Add BAU Representation of Financing and Financing-Related Policy Levers

[jrissman]

The EPS does not currently include a representation of financing (though it does spread out the construction costs of new power plants over a few years). An explicit, BAU representation of financing costs is probably necessary in order to add financing-related policies.

Some of the main things to do for the BAU case representation might be:

• Increase the overall cost of certain things, like new power plants, and perhaps new building components or industrial equipment
• Direct the revenue from those increased costs to the finance industry’s ISIC code
• Spread the total costs (not just the incremental financing costs) out over time, perhaps using some of the smoothing or time delay functions in Vensim

Then, policy levers would alter the approach above. For example, if a policy causes government to offer low-cost financing for certain power plant types, we’d need to reduce the incremental added (financing) cost and direct the financing revenue to government rather than to the finance industry’s ISIC code. We should generate a list of key financing policies (similar to this example) that we would be interested in representing.

[robbieorvis]

Some initial thoughts here:

Building out a simplified financing model is not too challenging, but limitations in Vensim mean we have to make some limiting assumptions.

The basic structure, using the power sector as an example, is:

1. Calculate capital cost ($/MW * number of MW)
2. Read in 4 new data points/files, probably subscripted by power plant type a. Equity share of financing b. Debt share of financing (a+b=1) c. Equity internal rate of return (%) d. Debt borrowing cost (%)
3. These are then used to calculate a weighted average cost of capital (WACC), which is a percent equal to the weighted average financing cost (debt share debt borrowing cost + equity share equity IRR).
4. We use this WACC to calculate the capital recovery factor, which is the construction annuity after incorporating financing costs (see here: https://en.wikipedia.org/wiki/Capital_recovery_factor) and multiply it by the upfront construction cost (note: this method does not use the discount rate).
5. We then add the fixed O&M, variable O&M, and fuel/non-fuel costs as we do now in the structure to derive an Cost per Unit New Elec Output (this is now a more proper LCOE than what we have in the model).

This approach is much better aligned with how other models estimate LCOE. The proposed structure would allow us to add at least one new policy: Improved lending rates (i.e. green financing) for certain power plant types that would lower the debt borrowing rate. We might also consider whether or not we would want to test changing the equity IRR too. It may be the case that certain policies can reduce the risk of equity investment, therefore lowering the required rate of return. Clean energy is kind of known as being less risky once built because there are no fuel costs.

The existing capacity and generation subsidy policies would continue to work as is, lowering the construction cost as fed into the CRF calculation and the annual operating costs.

There are remaining questions for me regarding other policies, like accelerated deprecation. While I'd like to to this in the EPS, I'm not sure it's really internationalizable and I'm also not sure we can do it in the EPS. The challenge arises in calculating discounted future cash flows in the EPS, because we can't have the model calculate a future cash flow beyond the model FINAL TIME and then take a discounted NPV. This makes the use of annualized forward cash flows impossible. We get can around this in the proposed structure because the CRF is an equation that converts a static annuity to a single year value as a percent of the upfront cost. But we already have to make some simplifying assumptions to get this working, e.g. that the debt term (i.e. tenor of the loan) is the same as the project lifetime, because we need the annualized amount to remain constant across years in order to make sense without having an actual discounted cash flow calculation (in reality, the tenor is probably likely to be shorter than the project lifetime).

It would be good to also solicit feedback from a couple of folks (like Uday at RMI or Jim at ACEEE) to see if there are additional policies that can be modeled within this framework or that we are missing, and to make sure the proposed approach is reasonable.

For non-electricity sectors, we could do something similar to the above, but with some simplifications. As a general rule, we only care about technologies that are financed, for example vehicles. In most cases the vehicle is likely have a small up front payment that lowers the portion of the capital cost that is financed, and then the remainder is financed through debt (probably 80-90% of the cost of the vehicle). We can use the CRF function to develop an annual levelized cost value. This is another case where we'd have to make some simplifying assumptions, because the loan tenor would probably have to be equal to the vehicle lifetime (I'm not sure how this relates to actual car loan tenor).

Since electricity and transport are the two sectors where we have economic choice in the model, we could probably focus on these for now and not worry about buildings or industry yet. I think we would need a choice mechanism based on prices in order to affect other sectors, though of course we could just calculate the change from BAU and find an elasticity to multiply the change in annualized costs by to see about uptake of more efficient technology or different fuel choices.

[jrissman]

Got it, this is really helpful! I do think I could use this framework to redo some of those power plant cost calculations, and getting feedback from others makes sense, once we have something to show them.

I agree with sticking to new power plants and new vehicles for now, and only considering building components and industrial equipment later.

Note that the buildings sector and industry sector do make various economic choices. Buildings already includes two elasticities- one that governs the efficiency of new building equipment/components purchased, and one that governs how much building components are used, both with respect to the cost to use those components (accounting for fuel cost changes and efficiency changes). There are also the macroeconomic feedbacks from the IO model. The only thing missing is price-driven fuel type switching. Industry has its own set of economic things it considers that can alter production levels by different industries. But industry has neither an elasticity that governs the efficiency of the newly-purchased industrial equipment, nor one that governs price-driven fuel type switching. Each of these things already has its own issue in the GitHub issue tracker:

Buildings - #10 (fuel type) Industry - #8 (efficiency) and #9 (fuel type)

These might involve tracking the physical "stuff" in these sectors (some measure of building components with age cohorts, and some similar measure of industrial equipment) like we track vehicles and power plants, but perhaps more abstracted. This might come with some downsides and trade-offs, so we haven't yet decided to move ahead on #8, #9, or #10. We will have to decide if we want to complete those three issues before we'd extend financing to those sectors. Without these three issues completed, financing costs won't affect fuel type in buildings and won't affect either fuel type or efficiency of new industrial machinery.

[robbieorvis]

Great! I asked @ssonniaa for her thoughts too, so she may have some additional ideas.

I hear you on buildings - I was just thinking like, as you note, we don't track the physical "things" so the economic choice structure is different.

@mkmahajan and I had an interesting conversation recently with LBNL about their buildings model, but our takeaway was that it would be prohibitively difficult to follow their methodology for updating the buildings sector.

I can add some thoughts on that approach in the separate issues, if desired.

[robbieorvis]

Two comments:

1) I think there is a way to actually account for the different borrowing times using some of the existing structure in the EPS and levelizing the sum of debt and equity costs of the projected lifetime generation. It is important to do this correctly because otherwise the generation capacity lifetime will play too large of a role in deciding a project's levelized cost. I will try to get a working structure and share it today.

2) We got a request to make sure we correctly allocated the policy costs to the correct economic actors (i.e. ISIC code) when we model finance policies. For example, if we modeled green financing (say through a green bank), we would want the financing to come from the government, not the financial services sector, whereas right now the financing would come from the financial services sector. We'll need to be careful to correctly allocate cash flows when we add in financing and related policies. Federal/financial service

[robbieorvis]

@jrissman see the attached spreadsheet and calculations on the "Cash Flow Calculations" tab. We can correctly account for different borrowing and equity timeframes by using a combination of the CRF formula and the PVordinary_annuity formula. Basically we are creating a present value of each stream of payments (debt, equity, fuel, variableO&M, and fixedO&M), discounting by the discount rate, then dividing by the lifetime expected MWh to derive an LCOE for each technology. This approach correctly discounts future cash flows and handles different equity and debt period/tenor. EPS Financing Model.xlsx

[robbieorvis]

One additional note here based on current model structure and a separate issue I logged related to soft costs.

To correctly model financing costs we first should take the overnight capital costs and split out hard costs and soft costs, then calculate the updated values of those, then resum them and then apply the financing calculations above.

[jrissman]

We discussed this today. In terms of cash flows, the general approach for power plant construction in the electricity sector would be something like:

• We continue to assume power plants are built in one model year, and we continue to decide what to build in that year
• However, the payment for that plant's construction in that year (to workers, etc.) comes from the finance industry, not from the electricity suppliers
• Electricity suppliers they pay back the finance industry over the following X years, with the amount dependent on the interest rates

This will prevent sudden shocks to electricity sector costs, which will also smooth out changes to electricity rates, when the electricity suppliers pass the costs on to consumers.

That's the physical, actual cash flows that create jobs, etc. Separately, in the code for deciding what plant types to build, the utilities will take into account an NPV of financing costs along with other costs. A new policy that lowers financing costs for certain power plant types would then be factored into the decision-making about which plant types to build upfront, and it would lower the amount paid back to the finance industry by the utilities in the X years after the plant is built. It would not change the amount the finance industry pays for the construction of the plant in the year it is built.

[jrissman]

Merging in issue #109, whose description is:

Currently in the EPS changes in power sector costs incurred as a result of policy are incurred in the same year they occur, with all costs passed through in a single year. In practice, utility bills are used to recoup these costs over the lifetime (more or less of an asset). So a power plants costs would be passed onto customers over ~30 years instead of all in a single year.

We should try to reflect this when modifying changes in electricity prices in combination with adding in financing policies in a subsequent EPS update. This would help smooth induced jobs effects and better reflect how rates/bills change in response to policy.

[robbieorvis]

Sounds right. Might make sense to discuss the proposed methodology in detail when you start drafting it up and before you implement it.

[jrissman]

This can be included in the electricity sector improvements in 3.5 (to finance power plants, transmission lines, etc.), but also financing should be available in other sectors (transport, industry, buildings). Policy levers are only needed when the financing is for green/clean technology.

[mkmahajan]

I've added a policy lever to decrease power plants' cost of capital, subscripted by power plant type.

We discussed addressing this issue sector by sector as we add in more sectoral detail in successive updates. I am leaving this issue open but moving the tag to 3.6, as I believe the power sector financing lever is all we will include in the 3.5 update.

[robbieorvis]

Just mentioning that in line with standard convention on LCOE, e.g. from NREL ATB, I am removing the part of the model that adjusts the Annual Financing Repayment per Unit Capacity by the Generation Capacity Lifetime. That part of the model has the unintended consequence of making high capital cost, long lived assets look far too cheap and the model is selecting them. You can imagine a situation in which you have a nuclear plant that is 5 times the cost of solar but lasts for 100 years. In that instance, the model would divide the repayment by 5 (20 years / 100 years), making them look equally attractive in the model. We want to avoid behavior where the model might pick a much more costly technology today because it might last longer. Or, if we wanted to do that, we would need to factor in future maintenance and upgrade costs, like we now do for power plant retirements.

In all the materials I have seen from EIA, NREL, Lazard, etc..., I have only ever seen an LCOE approach that looks at the financing period (which may vary from 20 to 30 years), but nothing that adjusts the way we do. I am therefore going to delete out this part of the model, and the annual financing repayment will be based on the capital recovery factor timeline.

[robbieorvis]

We still haven't implemented policy levers that affect this, but we could easily do it as part of forthcoming work in 4.1, so I'm tagging it for that.