BLS Jobs Situation Disturbing in September 2012

Written by Steven Hansen

The September 2012 BLS jobs report was a little disturbing because the data was mixed and inconsistent.

  • the real unadjusted jobs loss comparing the changes between August and September was average historically.
  • economic intuitive sectors of employment gave a mixed picture of the economy.
  • the inconsistency between the various surveys which form this BLS Jobs report are again large this month, and this report is worse than the ADP Jobs report issued Wednesday.

A summary of the employment situation:

  • BLS reported: 114K (non-farm) and 104K (non-farm private). Unemployment = 7.8% (down from 8.1%)
  • ADP reported: 162K (non-farm private)
  • Market expected: 120K to 165K (non-farm), 130K to 175K (non-farm private), 8.1% to 8.2% unemployment
  • Econintersect‘s Forecast: 150K (non-farm private).

The BLS reports seasonally adjusted data. This data is highly manipulated, and Econintersect believes the unadjusted data gives a clearer picture of the jobs situation.  Also, It should be pointed out that the BLS has issued a Preliminary Benchmark Announcement saying they are currently reporting seasonally adjusted non-farm private jobs 453,000 too low (details in ADP employment post).

This report is very inconsistent. The seasonally adjusted data says:

  • the unemployment rate fell implying the jobs situation was improving;
  • the jobs growth was below population growth rates implying a weak report;
  • the employment population ratio improved implying over 1 million jobs were born this month (yet only 114K was reported);
  • the unadjusted jobs growth year-over-year was 1.72% – the worst year-over-year improvement since March 2011 implying a weak jobs growth this month.

Non-seasonally adjusted non-farm payrolls fell 377,000 – worse than last year’s August / September fall, but seems historically average compared to times of economic expansion.

Historical Unadjusted Private Non-Farm Jobs Growth Between August and September (Table B-1, data in thousands)

/images/bls non-adjusted change.PNG

As always, the recent past data (last three months) is revised.

Change in Seasonally Adjusted Non-Farm Payrolls Between Originally Reported (blue bars) and Current Estimates (red bars)

/images/z bls1.png

Most of the analysis below uses unadjusted data, and presents an alternative view to the headline data.


The BLS reported U-3 (headline) unemployment fell 0.3% to 7.8% with the U-6 “all in” unemployment rate (including those working part time who want a full time job) unchanged at 14.7%.

BLS U-3 Headline Unemployment (red line, left axis), U-6 All In Unemployment (blue line, left axis), and Median Duration of Unemployment (green line, right axis)

Econintersect has an interpretation of employment supply slack using the BLS unadjusted data base, demonstrated by the graph below.

Employment-Population Ratio

The jobs picture when you view the population as a whole, has NOT improved since the end of the recession (regardless of changes in the unemployment rate).  This month the rate improved, but remains in the channel seen since the end of the Great Recession.

  • Econintersect uses employment-populations ratios to monitor the jobless situation. Changes in the base data effect our view of the economy. Below is the Employment / Population ratio – and what it is telling you is that jobs growth is keeping up with population growth (but seemingly no higher).
  • In the latest BLS report employment-population ratio improved 0.4 to 58.7. The employment-population ratio tells you the percent of the population with a job. Each 0.1% increment represents approximately 300,000 jobs. [Note: these are seasonally adjusted numbers – and we are relying on the BLS to get this seasonal adjustment factor correct]. An unchanged ratio would be telling you that jobs growth was around 150,000 – as this is approximately the new entries to the labor market caused by population growth.

Employment Metrics

The 3 year growth trend is up, and the short term trends are mixed depending on the periods selected – and this month the growth is slightly worse than last month.

Unadjusted Non-Farm Payrolls Year-over-Year Growth

The bullets below use seasonally adjusted data:

  • Average hours worked (table B-2) was improved 0.1 to 34.5. A rising number indicates an expanding economy if the employment is also rising. This number has been up and down in a tight range for several months.
  • Government employment grew 10,000 with the Federal Government Up 4,000,  state governments up 13,000 and local governments down 7,000.
  • The big contributor to employment growth this month were health care (44.5) and transport and warehousing (17.1).
  • Manufacturing declined 16,000.
  • The unemployment rate for people between 20 and 24 (Table A-10) fell from 13.9% to 12.4%. This number is produced by survey and is very volatile.
  • Average hourly earnings (Table B-3) rose seven cents to $23.58. Wages growth remains in a “less good” trend.

Private Employment: Average Hourly Earnings

Economic Metrics

Economic markers used to benchmark economic growth were ok, and well away from recessionary levels.

The truck employment grew (0.7k), and growth in this sector shows an expanding economy.

Truck Transport Employment – Year-over-Year Change

Temporary help fell 2.0K but remains in the growth range seen during the last 6 months.

Temporary Help Employment – Year-over-Year Change

Econintersect believes the transport sector is a forward indicator. Others look at temporary help as a forward indicator, and this is positive also.

Caveat on the use of BLS Jobs Data

The monthly headline data ends up being significantly revised for months after the initial release – and is subject also to annual revisions. The question remains how seriously can you take the data when first released.

The above graphic (updated through October 2011) is the month-over-month change in employment based on the original headline non-farm employment level and the current stated employment levels at month end. You will note some pretty drastic backward revision for a major economic release the market reacts to in real time.

Econintersect Contributor Jeff Miller has the following description of BLS methodology:

  1. An initial report of a survey of establishments. Even if the survey sample was perfect (and we all know that it is not) and the response rate was 100% (which it is not) the sampling error alone for a 90% confidence interval is +/- 100K jobs.
  2. The report is revised to reflect additional responses over the next two months.
  3. There is an adjustment to account for job creation — much maligned and misunderstood by nearly everyone.
  4. The final data are benchmarked against the state employment data every year. This usually shows that the overall process was very good, but it led to major downward adjustments at the time of the recession. More recently, the BLS estimates have been too low.

Econintersect has repeatedly pointed out questions about how the seasonal adjustment algorithms and data gathering methodology used by the BLS introduce uncertainty into interpretation of month to month changes in employment.

Econintersect believes the simplistic sampling extrapolation technique of ADP yields a far better picture of the employment situation than the complicated, convoluted Bureau of Labor Statistics (BLS) methodology.

ADP (blue line) versus BLS (red line) – Monthly Jobs Growth Comparison

Because of the differences in methodology, many pundits ignore the ADP numbers – while waiting for the BLS numbers. Although there can be a low correlation in a particular month, the different methodologies tend to balance out, and the correlations are excellent outside of the data turning points. We are now 16 months past the post recession turning point in employment.

However, there is some discussion that neither the ADP or BLS numbers are correct – as both are derived by a sampling methodology. The answer could be that there is no correct answer in real time – and that it is best to look at the trends. As has been noted, all eventually end up correlating.

The BLS uses seasonal adjusted data for its headline numbers. The seasonally adjusted employment data is produced by an algorithm. The following graph which shows unadjusted job growth – seasonal adjustments spread employment growth over the entire year. Employment does not really grow in the second half of the year and always falls significantly in Januarys.

Non-Seasonally Adjusted Employment – Private Sector

There is the proverbial question on what is minimal jobs growth each month required to allow for new entrants to the market. Depending on mindset, this answer varies. According to Investopdia, the number is between 100,000 and 150,000. The Wall Street Journal is citing 125K. Mark Zandi said 150K. Econintersect is going with Mark Zandi’s number:

  • If Econintersect used employment / population ratios to determine the number, the exact number seems to be between 140,000 and 160,000. The graph below uses the historical employment-population ratios to show jobs growth per month if the population was 300 million.

Historical Monthly Jobs Growth Comparison if Population was 300 Million

  • If Econintersect uses employment – population ratios, the correct number would be the number where this ratio improved. Using the graph below, the ratio began to improve starting a little after mid-year. This corresponds to the period where the 12 month rolling average of job gains hit 150,000.

Employment to Population Ratio

Note: The ratio could be fine tuned by adjusting to the ratio of employment to working age population rather than the total population. However, this would not change the big picture that an increase of somewhere around 150,000 (+/-) is needed for the growing population numbers. We have estimated 140k – 160k. The number might possibly be within the range 125k – 175k. Econintersect cannot find reason to support the estimates below 125k.

The question of how changing demographics impact the employment numbers is at the margins of analysis. Econintersect will publish more on this fine tuning going forward, both in-house research and the work of others.

Related Articles

All Employment Articles

2 replies on “BLS Jobs Situation Disturbing in September 2012”

  1. Another reporter gave this BLS page that gives the part-time numbers:

    The reporter said it indicated most of the increase in the household survey was due to part-time jobs, amounting to 572,000 jobs out of the 873,000. But this still leaves full-times jobs at 301,000 that is nearly three times higher than the employer’s survey at 114,000 jobs.
    Is it common to have this large a discrepancy between the household and employer surveys even when you restrict to looking at only full time jobs?
    There is also this Gallup jobs report:

    October 4, 2012
    U.S. Unadjusted Unemployment Rate at 7.9% in September.
    by Jenny Marlar
    “WASHINGTON, D.C. — U.S. unemployment, as measured by Gallup without seasonal adjustment, was 7.9% for the month of September, unchanged from 7.9% measured in mid-September but down slightly from 8.1% for the month of August. Gallup’s seasonally adjusted September unemployment rate was 8.1%, unchanged from August.”

    It is interesting that the Gallup unadjusted number is close to the BLS number released. Is it possible the BLS neglected to apply the adjustment this month?
    In any case, Gallup was predicting the adjusted number to be released by the BLS would be within 1 percentage point of the 8.1% they were measuring. I imagine then they were also surprised when the number released was all the way down to 7.8%, close to what the unadjusted number Gallup obtained.

    Bob Clark

  2. Bob,

    re unemployment rate: the bls determines this number by survey. they have a criteria for determining in the sample who is called unemployed. as it is no secret what the criteria is, it should come as no surprise that gallup and bls are close. both are determined by survey. however, i believe the criteria is wrong – and misleads politicians and the general population (as the headline number is way too low). the purpose of determining unemployment is to show employment slack. is not anyone who wants a job unemployed (which is not true with the unemployment numbers)?

    re discrepancy – 1/3 of the time it seems there is little discrepancy, 1/3 of the time it is noticeable, and then 1/3 of the time it is like the September jobs report – massive discrepancy. i do not like the bls jobs report because it seems very random and unprofessional. no corporation would allow a report to be issued with discrepancies without discussing them and providing likely “correct” interpretation.

    what this report has left us with is people spinning doo-doo everywhere.

Comments are closed.