How To Read A Jobs Report
The monthly jobs report from the Bureau of Labor Statistics (BLS) has been out for a few days. Having analyzed these reports for several years, I thought it would be good to talk about what they do and don’t do and how they are put together.
First, the BLS jobs reports are, with a few minor exceptions, focused on paid work. They do not examine whether this work is productive, unproductive, or even destructive of our society. They miss much socially valuable unpaid work such as caregiving, child rearing, mentoring, and yes, blogging, as well as paid work, principally the uniformed military. They do not measure job quality, whether it is meaningful or pays a living wage. Put simply, a job is a job is a job. All are treated the same from the CEO to the no-benefits temp making minimum wage, or effectively less.
Second, one should never take at face value any of the terms used in the BLS reports. All of them have specific definitions that can, and often do, vary widely from their usual, common sense usages, and the numbers themselves often require considerable explanation.
With that, let’s dive in.
As I am sure many of you are aware, the jobs reports are based on two surveys: a household or people’s survey that tracks employment and unemployment and an establishment or business survey that tracks jobs, hours, and wages. Employment and a job, coming from different surveys, do not mean precisely the same thing. In August, there were 156 million employed and 139 million jobs. Some of this discrepancy can be explained by the fact that the jobs (business) survey does not count farm workers or the self-employed (and some “independent” contractors may not be counted as well). On the other hand, the household survey for the purpose of employment counts those holding more than one job only once.
The household survey is much smaller and less accurate. It is based on a sampling of 60,000. The establishment survey is much larger and covers 144,000 businesses, 60% of them with more than 20 employees. The threshold for statistical significance, the confidence that we can say a number or a change in it means something, is 400,000 for the household survey and 100,000 for the jobs survey. The household survey could be improved by increasing its size and bringing in data from other sources, such as unemployment insurance claims, but this would cost some extra money, so end of discussion.
The “official” headline numbers of jobs and the unemployment rate are reported in seasonally adjusted terms. The one thing that can be said about seasonally adjusted numbers is that they almost never reflect what is actually going on in any particular month. That is because seasonally adjusted numbers are trendline. Unadjusted numbers are actual. In August, the actual unemployment rate (using the BLS definition for unemployment which we will shortly see has major problems with it) was 6.3%, not the 6.1% reported.
To recap, some of the things I have been talking about so far, let’s look at a couple of graphs. The first, despite the legend, charts jobs from the business survey. The blue line is the trendline, containing the “official” numbers, and the red line is what is happening in any given month. The trendline is fairly smooth, both because the business survey is so large, and also because revisions are updated into it.
As we can see, the graph shows the characteristic M-shaped pattern of jobs through the year: the spring rebuild, the fall off with the end of the school year, a second rebuild following the return to classes and approach of the Christmas shopping season, and finally the end of the year dropoff.
Now let’s look at the same graph for employment from the household data:
Reflecting the household survey’s smaller size and the fact it does not incorporate revisions on a monthly basis, the trendline is much rougher. The M-shape is still discernible, but you have to look harder for it.
The M-shaped fluctuations are where we actually live. The trendline is a convenient, less messy creation of economists and policymakers. The problem is that they treat the seasonally adjusted trendline numbers as if they were the actual numbers and do not mention at all the unadjusted, actual, numbers. This difference can best be seen in the January reports where, for instance, this year 144,000 jobs were reported created when, in fact, 2.8 million were lost.
The issue of unemployment is complicated, and a real treatment of it touches on several categories. The BLS uses an active jobseeker definition for unemployment. This is a very different concept from what we normally take to mean of being out of a job and unemployed.
If you are without a job and have looked for one in the 4 weeks preceding the reference week of the household survey (the week containing the 12th of the month), you are unemployed. As such, you remain in the labor force, which is the sum of the employed and unemployed. If you are without a job and would work if work were available but have not looked for a job in this time period, –because there are no jobs, the jobs are shit, or don’t meet your qualifications, you are said to have “left” the labor force, as if it were some voluntary decision on your part.
To all intents and purposes, you have ceased to exist. The BLS does have a couple of categories “not in labor force, want a job now” and discouraged workers “have looked for a job in the last year but not the last month, but these simply do not reflect the numbers and changes in them of those wanting jobs. The recession/depression that began in 2007-2009 was not your usual recession and was caused by forces (greed, criminality, stupidity) that we have not seen since the Great Depression of the 1930s. The BLS has not adapted to this new reality and updated its definitions and approach to unemployment.
There is one other place where you will show up. This is in the potential labor force as defined by the category the Civilian Noninstitutional Population over 16 (NIP). This is the root number of the household survey, but like all BLS categories, it is idiosyncratic in nature. Civilian means those in the uniformed military services, as I pointed out above, don’t make up part of it, even though they are engaged in paid work, and their numbers are significant. Noninstitutional makes sense when applied, as it does, to those living in nursing homes and care facilities, but less so when applied to America’s large, increasingly privatized prison system where inmates work, and their low wage scales benefit for-profit companies.
Nevertheless, this category contains many of those without jobs who have been defined out of unemployment and the labor force. And importantly, we can get an idea of their numbers. We can do this by looking at the participation rate, which is the ratio of the labor force to the potential labor force of the NIP and comparing the current labor force to a reference one. The reference I used for several years was 67%. This was the level of the participation rate during the heyday of the Clinton expansion where in the 45 months from October 1996 to June 2000, with the exception of a single month, the participation rate was at or above this figure. The assumption here is that 67% is an indication of a period where when people wanted work, they could find it, that is jobs were available, those who wanted them could find them, and so the number of those defined out of the labor force because they could not find jobs was very low.
Simple math tells us multiplying the participation rate by the potential labor force of the NIP gives us the size of the labor force.
Participation Rate = Labor force / Potential labor force
(Participation Rate)(Potential labor force) = Labor force
Multiplying the 67% participation rate into August’s potential labor force of 248.229 million would give an ideal labor force of 166.313 million. If we subtract this from the reported labor force, we get 10.354 million. This would be the number of those who wanted jobs and would work if jobs were available to them. Add this to the current number of unemployed and you would get the real number of unemployed and from this you could calculate the real unemployment rate.
There is, however, a major kicker to this approach, and this is that the Baby Boomer generation has begun to retire, decreasing our ideal participation rate. But there is more to the story than this. Baby Boomers began turning 65 on January 1, 2011. The participation rate has been going downhill since its Clinton era highs in 2000.
The economy (free trade pacts, offshoring, downsizing, recession/depression), not Baby Boomers, was responsible for declines in the participation from 2000-2011. The question is how much have Boomers leaving the labor force through retirement contributed to its decline since.
In looking at this problem, we are not concerned with all Boomers retiring but the number in excess of what we might have expected. This can be estimated.
I used the 11 year period from 2000 through 2010 to calculate an annual average of those over 65 and not in the labor force. I then subtracted this number for those over 65 and not in the labor force for each of the three years from 2011 on. This was the excess, which I could divide by the NIP, to give me the effect on the participation rate. To the end of the first quarter of 2011, the Boomer effect has decreased the participation rate 1.1%. The Boomer effect decreases the participation rate by about 0.03% a month. So for August, its effect is 1.25%.
What this means for us is that our ideal participation rate for August is 1.25% less than 67% or 65.75%.
.6575(248.229 million) =163.210 million
Subtracting August’s labor force level from this gives us an estimate of those not in the labor force but who would work if work were available to them.
163.210 million – 155.959 million = 7.251 million
Adding this to the number of unemployed for August (9.591 million) gives us the real number of unemployed, 16.842 million, and a real unemployment rate of 10.3%. This is a seasonally adjusted rate because I wished to compare it to August’s official unemployment rate of 6.1%. Since the beginning of 2011, the official unemployment rate declined by 3% but 70% of this decrease was a Boomer effect. We are seeing two things here. First, real unemployment remains high. Second, a majority of the decrease in unemployment in the last 3 1/2 years has come from Boomer retirements. Economists at the Fed seem to have some inkling of this even if they don’t have a real handle on it as can be seen by their moving away from a 6% unemployment threshold for raising rates. Politicians, as always, are willing to take credit for the decline in the unemployment rate even if they are responsible for very little of it.
You may think that a full time worker is someone who works 40 or more hours a week, but you would be wrong. The BLS considers anyone working over 35 hours a week a full time worker. Part time workers are divided into two main categories: part timers for economic and non-economic reasons. The media usually refers to these as voluntary and involuntary part timers. All of these are misnomers. Non-economic reasons include parenting and Social Security income limits. If you do not have or cannot afford daycare or if you need to work to supplement your Social Security, these are hardly voluntary or noneconomic from your point of view.
The Jobs Report
The concentration on seasonal adjustment (the trendline) makes for smooth graphs but obscures how dynamic the job situation is in the US. The trendline is also something in the nature of a projection. While the newly released August report shows seasonally adjusted 142,000 jobs were added to the economy. Unadjusted 327,000 were. Of these only 118,000 were in the private sector while the rest reflected early returns to the education system at the local government level. The unadjusted August number for private sector jobs is actually worse than the August numbers for the last two years by about 70,000 jobs. At least so far, this is an indication that the economy is running out of gas, and the 142,000 number is the BLS pulling back on its trend projections
Of course, this number may improve with the revisions the BLS makes over the next the next two months. Year to date, the economy is doing about 135,000 better than in 2013: 3.538 million vs. 3.403 million. Except for the bad winter and subsequent rebound, 2014 has been looking much like 2013, perhaps marginally better. The media spinners are right in one regard. We can not tell a lot from a single month (unless the sky is falling in).
We all know that job quality has been on the decline for some time. Indeed it has been going down for decades. Since 2000, this has been exacerbated by job scarcity. And since the meltdown in September 2008 and the start of the subsequent depression in 2009, job quality has dropped precipitously. This is an incredibly important issue to us both as individuals and as a society, but the BLS does not cover it. Instead we need to look at an amalgam of imperfect indicators: in what industries jobs are being created, hours, and pay.
While manufacturing is often looked to as the gold standard, it only accounts for a small portion of American jobs, 8.7%, and even these are not what they were in terms of pay and benefits. Most jobs that have been created in recent years, to be blunt, are shit jobs in retail, leisure and hospitality, and healthcare. Temp jobs also serve as a barometer. While still only a tiny fraction of American jobs (2927.4 million unadjusted in August), they are a sector which has shown steady growth since the beginning of the “recovery”.
Hours and wages are also places where we can gauge the quality of jobs and the overall health of the economy. These are given in seasonally adjusted terms. Traditionally, increases in hours and wages were indications of a strengthening economy. Tellingly, average weekly hours have been unchanged for all employees (34.5 hours) and for production and nonsupervisory personnel (33.7) for the last six months, and indeed both are unchanged from a year ago.
When the work week is unchanged, hourly wages can be used to measure wage gains, but when the work week fluctuates weekly wages are a better guide. Of course, these wages are nominal, that is they are not adjusted for inflation. The BLS has a different report which covers real (inflation adjusted) wages. For July, this showed that weekly wages year over year increased 0.3% for all workers and 0.9% for production and nonsupervisory personnel.
Production and nonsupervisory personnel comprise about 4/5 of all workers. We can use this fact to gauge what is happening in the top fifth (20%) which is not reported. The top 20% lost about 1.7% in wages from July 2013 to July 2014. The most likely explanation I have for this is Boomer retirement. Many Boomers are retiring at the top of their wage scales. Because so many of them are retiring, and being replaced by younger, cheaper workers, this is pulling down the overall wages of this group.
Of course, overall weekly wages have been largely flat for 35 years, since the late 70s.
The graph is for production and nonsupervisory personnel because data for them goes back this far. Real (inflation adjusted) wages are in blue, and nominal (what shows up on their paycheck) are red. The reason I am not charting the category “all employees” is because it is a much more recent, and one of the reasons for its creation is that it improved the overall wage picture by the inclusion of the top 20% in it. I should note that the Warren Buffetts and Bill Gates of the world do not show up in these data because most of their income is from capital gains not wages.
I should note too that these data are for “average” weekly wages. When looking at quality, and inequality, the median is much more useful. If Bill Gates walks into a bar with six people at it, the average worth of the group shoots into the billions while their median worth remains in the tens of thousands. Media and politicians, the rich and elites, prefer to focus on averages and aggregates because this makes things look much better, but this hides the fact that while it is doing much better for them, it is doing much worse for us in the lower 80%.
A jobs report to be understood and have some relevance to the state of the society in which we live must be deconstructed before it can be reconstructed into something that speaks to where we really are and how we are really doing. Meaningful information can still be gleaned from jobs reports, but it is largely up to us to dig it out.