Beware of Hedonic Adjustments to NFP!

I was as taken aback by today’s Non-Farm Payroll number as anyone. But I was curious as to this huge number could be such a surprising outlier, given all the wekaness we have been seeing.

Then, I dug beneath the headlines a bit, and discovered this small little hedonic adjustment to the NFP data:  The CES Net Birth/Death Model.

BLS notes " There is an unavoidable lag between an establishment opening for business and its appearing on the sample frame and being available for sampling. Because new firm births generate a portion of employment growth each month, non-sampling methods must be used to estimate this growth." In other words, they estimate the number of jobs created by new firms which the BLS has yet to sample.

Now, I have no idea how accurate or reliable this adjsutment actually is. But out of today’s headline NFP data, out of those 274k jobs, there was a +257k "Birth/Death" adjustment. Of the reported data, 93.8% of the gains have been hedonically adjusted. Note that this is a relatively new process, first introduced in 2000.

I did take the Under, so perhaps I am just "talking my bet." But if anyone can explain this mnore thoroughly to me, I’d greatly appreciate it  . . .


UPDATE MAY 7, 2004: 6:02am
Calculated Risk has a very thorough explanation as to the impact of this hedonic adjustment: 

He exhorts us not to mix: "seasonally adjusted numbers with "not seasonally" adjusted numbers. The actual BLS reported jobs created in April was 1.179 Million (before seasonal adjustment). Of these 1.179 million jobs, 257K came from the birth/death model (or about 22%).

If you follow the link to the birth/death model you will see:

"Note that the the net birth/death figures are not seasonally adjusted, and are applied to not seasonally adjusted monthly employment links to determine the final estimate.

"Finally, when the 1.179 Million jobs is seasonally adjusted you get 274 Thousand headline number.

This suggests that the impactof this hedonic adjustment is far less than initially implied.


CES Birth/Death Model

Frequently Asked Questions


Beware the Birth/Death Jobs Hedonics

Print Friendly, PDF & Email

What's been said:

Discussions found on the web:
  1. Damian commented on May 7

    I’m really confused by these numbers – so BLS reported that 1.179m jobs were created in April? I thought the number was 274k? Or is 274k the difference between April and the prior month? Where does this 1.179m jobs number come from? I can’t believe that we created 1.2m jobs in a single month…any explaination extremely appreciated.

  2. milt commented on May 7

    Seems absurd to have any statistical models regarding employment given that (a) every company no matter its size spends many hours on hiring or firing an employee (time spent at various management levels for each hiring/firing, HR’s time, accounting and tax departments’ time) and (b) it would take less than 10 minutes to update a BLS Webpage with monthly hiring/firing data; have accounting do it. Seems like the BLS is still using processes derived when only US Postal Service and telephones were available. There out to be a new regulation on Web based monthly updating of BLS employement data by every company and partnership in the US given the importance of this information in a our multi-trillion dollar economy. An economy that is tweaked and prodded by the Fed’s interest rate dictums.

    With that whine aside, I figure the April employment data provides the Fed with the cover it needs to increase interest rates regardless of the potential damage it does to the economy. Inverted yield curve – not the Fed’s problem. Twin deficits, no their area of responsibility. Inflation is their problem. Inflation is what they will use as the reason for increasing rates into the 4% range. Further, they say the economy is in a soft patch, and they called one last year, and they were right. So, I figure 4% to 4.5% is where we are headed this year.

  3. CalculatedRisk commented on May 7

    Damian, job creation / destruction varies widely based on a distinct seasonally pattern. As an example, in January over 2.5 million jobs were lost but the BLS reported a seasonally adjusted net job gain! See my post for a graph:
    Not Seasonally Adjusted Non-Farm Payroll

    No one ever reports the not seasonally adjusted numbers because they are useless unless you know the seasonal pattern. I noted them because some people were overlooking that the birth/death model was NOT seasonally adjusted.

    The headline number is the one to use (274K). Many economists dig through the numbers to find out what type of jobs are being created (see Kash’s Construction in the Economy):

    But it is important to distinguish between seasonally adjusted / not seasonally adjusted numbers (the swing are huge for NFP). It is also important to distinguish between establishment and household surveys.

    I know its confusing!

  4. spencer commented on May 7

    Early this cyle the birth/death model came under quite a bit of questioning. But in retrospect itseems to have worked out quite well.

    Milt, you are completely right on that the BLS ought to be able to get employment data quicker. Write your Congressmen suggesting they be given the budget to do it. In real terms the BLS budget is smaller then it was 20 years ago and the need for better data is a real pressing need.

Posted Under