All Episodes

[Episode #23] – Facts and Falsehoods in Energy Transition


Should we tweak our markets to keep nuclear plants alive, or forget about markets and pay for them another way… and do we really need them at all to keep the grid functioning? Is nuclear power really declining because of overzealous environmentalists, or are there other reasons? Is it possible to balance a grid with a high amount of variable renewables and no traditional baseload plants? Is cost-benefit analysis the right way to approach energy transition? How much “decoupling” can we do between the economy and energy consumption, and how can we correctly measure it? Why are we so bad at forecasting energy and economic growth, and how can we do it better? How will energy transition affect the economy?

We explore all of these questions and more, and try to separate fact from falsehoods in this wide-ranging interview. It might even change your mind about a few things.

Guest: Dr. Jonathan Koomey has been studying energy and climate solutions for more than 30 years. He’s a world-class researcher on the environmental effects of information technology, the economics of climate solutions, and exploring the future through computer modeling, among other topics. His latest book, the 3rd edition of Turning Numbers into Knowledge: Mastering the Art of Problem Solving, summarizes practical lessons he’s learned over the past three decades while doing analysis at the intersection of engineering, economics, environmental science, and public policy.

On Twitter: @jgkoomey

On the Web:

Recording date: July 14, 2016

Air date: August 10, 2016

Geek rating: 8

Chris Nelder: Welcome Jon to the Energy Transition Show.

Jonathan Koomey: Thanks very much Chris.

Chris Nelder: So from our Twitter conversations it's become clear that we share an interest in a number of diverse topics. So I wanted to explore some of them on the show. Maybe we should start with nuclear power since it's really been a hot topic lately. Nuclear power is getting pushed off the U.S. grid because it just can't compete with natural gas and renewables, under the current market design, and admittedly that market design does not give nuclear plants credit for generating carbon free power and it probably should. Particularly since most of our transition efforts are designed to combat climate change. So, just for starters what do you think about that, would pricing carbon be enough to keep nuclear plants in the black or is there a better market mechanism, or maybe should we just be going the route that New York is going in and applying a special subsidy just to keep nuclear plants around.

Jonathan Koomey: Well let's step back a little bit and talk about what's actually driving this. So many power markets are driven by short run marginal costs. And what has been done in those markets, which is not the whole United States, but in some important markets like California, they've created markets that are driven by the short run marginal cost of delivering power at any hour. And so if you're a power generator, and you want to build a new power plant, or you have an existing power plant, the way you are rewarded is based on the hour by hour cost in the marketplace, and that cost fluctuates a lot and it increases the risk for power generators and so for a highly capital intensive technology like nuclear power that creates risk for them because they would have to invest capital with the expectation that the markets would give them enough return to pay back. And the problem of course is that there's nothing guaranteed about markets, and so markets go up and down they vary a lot. And so just the very presumption that we need to use markets for power systems, it's an assumption it doesn't have to be that way. And so some of the mechanisms that you described are ways to tweak markets so you can put a tax on carbon that would benefit nuclear power but it would also benefit wind and solar and all other forms of zero carbon electricity. And so the issue is whether any one of those tweaks that you talked about will be enough, and that depends on both the specific power plant as well as the power system in which it's embedded. And so it's hard to answer in general. I think that it's clear that having some sort of price on greenhouse gas emissions is a sensible thing to do. And if we were to have the wisdom to do that and the politicians with the courage to do that, that that would create strong incentives especially in the power sector to move away from coal and natural gas and towards these lower emitting resources. So I'm not sure that that's necessarily a market solution that's needed. It could be, could be these tweaks that you've talked about, but it also could be that actually good old fashioned regulation for certain sorts of power resources may still be the most sensible thing to do where there is high risk, high capital costs. We still do this for transmission and distribution everywhere. It's just power generation where we've imposed this overlay of the market, and you know markets are social constructs. So there's nothing magical about a market, you can design markets well you can design markets poorly. And one of the prime examples of bad market design was the original California marketplace in the mid to late 90s that exploded because market manipulation mainly. But the point is that that was bad market design.

Chris Nelder: After the first aborted attempt to do deregulation you mean.

Jonathan Koomey: They tried to do it right, so they created a market, they set up incentives in a certain way, certain power generators decided that they could game that market. And they did. And that led to the collapse of the whole system. So there are good ways to design markets. There are bad ways to design markets, and there are terrible ways to design markets, and we need to make sure that if we're using markets, we're designing them in a way that will result in long term beneficial outcome for society.

Chris Nelder: So just to kind of shorten up maybe, or summarize what you just said, there's no reason why nuclear generators couldn't just be surviving on a cost plus basis, like all the rest of the rate based assets even in a deregulated or quasi-deregulated market.

Jonathan Koomey: Yeah, they could be. Now you know, some of those reactors may still be quite expensive.

Chris Nelder: Yeah that would be the next question, you know I don't actually know what a nuclear plan would cost on a pure cost plus basis.

Jonathan Koomey: Yeah, so an existing nuclear plant has marginal costs of between 1-3 cents a kilowatt hour although I haven't checked those numbers recently. They do change over time as reactors age. There's different things that have to be added and modified and so on. But it's generally high capital cost, very low marginal cost resource. But what is happening in part because of cheap natural gas, and also because of growth in wind and solar which also have very low marginal cost, tenths of a cent per kilowatt hour, you end up with some power market prices that are so low that just make it hard for a capital intensive resource to pay for itself.

Chris Nelder: Sure. OK. So moving along then to some of the other interesting topics about nuclear, there's a number of beliefs about nuclear power that persist today despite ample evidence to the contrary. And you've actually pushed back on some of those beliefs in your papers, so let's talk about a few of those. One is that the Three-Mile Island accident in 1979 was the main reason why nuclear power ran into resistance in the US and plant construction stalled and production from nuclear began to fall. What's your response to that?

Jonathan Koomey: Well, the most important thing to understand is that the nuclear industry was having big problems before 1979. So 40% of the reactors, the planned reactors that were cancelled in the United States were canceled before Three Mile Island. 40%. So that's a huge number. That tells you that there were some things going on that were completely unrelated to Three Mile Island and any regulatory response to that. I think the story that the industry likes to tell is that Three Mile Island lead to the regulators really clamping down, overreacting overregulating, overzealous enforcement, and so on and that that was what killed the industry. But the things that really killed the industry were largely in place before that.

Chris Nelder: What were those things? I mean what were the reasons why those 40% of plants got cancelled?

Jonathan Koomey: So there's declining demand growth for electricity: in the 60s you had pretty reliable growth of 7% per year in the United States. But after the first oil shock in the United States, you started to see efficiency standards for appliances, you started to see increases in electric rates that also affected people's desire to reduce electricity as you saw some industry shifting overseas. So the growth went from 7% per year which was pretty reliable. So reliable that do the forecast they would take a piece of semi-log paper put a ruler on it and draw the straight line. That's how reliable it was, it still varied a bit, but it was pretty reliable. But then you went in the 70s towards, you know, 4, 3, 2% growth, and then by the 80s it was you know 2% growth regularly, and then nowadays there is basically zero percent growth for electricity in the United States as a whole, from 2007 onwards. There's been a little bit of up and down, but you've seen basically a flattening of electricity demand growth. So that was number one. Number two, there were high interest rates in the 70s and that of course affected the capital intensive resource, like nuclear power. There were structural problems in the industry related to the way that plants were constructed. So they were starting construction on plants that were only 5 or 10 percent designed. Bad idea. Terrible idea. You also had changing public perceptions of the credibility of the nuclear industry because they would say things that ended up not being true, and not being true in a very public way, like we could never have a core meltdown in the United States or the power will be too cheap to meter, a number of different things right. So we all know those stories. The industry made claims that were not correct, and eventually public acceptance of nuclear power was affected by the fact that public events showed those statements to be incorrect. And then the fifth big factor was that the independent power industry in the United States started with PURPA in 1978. So what that law did for people who don't know is that it basically required the utilities to purchase power from independent power producers and it required them to do it at what was called avoided cost. Now it didn't define what avoided class was because the utilities and the regulators had to decide for each state and utility what that would be. That's a very situation specific thing but it required them to buy the power and avoided cost. And what ended up happening, I actually took a class from Bill who was at the California Public Utility Commission as the consumer advocate. He was there in the late 70s early 80s and when PURPA was passed, California put out, may have been a few years afterwards, but they put out a thing called Standard Offer 4 which was fixed prices for power escalating over time for independent producers and at that time there was still an expectation that there was going to be significant demand growth and that there was a need for more capacity. And, what happened was they put this out there thinking that might get a few bidders, and they ended up becoming completely oversubscribed, like many times more power was offered than they expected or could handle. And that was the beginning of a big shift where the independent power industry showed that it was able to produce reliable power for low cost. So those five things: declining demand growth electricity, high interest rates, structural problems in the industry, changing public perception of the credibility of the industry, and then the rise of the independent power industry. All of those things were in play or starting around the same time as Three Mile Island or before. All of those things had a huge impact.

Chris Nelder: Yeah, and four of five of those things is basically a market issue.

Jonathan Koomey: Yeah, absolutely.

Chris Nelder: OK. So moving beyond that, I mean among the myths and misperceptions about solar including the suggestion that the decline of nuclear power was mainly a political reaction to Three Mile Island have actually originated in recent years with the Breakthrough Institute, a pro-nuclear lobbying organization by any other name or characterization that's what they are. So you and I have both looked at data presented by the Breakthrough Institute on the cost of nuclear power and we both came to similar conclusions: they cherry picked the data, they made misleading comparisons to competing technologies, and they made a number of analytical errors and I'll link to our respective papers in the show notes about that, but can you summarize briefly what the main issues are with BTIs cost estimates?

Jonathan Koomey: So you are referring to a paper that came out in the journal Energy Policy. The first author was Lovering and she works at the Breakthrough Institute. And this paper came out earlier this year.

Chris Nelder: Well actually I was responding to a paper they put out almost three years ago. They have been doing this for a long time.

Jonathan Koomey: The most recent one came out earlier this year. And it was because I and my colleagues had done cost analyses of US and French nuclear reactors, some folks approached us and said you know what do you think of this, can you write a critique, so we did look carefully at the work that they did and the article is puzzling. It makes what I would consider to be a kind of amateur mistake in that they ignore interest during construction. So in the history of power plant generation and construction people have looked at capital cost in different ways. One of the ways that they look at it is in what's called overnight costs which is this fiction that kind of boils down the costs to, what if you could build this plant overnight? What is the cost in dollars per kilowatt? And that fiction worked very well for a long time when costs of power plants were predictable, and typically with smaller plants it's easier to predict how long it will take to build them, what it will cost and it's kind of a convention to show costs without interest in that way. So the industry has done this for a long time. They present these overnight costs and then they convert them to cost with interest during construction. And it's simply invalid to compare cost data without including the interest during construction because it takes somewhere between 5 and 15 years to build the nuclear plant. So of course the interest on all that capital building over time has a big effect on the cost of reactors, so you just can't ignore interest. That was a big methodological mistake in that particular study. And then what's even more puzzling is that the claims that they made about the data, they tried to to do a comparison across different countries, and they also tried to expand the data for a longer time period, so they took cost data for some of the very early reactors. There were demonstration reactors, there were turnkey reactors in the 50s and 60s in the United States and they tried to use those earlier data to argue that well look nuclear power costs don't always go up. Look at how these costs came down in these different countries in those early years. And that is true for those small reactors. It's true for the 50s and 60s when we had a very stable economic environment and we were first started to experiment with reactor technology. But what's not true is that those trends in the 50s and 60s say anything at all about the cost of building new nuclear plants or even the cost of building plants in the 70s and 80s.

Chris Nelder: Especially since all commodities repriced starting around 2007. I mean any any cost estimate before 2007 is just fundamentally invalid.

Jonathan Koomey: Yeah. So this was an unfortunate way to frame the discussion. I mean you had these reactors in the 50s and 60s. These were small reactors. They were tens to perhaps, you know a few hundred megawatts and in some cases. But the reactors being built in the 70s and 80s were a thousand megawatts each, completely different technology. And much bigger, and so a much more difficult construction challenge. And so it's just not valid. You can't use any cost trends from the 50s and 60s to say anything about current reactors or even reactors in the 70s and 80s. So that was the mistake and then they claim that the data show that costs sometimes decline, but in the modern era the only country data that they showed which showed any sort of a cost decline was South Korea. And you know first off these are without interest during construction so we can't you know really use them. But even if you accept them as valid, the South Korean data are not independently audited, they're of unknown quality. And until somebody independently looks at the data you just can't know whether those numbers make any sense. So there are a number of problems with this. And I think it misled some people, but I'm hopeful that Gilbert et al. rejoinder to the Lovering paper as well as our paper will have the effect of at least explaining to people why you can't make comparisons in the way that they did.

Chris Nelder: Yeah. So another myth about nuclear is the idea that renewables and DERs, you know distributed energy resources, this kind of catch all term, simply can't do the same job that nuclear plants can do in providing firm carbon free power, and so therefore they claim we'll always need for them to balance the variability of renewables. How do you respond to that?

Jonathan Koomey: The important thing to understand is that the individual characteristics of power plants are not the same thing as the characteristics of the system in which they are in embedded. So you can have a system composed of many small power plants that delivers reliable power even though some of those plants may have variable output over time, and you can compensate for that variable output in different ways. So one possibility is good forecasting so you can at least understand how the the variability comes about. And how to predict it. Another possibility is to have control of load. So most of the people who come at this from the perspective of nuclear power or large power systems, they come at it with the assumption that load is something that you just accept, it varies and we have no control over it. But there's no reason why that needs to be the case. And there's a long history of utilities trying to influence load with some success and there are controls on certain kinds of equipment like water heaters in residences, or air conditioners. There's actually incentives for commercial customers to make ice or store cold at night when electricity is cheap and then use that cold resource to cool off their buildings in the daytime to avoid using as much air conditioning. So there's many different possibilities for changing load, some of those involve storage, some of those involve computer technology, communications technology to better match the task we want to perform with the energy that we're supplying to do them, but demand is not necessarily a constant so that's another thing you can vary. And then you can use these resources in a way that spreads them geographically and also creates diversity of different kinds of resources so you have some solar, you have some wind, you have some biomass, and depending on the way the system is configured those can deliver reliable power, again combined with this better forecasting and the demand side changes and that, you know there's no reason why you need the system to have any big power plants. Now in some power systems it might be helpful. So I'm not personally opposed to for example keeping some nuclear reactors running in large power systems because replacing that zero carbon electricity is going to take some time. And so I'm, I'd be perfectly happy to have some of these reactors continue operating as long as it can be done cost effectively and safely. But the point is that we don't need them ultimately, we don't need to have that. Now, again it would be easier to achieve carbon reductions if we had some of those nukes still running. And so I'm not against doing that, but I think ultimately we can create a power system that does not require those large power plants.

Chris Nelder: Right, and so that reflects, you know a quite different set of priorities right. I mean the one question is, do you have to have those big baseload generators running in order to balance the grid. The answer is no. Do you want those things running in order to meet carbon emission targets? Maybe yes or at least for a while.

Jonathan Koomey: So there's also the storage issue that we didn't touch on yet, and that's an interesting one because the cost of batteries have been coming down 8% per year in the last decade or so. And that's a huge benefit. And of course batteries are not the only way to store electricity. I taught at UC Berkeley in 2011 as a visiting professor and two of my students there were from Germany and they came up to me after class and said you know they're doing something interesting in Germany they're taking old coal mines which are very deep, and they're using them for pumped storage. So they're running water through a turbine, and down into the mine, and then at night they pump the water back up to the top and put it in a reservoir. And so I think there's going to be a whole lot more exploration of innovative storage technologies that we haven't even thought of. There was another one I saw the other day where people were running electric trains up a mountain and then running 'em back down. You know they do that at night and then running them back down when they need electricity, and the regenerative braking on the train, train... it's kind of a you know it's just a faux train but it has regenerative braking on it. And so we can get you know a chunk of the electricity back at times when it's much more valuable.

Chris Nelder: Yeah I think that's a really cool project. I mean you know, it remains to be seen how scalable it'll be or you know what the round trip proficiency will be.

Jonathan Koomey: It's just an example of interesting stuff like that coal mines for pumped storage kind of thing.

Chris Nelder: Exactly, I'm definitely bullish on the opportunity for more innovation in the storage space. I think we've only really, we only really started getting serious about looking for that stuff about 10 years ago which was you know no time at all. Really when you're talking about basic research like this.

Jonathan Koomey: Yeah, so storage was kind of a sleepy area for decades. And now it's not. And that's a huge deal. The last thing I would say about this is that there's an assumption that there will be some costs to integrate these intermittent renewables and that's true. I think that's you know there's some cost above the direct cost of the solar or wind that you build. So just for context, your listeners probably know this but you're seeing now unsubsidised 25 year fixed price contracts for wind at two to three cents a kilowatt hour, and for solar we're seeing at five to six cents a kilowatt hour. And so those are fixed prices. And you're seeing of course very rapid declines in those fixed prices as well as the actual technology costs which is you know in solar we've seen 80 percent reductions in the cost of modules in the last five or six years. For wind it's more like 60 percent. But the point is that these are mass produced technologies that are declining in cost because we are learning how to do them better and faster and cheaper. And that also means that one other compensation strategy that we may well choose to adopt is to vastly overbuild the wind and solar and that some times of the day we might just not use the electricity, we might just let the turbines spin. But having that capacity constructed means that you know we can actually use the possibility of letting the turbines spin as our load following. So it's the capacity, it's like a capital intensive way to do this. But there's no reason why you couldn't do it, and given the costs of building these things you know being so low and falling so fast you could easily see part of the strategy being vastly overbuilding. And that's OK. That's OK because we can make a system that has zero emissions, that delivers electricity when people need it. And does it at a reasonable cost. And we don't, again I don't see that we need to have these big power plants. I think they could be helpful in some systems, in some cases but it's not necessary.

Chris Nelder: Yeah, I remember reading a paper, I can't remember who it was by, a couple years ago looking at that, you know that design idea of just massively overbuilding and being able to balance the grid just because there's so much excess resource and it's all interconnected and it still came in at a reasonable price.

Jonathan Koomey: Right. So there's that and then there's also transmission which we haven't talked about. So if you can build more transmission, which I think we should do in any case, what it does is it allows more tapping of resources in different regions, so the larger the geographic region, the more diversity you get from the resources and so that means you can take power from the Southwest to California at a certain time but maybe shift that power to the East at a different time. And so that is another coping strategy. It's another way of increasing the flexibility and reliability of the grid, and there was a recent study that talked about how DC power lines in particular are helpful in this regard because it means you don't have to create a larger system that's all synced up to 60 hertz on exactly the same cycle. You can have these DC high voltage lines connecting different systems and of course you create AC you know turn it into DC and then you create AC at the other end. But it creates kind of a buffer so you don't have to synchronize this vast grid. You can just bring the power in bulk across these DC power lines and that turned out to be a relatively cost effective thing to do.

Chris Nelder: So on the subject of grid balancing it's also often thought that natural gas generators are the only way we're going to be able to balance a grid when we have a large share of variable renewables because only gas generators are flexible enough to ramp up and down quickly and I think you've already essentially debunked that by your previous comments about how actually load can be flexible especially now with modern technology in a way that it wasn't in the past or in a way that we didn't try to use it in the past. But I think maybe a more interesting question is: where does that idea come from and why does that idea persist?

Jonathan Koomey: Well I think it's the same idea that talks about large baseload power plants. It's an old 70s thinking, it's people who...

Chris Nelder: So these are people that are just stuck in the 70s basically?

Jonathan Koomey: Yeah. I mean I personally would not object to having some gas plants around, again because sometimes things happen like sometimes there are weeks long lulls in wind, you know it depends on the place and the time and so on. But you would ideally like to have some means of you know beyond just the demand response, having some additional backup. And it creates an interesting market design problem because if investors are going to build those they're going to expect a return. But if for example utility were to say OK we're just going to keep some of these things around for those extra special emergency times and it's just going to go into rate base and we're going to treat it as our integration costs, our cost of service, we're not going to run them. The incentive is once you have a capital good, the incentive is to run it more hours. And if your goal is reducing emissions, you can't do that. But you need to change the incentives to make it possible for a utility to have this gas plant and not want to run it.

Yeah I mean if you're going to rape incident you can do that if you're an emergent generator and you're looking at the prospect of putting up a gas plant that's going to have a 5 percent capacity factor. That's probably not ever going to work.

Jonathan Koomey: Right. So that's why I think you to have different arrangements. So I think you can have some gas plants around for emergencies. The net result, the net effect on emissions will be very, very small. Because it will only run a tiny fraction of the year. What It will do is it will give you that additional insurance. It's a cost of that system integration that you know is probably worth paying.

Chris Nelder: Yeah. All right, well moving on a little bit here from the concept of nuclear and grid balancing to maybe something a little more theoretical. In a previous job I helped develop an expanded sort of cost benefit analysis which was designed to guide capital spending, to more effectively address climate change. So for example, we could price in externalities like the social cost of carbon or the value of ecosystem services. And it seemed to me that this approach, couching investments to combat climate change in the language and terms of conventional budgeting would be the best way to get comptrollers and treasurers onboard with those investments. And it's certainly the kind of thing that nuclear plant owners would like to see happen now for largely the same reasons. But you've actually proposed a different method which you call working forward toward a goal. So why don't you explain what's different about it and why you think that would be more effective.

Jonathan Koomey: Well let me just say that I think benefit cost analysis for short run decision making is actually valuable. Right. And so for companies in particular this is an important way to think about the problem. The paper that you're referring to is actually going after a bigger issue which is that in the economics of climate modeling world there's this basic sense that the way to understand climate as an environmental problem, and an economic problem is to analyze benefits and costs, and come up with a "optimal reduction strategy". So that paper was actually responding not to efforts like yours related to specific companies which I think is a very good thing and a very valid way to look at a problem. But actually looking at the way economics as a discipline, has analyzed the climate problem. And here's the bit, the way it normally works, and if you've been in an intro or graduate level environmental economics class you'll know that there'll be these two curves and one of them will be a curve that shows what the benefits are as you reduce emissions. And that's a declining curve as you move to the right, and then the other curve will be a curve of costs and the cost will increase as you get to higher emissions, as you move to the right, the point where those two things cross is the so-called optimal point where you end up with the marginal cost of doing something exactly equaling the marginal benefit to society of doing something about climate. And this way of thinking has actually been useful for things like criteria pollutants, so sulfur, where you can actually because we can measure effects and we know something about what the ecosystem damages are and the health damages are, you can actually calculate costs and benefits to some reasonable approximation. But the premise of the article you pointed to was that for climate we just can't do that.

Chris Nelder: So you don't have particular confidence that the way that we're going about estimating the so-called social cost of carbon is good.

Jonathan Koomey: I think I would make a stronger statement, I would say that it is literally impossible to calculate that. Think about it. Can you tell me what the cost of a nuclear plant or a solar plant will be in 2050? How can you possibly?

Chris Nelder: No I can't. And you know I think anybody who's done work in modeling the social cost of carbon would agree that you can't. But that's of course no reason why you shouldn't try and you know in the surveys that I looked, I mean on the one hand you've got like the EPA schedule which right now I think values carbon at about $40 a ton all the way up to a study from Moore and Diaz there at Stanford, that said, really it's more like $220 a ton if you allow the impact of climate change to actually damage the economy and reduce your long term GDP growth rate and you know there's so many factors that get into that. But I guess the idea is not that there's something particularly accurate or particularly good about valuing the social cost of carbon, but that rather if you have to go about getting budget decisions made that way then you might as well try to estimate it.

Jonathan Koomey: So again, for the purpose that you're describing which is a company or an institution who wants to make decisions, making some attempt at estimating the social costs of carbon makes sense because it allows you to operationalize these things. But again I'm critiquing a different thing. Which is a way of looking at the climate problem. And if it's impossible, as I claim to estimate costs and benefits for reducing emissions, reducing greenhouse gas emissions in 10, 20, 30, 40 years, then what does it mean to calculate your optimal path. It's impossible. There is no such thing as an optimal path. So it's basically a critique this way of thinking about the climate problem. And so what I propose instead was actually a more business framing of the problem. And the business framing is what you described which is working forward toward a goal you set a big strategic goal like we're going to keep the Earth from warming more than 2 Celsius degrees beyond pre-industrial times, or one and a half Celsius degrees beyond pre-industrial times. And then you figure out what you would need to do to meet that goal. Then you will implement things, see what works and what doesn't. You know, more of what works and less of what doesn't and try again, iterating until you are on a path to hit that goal. The economists love to come up with their optimal path but the world isn't like that and we can talk about why that is. But if you can't calculate an optimal path you need another way to think about it. This gives you a way to think about it and this is the way businesses as you know deal with big strategic challenges.

Chris Nelder: Yeah, in fact one of the recurring themes of the Energy Transition Show is that we can't really know what the ultimate market design, or the ultimate grid power mix, or anything is going to look like, once we get there. All we can do is say: is this a crazy path to take, this energy transition path, to this distill carbon ization path is this a crazy thing to do. And if the answer is 'No' and there are reasonable good powerful reasons to go down that path then you go, and you figure it out as you go and you dont try to be too clever about figuring out well 30 years from now this technology is going to be more expensive than that one and so we should buy this one instead now.

Jonathan Koomey: Yeah that's exactly right. You have it you have to try a lot of things, fail fast and just keep experimenting. The only way. That's the way the real world is, the world is path dependent. There is increasing returns to scale, there's information costs, there is risk that we don't understand, there are just a lot of things that are just non quantifiable and so that means we need to create the future.

Chris Nelder: And one of the things that we dont do very well and this is another subject you've explored is forecasting energy demand and that's really a knot of multiple topics kind of coming together. You identify uncertainty about economic growth as one major factor in that problem. But in itself, figuring out what the economic growth is going to be in the future rests on other questions about the relationship between economic activity and energy demand, or whether as it is often asserted we've actually reached a point of so-called decoupling where economic growth can continue while energy demand stays flat. So I want to try to unpack that a little bit starting with the latter issue. Now first of all, it's clear that energy demand particularly for electricity and oil, as you mentioned earlier, flattened out in the U.S. about a decade ago and GDP has continued to go up especially since we pulled out of the 2008-2009 slump. But personally I have never been particularly interested in that argument, the so-called economic intensity or energy intensity of the economy because GDP is such a desperately flawed and gross indicator and because it doesnt take into account structural changes in the economy like the offshoring of energy intensive manufacturing industries to other countries like China and it values the dollar if anything equivalently to a dollar of anything else. And you know to be sort of absurdest about it making the argument that the US and the world can continue with this decoupling indefinitely implies that we're heading toward a future where nobody makes anything. And the entire economy runs on us trading backrubs. So just to begin with, what are your thoughts about this decoupling argument?

Jonathan Koomey: Well, I would step back and think about the Second Law of Thermodynamics. What you can do when you look at an economy is you can figure out how much energy we use to deliver a certain service and we can also figure out what is the minimum theoretical energy that we would need to do that service most efficiently, and that allows you to calculate what's called the Second Law Efficiency of the economy. And typically the numbers for the economy as a whole come in around 5 or 10 percent. So the idea that we are somehow close to some optimum and that we can only make marginal changes in efficiency I think is false. I think that there are many opportunities for delivering services in ways that are less resource intensive, so as a concept there's no reason why we can't completely decouple. There's no reason. Now you've pointed out all these issues about GDP which are completely valid. GDP is a very bad indicator, but from the perspective of the overall efficiency of the economy there's no reason why we couldn't continue to become more efficient for a very long time. And of course as technology improves, and the tasks we desire to accomplish change then those opportunities will just grow. So I don't see that there's any reason to say we couldn't decouple. But the reasons that you pointed out with GDP and also the issues around what you're measuring. So, if you're just looking at the U.S. you need to account for the fact that a bunch of industry has gone off to other countries. But you need to do your sums properly, right you need to count everything that needs counting, and if you do that then I think it's you know it's certainly possible to decouple. I don't think there's an argument that we theoretically can't do it. I think the people who argue against it say well we've never done it in the past, or they say we've only done it at this rate. And I tend to be more immersed in the world of information technology where things happen faster and that's where I think the world is or at least the world is going. And so I think things will happen a lot faster than what has happened in previous transitions. But we'll see.

Chris Nelder: Yeah, I would agree with that. I don't think there's anything sacrosanct about the pace of things in the past applying in the future anything of that kind. I've just objected to energy intensity of the economy as being an overall gross indicator which I think is just so full of problems that it's almost not even worth talking about. But I am interested in what you were just talking about the sort of what did you call it the...Second Law Efficiency on a per-service basis. Now that to me seems like a useful metric.

Jonathan Koomey: Yeah, and that's how I think about the problem because if you think about the normal efficiency that people think of its First Law Efficiency, there's energy going into this device and there's a service coming out, or energy coming out. So if you're putting in fuel into a power plant you get electricity out, and you get some waste heat, and the sum of those outputs equals the sum of the imputs and so that that's the first law of thermodynamics, that gives you you know the difference between doing a normal water heater or a normal furnace at 80 percent efficiency, or a condensing furnace at 95 percent so your ability, you can't get above 100 percent in that world. But if you're defining the problem differently as how do we heat this house in the most efficient way. Well we know there are heat pumps, and we know there are ways to deliver heating and cooling much more efficiently than some of these older ways to do it. And that opens up possibilities that are illuminated by thinking about the Second Law Efficiency of delivering the task.

Chris Nelder: OK, so let's move on to the questions about economic growth. Many people who are concerned about the future availability of fossil fuels at an acceptable price, and of course I count myself among them see the repricing of oil starting in 2004 roughly as a major reason why the economy slowed down. Now I'm sure you're familiar with James Hamilton's work on that which showed an unambiguous net negative shock to GDP from a big upward spike in oil prices and which showed that the oil price shocks of 1973, 1979, 1980, 1990 and 2007 were all followed by economic recessions. So when we look at the likely cost of oil and natural gas in the future, particularly after 20 or 30 or so when I think the fracking boom will have faded away and I think it will continue to act as a brake on the economy. So in fact, one might even look at our current situation with negative interest rates around the world and QE to infinity and all of it still unable to generate any real inflation. And you look at the thesis of secular stagnation that Paul Krugman and Larry Summers have put forward. And you think yeah that's what entering into the era of high cost unconventional oil and gas gets you. So what do you think about all that, are high cost fuels contributing to this secular stagnation? And if so is there a way out of it?

Jonathan Koomey: I'm not sure that that they are the sole or even most important cause. And so the literature that you're talking about indicates of course that oil's real cost to the economy is actually a lot higher than what the prices is. That's important for everyone to think about, but it hasn't, as far as I can tell, it hasn't been shown that it's high prices of oil that have driven that result. There have been a lot of policy choices that have led to for example bubbles in the U.S. real estate sector, and the finance people then bundling those assets in imprudent ways and causing a financial crash. There's also in the United States there's a structural problem that they seem even with zero interest rate essentially people seem unable to get beyond the obstructionism of some of the legislators to fund infrastructure. Which has never been something that was controversial in the past but if you have zero interest rates then that means you can borrow this money for no cost. You can build roads and bridges and trains and transmission wires and power grids and everything else at zero interest.

Chris Nelder: You'd be crazy not to do it.

Jonathan Koomey: All that stuff will yield immediate job benefits, but it will also yield economic benefits down the line because the infrastructure will be up 80 won't have crumbling bridges you won't have potholes. You basically do the work that you need to do to keep the system running well, and so those kind of choices, the fact that we haven't made those investments I think have a lot more to do with secular stagnation than oil prices. I'm sure they contribute in some way but I just don't think that they are the only or even the most important cause and then the last thing I would say is then we need to get off of oil. And if we get off of oil then who cares. Let's transition to electricity for light vehicles. There's no reason why we can't do that.

Chris Nelder: Yeah I agree. I mean I take your point that this concept of secular stagnation might just sort of be more of an effect and a result of a failure to invest. But I also think there's potentially something to the idea that that the cost of running things has just gotten very high and that's naturally acting as a brake on the economy. And if you're standing there on the break even a low interest rate isn't enough to get the car moving again.

Jonathan Koomey: Yeah. So I think what you're pointing to is a great Ph.D. dissertation for someone who is really good with econometrics. And it's a kind of empirical question that somebody ought to be able to learn something about and say something sensible about.

Chris Nelder: I think so too. And you know what. That's what this show is all about we're just constantly throwing out ideas that would make great PhD dissertations and I'm really hoping, no seriously. And I'm really hoping that there are students out there listening to this stuff going. That's it. I'm going to do that. So OK, so to pop the stack of questions. And that brings us back to forecasting. So what are the main problems we have with economic and energy forecasting, why are we so bad at it?

Jonathan Koomey: I think that we are trying to do physics but we're really doing economics and economics is not physics. And the assumption behind physical systems is that there is a structural constancy to them, that you know the rules, the laws of physics as far as we know are the same here, as in Russia. The same here as as on Omega Centuri. And so that kind of structural constancy allows for reliable forecasting in a certain context. If I understand Newtonian mechanics I can predict when Jupiter will appear to rise from any point on the Earth and that going you know thousands of years into the future. So we can do that with high reliability, but you can't really do that with an economic system because you don't have structural constancy. Let me give you an example: so in 1973 when we had the first oil shock what happened was people and institutions changed their behavior. They changed how they dealt with energy, they change how they budgeted for it, they changed how they invested to deal with it and they changed their decision making. And if someone was running a model, which many people were at that time, trying to predict what would happen, they would be using parameters characterizing my decisions, and your decisions, and company decisions using data that came from the 60s. But of course data that came from the 60s are not a valid representation of what happened in the 70s because all hell broke loose in the 70s. Things changed a lot, and it's just, that's the best example I can think of where structural constancy, the assumption of structural constancy gets the modelers into big trouble. And so I just think fundamentally we're not dealing with physical systems, economic systems do not have structural constancy, elasticities change all the time with technology shifts and price changes and big shocks. And so those kind of changes make it very difficult to do accurate forecasts. And then there's one other big reason why it's hard to do accurate forecasting which is pivotal events. So let's say some giant proto-planet comes wandering into our solar system and changes the orbit of Jupiter. Well that could throw that prediction that we made it out when Jupiter was going to rise in 2021 completely off. And so it would cause a lot of other problems too it turns out. But the point is that that kind of unpredictable pivotal event like September 11th or the 70s oil shocks, or you know any of a number of other things...

Chris Nelder: Disco.

Jonathan Koomey: Disco - right. You can't predict these things. You can't predict when they're going to happen. You might be able to think, if this happens what would I do. Which is what the Shell scenario planners did in the late 60s which allowed them to respond better to the oil shock because they had already thought about it. And that's where it where I think we need to be going with this kind of modeling, we need to be thinking about ways to create scenarios and narratives that illustrate something about a certain path, and then allow us to think through how we would respond if a certain path comes to be. So the Shell scenario folks were able to imagine a world in which there was a big oil shock. OK what would we Shell do about that. And then they will have thought through what they should do so they won't be making decisions completely in a crisis mode. They'll actually have done some systematic thinking before. And that's where I think we need to go with this. We have to admit that we can't predict the future. Structural inconstancy, pivotal events make it impossible to do that. So let's figure out other ways to use these tools to allow us to think about the future so that we're prepared for whatever comes.

Chris Nelder: Yeah, you know it's often occurred to me that it really would have been great if we had just continued with the effort that was started in like you know 1970 with the Limits to Growth study. I mean there they were actually trying to take a multi-disciplinary systems approach, allow feedback loops between all parts of you know the economy. And I think that if we had continued with that kind of comprehensive modeling effort for the last 40 years we might be in a very different place today.

Jonathan Koomey: I think there's been a lot of resources expended on modeling, some of it is very multidisciplinary and very expansive in what it's trying to do. I think that the modeling community itself is insular. I think they believe in their heart of hearts that if you give them enough funding, and coffee, and graduate students that they can predict the future with accuracy. And the problem is that they need to accept that that isn't true, which is the kind of core underlying premise to their entire way of doing work.

Chris Nelder: So what can we do to improve our forecasting ability. I mean sometimes it seems like we like we have to do nothing less than completely break the grip of the priesthood of very serious people and economics and energy and other disciplines. And you know let a little fresh air into the world of forecasting, I mean I see a lot of very serious researchers out there modeling for example the future of transportation and assuming that future oil prices will follow a linear curve with zero volatility and if you look at the recent past well that's absurd. I mean even more absurd are the fuel consumption forecasts embedded in the IPCC carbon emission scenarios for example. But those beliefs are guarded by another priesthood of researchers. How can we transform the forecasting so it's more of a real I don't know a meritocracy? a la the efforts to find so-called super forecasters?

Jonathan Koomey: Yeah, I don't know whether you're going to have much luck with that but I would bring back this quote from Alan Kay, who was a pioneer in the computer industry. He said, "The best way to predict the future is to invent it." And to me that is the most succinct and powerful way to summarize the idea that we're creating the future. That the forecast can be helpful in kind of mapping the terrain, but that ultimately we have to decide the path that we want to go down and the choices that we make affect the options that we will have later. So if we decide to invest in nuclear power now, that will mean that there will be potentially the option to produce more nuclear power later, if we decide not to invest that much in it to build that capacity. You won't have that opportunity in 30 40 years but you may have the opportunity to do a lot more solar, a lot more wind, and that may drive the cost down. So there's all this path dependence that I think makes it imperative that forecasters, that they be modest and they understand the limits of their predictive power.

Chris Nelder: I couldn't agree more with that. So how do you think. How do you think the energy transition will affect the economy in the coming decades. I mean if the best we can do is to figure out, you know are we going down a crazy path or is this a worthwhile endeavor. We have to have some idea of how this could affect the economy. We don't have to be precise about it but we have to know that it will make sense or, or it will be very risky. I mean what's what's your sense of that?

Jonathan Koomey: Well the modeling exercise that I have seen or participated in all are premised on the idea that we can continue to deliver the same level of energy services as we have now and we expect to have decades hence. And so I think the most likely outcome for this transition is that there will be big winners and big losers. We know that the fossil companies are going to lose out, tens of trillions of dollars of market cap and assets that they counted on to exploit. They're not going to be able to exploit. On the flip side there's going to be a lot of winners of companies that decide they're going to be the ones who are investing for the future. And so I think there will be a lot of kind of tumults, there'll be a lot of changes. The actors who are most important now may not all be the same. The actors who are most important in 10 15 20 years. But I just don't see this as being fundamentally disruptive except if you're a fossil fuel company then it's really disruptive, but for the economy as a whole, there's no reason why we can't have cold beer and warm rooms and hot cooked food in the same ways as we'd grown to expect them over the years. We can do all of that with new technology just like we've done it in the past and we'll do a lot more of it with capital versus fuel expenses in the future. I think it'll be a lot more displacement of energy fuels with capital. And so that's part of the big picture transition that's going to happen, that means more fixed costs. That means we're going to have to modify the way we structure markets and the way people are incented and how they pay for different services. We're going to have think about new business models. That world is a totally different world, the world of high capital cost zero marginal cost resources is a totally different world and there will be a lot of change there. And even just that you know the power markets, they're kind of assuming that there is a marginal cost. But let's say you have a system that has pumped hydro, solar, wind, which I think Hawaii will get to. That's zero marginal cost and marginal cost doesn't have a meaning. Right you have to start to think about how we reward the people in the power sector so that they give us, society, what we want which is reliable electricity and the short run marginal cost paradigm may not be it.

Chris Nelder: Yeah. Now that leads into a whole topic about performance based regulation and so on that we touched on a couple of previous episodes and we probably shouldn't go down that rabbit hole today. But one thing that I still sometimes wonder about I'm not as worried about it as some people are but sometimes I wonder if in order to get to kind of this completely renewable future, or almost completely renewable future let's say an 80 percent renewables or something like that, if we will run into either shortages of raw materials or shortages of capital or even shortages of fuel to build it all because it is a pretty big infrastructure build out there we're talking about here.

Jonathan Koomey: Yeah I'm not sure if you look at the deep decarbonization work that E3 did. And almost any other study that's been done that I'm aware of, the numbers on costs are typically kind of you know half percent to 1 percent of GDP over some period.

Chris Nelder: So negligible.

Jonathan Koomey: So not you know important like in billions of dollars it's a lot less compared to the whole economy. It's not. Those estimates are pretty conservative when it comes to the learning rates that we've seen in solar and wind and efficiency. And so it's very likely things are going to cost less than that. I think there's an inherent pessimistic kind of high cost bias to many modeling efforts. Perhaps most. And so this is a whole other big area of discussion that we may not want to get into. But I think that you know the the IPCC modeling efforts and a lot of these other efforts there are rigidities built into these models that are not inherent to the economic system and that would not pertain if we actually go down this path in a real economy.

Chris Nelder: Yeah. Okay. Well I think we should talk about that. But in another podcast. We'll invite you back to talk about that. So before we wrap this up, I just want to talk a little bit about climate because you've also done some work on that. It amazes me how many different subjects you've covered here. So you've done some work on looking at the carbon emissions of data centers and I.T. So let's just start with by putting some numbers on that. How much energy are data centers and IT generally using now.

Jonathan Koomey: So data centers, we just did a report, Lawrence Berkeley Lab in collaboration with Eric Massenet at Northwestern. And other researchers at Carnegie-Mellon, me, we looked at data centers and total electricity use for data centers is about 1.5% of US electricity use. So those are the typical numbers for the U.S. and the World, kind of 1.5%. The other IT, so like your laptops and your printers and things like that. You know it's you know perhaps a couple of more percentage points, 2-3 percentage points added to that. And so the numbers that you see for all IT are typically in the kind of you know 3-6% range. And for some reason people have this I don't know whether it's an incentive or a tendency to believe that IT uses a lot of electricity. Part of that is just the coal industry wants you to believe that. So they spread stories like that which you know that's a whole nother set of stories that we can go into. But part of it is also people see that T. is important to the economy, so they think it has to also be important in the electricity sector. And it just isn't. And I think if you look at data centers as a high value activity that 1.5% of electricity is actually a pretty good use of that electricity because it allows us to have you know individuals to have these services and it allows companies to restructure, and to be more efficient and do all sorts of other... For example this podcast we're doing it over Skype. Didn't have to travel anywhere. In the privacy of my own home. So those kind of things turn out to be pretty important. And I think are likely to be more important than that relatively small percentage of electricity associated with data centers, and the surprising thing from this recent LBL work is that data center electricity hasn't really grown much since 2007. So that's pretty striking.

Chris Nelder: That is striking because the actual use of the Internet and you know IT devices, laptops and you know handhelds and everything has exploded since then.

Jonathan Koomey: Yes, so the services being delivered have gone up a lot. But the electricity use really hasn't gone up much at all. And the reason is that there is a lot of inefficiency in the way that data centers deliver those services. And what's been happening is a shift towards what's called virtualization. So they're able to put workloads on machines so that you have multiple operating systems running on one server and then you can move the operating systems or the applications around in real time, depending on how you can get the highest utilization and so on. So there's that. And then there's also a shift towards these much more efficient cloud computing facilities. So Google, Microsoft, Facebook, Amazon they're able to build these very cost effective highly efficient giant computing facilities. And you as a consumer are able to benefit from that because the costs are really low for those services but also for companies that shift work towards this public cloud or you're using the same technology or you know renting out some of this technology from these other companies they're able to capture a lot of those efficiencies and they're big, they're big you know we're talking about factors of 2 to 10 compared to the normal provision of IT within an enterprise.

Chris Nelder: That's really interesting. So I guess there's really no concern about being able to continue to service the load of electricity demand from data centers and IT in the future if it's that small and if it's...

Jonathan Koomey: It's single digit percent, you know if you just as a simple round number just call it you know 4-5% of electricity and well it's probably pretty good use of that electricity.

Chris Nelder: I mean I'm sure on a on a regional basis there are much bigger numbers right? Like a data center out in North Dakota probably consumes a great deal of that region's power for example.

Jonathan Koomey: Yeah there can be specific parts of Oregon probably where they're similar sorts of things. And so that's just a question of the local transmission distribution grid. Whether they can meet that, but it's pretty constant load. So it's relatively easy for the utility.

Chris Nelder: Interesting. Wow, that was a fast conversation there. Thanks Jon. I really enjoyed that. We'll definitely have to have you back to talk more about... talk about everything, yeah, I mean we only covered like a handful of the topics that you research today. So we'll just have to revisit all of that next time around.

Jonathan Koomey: Sounds fun. Thanks very much.

Chris Nelder: Thank you.