mkmahajan
commented
3 years ago Done in commit 1e3fab. I reviewed all files that allocate spending to ISIC codes and determined only the two files in the power sector (SoESCaOMCbIC and SoTCCbIC) needed to be updated.
When we updated to 3.2 we left in place many of the allocations of technology spending to ISIC codes, such as spending in the power sector on capital costs. However, some of the industries that were serving as catchalls, such as other manufacturing, are now much more granular and not necessarily related to the way in which spending happened previously. As an example, the other manufacturing industry, based on BEA data, is now mostly confined to furniture manufacturing, which obviously is not a category that would see spending from installing power plants.
The source data categories for spending that we relied on to estimate the shares are much more aggregated than what's in the EPS now. For example, this document from NREL shows what industry categories are included in "other manufacturing" in JEDI, our source document: https://www.nrel.gov/analysis/jedi/assets/pdfs/jedi-aggregation-scheme-base.pdf. Comparing this to what's in the EPS category for "other manufacturing" we can see there is a clear mismatch and that spending in "other manufacturing" in JEDI is likely to be in a distinctly broken out industry in the EPS, probably fabricated metal products.
We should revisit all the variables that allocate spending to ISIC codes and make sure they are properly mapped given all the new industry categories. This is somewhat urgent as it can have a significant impact on model findings related to GDP growth and labor impacts.