Indirect Multiplier Discrepancy
Greetings IMPLAN Staff, I am in the process of comparing the agricultural production and processing sectors in the 2011 Utah data set to the 2014 Utah data set. The discrepancy is in the 2011 indirect jobs created from the ag production and processing sectors, roughly 9,000 more indirect jobs were created than in 2014. Currently, I am trying to figure out why there was such a big bump in the indirect jobs created. Looking at the multipliers a little closer there are a few ag processing sectors that have significantly different multipliers. I have attached a file with highlighted sectors that I am questioning. Why was there such a big boost in the 2011 data set and then back down in 2014? Sectors that I am questioning the multipliers: 2011 2014 55-57 84-87 59-60 89-92 Any information or help would be appreciated! Thanks!!
Hello Karli, There have been two significant underlying changes to the agricultural data between the 2011 and 2014 data sets, which are outlined in our [url=http://support.implan.com/index.php?option=com_content&view=article&id=399]release notes[/url] and described in more detail below: In 2014 this included: • Incorporating BEA data into the farm sectors: We added a control of the sum of our state-level estimates to BEA’s national estimates for the value of crop sales. ERS, which is BEA’s primary initial source of cash receipts by commodity, estimates include adjustments for Commodity Credit Corporation (CCC) loans, and do not account for home consumption or inventory, all of which need to be addressed when estimating output based on cash receipts. We obtain estimates value of production for certain agricultural products from NASS; these values don’t require adjustments for CCC or inventory. BEA adds the value of intra-state livestock sales to its estimates, which should be included in output, so this tends to increase our estimates. We do not control individual state values to BEA values since we generally can obtain and process more current ERS and NASS data before they are incorporated into BEA’s data. Although BEA’s “other crops” category includes sugar cane, BEA does not produce any detailed estimate of sugarcane output, which is well-measured by NASS and ERS, so we do not apply the control to that IMPLAN sector. The Department of Agriculture’s National Agricultural Statistics Service, Economic Research Service, and the Census of Agriculture continue to be our primary data sources for estimating state- and county-level agricultural output. • In preparation for being able to report wage and salary employment separately from proprietor employment, we made improvements to the way we estimate Proprietor Employment and Income. In 2013 this included: • A new Census of Agriculture: The Census of Agriculture is released every 5 years; thus, there may be some sizeable changes in some farm sectors in some regions. The 2012 Census of Agriculture was released in 2014 and is incorporated into the 2013 IMPLAN data set. Census of Agriculture data are used to disclose data missing from USDA Economic Research Service (ERS) and National Agricultural Statistics Service (NASS) data sources. • Incorporation of NASS data: NASS Data for Agriculture Output: We use NASS sales and production data as a supplement to ERS sales data, where available, since the ERS sales data may omit inventory changes, home consumption, and production used in the production process of another agricultural good (e.g. hay used to feed animals). Large differences between the datasets tend to occur with products that are likely to be added or removed from inventory (grains) or consumed on a farm (hay, meat). • A new BEA Input-Output Benchmark: The BEA's Benchmark I-O tables are also released every 5 years. These tables set the course for IMPLAN's sectoring scheme, production functions, by-product coefficients, and market share coefficients. The 2007 Benchmark was released in 2014 and incorporated into the 2013 IMPLAN data set. This will cause changes across many sectors and regions. The benchmark is our only source of value-added data by farm commodity; other sources report value added data only for an aggregated “farm” sector. We use the benchmark value-added data, in conjunction with national employment data by farm sector from QCEW (scaled up to account for under-coverage) to estimate national ratios for wage & salary employment per dollar of output and employee compensation per dollar of output, and national farm count data by farm sector as a proxy for proprietor employment per dollar of output. Since these relationships shift only once every five years, comparing data from 2010, which uses the 2002 benchmark, and 2014, which uses the 2007 benchmark, can show considerable shifts. When large shifts in value added, output, and employment relationships occur, large shift in direct effects will occur. Regards, IMPLAN Staff
For further clarification, to avoid double counting the jobs generated by agriculture production, in the dairy and meat processing sectors we set the rpc's for ag production sectors to 0. So differences in how the ag production data was changed should not have affected the ag processing indirect jobs. When we looked at a few years both before and after 2011, we see 2011 indirect jobs for dairy and meat processing are significantly higher (more than double in some cases). For example for animal except poultry slaughtering the indirect employment multiplier was 9.4 in 2011, but .6 in 2014. I need to explain in a report why the indirect jobs are lower by 10,000 overall for ag processing indirect jobs. If I could explain this difference in the over 9 jobs in 2011 and only .6 in 2014 it would go a long ways towards explaining the overall difference. I realize that 2014 had a couple more sectors for animal processing activities than 2011. I know that IMPLAN is not a time series data set. However, this big of a change needs to have a reason. I included an Excel file with the relevant employment multipliers and output from both 2011 and 2014 if it is helpful.
Hello Ruby, If we zero out RPCs for agriculture production sectors 11 and 14, of which 89-92 make significant purchases – animal processing plants purchase livestock – then there is a large reduction in the indirect employment effects. Furthermore, when we do perform the RPC adjustment as described to a 2011 Utah file, we see a similarly significant reduction in the indirect employment effects. However, if you intend to measure the contribution of animal product processing industries, you need to zero-out RSCs for the byproducts of the industries whose contribution is being estimated. If you are zeroing out RPCs for industries like 11 and 14, this would be consistent with estimating the contribution of 11 and 14, but not of industry 89, for example. Could you describe your intended analysis in more detail? In addition, would you be willing to email copies of both of your models? If so, you can email them to email@example.com. With a clearer picture of both your study and your model, we will be able to provide a more complete response. Regards, IMPLAN Staff
Please sign in to leave a comment.