By Hugh, who is a long-time commenter at Naked Capitalism. Originally published at Corrente.
The Bureau of Labor Statistics continues to struggle with a model that does not correspond well to what is going on in the economy. Follow me. In the Establishment or survey of businesses, 171,000 jobs (seasonally adjusted) were created in October. In the Household survey, employment (seasonally adjusted) increased by 410,000 or about two and a half times the number of jobs created. Now the two surveys have very different levels of statistical significance: 100,000 for the Establishment survey and 400,000 for the Household survey, and they cover slightly different population sets, but it would seem a goal of its modeling that the two surveys converge as much as possible.
There are also other problems. In the Household survey last month, the number of involuntary part time workers (seasonally adjusted) spiked 582,000. As I wrote in a post here (Timing and Phantom Data), about 300,000 of that increase was phantom, the result of a statistical glitch in the seasonal adjustment. As I predicted, most of it (269,000) disappeared this month. Curiously, these phantoms may have been converted into voluntary part timers. Voluntary part timers increased 187,000 this month, but there is no way of knowing.
Nor is the larger and supposedly more accurate Establishment survey immune from problems. Look at the revisions (seasonally adjusted) for the two most recent months
August 96,000 > 142,000 > 192,000
September 114,000 > 148,000
The August jobs numbers were revised upward 92%; September, 30% in its first revision. August went from disappointing to very good. I mean what is the point of the modeling if the initial results signify nothing and that the final results may be either bad or good. With all those caveats, let’s turn to the numbers.
In October the potential labor force as represented by the non-institutional population over 16 increased 211,000 from 243.772 million to 243.983 million. The employment ratio increased a tenth of a percent to 58.8%. Multiplying this by the increase in the potential labor force gives us 124,000, an estimate of the number of jobs needed to keep up with population growth. So by this measure October job creation beat population growth by a moderate amount.
Seasonally adjusted, the labor force increased 578,000 from 155.063 million to 155.641 million. Unadjusted it increased 704,000 from 155.075 million to 155.779 million. This is unusual. There has not been this kind of unadjusted increase since the last Presidential election in 2008 and is similar to pre-recession numbers. Given these large increases, the participation rate (the ratio between the current and potential labor force) increased two-tenths of a percent to 63.8% both adjusted and unadjusted.
As already mentioned, the employed increased 410,000 seasonally adjusted from 142.974 million to 143.384 million. Unadjusted, it increased 706,000 from 143.333 million to 144.039 million.
The unadjusted number for the employed (706,000) is larger than the unadjusted number for the increase in the labor force (704,000). The labor force is the sum of the employed and unemployed. So this combination of events can only happen if the number of employed decreases by the difference between the two, and it does going from 11.742 million to 11.741 million. It’s unusual although not impossible that the whole of the increase in the employed should correspond to the increase in the labor force.
Seasonally adjusted, unemployment grew 170,000 from 12.088 million to 12.258 million. Since the labor force is 155 million, a tenth of a percent of this is 155,000 or about the same size as the increase in unemployment, explaining why the unemployment rate increased a tenth of a percent to 7.9% seasonally adjusted. Unadjusted, although the unemployment number was essentially unchanged, because of the increase in the labor force, the unemployment rate dropped a tenth of a percent to 7.5%.
The broader seasonally adjusted U-6 rate based on the unemployed (12.258 million, up 170,000), the marginally attached (2.433 million, down 84,000), and involuntary part time workers (8.344 million, down 269,000) decreased 14.7% > 14.6%. This U-6 seasonally adjusted corresponds to 23.035 million, down 183,000 from September. Unadjusted, the U-6 declined from 14.2% to 13.9%.
The BLS’ measure of its undercount, those who do not have a job, want one, but have not looked for one in the last month, declined 285,000 from 6.427 million to 6.142 million.
Because this measure does not reflect well changes in the economy, I have developed an alternative to it. In my alternate calculation, I compare the current labor force to where we would expect it to be in a solid economic expansion: labor participation rate of 67%. The difference between these two is my measure of the undercount.
.67(243.983million) = 163.469 million (where the labor force should be)
163.469 million — 155.641 million = 7.828 million (the real undercount)
This is a decline of 436,000 from the September figure of 8.264 million. This is the capture of the undercount that the BLS misses.
With this number we can now go back and calculate where the U-3 and U-6 really are, that is the real unemployment and real disemployment rates.
Real unemployment: 12.258 million (U-3 unemployment) + 7.828 million (undercount) = 20.086 million (down 266,000 from 20.352 million in September)
Real unemployment rate: 20.086 million / 163.469 million = 12.3 % (down from 12.5% in September)
Real disemployment: Real unemployment + involuntary part time workers = 20.086 million million + 8.344 million = 28.430 million (down 535,000 from 28.965 million in September)
Real disemployment rate: 28.430 million / 163.469 million = 17.4% (down from 17.7% in September)
The long term unemployed, those unemployed for 6 months or longer under the BLS’ restrictive definition of unemployed: without a job and have looked for one in the 4 weeks before the week in which the Household survey was taken increased 158,000 to 5.002 million but remained steady at 41% of all unemployed.
By race, unemployment among whites was unchanged at 7% and increased, after dropping in September, to 14.3% among African Americans.
In the Establishment survey, seasonally adjusted jobs increased 171,000 from 133.584 million to 133.755 million. The private sector gained 184,000 and government lost 13,000.
Unadjusted, jobs increased 911,000 from 133.881 million to 134.792 million. The unadjusted number though large is not unusual. About 2/3 of it comes from state and local schools and 1/3 is part of the annual build toward Christmas. Education at the local level added 455,900 jobs, and at the state level, 149,800: 605,700 total. The private sector gained 361,000.
Seasonally adjusted, most jobs were gained in areas where quality and pay are poor. 51,000 jobs were added to professional and business services, 36,400 to retail trade, 30,500 to healthcare, and 28,000 to leisure and hospitality.
The average work week for all private employees remained unchanged at 34.4 hours and is the same as it was a year ago. Average hourly earnings decreased one cent to $23.58 and weekly wages declined 35 cents to $811.15. Weekly wages for all employees have increased 1.55% year over year while the CPI has increased 2.1% October to September. Among production and nonsupervisory (blue collar) employees, average weekly hours declined one-tenth to 33.6. Average hourly earnings declined one cent to $19.79 and average weekly earnings dropped $2.32. Weekly wages in this group have increased 0.82% in the last year, again compared to an increase of 2.1% in the CPI. In other words, real wages are falling and they are falling faster in already lower paying jobs.
Last month, there was a particularly egregious mismatch between the number of jobs (114,000) and employed (873,000). We have this again this month, 171,000 vs. 410,000, on a reduced scale. Although the discrepancy is similar in absolute size, because the numbers are larger, the relative discrepancy is smaller and opposite in the unadjusted numbers: 911,000 jobs vs. 706,000 employed.
Beyond this, there was a significant drop in the undercount, representing an influx of workers from outside the labor force (as defined by the BLS). Since there was also a drop in involuntary part time workers and only a partially compensatory increase in voluntary part timers, that should mean that the 410,000 increase in the employed in the Household survey was almost all full timers. However, for a large increase in the employed, especially in the economic environment we have, we would expect a significant number of these to be in part time positions. This is even more our expectation given that wages and hours declined in the Establishment survey.
The picture is clouded further by the fact that the BLS numbers do not add up. In the A-8 table, voluntary part timers increased by 187,000 even as 269,000 phantom involuntary part timers disappeared somewhere. This yields a net decline of 82,000. But in the A-9 table, the BLS reports an increase of 144,000 part time workers and also an increase of 233,000 full time ones. Now first, these numbers do not add up to the 410,000 increase in the employed. Second, the A-9 is reporting an increase in part timers even as the A-8 is reporting a decrease. On the other hand, they would fit better with the wage and hour data from the Establishment survey. The shorter story here is that the part time data have always been something of a mess, but until the events of last month, this was a side issue. With the wild variation last month, this brought them to the fore and there continues to be a carryover into October. I guess the operating idea is that by the time of the next report, we will have forgotten about all this.
Overall, the jobs and employment numbers were good this month. The wage and hour numbers were not. There are, however, real problems both between the surveys and within the surveys which place significant question marks on what can be gleaned from any one jobs report.
Jack Welch seems to have sparked a discussion.
Nicely done…again. I always look forward to your analysis.
Good to see a sound analysis not tainted by a political point of view. Most pundits would make sure this said either Obama’s policies are failing or they are succeeding. (Gawd I’ll be glad when the election is over!)
I’m wondering how much BLS is taking these critiques to heart and trying to rectify the disconnects identified here. I hope they do because they appear aimed strictly at methodological issues, not a political slant.
There is a huge shift to part-time employment going on as businesses attempt to avoid the costs of Obamacare. This also explains why hours worked and hourly earnings are falling while the number of jobs increases. The jobs are being “created” simply make up for other employees whose hours are being involuntarily reduced. Mish does a good job of explaining this over at his blog.
This sounds intuitively just about right, but what is most interesting is that your numbers suggest the average non supervisory blue collar worker earns $34,577 annually assuming he gets paid for 52 weeks. Where is this enough on which to live anywhere but in your Mom’s basement? Why does anyone think that raising trivially the number of poor devils surviving on a pittance is something about which to crow? Moreover, as this average probably includes a small number of workers earning dramatically more, I would be interested to know the median blue collar wage before drawing conclusions about all this, although the word which first comes to mind is shocking, given the current price of luxury goods like apples ($1.50 each the last time I bought one) and bread (about $3.00 per loaf). Can someone explain how even employed people are managing to stay alive? And who is buying tickets to all these football games at which every stadium is packed to bursting?
Thanks Hugh for the excellent analysis as always.
Still keeping the list or has it grown too onerous?
It’s been a year since I updated it. It did indeed become too onerous. When I wrote my Bush scandals list, it was easier because the scandals were more discrete even if they were embedded in larger issues. With Obama, the scandals often take longer to develop. And it was taking me longer and longer to track down the initial links to them, tease them apart from each other, write them up, and keep them updated.
The simpler things are, the closer they are to the truth. This tells you about all you need to know.
Thanks for the analysis, but I´m from spain and in spain our Real disemployment rate is 23% and The disemployement rate of young people is more than the 40%.
A nice analysis.
What I used to recalculate Participation Rate from was that immediately after the 2001 recession or 66.6/7%. Some use the rate before employment took off which is still in the 66.something range. In the end U3 is still extremely high when it is calculated from when Particpation Rate started its decline.
Some blame the decrease in Participation Rate since 2001 on babyboomers retiring. I can’t see how such would be the cause of such a steep decline as too many of us took a bath in 2008 with our 401ks becoming 200.5ks. There has been an increase in those on disability. Many restaurant workers are being converted to part time to avoid the mandated healthcare insurance. I am sure the Tragets and Walmarts of the world are doing the same also. It would be nice to hear your opinion on this also.
I use 67% seasonally adjusted because for 44 of the 45 months from October 1996 to June 2000, the participation rate varied between 67.0 and 67.3. For me, that is close enough in the recent past to be valid for my purposes. But I understand other values can be chosen and other timeframes used.
And for anyone wondering why this is a big deal, it’s because a tenth of a percent now represents 244,000 people.
It’s interesting that you mention age groups because I was just looking at this on FRED. I graphed employment by age group from 2001 onwards. All of the groups from 25-54 were near flat at 30 million, plus or minus 2 million, throughout the decade, but the 55 and over group is an ascending constant slope line going from around 18 million in 2001 to 30 million today. It is now on par with workers in the 25-44 year old age groups. Only the 44-54 group is somewhat higher at about 33 million. I’m eyeballing this from the graph. I don’t know if this is a boomer effect or the result of the decline of private pensions, early retirements, and loss of value in 401ks, but it is quite marked.
the jobs reports is only as good as the numbers. considering that part of this is a telephone survey of respondents, and part is one of business. and with business, you will only get a response evetually if they are still around when they cab get around to doing it. the other is dependent on who answers the phone.
Hugh, last month, I was questioning the enormous amount of money going to community colleges and the tripling of the enrollment creating a place for the unemployed to go. This month phantom part time jobs disappeared. Could this be due to retail contraction? E.g. A Baja Fresh, part of the Wendy’s chain, but the Mexican Food brand, has closed shop and right across the parking lot, an older supermarket was shuttered, due to consolidation with a national chain selling off to leave the Philly Metro market. The new buyer had 3 stores within 5 minutes of each other. This was the oldest and it seems the odd man out. Most of the jobs in both places are part time bottom of the wage scale. Also, it was just announced, a Blockbuster location was closing, right across the street from this shopping center. And down the same street, a Barnes and Nobles announced it is to close in a month or so. A number of smaller chains, coffee shops, pizza joints etc have also closed this year along the same road. It s a major commercial corridor in an affluent suburb of Philly.
The main reason seems to be none renewal of leases. I don’t know on which side of the deal, whether rent concessions were offered or not, but closed is closed and the jobs are not doubling up in other locations. Blockbuster, bought up out of bankruptcy, seems to be just winding down the few bricks and mortar locations. But, the main point is, slack retail demand, over saturated locations of the big chains, all lead to loss of part time jobs. I’ve seen more this year, than in a while. As if, some were trying to hold out, the small time operators and the big chains were waiting for leases to expire. Is there a final comb out of small time and non performing locations causing part time jobs to just go away and not come back across the country?
I don’t know what the story is there. I am surprised that community colleges are receiving much funding. Our government tends to write off the unemployed rather than finding places, outside of prison, for them. So if enrollments are increasing it may be people desperately trying to acquire new skills, and I wonder how much debt they are taking on in the process.
Involuntary part time workers seasonally adjusted increased from 3.3 million in 2001 to 8.3 million today. Most of that increase took place between mid-2007 to the beginning of 2009. In 2009, they actually plateaued at 9 million and have gradually been decreasing since then.
Commercial real estate was one of the shoes that didn’t fall in 2008. One never knows but I would think that owners would be eager to see their storefronts filled. I think you are probably right about slack demand. Remember that almost all of the gains from the so-called recovery went to the 1%. So growth in demand is pretty limited unless you are willing to take on more debt or one of the lucky and increasingly few with stable, high paying jobs.
Paul Tioxin’s comment was anecodotal, but very interesting and I would like to see more comments as to store and restaurant closings. I’ve also gotten the impression that when the financial crisis hit, the retail and resturant operators did everything they could to carry on business as usual until the lease went up, at which point they were kicked out. If I’m right, the true impact of the downturn will become visible slowly as leases expire and landlords let their properties sit empty in the hope of landing a tenant with big pockets.
This is also anecdotal, but I’ve seen some indication of baby boomers hanging on to their salaried positions becuase their pension plans have gone south, and if they have hiring and firing power they make sure that any layoffs hit the younger cohorts in their workplaces. My suspicion is supported by some of the figures cited.
Thanks for the analysis Hugh.
Hugh, about your 67 percent labor force participation rate. It was indeed that high in the late 1990’s, but then dropped to 66 percent in the 2000’s (pre-crash), largely for demographic reasons, as more people moved into the 55+ age group which historically has a much lower LFPR than 25-54. You may want to lower your target rate slightly.
Regardless, your point is spot on: even U-6 doesn’t capture the depth of un/underemployment because the downturn has lasted so long.
I don’t quite understand your issue with the payroll revisions over the past three months. They are due to late sample reporters. If they late reporters have large changes, then we’ll get large revisions. I don’t see any way around this.
Anyone who puts too much stock into one report should take a step back and look for the longer-term trend, and also track QCEW data to gauge the size of payroll date revisions.
As I mentioned in a comment above, the employment in the 55+ group is the one group that showed consistent and dramatic growth throughout the 00s. I will try to look around for yearly estimates of the overall size of the 55+ group during this period. But unless the increase in size of this group was large, its participation rate has to have been going up. So while this group may have had a low participation rate in the 1990s and before, this may have significantly changed in the aughts.
As to your point about overreading the individual reports, you’re right. Seasonal adjustment, for example, looks like it works pretty well for jobs (Establishment survey) and is choppier for employment (Household survey) but for most subgroups its benefit is marginal and in some I would argue it is zero to negative. The overall trends qualified by an understanding of the underlying definitions and their limitations are what is important.
My point though is that all this, the Establishment and Household surveys, is a result of extensive extrapolation and modeling. So there should be a way, taking into account the slightly different makeup of the populations covered by the two surveys, to bring the two surveys into line with each other.
It also grates on me when politicians, reporters, and pundits report the seasonally adjusted numbers as if these actually represented the jobs created in the economy in that month. That information is really contained in the unadjusted numbers. Seasonal adjustments are all about trend lines.