic rise.

This first graph compares layoffs and discharges (blue) with the 4 week average of initial jobless claims (red):

You can see that, by the end of the recessions, layoffs were already declining, and continued to decline steeply over the next 3-8 months before reaching a “normal” expansion level. The turning point coincides exactly with the much less volatile, but more slowly declining, level of initial jobless claims.

The same has been the case this year, as layoffs and discharges already declined to their “normal” level in May, while initial jobless claims peaked one to two months later, and have been declining (slowly) ever since - a pattern that continued in September.

Next, here are hires (red) and job openings (blue):

In the past two recoveries, actual hires started to increase one to two months before job openings.

This year, both made troughs in April, but hires rebounded sharply in May and June compared with job openings:

Since then both have essentially leveled off.


Next, here are quits (green) vs. job openings (blue): 

In the past two recoveries, actual hiring started to rise slightly before quits made a bottom. After that, both rose more or less together (suggesting it is openings that leads to the increase in voluntary quits).

This year, both have moved together, both making a trough in April, and rising   equivalently since:


Finally, because seasonal adjustments might not be giving us a true picture because of the enormous moves during this pandemic year, here are job openings (blue), hires (red), and voluntary quits (green), measured YoY without seasonal adjustments for the recoveries after the 2001 and 2007-09 recessions:

Nte that, even taking out the seasonal adjustments, hires rebounded first following the 2001 and 2008-09 recessions. Quits and openings have moved generally in tandem with a slight lag.

The same pattern appears this year:

I have broken out layoffs and discharges separately below, because the their level in April and May of this year would obliterate all other variations (note: inverted so that fewer layoffs shows as positive):

This metric returned to normal almost immediately after both of the past two recessions, and did so again by July of this year, and is improving slightly measured YoY.

In short, the JOLTS report shows that the recovery in the jobs market from the pandemic lows is real, and as of September was continuing. 

With the pandemic largely once again out of control, and approaching emergency levels in many parts of the country, it is an open question to say the least whether this will continue. I have to think that responsible State governments are going to reinstitute lockdowns for at least long enough to bring the pandemic back under some semblance of control.

Coronavirus dashboard for November 9: Wow (and not in a good way)

 

 - by New Deal democrat

US total infections: 9,968,155*

US average last 7 days: 108,737

US total deaths: 237,570

US average last 7 days: 939

*I suspect that the real number is about 16 million, or about 5% of the total US population

Source: COVID Tracking Project

While we have been riveted by the 2020 election, the pandemic has continued to rage out of control in parts of the US, particularly in parts of the upper Midwest and northern Mountain States.

At its peak, NY had an average daily rate of 51 infections per 100,000 people. Now,  17 States have infections rates higher than that:

The worst is North Dakota, at 174 infections per 100,000 people. By contrast, the worst country on the planet, Czechia, had 117 infections per 100,000 people at its recent peak:

In addition to North Dakota, 5 other States have infection rates equal to or surpassing that of Czechia:

Next, here is the infection rate in 3 US States with large outbreaks: NY, AZ, and ND:

Whether due to a demographic shift in those who are getting sick (the early death rate in the US was largely a factor of so many cases in nursing homes), or due to better treatments in hospitals, the rate of deaths in the summer outbreaks (as shown by Arizona below) and *so far* in the recent outbreaks, has not come anywhere near the lethality in the early outbreak in NY:

At the same time, the current outbreak in ND is roughly 3x as bad as the summer outbreak in AZ. At its worst, the death rate in AZ was 1.14 per 100,000 people, vs. 3.93 in NY in April.

But, if the death rate now proves similar to that in the summer outbreak in AZ, which peaked at a 7 days average of 52.8 infections per 100,000 people daily, then ND, which currently has  a 7 day average of 173.6 infections per 100,000 people, can expect a death rate of 3.75 per 100,000 people daily, very close to the death rate of 3.93 per 100,000 in NY at its peak.

Another way of looking at this is to compare the infection and hospitalization rates for NY and ND, below:

So far, the hospitalization rate in ND is less than 1/3 of that of NY at its worst.

Comparing hospitalizations with deaths, below:

We see that deaths in ND are about 40% of those of NY at its peak.

But hospitalizations lag infections by roughly 3 weeks, and deaths lag hospitalizations by about another 2 weeks. Infections in ND have almost doubled in the past 3 weeks. A doubling of ND hospitalizations would put it at roughly 2/3 of the NY peak, with deaths expected to follow suit.

And of course there is no indication yet that ND’s infection rate has peaked. Nor, with he possible exception of South Dakota, is there any such indication for any of the other 16 States with infection rates similar to Czechia.

Sadly, it appears likely that by Thanksgiving there is going to be a full-scale emergency in parts of the upper Midwest and Mountain States similar to the one that engulfed NYC in April. And as the reality of his defeat slowly sinks in, it is highly unlikely that Donald Trump or anyone in his Administration will give a damn.

The 2020 Presidential and Senate races: a postmortem

 

 - by New Deal democrat

Way back in June I started writing nowcasts for the 2020 elections. Here’s what my very first map looked like: 

And here’s what the last one looked like:

When I pushed the “toss-ups”, by lowering the threshold from 3% to 1%, NC, GA, and FL also became “lean Biden” States.

Here’s how the election actually turned out, based on results through today (note: Alaska has only counted 50% of its votes, so the outcome there is still unknown):

 Two weeks ago, I wrote 

“The “blue wall” is extremely likely to hold. And if it does, Trump’s chances of victory are foreclosed.”

That turned out to be right on point.

Further, last week, and almost every week beforehand, I included a sentence like this:

“Even so, Biden still has 279 “solid” or “likely” Electoral votes, enough to win without any “leaning” or toss-up States.”

That turned out to be the most prudent caution, as on average the polls missed very badly this year. The below 2 maps contrast the actual results with Nate Silver’s polling average for each State, 1st for the Presidency and 2nd for the Senate races, using the following color code:

0 - 1.9% deviation - no color

2% - 3.9% deviation - light color 

4% - 5.9% deviation - medium color

6% or more deviation - dark color

Here’s the deviations from Nate’s numbers for the Presidency:

And here they are for the Senate:

The median deviation in his percentages for Biden was -5.2%. In not even a single State was the deviation 2% or more in favor of Biden. The median deviation in his percentages for Democrats running for the Senate was even worse, at -6.6%.

 

By contrast, Larry Sabato’s final Presidential forecast only missed on North Carolina:

The median forecast in Sam Wang’s model at the Princeton Election Consortium was also too optimistic, but he correctly always included a map of what would happen if Trump outperformed the polls by 3%:

And yet another alternative approach that was much better was the Cook Political Report’s “Swingometer,” which was explained thusly:

“How it Works: Start with the results of the previous election, adjusted for demographic change since 2016. Then, adjust the sliders below to see how shifts in turn out and support among five demographic groups could swing the Electoral College.”

Using their turn-out assumptions, this is the map they generated:

The map only misses Arizona, Florida and Georgia. Further, note that it correctly shows the “blue wall” holding, but being very close. Using the same criteria as I used for Nate Silver’s projections, here’s how close the “Swingometer” came in each State:

The median deviation from the actual 2020 results known so far is only 2.0% in favor of Biden. In other words, simply projecting the same demographics 4 years forward, and estimating increased (or decreased) turnout based on past turnout for each age, on average only 2% of voters either changed their mind  in favor of Biden, or groups favorable to Biden turned out an average of 2% more of the total.

Finally, to see how closely Trump’s results matched results in the Senate races, I applied the same metrics to the actual  Presidential and Senate results. Blue (Red) means a result more favorable to, or less unfavorable to (less favorable to, or more unfavorable to), Biden than the Democratic Senate candidate:

The median deviation between the Senate and Presidential results was only +0.3% overall. When we ignore the direction of the deviation, it is about 3.5%.

To summarize: there are some very serious shortcomings in Nate Silver’s model. Both he and the Economist (G. Elliott Morris) would be well advised to pick up Sam Wang’s approach and create maps for significant poll deviations. Further, the “Swingometer” looks like a very promising tool. 

In terms of the 2020 election, it looks like very few voters actually changed their minds in the past 4 years in their opinion of Trump. It appears that new voters, or increased turnout among some demographics were the biggest factor in the different outcome this year. And a significant percentage of “never Trump” GOP voters split their ticket, voting the the GOP Senate candidate downballot.