Update IO structure for on-road vehicles to account for changing input coefficients as vehicle sales types change

[robbieorvis]

Currently, the EPS allocates all changes in spending for on-road vehicles to the on-road vehicle manufacturing sector. However, the input coefficients for the sector are based on ICE sales (given historical sales shares), but the composition of input coefficients for BEVs (or FCEV for that matter) are very different than those of ICE, and many regions are evaluating targeted supply chain and labor practices to support this shift.

To support this research and more accurately capture shifting supply chains and labor dynamics, we should modify the IO model to account for this shift. There are (at least) two ways we could do this:

1) Allocate changes in spending by vehicle type to the indirect industries that are used to supply parts for those vehicles. This is akin to how we handle power plants in the electricity sector, where changes in capital spending are allocated to the industries that provide inputs to produce things like wind turbines and solar panels based on proportions from the literature. For the transportation sector, we would have different coefficients for the different engine/vehicle types for onroad vehicles, and spending on vehicles would then be allocated to each of those indirect sectors. This would be functionally equivalent to how we handle power plants, though would leave in question how we handle the onroad vehicle manufacturing industry. We would need just a single international set of coefficients to do this, so it wouldn't increase international data requirements.

2) The second option is to weight the DLIM values by the types of vehicles that are sold based on coefficients for different types of vehicles, this would continue to flow all the cash flow changes from changes in onroad vehicle purchases through the same ISIC code it uses today, but would modify the DLIM values based on how the types of vehicles change. If we were to go this route, we might consider doing a similar type of aggregation for the power sector instead of keeping things broken out as they are now, to be consistent across the model.

These changes are important as regions look at moving to 100% sales of EVs and electrifying vehicle fleets by 2050, which will have important and large labor implications. We have been asked this question by several experts in the field already, so it would be great to build in this dynamic.

[jrissman]

Unlike power plants, we don't want to pass through the revenue to the component suppliers (like battery makers and steel makers for EVs) because this cuts out the automobile makers. We probably want the approach that shifts DLIM (and other variables if needed) over time, maybe based on EV penetration rates.

[robbieorvis]

Update here: I have some new data from IMPLAN for WRI that actually provides a better approach. We can handle this issue, like we do in the power sector where we split spending on technologies into different ISIC codes depending on the type of technology (solar, wind, gas, etc... in power and we can do something similar for vehicles). For example, the table below shows how IMPLAN approaches this:

The WRI report actually has ISIC code spending allocations for basically all the clean energy technologies we are interested in, and could be an excellent resource for improving our model IO spending by creating allocations for each of the major technologies in our model. I'd suggest we move in this direction in the future.

[jrissman]

This sounds like your first suggestion, assigning some of the money to the upstream suppliers of the vehicle manufacturers (and some of the money to the vehicle manufacturers themselves), rather than trying to shift DLIM values over time.

Note that we cannot use any data from IMPLAN. IMPLAN is commercial software and has a license protecting its data (see points 3.3. and 4 in particular). We have never run or used IMPLAN, so we have never signed any license agreement with IMPLAN. But even if we haven't signed a license agreement, we still have to be mindful of IP ownership laws. So I think we must continue to avoid using any IMPLAN data in our tool.

The screenshot you have above says those data originally come from an Argonne National Lab report. Anything produced Argonne National Lab is free for us to use. If you can find the original Argonne National Lab source, we can use that.

[robbieorvis]

Yeah, guess I had already thought of this approach!

Regarding the splits, yes, the data is from Argonne, but referenced in a WRI report that used IMPLAN. I would like to figure out which Argonne report it is from, but haven’t had luck so far. I may message Michael Wang if we can’t find it.

From: Jeff Rissman @.> Sent: Friday, June 3, 2022 12:53 PM To: Energy-Innovation/eps-us @.> Cc: Robbie Orvis @.>; Author @.> Subject: Re: [Energy-Innovation/eps-us] Update IO structure for on-road vehicles to account for changing input coefficients as vehicle sales types change (Issue #205)

This sounds like your first suggestion, assigning some of the money to the upstream suppliers of the vehicle manufacturers (and some of the money to the vehicle manufacturers themselves), rather than trying to shift DLIM values over time.

Note that we cannot use any data from IMPLAN. IMPLAN is commercial software and has a licensehttp://implan.com/wp-content/uploads/IMPLAN-LICENSE-AGREEMENT.pdf protecting its data (see points 3.3. and 4 in particular). We have never run or used IMPLAN, so we have never signed any license agreement with IMPLAN. But even if we haven't signed a license agreement, we still have to be mindful of IP ownership laws. So I think we must continue to avoid using any IMPLAN data in our tool.

The screenshot you have above says those data originally come from an Argonne National Lab report. Anything produced Argonne National Lab is free for us to use. If you can find the original Argonne National Lab source, we can use that.

— Reply to this email directly, view it on GitHubhttps://github.com/Energy-Innovation/eps-us/issues/205#issuecomment-1146175291, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AK5N6SPAKOT3BVBLM3AHTATVNI2AJANCNFSM5JKWTRMA. You are receiving this because you authored the thread.Message ID: @.**@.>>

[robbieorvis]

Updating this to direct to the WRI paper with the code splits we might want to use: https://files.wri.org/d8/s3fs-public/2022-09/federal-policy-building-blocks-support-just-prosperous-new-climate-economy-united-states_0.pdf?VersionId=6S8uqr9_OIpnLpz8rqEZW8i4HKll7AsE

[robbieorvis]

I found the Argonne report that has what we need to determine component costs and assign them to relevant categories. Too big to upload but sourced from here: https://vms.taps.anl.gov/document-download/?name=BEAN%20Light%20Duty%20Vehicle%20Techno-Economic%20Analysis

That gives a full cost breakdown by component types. Additionally, EPA, EIA, and ICCT use a retail markup of ~1.5, which (per IMPLAN) is assumed go to 100% to the dealer.

[robbieorvis]

Additionally, we need to get from the MSRP (retail price) to the production costs. We can do that using "RPE" which stands for retail price equivalent. EPA has good data on this: https://nepis.epa.gov/Exe/ZyPDF.cgi/P100AGJ1.PDF?Dockey=P100AGJ1.PDF.

The retail prices are roughly 1.5x the manufacturing costs. In that extra 50% are a variety of things, which can each be assigned to the relevant ISIC code. From there, what's left are manufacturing costs. Here is what things get tricky.

We could simply divide up the manufacturing costs at that point into the relevant ISIC codes. Because ISIC 29 includes manufacture of vehicle parts, a lot of the non-battery costs for BEVs would still go to that sector with the battery costs going to the electrical equipment sector. These costs would have to be adjusted at this point for domestic content shares to ensure we are only accounting for domestically produced vehicles, batteries, etc...

The trickier next step is what to do about the DLIM/TLIM values for vehicle manufacturing, because the constituent parts will have changed. In other words, the remaining components of vehicle costs are different for BEVs, e.g. no transmission is needed. Maybe we ignore it for now, since we will already have adjusted for the biggest difference, which is the battery costs, and we just assume the breakdown of remaining vehicle manufacturing costs is roughly the same, especially since a lot of this comes from ISIC 29 already (the same ISIC that is vehicle manufacturing), since it is a producer both of vehicles and vehicle parts.

Once we have a good decision on this, the next step is how to handle labor requirements. There is a good evidence that vehicle assembly costs are roughly 1/3 lower for BEVs than ICE. On the one hand, if we remove the battery costs upstream, we will have removed more than 1/3 of the costs for vehicle manufacturing, and as a result we would see a big loss in vehicle manufacturing jobs. On the other hand, it's possibly that electrical equipment manufacturing just makes up for these and that we fail to correctly account for job changes. This is a very tricky thing to get right.

The better way to do all this would be to split vehicle manufacturing up into different types of vehicle manufacturing, by making some simplified assumptions around costs. For example, if batteries are 10,000 of a BEV costs but there's no transmission and a transmission is 7,000 of an ICE cost, maybe be we just zero out the transmission line in the make-use table and add that value multiplied by 10/7 to the line for the BEV vehicle manufacturing industry (which we would be making up). We'd be fabricating a BEV manufacturing industry to get the right multipliers. Because we could modify other aspects of this industry, such as as the labor intensity per unit output, we could get a lot better results by creating our own industry and calculating all the remaining values than trying to do it piecemeal above. But, this approach also requires a lot of work and data.

[robbieorvis]

At least according to the Argonne data I have for 2020, the first approach would reduce relative vehicle manufacturing costs by roughly 30%, which is pretty close to a 1/3 reduction. So actually, I think the first approach might be fine. To summarize:

Step 1: Take the amount spent on vehicles and assign the non-manufacturing costs to the relevant ISIC codes. This is equal to (RPE-1)/RPE * amount spent on vehicles. For remaining manufacturing costs, assign to the relevant ISIC code depending on vehicle type. An easy interim fix is to assign the battery costs to the electrical equipment industry (ISIC 27) and the remainder to vehicle manufacturing. Because we calculate battery prices directly in the EPS, we could use these values directly as a share of vehicle costs and spending. This yields roughly 30% lower spending per vehicle in vehicle manufacturing than ICE, which would yield roughly 30% fewer vehicle manufacturing jobs per EV. For PHEV, we might expect higher labor, because the costs are higher and the batteries are smaller, so less is removed. FWIW, other research suggests PHEV assembly costs are higher: https://www.epi.org/publication/ev-policy-workers/

We would do this for passenger and freight LDVs and HDVs, which we have the data for from the Argonne studies. We could implement the same structure for non-road but that will require additional research on costs of the manufacturing for different vehicle types.

The Argonne data is sufficient, I believe, to fully cover this approach for passenger and freight LDVs and HDVs. I may attempt an example of this to see if it works correctly.