Helping our clients measure job demand is central to what we do as a company, and something that we’re always trying to get better at. It would be great if there were some universal, objective “demand” number that we could apply to everyone everywhere, but like most problems, it’s not that simple. Different groups have different challenges. A college doesn’t have the same view of demand as a staffing firm. What to do?
In the past EMSI has used a number called job openings, an estimate based on national employment projections from the BLS. Openings is a combination of new and replacement jobs that many of our clients have used to measure job demand. It’s helpful in that it’s a more accurate picture of demand than looking at growth alone, since vacated jobs also need to be filled.
The downside is that openings takes a somewhat narrow view of what counts as a “replacement” job. If a dentist moves out of California to be a dentist in Vermont, that missing dentist in California isn’t counted by openings. Only workers who permanently leave their occupation through retirement, death, or changing careers factor in. The numbers are also modeled off of national projections, so regional nuances can get lost.
How else can we measure demand? Job postings have become increasingly popular in the past few years and we’ve talked about some of the pitfalls of using job ads as a substitute for labor market data. Job postings can be great for certain needs though, and as a supplement to other types of data. We’ll be talking about that in more detail in part two of this series with some interesting findings comparing “job postings demand” to jobs that actually materialize.
JOLTS and QWI
One of the most common approaches to measuring demand is found in the Job Openings and Labor Turnover Survey (JOLTS) from the BLS. JOLTS has been the go-to for measuring demand for some time, and it’s easy to see why. JOLTS produces monthly figures describing hirings, job openings, and separations for industries in the U.S. The job openings figure is especially attractive, with the survey describing openings as a specific position that exists with work available that can be full-time, part-time, or seasonal. The job would start within 30 days if the right candidate is found, and the employer is actively recruiting for the position.
On top of describing how many immediate openings businesses have, JOLTS also offers a fine level of distinction on how workers are leaving the industry; are they getting fired, laid off, quitting, or something else?
So what’s not to like? JOLTS promises monthly data on future job openings AND tells us why workers are leaving specific businesses. Sign me up!
Actually, there’s a catch. It’s easy to ignore the “S” in JOLTS and forget that we’re talking about a survey. JOLTS sends out voluntary survey forms to about 16,000 businesses to get their estimates on openings, hires, and separations. Contrasted with the 1.2 million businesses that OES surveys, and the over nine million businesses found in QCEW, that’s not a huge sample size. The BLS then takes those survey responses and models them out across all industries in the country. JOLTS isn’t alone in this practice, as other sources are also sample survey based and provide helpful indicators. Think Current Employment Statistics and the American Community Survey.
JOLTS is similar to the monthly jobs report, or “Employment Situation,” that the BLS also releases. It’s timely, and hugely influential, but at the cost of accuracy. More on that in a bit.
QWI (the last acronym we’ll burden you with today!) also gives us insight into how we can measure demand, and shows some real promise. Quarterly Workforce Indicators is published by the U.S. Census Bureau, and produces data on hirings and separations in the economy. It does this by connecting unemployment insurance forms from businesses with the unique Social Security numbers of each employee. Any time a worker shows up in a company’s payroll in one quarter when they didn’t show up in the previous, they are counted as a hire. Likewise, if for any reason a worker stops working for an employer, they get counted as a separation.
Data like this obviously takes some time, and isn’t perfect when we get it. QWI gets released at the industry level by county, and employs a number of other sources like QCEW, Business Dynamics Statistics, Census data, and individual tax returns. The cost of such complete coverage of all job movement in the economy is that the numbers take about a year to come out. One other drawback is that the state of Massachusetts has chosen not to participate in the program historically.
Job Churn and Its Implications
Figures on hires and separations allow us to see an aspect of the economy that’s previously been hidden below the surface of our data: job churn. Churn is a pretty straightforward concept, with important implications as we talk about measuring demand. Job churn is the movement of jobs in the economy that isn’t necessarily picked up by standard employment sources. To give you an idea of the scope of churn for the nation, take a look at these national numbers:
What can be tricky about churn figures like these is that it’s easy to confuse jobs with people. We aren’t saying that 71% of people changed jobs. We’re counting jobs, which means that in a given year, one job in the employed group could account for three, four, or 10 jobs in hires and separations. For example, a burger joint might show 40 jobs for 2013. But that doesn’t mean that only 40 people worked there that year. When someone quits and a new person is hired, that movement is hidden below the surface of traditional, yearly employment.
Churn is really noticeable (and expected) in industries like fast food and retail, but it isn’t unique to just low-wage, low-skill jobs. Manufacturing has had about a 40% churn rate historically, the information sector has been around 60%, and education a whopping 70% churn.
Job churn can be a helpful talking point and helps us understand job stability, but the real value of these numbers from QWI and JOLTS is the hires figures. If we’re trying to figure out what jobs are in demand, what better number to use than the number of hirings taking place?
Ultimately we want to find the best source for measuring hires. So we compared the actual headcount from QWI with the modeled estimates from JOLTS, and some interesting trends appeared. We found that all industry sectors except for professional and business services were underrepresented in the JOLTS figures, often by large amounts. Hotels and restaurants for instance showed 104% churn in 2011. The JOLTS numbers reported about 61% for the same year, missing about 5 million hirings.
It’s important to remember that we aren’t really pitting apples against apples here. QWI goes about collecting its numbers in a completely different way than JOLTS, and appropriately has different strengths and weaknesses. We’re talking about the difference between a headcount with QWI and estimates modeled from surveys with JOLTS.
One of the things JOLTS offers that QWI does not is that openings number. Being able to anticipate demand sounds really good. However, a look at historic JOLTS data revealed that the businesses surveyed have consistently reported more hirings than job openings.
As an exception, that can make sense. Say a software company reports only having five openings for the next month. The next week a flood of qualified developers and programmers submits résumés. That company is always looking for good developers and programmers, so it hires 15 of them, even though the “openings” didn’t exist beforehand.
But year after year the trend has stayed the same — always more hirings than openings. After a few years it takes some of the confidence out of that openings number. And that’s just within the JOLTS survey. When you compare to the actual, counted hires, you can see how much of the economy’s movement is missing from JOLTS. Even though they use different methods, both sources attempt to capture the number of hirings in the economy. We used this figure to do some direct comparison, while keeping in mind that JOLTS models out their survey findings and QWI waits for the dust to settle and counts noses. Notice how much lower the JOLTS figures are:
We’re still looking at the pros and cons of both sets, but our first impression has been that QWI offers more accurate numbers at the cost of timeliness, and JOLTS sacrifices good numbers for monthly updates. Historically, EMSI has focused on providing our clients with the most comprehensive numbers possible and to that end QWI seems like the source to lean on more heavily. In addition to the solid reliability, QWI also offers data on employment by race, ethnicity, and age groups that we haven’t yet tapped into.
We also aren’t writing JOLTS off by any means. There can certainly be value in using timely estimates to help inform projections and recent data that QWI hasn’t reported on yet, similar to how we currently use CES to keep our numbers up-to-date.
So how do job postings, or “real-time” data, fit into all this? Stay tuned for part two where we’ll compare some new EMSI data (churn and hires) with what the job postings reveal.