Saturday, October 18, 2014

Common Sense Shows the Household Survey is in Error, Under-reporting Employment by 1 Million (5 1/8%)

If UER-6 and household survey unemployment rate (now at 5.9%) showed  labor slack, a parallel status would be found in the insured unemployment data. Or the the ratio of insured labor force to total labor force would have changed significantly, dropping in similar fashion as the labor force participation rate.

But data  clearly shows that the ratio of insured labor force to total labor force has been unusually steady, ever since the large influx of women entered the work force in the 1970s onward.  The ratio of insured labor force to labor force does drop similar to  the labor force participation rate drop,  during the 2008 to 2009 business slowdown, but has corrected swiftly while the labor force participation rate has stayed unchanged.

The ratio of insured labor force to labor force has returned to levels experienced throughout the decade prior to the solvency event of 08.

The weekly initial claims and continuing claims data is accurate and timely. It is not a derived level or rate from a sample as it is comprehensive.  But for the ARIMA seasonal adjustments, the insured claims is consistent and accurate and factual.  It is what it is.  The BLS and CES uses the state data to calibrate and revise the household survey which by definition is  a sample, and the establishment business survey that derives payroll.  The household and establishment data can only be accurately portrayed with little margin of error every 10 years with the census - but even then that window of accuracy comes as hindsight after the census data is ordered and cleaned.  So BLS and CES depend on the above stable ratio to calibrate and revise the household and establishment data using the timely and comprehensive claims data.  At least they have done so for the 40 years prior to 2008. But in this solvency event with both the slowdown and then the recovery,  BLS and CES have not applied the claims data with the same rigor they have in the past.   If they had done so, using the 80% to 85% insured to total labor force ratio, the unemployment rate would have experienced much higher rates at the economic  nadir and then much lower rates as we recovered.  It is obvious that given the insured labor force to labor force ratio relative stability, the household data is now in error and will have upward revision of approximately 1 million employed.

That would change the current 5.9% unemployment rate to 5.1% unemployment rate raising the labor force participation rate accordingly and eliminating the unusual spread between UER-6 and UER-3.

Here I take the weekly complete census of insured unemployment rate and divide continuing claims by that rate to end with a derived household unemployment rate using insured unemployment and the stability of the ratio of the subset insured labor force to labor force.

 The last data point for both - August 2014 - is provided. The growing error in the household data by calibrating to the weekly claims census is obvious.  It also shows the error is approximately .8 % in the unemployment rate or approximately  1 million employed.

The magnitude of this error  is shown in the Beveridge Curve for insured unemployment and for household unemployment.

This shows the insured data coordinates to other impressions of the recovery like retail sales, consumer lending, durable goods and especially autos; while the household unemployment data is out of synch with other economic data, including  the JOBS data.

Detailed and complex contortions explaining why the household survey is accurate are more and more frequent, but none of these discussions even bother to explain the census based claims data - since it cannot be done -  it is simply dismissed or there is some sort of bizarro world parallel universe explanations that  claims data only reflects the rate of layoffs. That of course makes no sense when one considers the continuing claims data which drops only as employment picks up.

So the insured data Beveridge Curve is the accurate and complete depiction of the nations employment status.

The   reality is the employment levels are showing a near record if not record robust strength that can only occur with a consistent 4% real GDP. The unemployment rate when it is finally revised to the correct level will show that the current unemployment rate is about 5 1/8%.

This means that the FOMC is in the midst of one of the most significant policy errors in the history of the Federal Reserve since 1913.  That the whole premise of "forward guidance" is in error as the inflation rate is not related to the output gap, and then the output gap now being used is in error as right now there is no labor slack left and,  unless a new internet like technology flows into the economy as it did in the 1990s, heightening productivity, significant inflation given all the capacity being used will certainly occur.

But why is there no inflation now given a correct read of an extremely tight labor market? Where are the wage pressures that theory says should exist now.

That is easily explained pragmatically with obvious cause.  The 2008 solvency event hitting the "consumer of last resort" USA market made the USA specific event a global event.  That has created the illusion there is this Rogoff Reinhart "global economy" when there is not.  There is the massive hegemon USA economy and then the rest.

As the other sovereign economic units respond to the USA solvency crisis, almost all have resorted to a mercantalistic currency  policy.  Abenomics, China and now Germany via the Euro reversed the trade weighted decline of the dollar to a consistent 2 1/2% per annum enriching  of the dollar on a trade weighted basis. One can see when the dollar has a period of unchanged values then CPI then surges,  as we saw in the the 2nd Q 2014.  The lack of inflation is solely  dollar level caused and explains about 130% of the inflation that would have occurred now given the lack of capacity in the US economy.  If the FOMC persists in this gross misread of the current status - and perhaps Fischer does realize that as shown by his recent focus on exchange rates - and seeing that the dollar level has always been an administered rate, established by the US hegemon after considering security needs first then domestic needs second, the dollar will cheapen as it usually does, suddenly,  after some keystone forum as the USA asserts their power.  Then a calm to cheapening dollar will expose  the reality of the tight labor market  and little capacity of the US economy.  A large surge in inflation will occur to levels that may approach 6%. That is if the Fed carries on with ZIRP.

The BLS,  and the CES  - who does the heavy lifting and revisions for the BLS - very clearly go over the source of error in the household data and how it is repaired/revised by claims data:

"On an annual basis, the establishment survey incorporates a benchmark revision that re-anchors estimates to nearly complete employment counts available from unemployment insurance tax records. The benchmark helps to control for sampling and modeling errors in the estimates."

And the margin of error for the household data can be substantial:

"...the threshold for a statistically significant change in the household survey is about 400,000..."

(Both quotes from

Given the solvency event was a once in century trend event that experienced a dramtic sudden reversal, it is easy to see that a cumulative  error of three or four months has occurred of about 1 million employed.

The US current unemplyment rate is now about 5 1/8%.  It is very difficult to refute the above logic.  I cannot.

No comments:

Post a Comment