Aggregating IMPLAN sectors to NAICS sectors
I am running an analysis of reduced output effects on employment in Franklin County, PA. Because I previously have provided QWI employment figures to the client, I aggregated the IMPLAN sectors into NAICS sectors. I then went to customize my model and change all employment to QWI. However, there were two red flags.
First, total 2009 employment in IMPLAN was 70,239 while QWI has 51,586 for 2009 average. How can this discrepancy be so large?
Second, some of the aggregated IMPLAN employment numbers are similar to QWI, while others are very different. For example:
Real Estate and Rental: IMPLAN aggregated NAICS = 1207.8; QWI 2009 Avg = 258
Other Services (except Public Admin): IMPLAN = 5494; QWI = 1791
My concern is with the differences in output per worker in each of the aggregated IMPLAN sectors. Can you shed some light on these discrepancies? In a bit of a time crunch so I appreciate your timely response.
-Mark
-
I am going to assume that QWI (?) is similar to BLS QCEW (or ES202) data. QCEW data does not include self-employment. There are some minor QCEW non-coverages as well (railroad, religious organizations, elected officials, military, some agricultural workers). IMPLAN data will be more similar to BEA REIS data for comparison. -
OK thanks. I am going to dig deeper into the data sources. FYI - QWI (Quarterly Workforce Indicators): http://lehd.did.census.gov/led/datatools/qwi-online.html -
OK, assuming the main reason for the data discrepancy is due to the exclusion of self-employed and other workers, are there any concerns I should have by customizing my Franklin County model to reflect the BLS (QWI) employment numbers, and then modeling reduced output through industry changes? While for most industries IMPLAN's numbers are higher, in some cases employment figures from the BLS are higher than IMPLAN. I'm worried that output per worker will not be accurately reflected. -
If the only information you have is employment, you will want to edit all the industry data elements proportionately so the relationships do not change. That is, change the relationships for the data you have and not the ones you do not have data for. The existing industry relationships; however, are based on standard I-O employment definitions which you will be replacing with covered wage and salary employment only, which would theoretically, shift the definitions. The original employment levels have no effect on the indirect and induced effects as it is labor income that drives the induced effects and not employment. The trade flows are based on the data currently in the model. If you apply wholesale reduction to to industry employment/income/demand you will reduce the demand for commodities without proportionately reducing supply used locally, you will artificially increase the multipliers. Remember that the multipliers are based on the equation (I-A)^-1. I: is the identity matrix and is fixed. A: is the coefficient matrix and is based on per dollar of spending. The output of the region is not part of the equation. Industry output is used in the trade flow model, however, to determine the size of the original A coefficients. One other definitional difference between IMPLAN employment and QCEW data is that IMPLAN employment is adjusted to account for I-O redefinitions. The most important of which is to pull investment and new construction activity (such as putting up new utility lines and computer application development) out of industry sectors and into the appropriate investment/new construction sector. This particularly impacts hotels and lodging which has restaurant and gaming activity pulled out and put into the appropriate sectors. My suggestion is to keep the current formulation for impact analysis, and use their numbers to report historical levels of employment and wage and salary income.
Please sign in to leave a comment.
Comments
4 comments