Production Workers, All Other, and the Occupational Employment Quarterly

Despite the availability of better data, vocational experts continue to rely upon the Occupational Employment Quarterly (OEQ) from United Statistics Publishing (USP)  Confession is good for the soul and here is mine -- for years I defended the OEQ as the only game in town and we needed some source of data as a starting point.  That was more than a decade ago and now I know better.  Today we explore the OEQ and our favorite occupational group, production workers, all other in SOC code 51-9199.

According to the Bureau of Labor Statistics, this occupational group represents 241,910 jobs in the nation.   The Occupational Employment Statistics (OES) relies upon employer's surveys to estimate numbers of jobs in the nation within specific O*NET occupational groups.  The corresponding Census Code, 8965, represents the results of the Current Population Survey.   The Current Population Survey is the result of a survey of households, workers, asking them what they do.  BLS publishes a compilation of the current population survey stating that production workers, all other, represents 944,000 jobs.  That leaves a discrepancy of over 700,000 jobs between the two data sets, a discrepancy beyond the focus of this article.

In the fourth quarter of 2016, USP states that production workers, all other represents 760,983 jobs in the nation.  USP accurately states that the occupational group represents 1589 distinct DOT codes.  In the unskilled range of work, USP estimates the number of sedentary, light, medium, and heavy jobs:

Sed.         Light        Med.       Heavy +
24,903     193,957    88,598    20,593

Let's do the math, just for fun.

760,983 / 1589 = 478.91

The average number of jobs within production workers, all other, assuming the accuracy of the aggregate job numbers reported by USP is thus 478.91.  I rounded up, use your calculator to get a more accurate number.

If we assume 52 sedentary unskilled the DOT codes within the SOC code/OES group/Census code, we get 24,903.  How many sedentary unskilled DOT codes exist within 51-9199/8965?  The answer is 52.

Dividing 193,957 by 478.91 yields 405.  How many light unskilled DOT codes exist within 51/9199 and 51-3099 or census code 8965?  The answer is 405.

Dividing 88,598 by 478.91 yields 184.999 or 185.  How many medium unskilled DOT codes exist within 51/9199 and 51-3099 or census code 8965?  The answer is 185.

Dividing 20,593 by 478.91 yields 42.9997 or 43.  How many heavy unskilled DOT codes exist within 51/9199 and 51-3099 or census code 8965?  The answer is 43.

USP states that the semiskilled and skilled ranges of work aggregate two 432,933 jobs.  Dividing that number by the average number of jobs per DOT code comes out to 904.  Adding together the results of our divide and conquer request from the sedentary, light, medium, and heavy ranges of work brings the total number of occupations to 1,589.   By doing the math, we have ascertained that USP uses an aggregation methodology that starts at the aggregate number of jobs within the Census code.  The methodology assumes that every occupation within the group represents the same number of jobs.  Statisticians call this aggregation error.

If we take a look at USP's other publication, The specific Occupational-Unskilled Quarterly (SOEQ), we can readily ascertain that the number of jobs described in the OEQ as belonging in 51-9199/8965 is wrong.  Adding up the 52 sedentary occupations totals 14,432 jobs.

If USP has used a valid measure for estimating the number of sedentary unskilled jobs, then USP would have the same result in both publications.  USP does not have the same result in both publications because the two publications use a vastly different methodology for estimating the number of jobs.  The OEQ uses a frank aggregation, dividing the number of jobs by the total number of DOT codes and then multiplying by the number of DOT codes within a specific classification, e.g. sedentary and unskilled.

The SOEQ uses the industries to estimate the number of sedentary unskilled jobs.  But the SOEQ reports that 11 different DOT codes have exactly the same number of jobs, 193.  There are several sets of pairs were two occupations have the same number of jobs.

What is clear is that it is unreasonable to rely upon the OEQ to estimate the number of jobs.  USP does not start with the BLS job number reported in the OES.  USP does not start with the number of jobs reported in the Occupational Outlook Handbook.  The starting point for the number of jobs is unreliable; the methodology is invalid; and the other publications from USP demonstrate that the OEQ is not substantial evidence.