Following on from the previous article regarding the misuse of metrics, this article is a guest post by Keith Pickering. More of his analysis and commentary can be found at Daily Kos.
A few thoughts on LCOE, Levelized Cost Of Electricity.
The first thing to realize that LCOE is, and always has been, an investment tool, designed for investors, to aid investors in energy markets make investment decisions. And when LCOE is used for that purpose, it is (usually) appropriate.
The problem comes when we want to use LCOE to make public policy decisions, which can (and usually do) have a different set of decision parameters than financial investment. One obvious difference is in asset lifetime.
For example, the US Energy Information Agency publishes LCOE estimates every year, and while they do a pretty bad job of explaining how they compute things, one thing they do say is that for all energy types they use a lifetime of 30 years. Why? Because banks don’t make loans for longer than 30 years, that’s why. Now if you’re considering whether to loan money to an energy project, that 30 year lifetime makes perfect sense. But if you’re planning an energy infrastructure for half a century or more, the 30-year lifetime in your LCOE calculation will systematically undervalue long-lifetime assets (like nuclear and hydro) and systematically overvalue short-lifetime assets (like wind.) Using a 30-year lifetime implies, essentially, that generating assets with a lifetime of more than 30 years will have zero asset costs during their lifetimes beyond loan payoff. Essentially that pretends that the electricity cost would *drop like a stone* to extremely low levels at the 30 year point. But those really low future electricity costs are *never reported* in LCOE; the assumption is just left out there, unmentioned.
Another thing to realize is that a key component in all LCOE calculations is the “discount rate.” Basically, the discount rate is the annual rate of return investors would expect to get on a properly valued asset. If the discount rate is high, investors want their money back right away. High discount rates value the present highly, while discounting the future strongly. High discount rates therefore penalize technologies that rely heavily on long-term fixed assets (once again, hydro and nuclear.)
Discount rates are used elsewhere too, for example in computing the Social Cost of Carbon (SCC). If you’re taking the long view, a low discount rate values the future more highly. For that reason, climate hawks like to use low discount rates when computing SCC, because that computation raises carbon cost. The carbon we emit today will continue warming the earth for centuries, and will continue to cause damage for that entire time. The lowest possible discount rate will capture (some of) that future damage and value it when computing SCC. The US government currently uses a 3% discount rate when computing SCC. And even that may be too high, when you consider the entire lifetime of CO2 in the air.
To be consistent, then, us cliamte hawks should also press for an equally low discount rate when computing LCOE; that is the socially responsible way to value the future in the face of long-term climate change. But EIA uses a Weighted Average Cost of Capital (which is the discount rate by another name) of 5.5%, nearly twice the rate used in computing SCC. That doesn’t mean it’s wrong; for an investment tool, it’s appropriate. But again, if you want to use LCOE for policy purposes, there are other things to consider.
The investment management company Lazard publishes their own LCOE results every year, and every year the low-low LCOE of wind is caressed and trumpeted by certain wind-loving types. It’s no coincidence that Lazard is heavily involved in wind energy stocks, and has skin in the game as far as wind energy is concerned. The Finnish blogger Jani-Petri Martikainen has already cataloged some of the many thumbs Lazard puts on their scale to favor wind, and it’s no surprise that jacking up the discount rate is one of them: Lazard’s rate is a whopping 9.6%, which immediately rockets high-asset technologies (like hydro and wind) way up in price. Then they lower their LCOE (for wind only) by assuming hugely unrealistic (55% !!) capacity factors for wind. The net result is that Lazard’s bottom-line wind numbers look about like EIA’s (so they can reassure their customers that they’re doing it right) while all other technologies are way too high. It’s utterly deceptive, but they apparently hook the investors they’re trolling for.
Another good LCOE resource is the OpenEI Transparent Cost Database, which is a meta-analysis of everybody else’s LCOE, but with homogenized parameters for tax rate, discount rate, and capacity factors for the various technologies. Unfortunately it looks like it hasn’t been updated in more than a year now, but it does have everything spreadsheeted out, which lets you examine the calculations and play around with the assumptions. With OpenEI’s standard parameters, nuclear already looks appropriately cheap, even in the first thirty years. And if you count the second thirty years of expected plant life, it’s no contest.
One final word about cost. You often read about some contracted electricity price for some new installation (typically solar) that is impressively low. These Power Purchase Agreements (PPAs) are common in the industry, but you should be aware that PPA price is always lower than LCOE. That’s because a PPA does more than transfer energy: it also transfers risk, from the seller to the buyer. If you want to build a new generator (of any type), you’re taking different financial risks: the risk that the project will never get built, and the risk that you won’t be able to sell the electricity, or not for the price you need. Banks understand these risks and set the interest rate on the loan accordingly. When a PPA is signed, the first part of that risk (that it won’t get built) has already passed, because as a general rule a PPA isn’t signed until the generator is already built. And when a PPA is signed, the second risk component (not being able to sell it, or for the right price) has also been eliminated. With a newly signed PPA in hand, the generator owner can re-finance his loan to a very, very low rate, because at that point the risk is almost completely gone. The buyer of the electricity (the other party in the PPA) has assumed the risk that the price he’s paying on the PPA will be lower than the price he could have gotten elsewhere on the wholesale spot market. Because the buyer is assuming that risk, he expects a lower price than he otherwise would have gotten; and because the seller is shedding that risk, he’s willing to sell at a lower price too. Generally, a PPA price is about what LCOE would be if the discount rate were close to zero.
Many thanks to Keith for this clarifying commentary. I added the description of Hornsdale wind farm to help illustrate it with real world Australian context.
For the reasons mentioned above, and as it’s so heavily relied on by anti-nuclear campaigners, I avoid using Lazard’s analysis on this blog. But its latest edition included this piece of important guidance which many would be wise to take on board.
Even though alternative energy is increasingly cost-competitive and storage technology holds great promise, alternative energy systems alone will not be capable of meeting the baseload generation needs of a developed economy for the foreseeable future. Therefore, the optimal solution for many regions of the world is to use complementary traditional and alternative energy resources in a diversified generation fleet.
I think we should not discount for the first 20 years of a project. Beyond that the future is unknowable and 0% is less controversial than other numbers.
When the Heywood Interconnector was out did the ACT still claim to be getting power from Hornsdale? I understand they don’t claim the LGC. If pricey gas and the Hazelwood closure doesn’t affect them how come ACT retail prices went up 19%?
I haven’t read Finkel’s original report but perhaps intermittent generators should contract for say 2 hour blocks so they have to find their own storage as well as frequency correction.
Interesting analysis. Just one aspect I’d quibble with:
Actually the 30 year lifetime assumes there’s zero value after 30 years. If an asset is built to last 60 years, there’s no allowance for that since capital costs will have been amortized over 30 years, thus magnifying any erroneous application of higher discount rates.
Think of it as a mortgage that has to be paid off in 30 years when that asset is still generating revenue for 60 years. Such a shorter mortgage term combined with unfairly high discount rates, increases the burden on the investor and thus makes the investment less attractive.