The US Jobs Market Is Much Worse Than The Official Data Suggest: The Full Story | Zero Hedge
Following Friday's disappointing payrolls report, yesterday we showed another even more troubling fact about the state of the US labor market: since 2008, over 93% of the total 6.7 million net jobs "created" in the past decade, have been statistical, existing simply inside an excel model somewhere in the US Department of Labor, as a result of the BLS' favorite fudge factor, the Birth/Death adjustment.
Unfortunately, that's just the tip of the iceberg for why the US labor market "recovery" is perhaps the biggest 'fake news' of the US economic narrative, and as a comprehensive recent analysis issued by Morningside Hill reveals, the state of the US jobs market is far worse than the official data suggest.
Here is the real story.
The US jobs market has been described as the backbone of the recovery – 80 months of continuous jobs growth with unemployment hitting 4.3% - the lowest since 2001. However the perceived strength in jobs creation is at odds with other economic indicators. President Trump ran on a campaign that repeatedly touted “jobs, jobs, jobs.” His emphasis on jobs creation and bringing employment back to America struck a chord with voters. Trump’s election in itself contradicts the popular narrative that the US jobs market is tight and robust. Wages, disposable income and real earnings growth along with low productivity and overall slow economic growth all challenge the BLS’s jobs numbers and thus Wall Street’s perception that the jobs market is tight.
Since the monthly jobs report is eagerly awaited as the most important piece of economic data for financial markets, it warrants a deep dive in order to understand what is going on under the hood. Before we delve into the data, here are some highlights of our findings.
The Bureau of Labor Statistics (BLS) has been systemically overstating the number of jobs created, especially in the current economic cycle.
The BLS has failed to account for the rise in part-time and contractual work arrangements, while all evidence points to a significant and rapid increase in the so-called contingent workforce.
Full-time jobs are being replaced by part-time positions, resulting in double and triple counting of jobs via the Establishment Survey.
A full 93% of the new jobs reported since 2008 and 40% of the jobs in 2016 alone were added through the business birth and death model – a highly controversial model which is not supported by the data. On the contrary, all data on establishment births and deaths point to an ongoing decrease in entrepreneurship.
Jobless claims have recently reached their lowest level in 43 years which purportedly signals job market strength. Since hiring patterns have changed significantly and increasingly more people are joining the contingent workforce, jobless claims are no longer a good leading economic indicator. Part-time and contract-based workers are most often ineligible for unemployment insurance. In the next downturn corporations will be able to cut through their contingent workforce before jobless claims show any meaningful uptick.
Overall, we have found the headline jobs number, unemployment rate and jobless claims to be poor macroeconomic indicators, since they have failed to account for significant shifts in labor market dynamics.
Who computes the jobs report and how
The “nonfarm payroll employment situation” report (NFP) or the “jobs report” as it is widely known, is calculated monthly by the Bureau of Labor Statistics (BLS) and is released every first Friday of the month. The two most important statistics in the report are the number of jobs added each month and the unemployment rate. The data for these statistics are collected through two separate surveys.
Number of jobs added
The number of jobs added to the economy comes from the Establishment Survey. Basically the BLS sends out the following survey to a sample of about 146,000 businesses.
As per the form, the government gathers the following information: how many employees worked at the firm in that month along with the total employee compensation and hours worked.
By surveying just a small fraction of US businesses, the measurement errors around this survey are significant. The 90% confidence interval is +-120,000 jobs. Thus in the March 2017 report which had +79,000 jobs, 90 out of 100 samples fall within the interval of -41,000 jobs to +199,000 jobs. In other words, statistically it is impossible to say with any confidence that the economy added any jobs that month, since the number crosses the zero bound and could therefore be negative.
Here is the official BLS data for the jobs added over the past 11 years, and a monthly breakdown for 2016.
The main problems with this survey are the double and triple counting of jobs because of a structural shift in labor dynamics and the business birth and death adjustment. We will proceed to tackle both in the following pages.
Double and triple counting of jobs in the Establishment Survey due to the transition to a “gig” economy
The shift to part-time and contract-based work has been a disconcerting development over the past years. This unprecedented growth in part-time work relationships has been dubbed the new “gig” economy, capturing the ad-hoc nature of these work arrangements. Increasingly employers prefer to add temporary part-time workers instead of full-time employees. The rationale is to save costs by paying less for full-time employment perks like healthcare and paid vacation leave among others, while increasing the flexibility for the employer to downsize the workforce if needed. This shift started a few decades ago and accelerated after the financial crisis of 2008.
This new “gig” economy is taking many forms – from Uber and Lyft drivers to multinationals like Amazon and HP – new employees are hired as contractors. A few decades ago part-time and contractual work was mostly utilized in seasonal industries like retail, food services and construction. It has recently spread to manufacturing, technology, airlines, telecoms, financials – it is a structural shift in almost every sector.
As more people are losing their full time jobs they often have to compensate by working multiple part-time jobs to make up for the lost income. For example a laid-off manufacturing employee might need to work 2 or 3 part-time jobs as a waiter at a restaurant and as an office clerk at a healthcare facility. When the BLS does their jobs count in the Establishment Survey they get 1 lost manufacturing job and 2 new jobs in services. Thus, the same employee is double-counted as if 2 new jobs were created for the single job that was lost.
Here is how the BLS treats multiple job holders according to their own definition:
“Establishments report the number of persons on payroll during the pay period that includes the 12th of the month. A person working multiple jobs at different establishments is counted once at each establishment. A person working different jobs at the same business establishment is counted once.”
This creates a perverse environment – as people lose well-paid full-time jobs with benefits, multiple new part-time jobs are created in the economy. However these new jobs are low paid and insecure positions. While the average worker in the economy is worse-off, this dynamic gives a false sense of economic strength.
Needless to say, part-time jobs are far inferior to full-time positions. Contingent workers are facing many challenges compared to full-time employees:
Low pay: contingent workers have a median hourly pay around $13 compared to $18 for full-time workers amounting to 38% lower remuneration.
Job instability: 28.5% of part-timers were laid off in the previous year compared to 8.2% for full-time employees.
Low access to health insurance: only 18% of contingent workers had employer-provided health benefits compared to 56% of regular workers.
High poverty rates: 33.1% of core contingent workers earn $20,000 or less per annum while only 10.8% of full-time workers fall within that category.
High reliance on public assistance – part-time employees rely on programs like the Supplemental Nutrition Assistance Program for example.
We would now like to consider the data behind these dynamics of a growing contingent workforce.
US Government Accountability Office Survey
A 2010 report by the US Government Accountability Office (US GAO) states that the actual number of contingent workers in the US economy is a full 37.1% of all employees. More importantly, this number has been rapidly growing. In 2006 the contingent workforce was only 32.4%, which means that in 4 years the number has grown by 4.7%. As the table below shows, most of the increase came from the standard part-time workers category.
The 4.7% increase translates to 9 million workers shifting to part-time employment in just 4 years. Unfortunately, 2010 was the last year in which the US GAO conducted a survey on the contingent workforce, so there is no current data available. We believe it is safe to assume that the percentage of contingent workers in the US economy has further expanded.
Harvard-Princeton study on employment
A new research paper by economists from Princeton and Harvard found that 94% of the jobs created in the current economic cycle were temporary positions. Lawrence Katz, a Harvard University professor and former chief economist at the US Department of Labor and Alan Krueger, a Princeton University professor and a former chairman of the White House Council of Economic Advisers, found that most of the jobs created under the Obama administration were temporary positions.
The study demonstrates that the percentage of workers engaged in alternative work arrangements (defined as temporary help agency workers, on-call workers, contract workers, and independent contractors or freelancers), rose from 10.1% in 2005 to 15.8% in 2015. Unfortunately, those newly created part-time positions account for the vast majority of all new jobs created over the past decade. To convert percentages into actual numbers – more than 9 million out of the 10 million jobs created in that time frame were in fact part-time. The charts below highlight some of the key findings of this study.
The Business Roundtable, a lobbying group composed of CEOs of the largest companies in Corporate America, release a quarterly CEO Economic Outlook Survey. The group includes companies with combined revenues of $7 trillion. They provide their hiring outlook in the CEO employment sub-index. As we can see from the chart below their outlook has been consistently far less optimistic than the numbers we have seen in the official payrolls reports. The blue line tracks the official BLS data, while the black line represents the CEO Employment Sub-Index.
The BLS’ own measure of part-time employment
The Establishment Survey does not include an estimate for part-time versus full-time employment, but the Household Survey does include one and here is the data:
As you can see in the chart above, according to the BLS there has been no significant increase in the contingent workforce. We did not have to look far to understand why the BLS has failed to gauge the rise of the “gig” economy.
The way the BLS measures part-time employment is based purely on the number of hours worked:
“Full time is 35 hours or more per week; part time is 1 to 34 hours per week.”
Therefore, as long as one can tally 35 hours of weekly work regardless of the number of jobs they hold, they are counted as full-time employees. Needless to say, many of the part-time employees need to maintain anywhere from 1 to 4 part-time jobs to survive financially, easily reaching the 35 hours per week threshold. They are all counted as full-timers by the BLS.
The controversial birth death adjustments
In order to account for jobs created or lost by new business formations or bankruptcies each month, the BLS introduced the birth/death adjustment. It started during the Reagan administration as Reagan was complaining that the bureau was undercounting the jobs he created. The birth-death model used to have a terrible name – the “bias adjustment factor.” This adjustment is computed using a model based on probability-based sampling methodology.
The table below shows the number of jobs that were added through birth/death adjustments over the past 17 years and the percentage of jobs added through the birth/death model
Let’s analyze the data.
Before 2003 few jobs were added through the adjustment, despite the fact that net business formations were much stronger back then (see data below).
Then, what strikes us as odd, is that according to the BLS in the depths of the 2007-2009 recession, the birth/death adjustment continued to add a lot of jobs – 904,000 jobs were added in 2009 alone. One would assume that in the nadir of the Great Recession when business defaults skyrocketed, the birth and death adjustment would be a net negative and subtract from the overall jobs number instead of adding to it.
Lastly, it turns out that a full 30% of jobs created since 2010 or 4.5 million out of 15 million jobs were added via the birth/death adjustment. It is also interesting to note that 40% of the jobs added in 2016 came through the adjustment.
The reason the BLS wanted to include this adjustment was a perception that they were undercounting jobs created through new start-up business formations (that were too young and too small to show up in the Establishment Survey). Those start-ups would eventually appear in their data, but with a few months’ lag. Therefore, if there was a steady supply of new start-up businesses and no sudden shifts in the trend, no adjustment would be necessary. Logically, it would only make sense to apply the adjustment if there is a significant increase in the rate of start-up formations, which has not materialized. On the contrary, multiple studies track a consistent decline in new business creation. Literally every study we have found documents the consistently deteriorating entrepreneurial environment in the US.
The following charts trace a clear downward trend in both employment gained from private sector births and the number of business births per year. Notice the suppressed level of births after 2008.
Furthermore, self-employed persons as a percentage of the working age population and the number of jobs created by establishments less than one year old are also declining.
A study by Harvard Business School entitled “Problems unsolved and a nation divided” summarizes the findings of its multi-year long project called “The US competitiveness project.” The study is a “fact-based effort to understand the disappointing performance of the American economy.” We found this project to be well worth the read and have selected the following chart (below to the left) depicting the multi-decade slowdown in new business formation. Further supporting the Harvard study findings, a Brookings Institution paper called “Declining business dynamism in the United States: a look at states and metros” shows that business formations slowed down and business deaths accelerated after the crisis of 2008 (below to the right).
Below to the left we have a chart from the Economic Innovation Group showing the net annual change in the number of US firms. Notice the significant slowdown after 2008, including 3 negative years. This is clearly not captured by the data from the Bureau of Labor Statistics. Below to the right we have a few charts from the Wall Street Journal summarizing some data points that confirm these trends.
With the data on new business formations and deaths in mind let us now go back to the BLS’s official birth / death adjustments. We have charted the net jobs added through the BLS model and ran a linear trend line to see if it captures the deteriorating entrepreneurial environment. In the chart below, the upward-trending line representing net jobs added through the adjustment is in complete dissonance with all the other data.
The Bureau of Labor Statistics (BLS) seems to be alone in its belief that the entrepreneurial environment in the US is improving. We believe that the BLS has been artificially inflating the monthly payroll numbers via the birth and death adjustment. This overstatement is not trivial in nature – the adjustment added 30% of all jobs reported since 2010.
Employment by companies in the S&P 500 index as a proxy
Instead of looking at employment surveys, Deutsche Bank compiled employment statistics from the S&P 500 index. The downside to using the S&P 500 as a proxy is that the index is clearly skewed towards very large businesses and it excludes small and medium enterprises. On the flipside, we are looking at actual data instead of relying on surveys. Since the index employs almost a fifth of the country’s workers (about 17%) it can offer a glimpse into the large business employment situation. As public companies report the total number of employees every quarter, Deutsche simply collected their employment numbers. According to their research, employment growth in S&P 500 companies just turned negative for the first time since 2009.
The unemployment rate is calculated from a separate measure called the Household Survey. The BLS calls about 60,000 households and asks if any members have worked for at least 1 hour during the month or if they have been actively searching for a job. If any household members have worked for at least 1 hour they are classified as employed, if they have been actively searching for a job, they are classified as unemployed.
It is interesting to note that in order to be counted in the labor force, a household member should have either been actively looking for a job or have worked for an hour in the past month. If neither of these conditions is satisfied, the member is not counted towards the labor force and they are not considered unemployed. To be classified as actively seeking employment one must meet the following criteria:
Contact employers or employment agencies directly and have job interviews within the last 4 weeks;
Submit resumes or fill out job applications within the last 4 weeks;
Place or answer job ads within the last 4 weeks.
With such strict criteria for being classified as unemployed many individuals who are looking for a job are not counted in the labor force, thus making the headline unemployment number look artificially low. People who have been unemployed know that there can be long stretches of time when they don’t place or answer job ads. But after 4 weeks have elapsed they simply get removed from the labor force despite the fact that they are still looking for a job.
On the other hand, the criteria to be categorized as employed are as relaxed as possible – just 1 hour of paid work is sufficient. No signed work contract or any formal arrangement is required. This means that if my daughter mowed the neighbor’s lawn or walked their dog for a few bucks she would be counted as an employed person. However, if I am looking for a job, but did not place or answer any job advertisements in the past 4 weeks, I will not be counted as unemployed.
We would like to focus on two measures of unemployment. The first one is the headline unemployment rate, which the BLS calls U3 and is the official rate you read about in newspapers. The second one is the broader unemployment rate (U6) which adds marginally attached and discouraged workers to the U3 rate. The chart below portrays the large difference between the measures with the U6 rate being close to 10% while the official rate is at 4.5%.
The major difference between these two rates can be explained by the labor force participation rate. As we explained, unemployed persons are easily excluded from the labor force given the strict criteria set by the BLS. These individuals are therefore excluded from the labor force participation rate. As depicted in the chart below, currently about 63% of the working-age population is counted toward the labor force, down from 67% a few years ago. Over the past decade this number has been steadily decreasing. It has recently reverted back to its level from the 1970s when women were underrepresented in the labor force.
The current very low unemployment rate of 4.5% is at least partly explained by the low labor force participation rate.
Slow wage and disposable income growth
A strong labor market is usually characterized by robust growth in wages and disposable income. This is simply not the case in the current cycle. Nine years into the recovery, both wage and disposable income growth in the US are still trailing previous cycle levels.
In addition, data from the Economic Policy Institute breaks down wage growth by educational level. While the employees with advanced degrees were able to preserve their levels of compensation, all other categories have seen their wages decline in the current cycle.
The chart above to the right represents the total wages and accrued salaries for US workers as a percentage of GDP. After peaking in the 1950s and 1960s, the trend has been downward sloping
The transition to a “gig” economy is a global phenomenon
The rise of part-time work arrangements has not been limited to the US. According to a study by McKinsey Global Institute between 20% and 30% of employees in Europe are contingency workers.
The following charts represent the worsening employment-to-population ratios for a few advanced economies, including the US. This is especially interesting for economists who blame the declining labor force participation rate on deteriorating demographics. Demographically challenged Japan has seen an improving employment-to-population ratio, unlike the US. The second chart dispels another popular myth – having a higher education does not lead to secure employment.
Part-time work arrangements are set to expand further. A new report by the Economist Intelligence Unit entitled “Global firms in 2020: the next decade of change for organizations and workers” has conducted a survey of 479 senior executives on how they see their businesses developing by 2020. They expect more positions to be automated or outsourced, they foresee an increasingly contingent workforce with more mobile or flexible work hours. 62% expect a growing proportion of workers to be contract-based while only 12% expect a higher share of permanent staff. A flexible workforce will make it easier to scale up or down as business needs dictate - “just in time” resourcing. As Robert Orth, HR director for IBM, explains, it is not easy to forecast HR needs, especially in high-tech fields where skills have a short lifespan. The goal is to build a business model “that is flexible enough that even if you don’t get the forecasting exact, you can find and move skills and capability at shorter notice.
Why is the US Government still using surveys to assess employment?
We find it interesting that in this day and age of infinite computational power, the internet, “big data” and electronic services and payments, the US government is still using phone surveys to assess the jobs market.
Despite being a non-US fund based in Europe with non-US investors, Morningside Hill still has to file FATCA documents with the US Internal Revenue Service (IRS). While the IRS has up-to-date information on a European fund and its non-US investors, another branch of the government, the BLS, has to call random households to extrapolate employment numbers in the States.
“In today’s computerized world why do sampling surveys at all, why not log on to the social security metrics and computers and ask them how many people they collected the FICA taxes from that month?”
Erica Groshen, Commissioner of the Bureau of Labor Statistics [answers with a chuckle]:
“Actually, that information takes a long time to get from these administrative data sources.”
We are not quite sure if we can take this answer seriously but we will leave it at that.
Usually government agencies are under constant political pressure to report better-looking statistics. The rationale is fairly obvious – it makes politicians look better. The BLS is no exception. Over the years it has been subject to many changes in practice and methodology that have always resulted in better-looking numbers. Keep in mind that the chief of the Bureau is directly appointed by the president.
Reagan’s administration insisted on including some sort of an adjustment for newly created businesses which resulted in the birth and death adjustment. Then under pressure by the Clinton administration, the BLS redefined the workforce to exclude all but a small percentage of discouraged workers. This resulted in the statistical disappearance of about 4 million unemployed persons, now classified as long-term discouraged. Clinton also reduced the sample size of the household survey, thus excluding a disproportionate share of inner-city households who were less likely to hold jobs. Furthermore, George W. Bush’s administration introduced the probability-based sampling methodology which, as we have seen, increased the birth and death adjustment figures.
Alterations like these have driven economics-data skeptic John Williams to describe them as “Pollyanna creep” – the inclination of statistics agencies to make adjustments that over long periods of time make the data look better than the economic reality.
Initial jobless claims is another widely followed economic indicator which represents the number of people who are filing to receive unemployment insurance benefits, as reported weekly every Thursday by the US Department of Labor. It is important to note that this number includes only new filings i.e. individuals who have just lost their employment. Therefore, the number has been an important early-warning indicator – when it increases it portends trouble in the economy as layoffs surge.
In the past few years, the jobless claims indicator has been stellar. As visible in the chart below, the claims recently fell to 234,000 per week – the lowest number since 1973, even though the US population has increased by 50% since then.
Is it possible that while the US population increased by 50% in 43 years, the jobless claims remain at the exact same level? Above to the right we have posted a typical explanation by the FT, which mirrors most other media – the low number is a sign of a tightening labor market.
It is hard to believe that today’s sluggish economy (growing at less than 2%) is the strongest we have seen in 43 years. As our research reveals, there are structural reasons for this indicator’s subdued levels. As employees move to the “gig” economy with contingent work arrangements, fewer people are eligible for unemployment insurance. Here is a list of factors that suppress initial jobless claims:
The economy has added mostly part-time jobs and fewer workers are eligible for unemployment benefits.
Employees in contingent work arrangements tend to hold multiple jobs to maintain financial solvency. Even if they lose one of their part-time positions, they often do not qualify for unemployment insurance.
Consequently, the low number of initial jobless claims is a misleading indicator. In the next downturn corporations will be able to cut through their increased contingent workforce, before they need to lay off any full-time workers. Therefore, when economic conditions deteriorate, initial jobless claims can remain low for a protracted period of time.
There has been an ongoing macroeconomic debate on the BLS’s job numbers and why they are at odds with so many economic indicators. After presenting the evidence on new business formation dynamics, we believe this table merits another look.
We won’t call the jobs added through the birth and death model fictional, but they can be described as hypothetical, unaudited data, with no support from actual business birth / death statistics.
Beyond the birth and death model, no one knows how many of the BLS jobs were double and triple counted part-time positions. According to some studies (Harvard-Princeton, etc.) most of the jobs added after 2005 were indeed part-time jobs. Therefore if, according to official numbers, we added a net 6.7 million jobs over the past 9 years and most were part-time jobs, while the birth and death model added 6.3 million of these jobs, this means that the actual number of full-time jobs has declined significantly. By this measure the jobs market has never recovered from the 2008 recession. In this case, US workers have every right to be discontented. This may explain why so many of them voted for Trump and his ‘bring back the jobs’ platform.
The NFP jobs and the initial jobless claims reports are flawed indicators. If tomorrow General Motors lays off 1,000 full-time workers and they end up working two part-time jobs each, the BLS will report a net gain of 1,000 jobs. If, further down the line, those same workers lose their part-time jobs, they will not be eligible for unemployment insurance and will not show up in the initial jobless claims report.
Fretting over whether the next jobs number will be 160,000 or 211,000 adds little value to any fundamental analysis.