by Aegidius25
The Bureau of Labor Statistics (BLS) put out its most recent report on the unemployment situation recently. I’ve long had a problem with the numbers this government agency publishes. Over the years they have been brazenly manipulated and much of the data it purports to show isn’t much more than a glorified guess. The report is made up of at least two elements; first the household survey from which it gets the official unemployment rate at 5.4% in July 2021 (-0.4% in July) and second a so-called establishment survey.
The first is just what it sounds like, an in person or telephone survey of US households compiled to figure out what proportion of workers are employed. By contrast the establishment survey is a combination survey and guesstimate. The first element surveys about 144,000 business establishments and government agencies compared to the about 32,500,000 businesses that used to be estimated to exist in the US; collecting data on numbers of workers at each firm, their hours worked and earnings. The second element is essentially a fancy formula used to guess at the number of new business establishments created or destroyed in the unsurveyed segment of the economy. (Note: It is likely the number of surveyed firms has also declined with the downturn.) From here the BLS comes up with an average number of jobs created or destroyed with all of these firms, giving us by far the biggest part of this figure in this survey (+938,000 in July).
According to the latter segment of this “survey” the economy saw increases in employment in leisure and hospitality (+380,000), education (+261,000), transportation and warehousing (+50,000), professional and business services (+43,000) and other services (+39.000) among others.
But these numbers don’t capture the whole picture. The BLS’ methodology and the size of their data sets bring much of this information into question. First because the unemployment rate from the household survey only takes into account workers considered in the labor force. This means workers who have been out of work for a period considered too long by the government and whose unemployment benefits have run out are no longer counted in this figure, affecting the laborforce participation rate at 61.7% in July (+0.1%).
Clerical errors have also abounded since the start of the pandemic. “As in previous months, some workers affected by the pandemic who should have been classified as unemployed on temporary layoff were instead misclassified as employed but not at work. However, the share of responses that may have been misclassified was highest in the early months of the pandemic and has been considerably lower in recent months. Since March 2020, BLS has published an estimate of what the unemployment rate might have been had misclassified workers been included among the unemployed. Repeating this same approach, the seasonally adjusted unemployment rate for July 2021 would have been 0.3 percentage point higher than reported. However, this represents the upper bound of our estimate of misclassification and probably overstates the size of the misclassification error.” (Employment Situation for July 2021 p. 6)
Then there is the issue that all data in this report is SEASONALLY ADJUSTED. This sounds like a means to take seasonal factors such as weather into account when calculating them, but ts really a practice by which long-run averaging is used to disguise volatility in the data. The downside to this is that it can misrepresent the actual situation. Say the economy’s been in a slump for a prolonged period and the average of jobs lost each month is -300,000. Then a number above -300,000 for a given month even if still negative at say -100,000 could still be counted as positive. The same goes for positive numbers; with an average of +300,000 jobs a month a number below that, say +200,000, could be ADJUSTED to show a less positive or even negative number. This is to some extent an oversimplification of the process but gets across the manner in which such figures are constructed. And the same sort of method is also applied to the unemployment rate and all other economic date from the government. So we can see how unreliable much of this data is.
In fact in July’s report this is stated outright when the bureau writes “[s]taffing fluctuations in education due to the pandemic have distorted the normal seasonal buildup and layoff patterns, likely contributing to the job gains in July. Without the typical seasonal employment increases earlier, there were fewer layoffs at the end of the school year, resulting in job gains after seasonal adjustment. These variations make it more challenging to discern the current employment trends in these education industries.” (The Employment Situation July 2021 p. 3). On an unadjusted basis the education sector actually lost about 900,000 jobs while the overall job change for the month was nearer -100.000 not +900,000
The manipulation is made clear by the fact that “[i]n accordance with annual practice, the establishment survey data released . . . have been benchmarked to reflect comprehensive counts of payroll jobs for March 2020. These counts are derived principally from the Quarterly Census of Employment and Wages (QCEW), which counts jobs covered by the Unemployment Insurance (UI) tax system. The benchmark process results in revisions to not seasonally adjusted data from April 2019 forward. Seasonally adjusted data from January 2016 forward are subject to revision. In addition, data for some series prior to 2016, both seasonally adjusted and unadjusted, incorporate other revisions.” (January BLS Report p. 6) So that “[d]ata users are cautioned that these annual population adjustments can affect the comparability of household data series over time.” (The Employment Situation January 2021 Report p. 7).
Additionally “At 10:00 a.m. (ET) on August 18, 2021, the Bureau of Labor Statistics (BLS) will release the preliminary estimate of the upcoming annual benchmark revision to the establishment survey data. This is the same day that the first-quarter 2021 data from QCEW will be issued. Preliminary benchmark revisions for all major industry sectors, as well as total nonfarm and total private employment, will be available at www.bls.gov/web/empsit/cesprelbmk.htm.The final benchmark revision will be issued with the publication of the January 2022 Employment Situation news release in February 2022.” (The Employment Situation June 2021 p. 6).
A better if still imperfect measure of the jobs situation can be found in a measure in the BLS report called the Employment-Population Ratio. This number, if still seasonally adjusted, tries to compare the actual number of people with a job to the entire working age population of the United States. Unfortunately it fails to take into account the number of say stay at home mothers, who actually are not in the labor force, but in our modern world this is such a rarity that statistically it shouldn’t impact the overall data too much. However neither does it take into account the number of retirees. This is easily corrected by finding what percent of the population is retired and subtracting it from the remaining number of people without a job. In this month’s BLS report the EMPLOYMENT-POPULATION RATIO IS A SEASONALLY ADJUSTED 58.4 (+0.4 in July). THAT MEANS 41.6% OF PEOPLE 22 AND OLDER ARE WITHOUT A JOB. IF WE SUBTRACT THE ABOUT 14% OF THE POPULATION THAT RECEIVES SOCIAL SECURITY WE GET A RATE OF JOBLESSNESS AT 27.6%. Though this figure isn’t modified to count disabled adults who’d like to work, elder workers collecting Social Security but who still have to work, or people who really aren’t disabled but added to the Social Security rolls during the Great Recession to keep them out of the unemployment figures. WHEN WE CONSIDER ALL THESE PEOPLE THE ACTUAL REMAINDER FROM THE EMPLOYMENT-POPULATION RATIO IS PROBABLY NEARER 30.8. THAT IS TO SAY ABOUT 31% OF WORKING AGE AMERICANS ARE WITHOUT A JOB COMPARED TO A MINIMUM OF 38% AFTER THE START OF THE PANDEMIC. Below is a link to this month’s employment report for any that are interested. www.bls.gov/news.release/pdf/empsit.pdf