mkmahajan
commented
11 months ago I've been chatting with @robbieorvis and @minshudeng about the need for this type of fix for some custom China modeling that is ongoing. I believe they need this fix on a custom side branch soon (but that we could then officially add it later to the official development queue).
I'd propose the structural edits below for the custom China branch. This will actually be separate from the material efficiency lever, because it is representing a different kind of policy. The intent is to represent actions China would take to shift its economy from more manufacturing-based to more service-based. So rather than reducing demand for goods, we'd be encouraging the growth of certain service industries while reducing output from manufacturing industries. This type of policy makes sense for an economy like China's where the government has more control over industrial production but would likely not be as relevant in many EPS regions. Based on input from our partners, the best way to represent this is a shifting of Value Added from a certain set of industries to another set of industries.
- We’d set up a new policy lever that allows you to set a percentage decrease in Value Added for different ISIC codes.
- We'd then sum up the total amount of Value Added to shift. We could set up an input data file where you can select which ISIC codes you want to shift the employees into, then allocate the total amount of Value Added (we might want to set this up like the TDM lever where we select the percentage of the total that should go to each recipient ISIC code).
- We'd then calculate the corresponding percent change in domestic output for affected industries (both the decreases and increases) using 'Within Industry Value Added per Unit Output.' The percent change in output would be factored into the 'Percent Change in Nonfuel Production due to Policies' variable, alongside the changes from the Material Efficiency lever.
- Because we're calculating changes in output, we need to determine how that translates into changes into imports, exports, and/or domestic consumption. My understanding is that we want to direct this policy towards changes in imports and/or exports. (If we wanted to change domestic consumption, we'd use the Material Efficiency lever). If we are decreasing output from manufacturing industries, it might make the most sense to subtract the changes from exports. If we zero out exports, then perhaps we want to move on to increasing imports. That means we need to calculate the change in exports first. In the export calculations below, we would subtract out the max of either the change in manufacturing output or the total remaining exports from 'Change in Exports of Nonenergy Products.' Note that this section only applies to ISIC codes that correspond to industry categories, as it's taking place on the Industry - Main page.
- For each industry category, we'd also need to determine whether there is any remainder to apply to an increase in imports. The policy effect we're calculating is distinct from price-driven import substitution, so I think we'd want to handle this downstream from the Armington elasticity calculations. We could add this new effect into the 'Change in Nonenergy Industry Product Imports as Share of Domestic Production' variable circled below.
- We'd also need to add this percent change in imports into our domestic content calculations. In the two places we use the variable 'Percent of Purchases of Domestically Produced Goods Shifted to Purchases of Imports' below, I think we'd need to also apply the newly calculated percent change in imports due to our new policy lever as a separate adjustment.
-
We'd need to add in the changes in output to the Direct Policy Effects on Industrial Output calculations below. Since we're working with changes in domestic output, we could add it directly to the variable 'Domestic Nonenergy Industry Expenses Passed Through to Buyers by Producer ISIC Code.' There's no need to filter it through the Domestic Content Share of Consumption by ISIC Code (however, we did still need to account for this policy in the domestic content calculations since that is used again in the induced impacts calculations).
-
Finally, I think we need to account for increased tax revenue from the non-industry category ISIC codes. I see that we are capturing the increases from indirect and induced changes in these ISIC codes below. But we need to add in the direct change for non-industry category ISIC codes for just this new policy here. The direct changes from other policies are already accounted for in the sector specific changes in nonenergy industry revenue.
I think the above steps account for a shifting economy in all of the following: energy use/emissions, direct IO impacts, indirect IO impacts, induced impacts, import/export accounting, and tax revenue. However, it's definitely possible I'm missing something or not seeing potential complications from the above approach. One thing I'm not positive about is how this type of policy should interact with our passthrough effects calculations. It would be great to get @jrissman's thoughts on the proposed steps above before I start any custom edits.
jrissman
robbieorvis
The current design of the demand reduction does not work well with the IO model and also fails to capture important indirect effects.
Currently, the demand reduction lever feeds directly into the Percent Change in Nonfuel Production to Policies and flows through to the rest of the model (like the IO structure) from there.
The first issue with this approach is that it fails to capture indirect effects that result from improved material reuse and recycling. I believe that the correct way to do this would be to adjust the indirect multipliers by the percentage reduction in material demand, which would carry it through the IO model. It may be the case that rather than applying this at all in the industry sector, we ought to be applying it in the IO model, like how we handle the buy-in region policy. It seems fairly consistent with that methodology.
The second issue pertains to whether or not there would be a reallocation of the labor force (and possibly GDP and Output) that would result from this policy, both from lower material costs for produced goods but also from re-employment of workers no longer needed. For example, if the input demand for iron and steel were to drop by 25%, it is likely that some of the existing domestic supply for domestic demand would be exported, that there would be some increase in domestic supply, that the workers would go to work in other sectors, and possibly that domestically produced goods that now require less iron and steel might be slightly more economic and demand might increase. This is kind of like the reverse of passing through energy policy costs to the industry sector... in this case we would passing through lower production costs with a resulting increase in demand and shift away from imports. This is a lot of factors to consider, clearly. However, failing to capture some of these effects has negative outcomes in the modeling and in framing results.
Most notably, introducing a material demand reduction policy can have a very negative impact on jobs, because it directly eliminates jobs and does not find anywhere else for them, either by reallocating them or by trying to estimate demand changes in other industries or in exports. This is not necessarily true and also not really a helpful narrative. One idea would be to design the material demand reduction policy lever like the TDM policy lever, where there are user defined pathways to reallocate output. This feels a little forced, but I struggle to find another option, unless we can somehow estimate the change in producer prices elsewhere in knock on demand effects.
The indirect impact issue seems more tractable. Ultimately, I think we need to tackle at least these two pieces in some way, in order to improve the way this policy is handled in the model and in order not to present results that are unnecessarily negative when looking at this policy. Since we already have the indirect tables though, it seems possible to come up with an elegant solution. For example, we know for each dollar of output from fabricated metal products how many dollars of input are required from iron and steel. Let's say the coefficient is 0.2. A 25% percent improvement would mean this would drop to 0.2 to 0.15, and we know the total sum of the inputs (let's say it's 1, we could figure out the change in cost to produce one unit of a commodity given the improvement in the input requirements for iron and steel). In this case that means a 25% improvement in demand for iron and steel would drop the producer prices for fabricated metal products by 5% (this assumes that the DLIM value is capturing the entire cost to produce and not just the non-value added component, but since we have output and value added, we can recalculate the change as needed if necessary).