Survey based employment data is suspect.

The Household Survey published by the Bureau of Labor Statistic (BLS) has heavy cross currents from both seasonal adjustments (the X13 ARIMA machine and the churn as folks leave enter or pause in the defining their labor status which produce very large margin of errors which seem to be smoothed or even manioulated.

The unemployment headline rate (UER) was capped at no more than 10% during the worst of the 2008 to 2010 crisis - the likely true peak in UER was around 13% to 14%.

When the economy improved rapidly - as is the norm - the drop in UER seemed near contrived and discrete. During the Yellen Chairmanship at the Fed while Groshen headed up the BLS - there seemd unusual coincidence between what the monthly survey data and the Yellen Fed announced objectives. This seemed to intensify during the 4 quarters before the Novemebr 2016 election.

So using the Household Survey and the other popular survey, Non Farm Payroll (NFP) does not provide useful insight as to the current state of the US economy. (I do not consider the recent December 6th NFP 266,000 as reliable and is even suspect and contrived, despite the popular press reporting.

The state unempoyment claims data is not survey based, but a census.

The state unempoyment claims data is not survey based, but a census, and no statistical error. Whats more the seasonal X13 ARIMA adjustments, not being survey, comes out quickly if there is an unusual seasonal adjust anomaly. ICSA is no longer effective in terms of signaling when the recovery phase is slowing or even starting to reverse. It becomes confirming, however.

The “Continuing Claims” (CCSA) data, how many are maintaing their unemplyment insurance payments is useful during a recovery to plateau and even the start of a slowdown, while it is not as useful when the recession starts.

Therefore a better and more current measurement of how the recovery is doing, at this stage of a recovery, is to focus on CCSA.

CCSA is plotted above and it clearly shows the business cycle phases. It is the most sensitive and indicative of the economy during the late stage recovery phase.

The year over year % change in CCSA levels is a good way to follow this data array.

When the time axis is shoterned to just the last year, it is clear “something happened” and the reduction of claims typical in a recovery has reversed until in summer of 2019 the number of continuing claims has grown, a first since 2009. The year over year percentage has just surged to over 5%, perhaps indicating problems with the Holiday sales.

Like CCSA, the non seasonally adjusted continuing claims is also plotted (CCNSA)

The year over year % change in CCNSA levels is a good way to follow this data array.

When the time axis is shoterned to just the last year, it is clear “something happened” and the reduction of claims typical in a recovery has reversed until in summer of 2019 the number of continuing claims has grown, a first since 2009. The year over year percentage has just surged to over 5%, perhaps indicating problems with the Holiday sales.

Using A Sahm Rule Process on CCSA

Sahm Rule is a signal for the start of the recession using the unemployment rate U3 (UNRATE) where the the rolling average of the last 3 months is divided by the last 12 months minimum. If the results is over 5% the start of a recession is signaled.

Real Time Sahm Rule on FRED

“https://fred.stlouisfed.org/series/SAHMREALTIME”

I take this and modify it, given lack of confidence in the U3 rate and use the seasonally adjusted levels of insured unemployment -continuing claims (CCSA) - with the rolling average for the last 3 months of continuing claims is divided by the rolling minimum over the last year of data.

While Claudia Sahm has the trigger at 5%, the continuing claims trigger seems to be 8%.

It is interesting to consider why does a Sahm Rule “work”. It is really a complex and subtle question. My reasoning is that given the US “steady state” is year in year out growth in Labor Force and population, and a rather constant Employment Population ratio - then when it isnt so it is always a regime blow out, a stochatic “jump diffusion”. Therefore if the point can be indetified where a stochastic jump is occuring, it is certainty that it will play out in a recession. Think of the forces required to cause a jump, to force the US economy out of that rather steady state. They are massive and always do produce a jump. Minsky’s instability.