QWI Separations Methodology

Emsi publishes occupation separation counts using a combination of Emsi jobs data, separation rates from the Bureau of Labor Statistics (BLS), and industry-based separations data from the Census Bureau’s Quarterly Workforce Indicators (QWI) program.

Sources

• Census’ Quarterly Workforce Indicators (QWI): original source of separations data. Data is reported by businesses and therefore by industry. Emsi’s separations process will transform these industry figures into occupation figures.
• Emsi regional industry by occupation decline
• BLS separation rates. We use the separations rate (not separation counts), and we use the total rate (sum of Labor Force Exists rate and Occupational Transfers rate).

Methodology

1. Use Emsi occupation decline together with BLS occupation separation rates to model, for each industry-area combination, the pattern of occupational separations.
• Calculate Emsi decline between each consecutive year pair (e.g. 2016 and 2017) for each county by industry by occupation node by subtracting later year employment from earlier year employment. If there is growth rather than decline (e.g. 150 jobs in 2016 and 210 jobs in 2017), decline = 0.
• Calculate the number of BLS-based separations for each county by industry by occupation node. This is done by multiplying employment by the BLS separation rate for that particular occupation.
• Sum Emsi decline and BLS-based separations for each county by industry by occupation node. This results in model separations.
• Normalize model separations along the occupation dimension so that the sum of all model separations within a particular county- industry node equals 1.0. In lay terms, this means that the model separations figures for each county-industry node are turned into percentages. In other words, the result of this step would show that in Latah county, 35% of the model separations in General Medical and Surgical Hospitals are for CNAs, 20% of the model separations are for Orderlies, 5% of the model separations are for Security Guards, and so on. We now have normalized model occupation separations we can use to break out our chosen source of separations data, QWI industry separations.
2. Apply the normalized model separation rates derived at the end of Step 1 to QWI industry separations.
• Multiply the normalized model separations by the corresponding industry separations from detailed Emsi industry data (somewhat processed version of QWI industry separations). This yields the estimated number of separations made in each county by industry by occupation node, using the QWI Separations number as the basis for the overall industry separations count, and using the model separations breakout from Step 1 to break the QWI industry separations figure out into occupational separations figures.
• Since the goal is simply occupation separations not broken out by industry, all separations are summed by occupation. For instance, occupation separations for Loggers are summed from all industries. This yields a final Emsi separations figure for Loggers (and each other occupation) in each county, for each year.

Example

To illustrate how the process works, we can walk through an example using a fake county in which there are two industries and three occupations.

Calculate job decline between years. If change is negative, decline = change. If change is positive or 0, decline = 0:

Industry	Occupation	2016 Jobs	2017 Jobs	Decline
IA	O1	50	25	25
IA	O2	350	300	50
IA	O3	100	110	0
IB	O1	200	190	10
IB	O2	300	300	0
IB	O3	80	70	10

Calculate the number of separations for each industry-occupation combination by multiplying the BLS’s separation rate for the occupation by the occupation’s employment:

Industry	Occupation	2016 Jobs	BLS Separation Rate	BLS-Based Separations
IA	O1	50	0.08	4
IA	O2	350	0.06	21
IA	O3	100	0.11	11
IB	O1	200	0.08	16
IB	O2	300	0.06	18
IB	O3	80	0.11	9

Sum Emsi decline and BLS-based separations for each industry-occupation combination. This results in model separations for each county-industry-occupation combination:

Industry	Occupation	Emsi Decline	BLS-Based Separations	Model Separations
IA	O1	25	4	29
IA	O2	50	21	71
IA	O3	0	11	11
IB	O1	10	16	26
IB	O2	0	18	18
IB	O3	10	9	19

Turn occupational Model Separations into a percentage of separations for each industry:

Industry	Occupation	Model Separations	Total Industry Model Separations	Percent of Model Separations
IA	O1	29	111	26%
IA	O2	71	111	64%
IA	O3	11	111	10%
IB	O1	26	63	41%
IB	O2	18	63	29%
IB	O3	19	63	30%

Multiply QWI industry separations figures by the percent model separations figures, breaking each industry’s separations out into separations by occupation:

Industry	Occupation	QWI Separations	Percent of Model Separations	Occupation Separations
IA	O1	75	26%	20
IA	O2	75	64%	48
IA	O3	75	10%	7
IB	O1	175	41%	72
IB	O2	175	29%	50
IB	O3	175	30%	52

Since the goal is occupation separations without regard to industry, separations for each occupation across all industries in the county are summed (e.g. all separations for O1, from all industries, are summed):

Occupation	Final Emsi Occupation Separations
O1	92
O2	98
O3	59

The end result is separations by occupation, particular to each area in the US.

Why are Emsi Separations so High?

Emsi’s separation counts are significantly higher than separation counts based solely on the BLS’s Separation Rate. Separation counts could be calculated by simply taking the national replacement rate as published by the BLS for each occupation, and multiplying it by the job counts for that occupation.

BLS Separations only take into account “Occupational Transfers” and “Labor Force Exits.” Occupational transfers are people who leave the job category (SOC) they had been working in and begin a job in a new job category. An example of this would be a Registered Nurse (SOC 29-1141) becoming a Pharmacy Technician (SOC 29-2052). Labor Force Exits are people who leave the labor force entirely, as in retirement.

Excluded from BLS separation counts are job transfers within a job category, as in a Registered Nurse leaving a job at a hospital and instead working as an RN at a local physician’s office. This is neither an Occupational Transfer nor a Labor Force Exit, and is therefore not counted in the BLS’s separation rates. These types of job switches within a SOC are very prevalent, and so a Separations metric that includes them will be by definition much higher than a Separations metric that excludes them.

The Separations numbers captured by the Census’s Quarterly Workforce Indicators (QWI) dataset include separations due to labor force exits, occupational transfers, and normal job-switching (churn). It is a more robust measure of separations, and since its goal in Emsi software is to be a part of a churn/turnover metric, it is necessary to use this more robust measure of separations, which will necessarily be much higher than a separation metric calculated using only the BLS Separation rate.