Did the BLS Really Cook the Books?

October Jobs Numbers Have Been Released:
October 2012 Jobs Reports Summary

How Can the Unemployment Rate Have Decreased .3% When We Added Only 114,000 jobs?  And 873,000 more people are now employed?  But only 114,000 jobs?   Something is fishy...

Well, no.  There's nothing fishy about these numbers, and I'm happy to explain what they do and do not mean...   

1.  First, the jobs numbers come from a completely separate source than the employment/unemployment rate numbers. 

The jobs numbers come from the "Establishment" survey of employers, both private employers and government employers.  This is a sample survey, as real data from every employer in the country would be impossible to produce on a monthly basis.  

The employment/unemployment rate numbers come from a survey of the population, the Current Population Survey or "Household" survey, which is conducted monthly by the Census for the Bureau of Labor Statistics.  60,000 households are interviewed monthly and asked questions about the employment/unemployment status of every person in the household.  Data from young people under age 16 are not counted.  This is also a sample survey as it would be impossible to interview every household in the country every month.    

2.  The Current Population Survey counts groups of people that the jobs report (the Establishment Survey) does not count; in particular, agricultural and self-employed workers.  

But the sample survey size of the Current Population Survey is smaller than that of the Establishment Survey, so it is more volatile.  The unemployment rate often does go up and down .2% to .3% of a point and the numbers of employed and unemployed people go up and down as well-- while the jobs numbers tend to go either up, as they have now done for almost three years, or down.  My private jobs graph (see bottom of article) is a nice relatively smooth curve; the unemployment rate graph zig zags.

3.  Over time, these numbers rise and fall in parallel, but they are not equal.  The number of jobs is usually about 93.5% to 94.2% of the number of "people employed". 

When the economy is better, this percentage tends to go up.  I don't know why, but I would speculate that self-employed people, such as consultants, are wooed by employers offering more job security and benefits.  When the economy takes a dive, people who may have a hard time finding work create and work in small micro businesses.  More about this some other time; my point here is that there will be more "employed people" as a whole than jobs.          

4.  Seasonal adjustments.

Ah yes, one of my favorites.  The numbers that we hear bandied about on the news are seasonally adjusted numbers.  If you don't understand the concept, please read HERE.  Some months the seasonal adjustments seem hinky to me.. just a good word.  Not skewed, but just not quite what I would expect.  No, nobody is cooking the books; but there is something about the statistics involved in the seasonal adjustment process that I just don't understand... yet.

In any event, sometimes the jobs numbers are seasonally adjusted up while the numbers of people employed are adjusted down.  And vice verse.  That happened this month.  Often looking at the unadjusted numbers clarifies things.      

5.  Underreporting of jobs data..

We've added 32,000 a month in benchmark revisions through March 2012; there's a good chance that jobs are still being under reported.  More about these revisions HERE.

6.  Calendar variance.

I wrote HERE about the effect of August 12th falling on a Sunday this year and what it meant for the August employment numbers.  I was pretty sure that those jobs numbers would go up this month because they were unrealistically down last month, and I was right.


7.  Teachers coming back?.

Added 10/8/12:  
I left this last night with a discrepancy of about 122,000 employees.  This morning I remembered that government jobs, which had gone down just about every month for two years, had shown an increase in the last couple of months.  I started to look at the changes in the various government sectors; in particular, in the local education (teachers) segment.  

A preliminary look at unadjusted government jobs seems to indicate that we have added back 100,000 to 130,000 jobs at the "Local government:  Education" level.  Again, we will know more with the October numbers.

The "local government:  education" sector usually loses jobs (unadjusted numbers) between May and July and jobs are added back to this sector between August and November.  In September, we still have more job losses than gains in this sector.  In 2010 we were down 670,000 in this sector.  In 2011, 641,000.  But this year we are only down 518,000 jobs in this sector  compared to May.  In  unadjusted "raw" numbers, this is a difference of 120,000 jobs.  In adjusted numbers, there is a difference of 114,000 jobs.  We have 10,000 more jobs in "local government: education" than we did a year ago.  That might not seem like much, but the total of jobs in the "Local government: Education" sector has been lower each September than the previous September for 3 straight years.. until 2012.

However, I think I've already counted these extra teaching jobs in education in looking at jobs without seasonal variations, so I'm not sure that this would count towards filling up the final 120,000 jobs in the discrepancy.    

8.  Comparing spring into fall employment numbers across years:
In "raw" unadjusted numbers of people working, numbers generally start to go up in March, peak in July, go down in August, sometimes go up and sometimes down in September, and then go up firmly in October.  In terms of unadjusted numbers of people working in September compared to March, 2012 is a strong year, the strongest year since 2006.  We are up 1,921,000 in "people employed" in unadjusted numbers since March.  In 2006, we were up 2,238,000 "people employed" in September vs. March.    

My point here is that the increase in the number of employed people is simply not that off the wall when we compare these numbers to those of previous years.      

O.K.... so let's put numbers to some of these factors:  

4.  Seasonal adjustments:  

The number of jobs in seasonally adjusted numbers increased by 114,000.

But in unadjusted "raw" numbers, the number of jobs increased by 574,000.  

But:

The increase in seasonally adjusted numbers of people employed was  873,000.

The increase in unadjusted "raw" numbers of people employed was  775,000.

Well.. right off the bat, 775,000  more people employed in "raw" numbers is a lot closer to 574,000 more jobs in "raw" numbers than 873,000 is to 114,000.

Over time the adjustments to "jobs" vs. "household" numbers even out, but in any one month, they can be.... hinky.  

So now instead of explaining a discrepancy of  761,000 jobs vs. employed people, we are looking at a difference of 201,000 (775,000 people employed minus 574,000 jobs)  in unadjusted numbers. 

2. & 3.  The usual difference between the two reports:

94% of 775,000 is 728,000, which would be the expected increase in the number of jobs in raw unadjusted numbers with an addition of 775,000 employed people.  

We're down to a discrepancy of 154,000 (728,000 expected jobs increase minus 574,000 real jobs increase). 

5. & 6.  Underreporting of jobs data and calendar variance:

Benchmark revisions showed that we have underestimated job growth by 32,000 jobs a month from March 2011 through March 2012.  As we are adding jobs, there is a good chance, a very good chance, that we are still undercounting jobs.  Let's just assume that we are still underestimating jobs by about 32,000.   That would take the discrepancy to around 122,000, which is completely reasonable when we are dealing with a labor force of 155,000,000.  

7.  Hmmm.... Teachers?

As I wrote above, even though it appears we will have more teachers this fall than we have for the past three years, this number will probably not explain the final 120,000 discrepancy.  But it is a reason why may have more jobs and more people employed this September vs. the last couple of years.
 

That's it.  Let me make this clear:   

The purpose of this exercise is to demonstrate how there can be such a discrepancy, a logical but large discrepancy, between the number of jobs and the number of people employed in any one month.  As usual, we'll learn more when we see the data from October and November.

Not that this will convince ANY wingnut conspiracy type.