Home > Counter-Cut, Political > Cooking the Books?

Cooking the Books?


The reported 8.3% unemployment figure for January 2012 could be as high as 15.4%, based on the Government’s own numbers.

The source data used to calculate unemployment is subjectively derived, although it is clothed in a veneer of objectivity.  As such the data collection process could be easily manipulated for political purposes without notice.  There is no indication it was manipulated, however.

Total workforce estimates are not verifiable and in fact disagree by about 12% depending on methodology; the inconsistencies are not readily reconciled.

Published BLS analytical results are internally inconsistent between tables and sometimes within tables themselves, indicating a lack of rigor in analysis and/or presentation.  As such the results could be easily manipulated for political purposes without notice.  There is no indication it was manipulated, however.


There has been great rejoicing of late over the reduction in reported unemployment rates from 9.1% in January 2011 to 8.3% in 2012[1].  This reduction has occurred over a period of a few months, although there have been no structural changes to the economy.  What is driving this reduction?  Are these numbers real?  All is not as it looks.

Data Collection Methodology

Contrary to popular belief, the BLS estimate of total unemployment is independent of unemployment benefit rolls.  The source data is subjectively created.  Every month the Labor Department’s Bureau of Labor Statistics (BLS) randomly surveys about 60,000 households.  This survey, the Current Population Survey (CPS), collects information on labor force related activities, such as job holding, job seeking, and demographics.  The interviewer subjectively determines employment status for each respondent over 16 years of age based on the responses received from the interviewee.

BLS estimates that the accuracy of its survey yields an error of ±3% of the reported unemployment rate.[2]  While statistically true, the subjective determination of employment status could introduce serious errors and be used to skew reports for political gain without affecting the claim of 3% accuracy.

Recent Changes in Data Collection Methodology

Prior to January 2011, the maximum allowed unemployment duration was 2 years; subsequently it was changed to 5 years.  According to the BLS, this change had no impact on unemployment rates, but it did change the mean unemployment duration (ie. time between jobs).[3]  Although the mean unemployment duration doubled between January 2009 and January 2010, the extension of maximum allowed unemployment duration to 5 years did keep the long-term unemployed in the “unemployed” category instead of dropping them out of the workforce entirely.

In January 2012, the BLS began to use updated population estimates based and the 2010 census; prior to that it was using population estimates based on the 2000 census. According to a BLS analysis, this change had no noticeable impact on the total unemployment rate calculation, but that calculation ignores those who fell out of the labor force and gave up looking for a job.[4]

Independent verification of source data

In addition, the Labor Department also surveys employers (instead of individuals) to get a count of total employment from the employers’ perspective; this is the Current Employment Statistics (CES)[5] survey.  The CPS survey reports a total labor force of over 154,000,000, while the CES survey typically reports only 131,000,000.

There is no quick and easy way to reconcile the two reports.  A brief scan of the “BLS Handbook of Mehods”2 indicates that the total CPS labor force is derived analytically using models maintained by BLS.  The method is subjective and potentially subject to manipulation, although there is no indication of such manipulation.

Such a serious un-reconciled discrepancy in our checking accounts would be a huge red flag, but it is virtually ignored by the BLS.  The discrepancy calls into question the fundamental dependability of the numbers reported.  Clearly, much additional research is required.

Data Analysis

The unemployment rate is derived by dividing the total unemployed (from the CPS survey) by the total labor force (derived from the CPS or CES survey).  Whether we use CPS or CES makes a significant difference.  Many businessmen and economists think the latter is more accurate because the CES definition of “employed” is more consistent with what most people regard as “employed.”[6]  Nevertheless, the lower CPS based figure is the one most commonly reported in the news.

But why the change over time of the CPS based rate itself?  That is derived from the way BLS calculates the unemployment rate.  We have to compare January 2011 and January 2012 numbers to understand why (all figures in thousands):

TABLE 1[7]

BLS figures

Adjusted figures

Jan 2011

Jan 2012

Jan 2011

Jan 2012

1 Total population





2 Total labor force (CPS)





3 Not in labor forceAs % of total population









4 People who gave up in 2011


Not counted



5 Unemployed





6 Unemployment Rate (CPS)

9.1% [9]

8.3% 9

9.1% 9

8.9% [10]

7 Total labor force (CES) [11]





8 Unemployment Rate (CES)





9 Want a job[12]

Not counted

Not counted



10 Unemployment Rate (CPS)





11 Unemployment Rate (CES)





Clearly the unemployment rate (Rows 6, 8, 10, 11) remains effectively the same in January 2012 as it was in January 2011 once the people who gave up looking for a job in 2011 were moved from row 3 to row 4 (See “Adjusted Figures” columns).

Rows 7 and 8 are included to show the impact of using total workforce data from CES instead of CPS.

BLS data do not always appear to be consistent.  In doing research for this essay, there were inconsistencies between various tables, and even inconsistencies within a given table.  One such area was regarding statistics related to people wanting jobs.  It isn’t clear whether those people were already included in the “Unemployed” category, or part of yet another uncounted category.  Rows 10 and 11 are provided to illustrate their potential impact on unemployment rates.  Nevertheless there is an insignificant change from January 2011 to January 2012.


Whether we use the CES or CPS total employment numbers, a closer look at how unemployment is calculated indicates that unemployment rates remained virtually flat during 2011.  Estimates for January 2012 could range from a low of about 9% to more than 15% depending on the methodology and assumptions used – but not the officially reported 8.3%.  All is not as it looks.


[7] All data from http://www.bls.gov/news.release/empsit.a.htm, unless otherwise noted in Table 1

[8] “Not in the labor force” increased 35.8% to 36.3% of the total population between January 2011 and January 2012.  If the actual percentage of the total population “Not in the labor force” is relatively constant year to year, the increase represents the people who gave up in 2011 as shown.

[9] = Unemployed / Total labor force

[10] = (Unemployed + People who gave up) /  Total labor force

  1. No comments yet.
  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: