The EPS uses input-output tables and related tables (jobs, value added, etc.) published by the OECD to power its input-output model for the U.S., and the OECD has these tables available for over 60 countries and regions. The tables are a single-year historical snapshot. This introduces some inaccuracy, as in the BAU case, we would expect some ISIC codes to grow in importance and others to shrink. Therefore, some of the I/O model's assignments of indirect and induced impacts (jobs, value added, employee compensation) may be less accurate as we approach 2050. For example, today, 100 units of increased output by the machinery-making industry might require (cause) 5 units of increased output by the fossil fuel extraction industries, but maybe in 2050, 100 units of increased output by the machinery-making industry might cause zero units of increased output from the fossil fuel extraction industries.
Any or all of the pink (time-invariant) input variables on the "Input-Output Model" sheet would be more accurate if they were replaced with future time series versions, but the most important ones would be:
io-model/BObIC
io-model/BEbIC
io-model/BVAbIC
io-model/BECbIC
io-model/DLIM
These five variables must be done as a set, along with io-model/TLIM, because they will not be accurate if they do not use harmonized assumptions.
If this is done, the structure under the "Time-Series Totals" header might be able to be simplified. BUT keep in mind that the approach must work for all countries/regions, so if future time-series IO data is only available for some regions and can't be generalized for all EPS geographies, it likely is necessary to preserve the existing structure under the "Time-Series Totals" header to accommodate those other regions. If the existing structure is preserved, io-model/BPEaCP and io-model/BGDP should use values matching the time-series totals for regions that do have time-series data in the variables listed above.
Time-series projections of these IO data may not be available, so it may be impossible to address this issue, but we log it here in case it proves tractable at some point.
The EPS uses input-output tables and related tables (jobs, value added, etc.) published by the OECD to power its input-output model for the U.S., and the OECD has these tables available for over 60 countries and regions. The tables are a single-year historical snapshot. This introduces some inaccuracy, as in the BAU case, we would expect some ISIC codes to grow in importance and others to shrink. Therefore, some of the I/O model's assignments of indirect and induced impacts (jobs, value added, employee compensation) may be less accurate as we approach 2050. For example, today, 100 units of increased output by the machinery-making industry might require (cause) 5 units of increased output by the fossil fuel extraction industries, but maybe in 2050, 100 units of increased output by the machinery-making industry might cause zero units of increased output from the fossil fuel extraction industries.
Any or all of the pink (time-invariant) input variables on the "Input-Output Model" sheet would be more accurate if they were replaced with future time series versions, but the most important ones would be:
io-model/BObICio-model/BEbICio-model/BVAbICio-model/BECbICio-model/DLIMThese five variables must be done as a set, along with
io-model/TLIM, because they will not be accurate if they do not use harmonized assumptions.If this is done, the structure under the "Time-Series Totals" header might be able to be simplified. BUT keep in mind that the approach must work for all countries/regions, so if future time-series IO data is only available for some regions and can't be generalized for all EPS geographies, it likely is necessary to preserve the existing structure under the "Time-Series Totals" header to accommodate those other regions. If the existing structure is preserved, io-model/BPEaCP and io-model/BGDP should use values matching the time-series totals for regions that do have time-series data in the variables listed above.
Time-series projections of these IO data may not be available, so it may be impossible to address this issue, but we log it here in case it proves tractable at some point.
Suggested by Jim Barrett in July 2020