Re:Dazed & Confused: Notes on Labor Force & Wage Growth
Earlier today Stephen Williamson, of the St. Louis Fed, posted an excellent response to the vocal disappointment of many economic commentators over the low headline number of payroll growth in May reported on the Employment Situation report, rebutting claims of labor softening and illustrating that the labor market is actually quite tight. While the review of measures is excellent, Williamson regrettably omitted a discussion of wage inflation, which I will mostly focus on here.
As a background, tightening labor market is something my colleagues Matt Busigin, Conor Sen and I have been arguing-for for literally years now. Matt with his call that the US would experience a continuing normal labor recovery in Oct, 2012 and I in Dec, 2013 arguing that we were on the eve of wages leading to corporate profit compression, and then again in January 2015 in the unambiguously-titled “Labor: To the Moon!” For years we’ve publicly called attention to the fact that labor conditions are at levels that are coincident with increasing labor bargaining power (May 2014; Feb 2015) and that alternate measures of compensation are showing very robust wage growth (Aug, 2014; Mar, 2015; Aug, 2015) especially as entry-level positions became hard to fill by 2015 leading to low-prerequisite jobs seeing faster wage growth, which has been visible in the skew, much to the chagrin of commentators who insist “unskilled labor” (there’s no such thing) will be made obsolete by robots *eyeroll*. When final demand underwhelmed in late 2015, we showed it was most likely noise due to temporary drag from an increase in effective tax rates, not the start of a decline. I’ll stop with the victory laps, for now.
Labor Force Growth
One of the key points is Williamson’s piece is that we should not be concerned about low headline levels of employment growth as the pool of potential hires continues to shrink, a view he first publicized in October 2015.
With the working-age population (ages 15–64) growing at a low rate of about 0.5%, if the labor force participation rate failed to increase and the unemployment rate stopped falling, payroll employment could grow at most by 60,000 per month, as I saw it last October.
My colleagues have shared this view, too (April 2015), but I am far less sanguine about the lack of growth in the labor force and believe unemployment can head lower still and that we’ll experience reduced incidence of retirement until the next recession is upon us. I also think that tight labor and rising real wages could possibly prevent exits and possibly even drive some re-entries into the labor force.
For one, teenagers still experience low employment rates. This could be assumed to be a secular decline, but given rising wages for low-prerequisite jobs, I do not expect the 16–19 unemployment rate to be in a “new normal.” In fact, while white teen participation remains low, black teen employment-to-population ratios are re-entering historical levels. The decline in labor force participation of tertiary education age individuals of ~15 percentage points and secondary education age individuals by ~20 percentage points has no clear explanation. Rising enrollment in tertiary education would leave the simultaneous decline in 16–17 participation to be explained. Likewise, availability of student loans for university-age students would not explain the decline in the 18–19 group. And while participation has kept mostly steady among the prime-age 35–54 years, it remains depressed for 20–24 individuals. What is clear is that, from age 16 to age 24, incidence of labor force participation increases as a function of the subject’s age, remains steady until age 55 and then starts to decline rapidly.
In contrast to the decline in participation by the younger population, the elderly population has steadily increased their participation rate, with the biggest percentage point gains concentrated in the most elderly workers. This is why I think that expectations of demographic stagnation of participation are premature, especially in a tight labor market where experience and seniority may be able to earn a premium or lead to increased job security.
Which leads me to the following chart, projected population changes broken down by age group. Between 2015 and 2020, the Bureau of the Census estimates that population will increase by 12 million individuals in 2 age groups: 24–44 (+4.861mm) and 65–84 (+8.188mm). Using estimated 80% & 20% rates of participation respectively gives us an additional 5.526mm participants, an increase of 3.5% over the 2015 annual figure of 157mm. This alone translates to a 0.70% growth rate, which is significantly higher than Williamson’s 0.5% figure. If we were to see 18–24 participation increase even modestly, labor force growth could easily surpass that. A 1% increase in the 18–24 participation rate over a period of 5 years would lead to an additional 0.04% in the annual growth rate of the labor force (~0.20% over 5 years). Due to the incidence of participation increase during expansions prior to the mid 2000s, I think this is a very plausible scenario and I could easily see the labor force growing at 0.85% until the next recession, with employment growth exceeding that by 0.15%–0.20%. Unemployment bottoming between 3.7% and 4% would be well within levels experienced in prior expansions.
Before I move on to wage growth, I want offer one explanation to the precipitous decline in labor force participation of the younger population. It is my belief that the overlap of 2 large generations in the labor force during the 2000s led to a surplus of candidates, tipping pricing power to employers, and leading to stagnation in “ labor share.” My preferred method of illustrating this claim is what I call “Labor Intensity of Production,” which is the percentage of Gross Value Added that is ultimately paid as wages and benefits (Compensation of Employees). A labor supply shift created by the overlap of two out sized generations in the workforce combined with the incidence of two large recessions (2001 & 2008) could lead to the decline in bargaining power necessary to push this measure to levels that are severely depressed by historical standards, even during the 2001–2007 and 2009-present expansions.
A decline in bargaining power of employees combined with a large pool of potential applicants would lead to diminished demand for the most marginal workers: Youths. Teenagers, after all, are the worst; why would anyone ever hire one unless there was no other alternative? If the labor market really is tight — and I believe it is — and wages keep rising at their current pace, I would not be surprised to start seeing teenagers re-join the labor force. Wage gains at the lower-end, after all, have had the most growth lately. If clearing entry-level wages for low-prerequisite adult labor rise (grocery stores, retail, quick-service restaurants, etc) rise significantly, I fully expect employers to resort to hiring teenagers, as they’ve always done, to protect margins or fill vacant positions.
Wage Growth!
One set of measures that was excluded from Williamson’s post were those surrounding wage growth. The inability to control for a multitude of variables make measures of wage growth imprecise at best, and, at worst, completely useless. The establishment's survey average hourly earnings is closer to the latter, in my opinion, as I wrote last year, so I tend to either use the Employment Compensation Index from the National Compensation Survey or concentrate on aggregate measures of of income.
Average Hourly Earnings and Average Hourly Earnings of Production and Nonsupervisory Employees are not measures of wage rates. They are aggregate measures of the amounts paid to workers which may include distortions such as overtime or holiday pay or fluctuations in the proportion of hours worked by higher or lower waged workers. They are calculated by dividing the sum of the payroll by the total hours worked for a cell and then calculating a weighted average across industries. It’s important to note that these series cannot be used as a measure of wage inflation as they do not control for overtime or the composition of the workers. For example, in a rising wage environment that coincides with an increase in low wage employees and an increasing number of hours worked by those employees, average hourly earnings could fall even though a measure that controlled for occupation would rise. Special bonuses and non-cash compensation are not reflected in this measure.
The ECI subseries for “Total compensation for All Civilian workers in All industries and occupations,” suggests that from Q1 2015 to Q1 2016 total compensation rose 2.4%, an increase which exceeds the current level of price inflation (more on this later). Adjusting for inflation, wages are rising, and much faster than in the 16 quarters from Q1 2010 to Q1 2014. This is notable because it was in March 2014 that Average Weekly Hours reached 34.6, their maximum level during this expansion.
The shift up in the trend during 2014 can also be seen in an alternate measure of income, real wages & salaries disbursed per employee, which clearly shows that there was a jump event in Q1 2014 during which the growth rate of income per employee shifted upwards almost 1.5%. The simultaneous shift up in average real income and the ECI as hours peaked is consistent with peak-utilization of the labor pool and higher wages required to attract suitable candidates.
This increase, in turn, appears to have arrested the downtrend in the labor force participation rate of the 25–54 workforce, where participation rates are once again increasing after a 7-year downtrend. The reason wage inflation hasn’t risen further is likely that the population is responding to growing wages with increased participation.
This is most clearly seen in alternative measures of unemployment, where non-participants that expressed desire to work as a percentage of the population has essentially normalized. These will be the first population that is re-absorbed into the labor force. once these measures finish normalizing it is likely that we will once again experience another shift in the income growth rate in order to attract additional participants (remember teens?).
Given that the absorption rate of new entrants (% of new entrants that transition directly to Employed) is near all-time-highs, it might not take very long to exhaust this shrinking pool of slack.
The May 2016 Job Report: One-off or New Normal?
I will admit to being surprised by the May jobs report. Even with the Verizon effect, I was expecting a much higher number, closer to 110k. It is my expectation that through a combination of a a higher headline and some upward revisions the payroll employment level will rise significantly over the next couple of months. One of the reasons for this expectation is that the most real-time measure off aggregate income disbursed, withheld employment income and tax deposits (the money that comes out of your paycheck before you get it) experienced on-trend growth during the month of May. Withholding can be volatile due to the presence of multiple patterns in the data (a combination of seasonality of employment, a different seasonality to deposits, and a different seasonality to pay periods) but with a sufficiently well-calibrated ensemble of adjustments it is possible to generate a seasonally adjusted value. In may the year-over-year growth rate of withholding exceed 5% while the monthly growth was 1.12% for May and (1.99% for April).
Comparing the seasonally-adjusted withholding level with wage & salaries disbursed allows us to estimate the effective tax rate for the periods preceding the current one to estimate the level of wages & salaries that the withholding would correspond to.
Absent a significant increase in the effective tax rate for the month of May and using the effective tax rate of the last 6 months, May’s withholding deposits would suggest wages and salaries disbursed of ~8,311 billion (SAAR) which translate to 6.5% growth over year ago levels. Using May’s headline, the total number of payroll jobs added over the last year has been 2.398 million, or ~1.695% of the 141.496 payrolls estimated for May 2015. If we assume hours unchanged and that wages & salaries are a function income growth and employment growth, we could solve for income growth, which would solve to 4.725%. Unless we are experiencing a similar shift to the one experienced in Q1 2014, this does not seem likely. This is why I am expecting a significant revision to May and/or a much higher headline number in June. The Treasury data is cash data that is free of revisions, and the money being deposit indicates that either payrolls need to be revised up, or the ECI is about to give us a big surprise.
While I don’t disagree with Williamson or with my colleagues about the arithmetic necessity for employment growth to slow as employment:population ratios rise, I do believe there is significant underestimation of the possible increases in participation, especially as wages rise. To anyone suggesting that the long-term unemployed / non-participants are not sufficiently skilled for any job I would argue that widespread labor shortages never start with middle-management and professional workers. Organizations can always train young recruits, promote up, or poach from competitors at the higher wage levels. It is at the entry level where shortages occur, once you exhaust the supply of entry-level labor, the only solution is to raise wages to what is necessary to compete with higher-prerequisite employers for the same labor pool or to bring in participants which may not have been interested in joining the labor force at lower wages (teens!).
Productivity
Lastly, Williamson presents a chart of the level of real GDP to payroll employment and suggests there is a shocking and concerning lack of productivity. He is right that this is indeed something to be concerned about, but wrong in that it represents a general lack of productivity growth. Using NAICS Value-Added by Industry data, we can compute the labor intensity of production for each industry type by dividing dollar compensation of employees by dollar value-added. Measuring from the peak of the prior expansion to minimize the difference in conditions at time of measurement, I measure the change percentage points of beginning and ending labor intensity. A lower, more negative, number signifies an industry where value-added grew faster than employee compensation and a higher, more positive, number the opposite. As can be seen, most sectors experienced significant declines.
A more likely explanation for the decline in output per employee is that, with a surplus of labor, adding additional workers was easy and capital investments into labor-saving technology were not needed. Also possible is that as efficiently-produced goods became cheaper, expenses were more likely to be diverted to more labor-intensive services, which necessitate a higher level of payroll growth per additional dollar of real output.
As labor continues to tighten and incomes continue to rise from their still-low levels, the economics of labor-saving capital investments will improve, leading to renewed producitivty growth by freeing lower productivity employees to be employed in higher value-added tasks.
Conclusions
- Expectations of low labor force growth are likely overstated. A 0.70% annualized rate of growth over the next 5 years is a credible base case and with sufficient wage inflation, the labor force could expand by another 0.15%-0.25% per year.
- Barring some unanticipated aggregate shock, assuming the current rate of compensation growth continues, and a 0.2–0.5% range of unemployment reduction would, when combined with higher estimates of labor force participation, lead to a range of something closer to 107–172k payrolls per month. My personal expectation lies somewhat in the middle at 144k/mo
- As wages growth accelerates, we should see an increase in labor-saving capital investments that will lead to productivity increase. Productivity is not dead, it’s just that labor has been to cheap for too long and that has reduced investment in productivity.
