Regionalization of National Staffing Patterns

Emsi develops occupation job counts data by taking Bureau of Labor Statistics’ Quarterly Census of Employment and Wages (QCEW) industry job counts and applying staffing patterns to transform industry jobs to occupation jobs. Click here to read more about the occupation jobs process.

Since OES only offers a national staffing pattern for each industry, Emsi must take the staffing patterns and “regionalize” them, allowing each region’s distinct industry-occupation mix to impact the staffing patterns, tailoring them to the realities of the region.

We begin with the national staffing patterns and regionalize to the state level using OES state-level occupation job counts and QCEW state-level industry job counts. After regionalizing to states, we repeat the same step one more time, regionalizing from the state level to the metropolitan/non-metropolitan areas inherent in OES data. For this step, OES metropolitan/non-metropolitan occupation job counts are used in conjunction with metropolitan/non-metropolitan industry job counts from QCEW (offered at the county level and easily summed to match OES at the MSA/non-metro level).

Walkthrough

We begin with three elements, the national OES staffing pattern, which shows what percent of each industry (column) is staffed by each occupation (row); the state-level occupation job counts from OES by occupation; and the state-level industry job counts from QCEW by industry.

State-level industry jobs from QCEW sum to 8410 (3000 + 4320 + 1090), and state-level occupation jobs from OES sum to 8110 (6500+760+850). We first adjust OES to match the QCEW total, since QCEW job counts are our gold standard. This is done by proportionally adjusting OES job counts to match QCEW. For the above case, QCEW is 4% higher than OES, so each OES SOC is bumped up by 4% (e.g. 6500 * 1.04). After this adjustment, the new OES state-level job counts sum to match QCEW state-level job counts:

We can now combine the state-level industry and occupation job counts with the initial national staffing pattern:

Like the first staffing pattern shown above, this matrix shows each column totaling to 100%, representing the occupation percent breakout for each industry. However, rows do not currently total to 100%. We apply a three-dimensional, hierarchical proportioning algorithm to the above matrix. The algorithm is a proprietary implementation of the proportional algorithms commonly found in literature. This algorithm adjusts the percentages across rows and down columns, repeatedly, until both rows and columns sum to 100%. The result is a percentage matrix (a staffing pattern) that fits the regional industry and occupation job counts.
At the end of this process, we have regionalized staffing patterns for each state in the nation.

The last step is to further regionalize each state’s staffing pattern to the metropolitan and non-metropolitan areas within it. This is done by repeating the regionalization process above; however, this time QCEW industry job counts and OES occupation job counts for the metro/non-metro areas within the state are used to create the initial matrix. The same hierarchical proportioning algorithm is used to create final staffing patterns for the metro/non-metro areas. County-level QCEW industry job counts are then sent through the appropriate regionalized staffing pattern, and the result is occupation job counts by county.