On March 18, EMSI hosted a webinar discussing our new guidebook, Contextualizing Real-Time and Traditional Labor Market Information. Josh Wright, director of marketing and PR, and Matt Gaither, training and Certification manager, discussed the report's key findings, including the key strengths and weaknesses of both data types, examples of occupations that are commonly underrepresented (or overrepresented) in job postings, and how to combine real-time and traditional LMI to provide context for labor market analysis.
For more information visit: www2.economicmodeling.com/contextualizing-real-time-report
4. Topics
• Defining job posting
analytics (JPA) and LMI
• Practical ways to use
JPA and LMI in tandem
www2.economicmodeling.com/contextualizing-real-time-report
5. Topics
• Defining job posting
analytics (JPA) and LMI
• Practical ways to use
JPA and LMI in tandem
• Overrepresented and
underrepresented
occupations in postings
www2.economicmodeling.com/contextualizing-real-time-report
6. Topics
• Defining job posting
analytics (JPA) and LMI
• Practical ways to use
JPA and LMI in tandem
• Overrepresented and
underrepresented
occupations in postings
• EMSI’s approach
www2.economicmodeling.com/contextualizing-real-time-report
7. Topics
• Defining job posting
analytics (JPA) and LMI
• Practical ways to use
JPA & LMI in tandem
• Overrepresented and
underrepresented
occupations in postings
• EMSI’s approach
www2.economicmodeling.com/contextualizing-real-time-report
8. Why This Report?
• Job posting data is widely used
www2.economicmodeling.com/contextualizing-real-time-report
9. Why This Report?
• Job posting data is widely used
• EMSI integrated JPA and LMI in Analyst
last summer
www2.economicmodeling.com/contextualizing-real-time-report
10. Why This Report?
• Job posting data is widely used
• EMSI integrated JPA and LMI in Analyst
last summer
• Recent reports have emphasized using
JPA and LMI in conjunction
www2.economicmodeling.com/contextualizing-real-time-report
11. Why This Report?
• Job posting data is widely used
• EMSI integrated JPA and LMI in Analyst
last summer
• Recent reports have emphasized using
JPA and LMI in conjunction
• But how?
www2.economicmodeling.com/contextualizing-real-time-report
12. Key Points
• JPA and traditional/structural LMI are
complementary
www2.economicmodeling.com/contextualizing-real-time-report
13. Key Points
• JPA and traditional/structural LMI are
complementary
• Each answers questions the other can’t
www2.economicmodeling.com/contextualizing-real-time-report
14. Key Points
• JPA and traditional/structural LMI are
complementary
• Each answers questions the other can’t
• Data is ultimately a tool: need to know
what it’s good for, what it’s not good for
www2.economicmodeling.com/contextualizing-real-time-report
15. Key Points
• JPA and traditional/structural LMI are
complementary
• Each answers questions the other can’t
• Data is ultimately a tool: need to know
what it’s good for, what it’s not good for
• Key link between JPA and LMI: hiring data
www2.economicmodeling.com/contextualizing-real-time-report
16. Data collected from public sources
used to describe trends and
projections for standardized
industries and occupations.
LMI is…(Labor Market Information)
17. Online ads from companies trying
to attract applicants
Job Postings
18. Data is Like a Map
www2.economicmodeling.com/contextualizing-real-time-report
19. Data is Like a Map
www2.economicmodeling.com/contextualizing-real-time-report
24. Job
Postings
• From companies trying
to attract applicants
• A small sample of the
potential labor force
LMI
• From companies filling in
government forms
• Total labor force
www2.economicmodeling.com/contextualizing-real-time-report
25. Intentions vs. Actions
• JPA primarily measure intentions, and
only those of employers who advertise for
jobs online.
www2.economicmodeling.com/contextualizing-real-time-report
26. Intentions vs. Actions
• JPA primarily measure intentions, and
only those of employers who advertise for
jobs online.
• LMI measures actions and tracks all
payroll employment.
www2.economicmodeling.com/contextualizing-real-time-report
27. Intentions vs. Actions
• JPA primarily measure intentions, and
only those of employers who advertise for
jobs online.
• LMI measures actions and tracks all
payroll employment.
• This has several ramifications …
www2.economicmodeling.com/contextualizing-real-time-report
28. Ramifications
1. Postings aren’t simply a faster version of
LMI, though they can be a leading
indicator of job growth in some sectors.
www2.economicmodeling.com/contextualizing-real-time-report
29. Ramifications
1. Postings aren’t simply a faster version of
LMI, though they can be a leading
indicator of job growth in some sectors.
2. Postings data by itself can’t give you a
complete picture of the workforce.
Without context, it’s hard to distinguish
between substantive trends and noise.
www2.economicmodeling.com/contextualizing-real-time-report
30. LMI Isn’t Cure-All
Traditional LMI doesn’t provide all the
answers in isolation, either
www2.economicmodeling.com/contextualizing-real-time-report
33. LMI Strengths
• Thorough
• Standardized
• Lots of detail on set industries (NAICS)
and occupations (SOCs)
www2.economicmodeling.com/contextualizing-real-time-report
34. LMI Strengths
• Thorough
• Standardized
• Lots of detail on set industries (NAICS)
and occupations (SOCs)
• Insight for strategic decisions
www2.economicmodeling.com/contextualizing-real-time-report
35. LMI Weaknesses
• Not easy to collect and integrate
www2.economicmodeling.com/contextualizing-real-time-report
36. LMI Weaknesses
• Not easy to collect and integrate
• Inconsistent focus
www2.economicmodeling.com/contextualizing-real-time-report
37. LMI Weaknesses
• Not easy to collect and integrate
• Inconsistent focus
• Non-disclosed data
www2.economicmodeling.com/contextualizing-real-time-report
38. LMI Weaknesses
• Not easy to collect and integrate
• Inconsistent focus
• Non-disclosed data
• Not necessarily timely
www2.economicmodeling.com/contextualizing-real-time-report
39. LMI Weaknesses
• Not easy to collect and integrate
• Inconsistent focus
• Non-disclosed data
• Not necessarily timely
• Inflexible
www2.economicmodeling.com/contextualizing-real-time-report
40. LMI Weaknesses
• Not easy to collect and integrate
• Inconsistent focus
• Non-disclosed data
• Not necessarily timely
• Inflexible
• Poor connections to businesses
www2.economicmodeling.com/contextualizing-real-time-report
41. JPA Strengths
• Leading indicator
www2.economicmodeling.com/contextualizing-real-time-report
42. JPA Strengths
• Leading indicator
• In-demand skills and credentials
www2.economicmodeling.com/contextualizing-real-time-report
43. JPA Strengths
• Leading indicator
• In-demand skills and credentials
• Valuable for “now” decisions
44. JPA Strengths
• Leading indicator
• In-demand skills and credentials
• Valuable for “now” decisions
• Way go gauge employers’ effort
45. JPA Strengths
• Leading indicator
• In-demand skills and credentials
• Valuable for “now” decisions
• Way go gauge employers’ effort
• Who’s hiring
47. JPA Weaknesses
• Limited in scope
• One posting doesn’t = one opening
www2.economicmodeling.com/contextualizing-real-time-report
48. JPA Weaknesses
• Limited in scope
• One posting doesn’t = one opening
• Biased towards high-skill fields
www2.economicmodeling.com/contextualizing-real-time-report
49. JPA Weaknesses
• Limited in scope
• One posting doesn’t = one opening
• Biased towards high-skill fields
• Geographic differences
www2.economicmodeling.com/contextualizing-real-time-report
50. JPA Weaknesses
• Limited in scope
• One posting doesn’t = one opening
• Biased towards high-skill fields
• Geographic differences
• Not a good measure of hiring demand
www2.economicmodeling.com/contextualizing-real-time-report
51. JPA Weaknesses
• Limited in scope
• One posting doesn’t = one opening
• Biased towards high-skill fields
• Geographic differences
• Not a good measure of hiring demand
• Not the only businesses recruit
www2.economicmodeling.com/contextualizing-real-time-report
52. Importance of Hires
• Link between LMI
and JPA
• By comparing hires
to postings by
occupation, we
have information
that’s greater than
the sum of its parts
www2.economicmodeling.com/contextualizing-real-time-report
53. Applying JPA and LMI
• The report goes into applications for:
• Colleges and universities
• Workforce development
• Economic development
• Private sector
www2.economicmodeling.com/contextualizing-real-time-report
55. Workflow Example
• Start with industry data and finish with
postings
• Look at Activities Related to Credit
Intermediation industry
• Phoenix MSA
www2.economicmodeling.com/contextualizing-real-time-report
57. From Industry to Occupation
…
• Personal financial advisors part of this
industry
• Make $28/hour
• Grown 32% since 2010
www2.economicmodeling.com/contextualizing-real-time-report
58. … to Job Postings
• Edward Jones posting more intensely
than other employers
• Certifications: Certified Financial Planner,
Series 6, etc.
• Skills: Finance, Investments, Marketing,
Business Development
www2.economicmodeling.com/contextualizing-real-time-report
59. Why Is Context Important?
• If we had just looked at postings, we
would have seen the skills and
certifications for personal financial
advisors.
www2.economicmodeling.com/contextualizing-real-time-report
60. Why Is Context Important?
• If we had just looked at postings, we
would have seen the skills and
certifications for personal financial
advisors
• Wouldn’t have known if it was an
important occupation, and what industries
it’s tied to
www2.economicmodeling.com/contextualizing-real-time-report
61. Representation in Postings
• Postings data comes from job
advertisements posted online.
www2.economicmodeling.com/contextualizing-real-time-report
62. Representation in Postings
• Postings data comes from job
advertisements posted online
• Means some occupations are naturally
going to be underrepresented (or
overrepresented) in postings
www2.economicmodeling.com/contextualizing-real-time-report
63. Representation in Postings
• Postings data comes from job
advertisements posted online
• Means some occupations are naturally
going to be underrepresented (or
overrepresented) in postings
• See this clearly when comparing postings
to hires
www2.economicmodeling.com/contextualizing-real-time-report
68. Takeaways
• Professional, high ed. level occupations have better
representation in postings
www2.economicmodeling.com/contextualizing-real-time-report
69. Takeaways
• Professional, high ed. level occupations have better
representation in postings
• Middle- and lower-skill jobs are mostly (but not
always) absent from postings
www2.economicmodeling.com/contextualizing-real-time-report
70. Takeaways
• Professional, high ed. level occupations have better
representation in postings
• Middle- and lower-skill jobs are mostly absent from
postings
• EMSI’s postings & hires does the hard work
www2.economicmodeling.com/contextualizing-real-time-report
71. EMSI’s Approach
• Current, complete picture of local labor
markets
• Suppressions removed
• Self-employed added
www2.economicmodeling.com/contextualizing-real-time-report
72. EMSI’s Approach
• Current, complete picture of local labor
markets
• Suppressions removed
• Self-employed added
• Bedrock LMI based on 90+ publically
available sources
www2.economicmodeling.com/contextualizing-real-time-report
73. EMSI’s Approach
• Current, complete picture of the labor
market
• Suppressions removed
• Self-employed added
• Bedrock LMI based on 90+ publically
available sources
• Supplement with job posting analytics
www2.economicmodeling.com/contextualizing-real-time-report
74. EMSI’s Approach
• Current, complete picture of the labor
market
• Suppressions removed
• Self-employed added
• Bedrock LMI based on 90+ publically
available sources
• Supplement with job posting analytics
• Link with hiring data
www2.economicmodeling.com/contextualizing-real-time-report
Hello, and thanks for joining us today. My name is Josh Wright, and I’m the director of marketing here at E-M-S-I.
In this webinar, my colleague Matt Gaither and I are going to discuss EMSI’s new guidebook, Contextualizing Real-Time and Traditional Labor Market Data. Your phone lines have been muted, but please feel free to ask questions at any point in the chat box. We’ll get to them during a Q&A period at the end of our discussion.
The topics we’ll cover today include defining traditional labor market information and real-time data (what we refer to as job posting analytics, or JPA);
applying JPA and traditional LMI in tandem
and discussing over and underrepresentation in postings.
We’ll also go over EMSI’s approach to combining JPA and LMI.
That second bullet, using both datasets in tandem, is a good segue to discuss why we wrote this report.
The use of job posting data by local workforce practitioners is now widespread.
Last summer EMSI introduced job posting analytics in our labor market research tool, Analyst. The feedback from clients has been excellent.
In evaluating job postings, Georgetown University, Jobs for the Future, The Conference Board, and others organizations have concluded that JPA is best used in conjunction with traditional labor market information.
But how can local practitioners use these two datasets together? That’s the question we wanted to answer in our JPA and LMI guidebook.
Here are the key points we make in the report. As we hit on a second ago, JPA and traditional (or structural) LMI are complementary. They work well together.
And each answer questions the other can’t. But data ultimately is just a tool.
As with any tool, it is important to understand what particular types of data are good for—and what they are not so good for.
We’ll get into that in a minute, but we want to emphasize the key link between job postings and traditional LMI is data on actual hiring by occupation. Hiring data is what provides the necessary context for postings.
To go deeper on LMI and JPA, I’m going to turn it over to Matt.
More and more they’re being aggregated and sorted by third parties, who try to make analytical data out of the info found in the ads.
Here’s how to think about how those two work together: Data is like a map. It’s an abstraction of reality. It’s good for telling you certain things, and worthless for telling you other things. In this case, LMI helps us get a great structure in place like this map of the entire US. This map is good for showing where the borders are, the shape of the country, and how big Texas is compared to Michigan. It’s bad for telling us what the streets look like in Seattle.
This map is good for knowing where the states are in relation to one another. It’s great at telling us how big Texas is compared to Michigan, but it’s worthless for telling us how to drive from Portland to Seattle.
[Can I get some major cities (Seattle, LA, Chicago, NY, Houston, Atlanta), state borders, and the LMI text under the map? Simple mountains showing the rough locations of the Rockies and Appalachian ranges]
This is a map of Washington state. It’s good for showing us the major cities in the area and the major highways connecting them. We can also see where there are national forests, mountain ranges, and rivers. But this map doesn’t tell me what the houses look like in Seattle, or how long it might take to drive from Yakima to Phoenix. As the scale changes, you lose stuff on both ends, fine detail and big picture. But each map is still useful for certain things.
Finally, here’s a Google StreetView image. The level of detail in this is amazing! I can tell you what kind of tree is growing in the front yard, or if your driveway needs to be repaired. But at this level, we lose a lot of the context because we don’t know where we are in relation to other things. Google has got the right idea by giving you that context with an address (top left) and a minimap (bottom left). The point is that none of these maps are better or worse than the others, they just specialize in different things.
Google streetview is REALLY detailed and shows us what a street in Seattle looks like. But that map isn’t good for other things, like knowing where the closest coffee shop is, or how long it will take to get to Portland. And as cool as Street View is, it’s not available everywhere. Job postings are a lot like this. They can provide astounding detail, but it’s often not clear where you are in the big picture. And they’re not available everywhere for every job.
Different maps for different purposes. Different data for different questions.
The point is that you wouldn’t say one of these maps is better than other one. They’re useful for different things, and we think LMI and postings are both important enough to be used together.
This map is good for knowing where the states are in relation to one another. It’s great at telling us how big Texas is compared to Michigan, but it’s worthless for telling us how to drive from Portland to Seattle.
(Jump up and down on the fact that postings are voluntary, not standardized, and not regulated, ie anyone can post as many jobs for whatever they want, for as long as they want. LMI is grounded in reality, and waits for jobs to happen before they’re counted)
[Break this into 4-5 slides, adding bullet points with each slide]
Here’s another way to think about it: Job postings primarily measure intentions, and only those of employers who advertise for jobs online; traditional LMI measures actions and tracks virtually all payroll employment. In other words, postings show what some employers are intending to do, and LMI shows what all companies have already done.
Here’s another way to think about it: Job postings primarily measure intentions, and only those of employers who advertise for jobs online; traditional LMI measures actions and tracks virtually all payroll employment. In other words, postings show what some employers are intending to do, and LMI shows what all companies have already done.
This has several ramifications:
• Postings aren’t simply a faster version of traditional LMI, though they can be a leading indicator of future job growth in some sectors (particularly at the national or state level).
• Postings data by itself can’t give a complete picture of the workforce. Without context, it’s hard to distinguish between substantive trends and noise.
This has several ramifications:
• Postings aren’t simply a faster version of traditional LMI, though they can be a leading indicator of future job growth in some sectors (particularly at the national or state level).
• Postings data by itself can’t give a complete picture of the workforce. Without context, it’s hard to distinguish between substantive trends and noise.
That said, traditional LMI doesn’t have all the answers, either. We’ll get to its weaknesses in a moment. But first, let’s look at the strengths.
No. 1, the data collection process from government agencies is measured and comprehensive. Perhaps the best example is the QCEW program from the BLS. This program is the gold standard for detailed industry data and is based on the administrative records of more than nine million businesses, covering 97% of payroll jobs (those in state or federal unemployment insurance programs).
Second, with organized coding systems, LMI lets you analyze historic data and projections for specific industries or occupations.
LMI provides granular data on industry and occupation job counts, business establishments, wages and salaries, hires and separations, the age, gender, and ethnicity of the workforce, demographics, and much more—down to the county and ZIP code.
LMI offers accurate, reliable, and comprehensive insight for long-term strategic planning and investment.
Collecting the raw data necessary to provide a complete picture of a local economy is hugely time-consuming since most LMI is housed on numerous state and federal sites—all with multiple data programs on separate databases. (Which is why EMSI combines 90+ sources and offers comprehensive data in a single package.)
Some sources offer detailed industry detail but limited geographic detail, some are survey-based and some are not, some are published annually and others quarterly, and so on. Analysts must laboriously stitch together multiple sources with varying scope and detail in order to arrive at a full picture.
Due to government privacy policies, public data sources cannot publish detailed local data that could possibly be tied to one or two specific business establishments. Government statistical agencies must suppress these elements, usually along with multiple other pieces of data that could be used to fill in or calculate the undisclosed data. This is a necessary limitation of public data that frustrates intensive labor market research. (Which is why EMSI uses mathematical algorithms to unsuppress local data, a process that state labor market agencies that subscribe to our data tools have regularly vouched for.)
Because of rigorous collection processes, most traditional datasets come with a significant time lag—anywhere from a few months to a few years.
Most of the information collected from public sources is available only in various plain text, database, or spreadsheet formats. On top of that, the coding systems (NAICS for industries and SOC for occupations) can leave analysts and researchers wishing they had more detailed (or different) job titles or industry codes.
Lastly, LMI can be hard to connect to the actual businesses on the ground. You might find that the pharmaceutical industry is booming, but the data itself doesn’t tell you WHO that is. One industry often represents multiple businesses. That confusion is part of the government’s commitment to keeping company-specific data private.
These aren’t just minor nitpicks. It would be great for colleges to know what skills employers are asking for when they design curriculum. It would also be great for workforce boards to know the detailed jobs that look promising. And everyone would love to be able to better connect the reliability of LMI to the actual businesses doing the hiring. Which brings us to job postings …
To start with the strengths, job postings can serve as a leading indicator of future job growth, particularly in large labor markets or broad employment sectors, and can help identify emerging trends and skills in the workplace.
Postings in aggregate form show the most sought-after skills and credentials from businesses that advertise for jobs online.
Postings have virtually no time lag (hence the term “real-time LMI”), so they can give insight into employers’ immediate skill needs.
Postings are a good way to see how hard employers are hunting for specific talent.
It’s helpful for practitioners and researchers to know which employers are currently hiring, or at least thinking of hiring.
Moving onto to the limitations of job postings … First, since they’re online advertisements, postings predominantly come from businesses in industries that post jobs on the Internet (think technology-intensive and professional services industries and not construction or manufacturing).
One posting may represent 30 jobs, or 30 postings may represent one job. Some companies post jobs to stockpile résumés; others intend to hire multiple people based on one advertisement.
Postings tend to be skewed toward high-skill or high-wage occupations with advanced educational requirements (IT jobs, managers, etc.), partly because employers are more willing to spend advertising dollars to find high-skill workers.
Urban areas have a higher concentration of postings than rural areas, often because of the differences in industry mix.
Not all occupations are represented equally in online ads (and some aren’t represented at all), so job postings aren’t an accurate measure of the hiring demand for every field, particularly when looked at in isolation. Postings give insight into employers’ effort and intentions, not their actual hiring patterns.
Job postings represent only one of many ways businesses recruit. Firms are increasingly using social networks, employee referrals, talent networks, career fairs, and other avenues to find talent.
Since job postings are voluntary and unregulated, there’s no way to know if a job posted turns into a job hired. But we can get very close to answering that question by comparing the number of hires to the number of postings. In other words, the link between traditional LMI and job postings is data on actual hires by occupation.
The hires data that EMSI presents alongside postings for every occupation forms the backbone of the application section of our report. We’re not going to get into specific applications for different groups, but we encourage you to check out second chapter in the report and dig into examples we show for colleges and universities, workforce development, economic development, and the private sector.
Also in Chapter 2, we outlined potential workflows that bring LMI and JPA together. We’re going to walk through one of those workflows here.
In this example, we start with industry data—the backbone of all economic analysis–and finish with postings.
Activities related to credit intermediation, an industry that includes mortgage brokers and financial processing clearinghouses, accounts for about 22,000 jobs in the Phoenix metro area with a concentration six times greater than the national average
We can also see which occupations make up the industry and can even zero in on personal financial advisors to discover that they make a little over $28 per hour and have grown 32% since 2010 in Phoenix—nearly three times the national growth rate. All of this history, context, and occupational detail are possible thanks to LMI.
Moving onto job postings: We see that Edward Jones is posting a lot of jobs in Phoenix. By de-duplicating repeat postings, we can see that Edward Jones seems to be putting serious effort into finding those workers, with a posting intensity of 7:1, or 7 total postings for every 1 unique (or de-duplicated) posting.
Activities related to credit intermediation, an industry that includes mortgage brokers and financial processing clearinghouses, accounts for about 22,000 jobs in the Phoenix metro area with a concentration six times greater than the national average
Activities related to credit intermediation, an industry that includes mortgage brokers and financial processing clearinghouses, accounts for about 22,000 jobs in the Phoenix metro area with a concentration six times greater than the national average
As we said earlier, postings data comes from job advertisements posted online. This means that some occupations are going to be underrepresented and others overrepresented in postings, based on how businesses recruit talent. We can see this clearly when comparing postings to hires
As we said earlier, postings data comes from job advertisements posted online. This means that some occupations are going to be underrepresented and others overrepresented in postings, based on how businesses recruit talent. We can see this clearly when comparing postings to hires
As we said earlier, postings data comes from job advertisements posted online. This means that some occupations are going to be underrepresented and others overrepresented in postings, based on how businesses recruit talent. We can see this clearly when comparing postings to hires
In the report, we talk about skilled trades occupations nationally. And when you go to do some analysis on YOUR region with the jobs and people you’re working with, the actual number of job postings compared to the number of hirings that take place differs wildly. Here we can see average monthly postings fro mechanical engineers in Seattle is twice average monthly hires.
[Pop out Mechanical Engineers, Avg Postings, Avg Hires. Break into two slides, one for engineers next one for welders]
And staying in Seattle, we see the opposite for welders – 176 average monthly hires vs. 40 average monthly postings.
[Pop out Mechanical Engineers, Avg Postings, Avg Hires. Break into two slides, one for engineers next one for welders]
Here are a couple examples in the report. From November 2013 to October 2014, Houston was the top metro area in the nation for welder job postings. Employers posted an average of 200 unique job ads for welders, more than double the number in Chicago or Los Angeles. Yet all that posting activity in Houston pales in comparison to the number of welding hires—an average of more than 1,000 per month. So for every posting, companies hired five welders.
Unique monthly postings for web developers in the New York City metro outnumbered average monthly hires 8 to 1 from November 2013 to October 2014. This could signal employers can’t find the developers they need, but also consider that there were nearly five times as many postings as hires nationwide for web developers over this time.
Employers in New York City haven’t posted as intensely for web developers as the metro average for all occupations; if they had, it might suggest they’re facing a talent shortage. But as it is, New York City appears to be oversaturated with postings for developers
But it’s still case by case, job by job, region by region.
Two ramifications: we really don’t get anywhere near the true demand for these jobs by looking at postings, AND because there are less postings out there, we get less detail like requested skills and companies currently posting.
The good news is that by showing you postings alongside hires, you always know the context for job postings in a particular region.
EMSI strives to give our clients a current, complete picture of labor market activity by region. We provide foundational labor market information that represents nine million business establishments
We supplement this foundational LMI with job posting analytics from over 15,000 websites – a number that’s growing.
EMSI strives to give our clients a current, complete picture of labor market activity for any region. We provide foundational labor market information that represents nine million business establishments. By combining 90+ sources, we’re able remove suppressions (or non-disclosed data) typically found in detailed local industry data and add in self-employed workers and other proprietors.
And we contextualize JPA with data on hires. This data comes from EMSI’s proprietary database and is based on the U.S. Census Bureau’s Quarterly Workforce Indicators. EMSI combines QWI with other datasets to produce up-to-date hires and separations data for all occupations using staffing patterns that show the percentage occupational makeup of jobs within each industry.