Saturday, July 4, 2020

3/7/20: Labor Market COVID19ed


I have been running a regular update on my 'America's Scariest Charts' covering labor markets developments (see most recent one here: https://trueeconomics.blogspot.com/2020/07/2720-americas-scariest-charts-updated.html). These charts rely heavily on two data sets: Non-Farm Payrolls data (monthly frequency), and initial unemployment claims (weekly frequency). I ignored for now two other data series:

  1. Average duration of unemployment: this is, of course, rising, but from low levels, as the COVID19 crisis is still relatively young; and
  2. Continued unemployment insurance claims: these data have been also proximate to the initial unemployment claims through the period of February-May.
Now, with two months of some jobs recovery, it is worth looking at (2) above. So here they are: continued unemployment claims charts:

Let us start with history:


There is no scaling in the above chart, just the numbers of people claiming unemployment insurance on continued basis. Which is telling: in recessions, these rise; in recoveries they fall. You can see that the lowest unemployment claims tend to happen some months before the onset of the subsequent recession. And recoveries take long. Of course, in the 1970s, there were fewer people in the labour force and, therefore, the absolute numbers of the unemployed were also lower.

Which means, it is worth rescaling each episode of rising and falling unemployment claims to the pre-recession levels ('norm') and to the recession peaks, taking into the account how long does it take unemployment claims accumulated during the recession to drop back to the levels of pre-recession claims.

So, methodology. I define 'normal' unemployment claims level as being the lowest level attained in the 12 weeks preceding each recession. This is set as an index of 100 for every recession. We look at the period of 6 weeks prior to the onset of the recession to identify the starting level of recession in terms of unemployment levels (these are weeks -6 through 1 on the X-axis). We then look at subsequent weeks (non-negative values on the X-axis) and plot index of unemployment claims (current unemployment claims normalized to the 'normal' level) through the recession and into the recovery, mapping these until one of the two events occurs:

  1. Either the index returns back to 100 - meaning unemployment claims finally are restored to the level of the pre-recession levels or the 'normal'; or
  2. In cases where this does not happen, until the onset of the next recession.

First, let us do this for all recessions since 1967 (when the data starts) through the end of 2019 - ignoring for now the COVID19 period.

Lots of interesting stuff in the above.

  • 2008-2009 Great Recession was long - longer than any other recession - in terms of labor markets recovery to 'normal' levels of unemployment claims. It was also sharp - second sharpest on record - in terms of the mass of unemployment claims at the peak of the recession. 
  • Legacy of the 1990-1991 recession was also painfully long, but shallower on the impact side (peak levels of unemployment claims). 
  • Epic 1973-1975 recession was horrific: it had a long lasting impact on unemployment claims and, in fact, it never got the point of returning unemployment claims levels back to the pre-recession 'normal'. 
  • We normally think of the 2001 recession as being 'technical' - caused just by the gyrations in the stock markets, aka the dot.com bubble burst. But in reality it too was pretty long in terms of its impact on the unemployed and it was pretty sharp as well.

And so on... but now, time to bring in the COVID19 pandemic. Let us start by just plotting it with the rest of the data. Boom!


The COVID19 pandemic made so many people claim unemployment insurance - on continued basis, not just one-off first time claims that anyone can file - that you can no longer meaningfully consider the rest of the recessions in comparison. In data analysis, we say that COVID19 pandemic is an influential outlier - it distorts our analysis of all other recessions. In this case, it is useful to use logarithmic scale to visualize the data. So here it is:

 
Even with log-transform, as above, the COVID19 crisis is off-the charts! No one has ever seen anything like this. Which we know.

The recovery from the pandemic has been sharp as well (steeper slope than in other recessions), but both charts above highlight the fact that whilst the U.S. economy is restoring some jobs during May-June re-opening period, the process of restoring these jobs is slow. Unlike what you hear from the White House and the Republican Party and its media, we are not in a 'tremendous recovery' and there is no 'roaring growth'. There is a mountain of pain that is being chipped away. At a current rate of 'chipping away', it can take the jobs markets some 10 months to come back to the pre-COVID19 'normal'. But that assumes that there will be no permanent or long-term jobs losses from the pandemic and the aftermath of the pandemic. It also assumes that the second wave of COVID19 infections that the U.S. is currently experiencing is not going to lead to such losses of jobs, and will not result in return to April-May levels of restrictions, and will not trigger a third wave of pandemic in Autumn. All three 'ands' must hold to get to that 10 months recovery. 


Stay tuned, I will be updating this chart as we go.

No comments:

Post a Comment