Would the real US employment figure please stand up: the Birth/Death adjustment debate.

A number of commentators have questioned the underlying momentum in the US economy.  If we assume that today’s birth death adjustments are based on latest data as of Q4 2014, and 2014 Q4 showed the largest increase in employment since 1983, and the latter half of 2014 represented a relatively strong period of economic growth in the current cycle, then just maybe, if the cycle is turning, then there is a risk that current employment data may be increasingly wide of the mark.

btwnnboomandbust2

In the latest report on US employment we see private jobs growth averaging 222,000 a month over the last three months.  On the face of it, employment growth suggests that all is well with the economy despite the slowdown in manufacturing and world trade. 

The CES (Current Employment Statistics) from which changes in employment are calculated is drawn from a sample of firms: about “143,000 businesses and government agencies” drawn from a “sampling frame of unemployment insurance tax accounts which cover approximately 588,000 individual worksites”.  

What the CES is slow to capture is the impact of firms going out of business and the hiring generated by new firms.  Firms going out of business are ignored, but an estimate of the difference between net employment generated by new firm hiring and old firms going out of business is added to non seasonally adjusted employment data (the residual).  A 5 year time series (of births and deaths) is used to calculate the difference between new firm/dead firm employment.   Prior to 2011 this model was used to update birth/death assumptions annually, but procedures post 2011 now do this quarterly.  However while the procedures are now updated quarterly, given the way in which the model is applied, there is a time lag that can be as long as 12 months.  

The quarterly forecasting methodology will produce initial forecasts that extend, at most, 12 months. For example, initial quarter 1 forecasts, using input data through March 2010, would forecast net birth/death components through March 2011. When quarter 2 QCEW data is made available, initial forecasts would be made for the months April – June 2011. Similarly, forecasts for July – September 2011 and October – December 2011 would be made when new quarter 3 and 4 QCEW data, respectively, become available.

What this means is that the assumption regarding employment growth due to birth and death assumptions are not going to be exactly what is happening at the present point in time.  The differences are wider at turning points in the economy (i.e. as the economy enters recession) and more so at economic cycle peaks: while the post 2011 adjustment to the model may have minimised the errors in calculating births and deaths, historical analysis of this adjustment does not seem to apply to economic cycle peaks as much as troughs – see table 1 of hyperlink.

A number of commentators have questioned the underlying momentum in the US economy.  If we assume that today’s birth death adjustments are based on latest data as of Q4 2014, and 2014 Q4 showed the largest increase in quarterly employment since 1983, and the latter half of 2014 represented a relatively strong period of economic growth in the current cycle, then just maybe, if the cycle is turning, then there is a risk that current employment data may be increasingly irrelevant.

image

I have also done some numerical analysis on the data, in particular the relationship between changes in employment (here I have used private employment from the CES reports) and changes in real GDP.

image

As we can see in the above chart changes in GDP have historically exceeded changes in employment, with the exception of a few periods:

image

We can also see this metric from a different perspective:

image

And more specifically, here we identify historic periods where economic growth is slowing and is lower than or equal to the current annual growth rate and look at employment growth in that period and then compare this to longer term average growth rates – in this case annualised growth rates over rolling ten year time frames.

image

And here we have the annual rate of growth of the following: the difference between the cumulative increase in real GDP and change in private employment as a % of the change in employment.   Essentially when employment is growing faster than real GDP we should be able to pick this up as a negative annual rate of change.   We then find those instances where employment growth is greater by this measure and show the annual change in GDP for those –ve data points.  

image

The current period of employment growth exceeding real GDP growth is fairly rare, except for the late 1990s.  And the following shows a further detail added to the above chart:

image

BLS references

http://www.bls.gov/web/empsit/cesbd.htm

http://www.bls.gov/ces/ces_quarterly_birthdeath.pdf

http://www.bls.gov/news.release/cewbd.t08.htm

http://www.bls.gov/bdm/

http://www.bls.gov/news.release/cewbd.nr0.htm

http://data.bls.gov/cgi-bin/surveymost?bd

http://www.bls.gov/web/empsit/cesbdhst.htm

Leave a Reply