Photo: Peterson Institute for International EconomicsTrevor Houser is a green energy wonk’s green energy wonk. In the last few years, he has produced, among other things, congressional testimony on greening the stimulus bill, an astute take on the market failures around energy efficiency in buildings (PDF), and a comprehensive assessment (PDF) of the Copenhagen Accord and the state of international climate negotiations. He also has a post on Grist taking on the CBO over jobs numbers.
Houser is the head of the Energy and Climate Practice at the Rhodium Group research consultancy and a visiting fellow at the Peterson Institute for International Economics, where last week he and his colleagues produced the first comprehensive assessment of “The Economic, Employment, Energy Security and Environmental Impact of Senator Kerry and Senator Lieberman’s Discussion Draft [of the American Power Act]” (PDF). It got some good press coverage (including a post from me), thanks in part to its rigor and in part to some intriguing conclusions.
I chatted with Houser about some of those conclusions, and about the features and flaws of the economic modeling that so dominates climate policy discussions. (Yes, it’s wonky.)
Q. How does your analysis differ from the ones we’ve seen from the Environmental Protection Agency (EPA), the Congressional Budget Office (CBO), and the Energy Information Administration (EIA)?
A. Our analysis is similar to the EIA’s, since we use their model, with one important exception. We treat the change in investment that comes from pricing carbon differently than they do.
EIA’s model, as well as EPA’s and others, assumes that over the lifetime of the program, changes in investment between clean energy power generation and fossil fuels will net out, so they don’t focus a lot of attention on quantifying the changes in investment that you see in a high-carbon vs. low-carbon future. In the U.S., those changes are pretty significant.
In an emerging economy like China or India, the choice is between, Do I build a new coal-fired power plant or do I build a windmill? In the U.S., the question is, Do I continue with a coal-powered energy plant (95 percent of which were built before 1987) or do I replace that coal fleet with low-carbon generation? This is slightly a U.S.-specific case, but when we model something like the American Power Act, we see a $20 billion average annual increase in power-sector investment, from a baseline of $18.5 billion to $41.1 billion. So we capture the employment impacts of that change in investment.
Q. Is that the main reason your analysis shows a net employment increase, which is different from what most models show?
A. Right. If an economy is operating at full employment — if we had 5 percent unemployment instead of 10 — economic theory suggests the increase in investment would be inflationary. It would result in higher prices of goods, because you’d have competition for capital and labor, trying to draw them away from other industries. But that’s not the case; we’re at 10 percent unemployment.
So what’s most different about our study is we analyze the impact of energy and climate legislation on the U.S. given that we’re in recession. The conventional wisdom has been that we can’t afford to do this now because the economy is emerging from recession and unemployment’s still high. Our finding suggests the opposite: Energy and climate policy can have a stimulative impact on the U.S. economy in the first decade, albeit a modest one.
Q. How long is the U.S. expected to stay in recession? Are there official projections?
A. We use the projections from EIA, which come from the consulting firm Global Insight. They see a slow recovery in the U.S. economy and an even slower recovery in employment. We’re not staying at 10 percent for the next decade, but we don’t get back to 5.5 percent unemployment, which is what they see as full employment, until 2020.
Q. Over the years, environmentalists have charged that EIA’s model radically overestimates how much petroleum will be available (among other complaints). Do you share any of these familiar criticisms of models like EIA’s?
A. Long term, there are a bunch of problems with all models. They all fall short in their presentation of reality for different reasons. Their robustness in capturing long-term global petroleum supplies we felt was less of an issue for this analysis because we’re looking at the relation between the policy and the baseline, particularly in the next decade or so.
Long-term oil supply and oil prices that EIA projects currently are not terribly different from what the International Energy Agency forecasts — fairly tight global oil markets. There doesn’t seem to be a ready alternative in terms of a modeling framework if you have a different point of view about long-term crude oil supplies.
Q. Another familiar criticism of typical macroeconomic models is that they don’t do a very good job of capturing the benefits of energy efficiency. Did you do anything in your analysis to try and compensate for that?
A. CGE [computable general equilibrium] models — the equilibrium models that EPA uses, for example — do the poorest job of capturing the potential of energy efficiency. They assume that investment is naturally flowing to its most productive use. NEMS [National Energy Modeling System], the EIA’s model, is a little bit different. It’s an engineering model, so you can actually change the efficiency of different types of technologies through codes and you can see the resulting change in energy prices and how that mitigates the impact for consumers.
What it still doesn’t capture is if investment in energy efficiency is providing a higher rate of return than an investment in other sectors of the economy, what that does to long-term growth in the capital stock — you have investors making more money, which they can then reinvest in other parts of the economy. So that part we weren’t able to capture.
We did model transportation efficiency provisions in the bill, the building codes that Sens. Kerry and Lieberman have indicated will be brought over from the ACELA bill out of the Senate Energy Committee, and the use of allowance revenue for investments in commercial and residential building efficiency. That does important work in terms of mitigating the price impact on consumers. But the broader fact you’re talking about — the kind of change in the productivity of investment — we don’t capture.
Q. One of the most surprising outcomes of your analysis — Michael Levi commented on this — is the large percentage of new employment in the next 10 years that comes in the nuclear sector. What explains that?
A. There two things, primarily. The first is that we see more deployment of renewables than anything else between 2011 and 2030. There’s 106 GW of renewable capacity, driven entirely by the cap.
Only 24 GW of that is additional above business as usual because our business as usual includes the [American Recovery and Investment Act]. In the EIA analysis of the stimulus bill, you get pretty healthy increases in renewable energy out to 2020. So between now and 2020, which is the window in which we asses the jobs impact in detail, you don’t have much renewable energy growth above the baseline (although you have very aggressive renewable energy growth). The baseline assumes that stimulus provisions are successful in deploying fairly large amounts of wind power and solar power.
Q. So there are renewables jobs, they are just already incorporated into the baseline.
The second thing is that our modeling shows pretty big nuclear capacity deployment — 78 GW of nuclear over the next two decades, and 68 is additional to business as usual. That’s for a couple reasons. First, there’s the capital cost assumptions around nuclear power that EIA uses. There’s a lot of debate about what the right capital cost of nuclear power is. We don’t have a lot of good reference points.
We used the 2009 version of the Annual Energy Outlook. Two weeks ago, EIA released the 2010 version. We didn’t have time to run the model on that full version, but the one thing we ported in was its capital costs, because they had made a significant revision in nuclear power costs — 16 percent revision in the capital costs of nuclear, upwards relative to 2009. Even with that, you still see a lot of nuclear power, in part because of an increase in loan guarantees of $36 billion and a 10 percent investment tax credit for nuclear power out to 2025. There’s a lot of nuclear incentives in the legislation.
Now, things the model doesn’t capture could very well slow the deployment of nuclear power — for instance, supply-chain restraints for labor or equipment. That’s not something the model’s equipped to capture. We’re going to do some alternative scenarios in the next couple weeks where we put exogenous constraints on nuclear deployment.
There are a lot of assumptions in the EIA model that we don’t think are correct. EIA’s model is very public and frequently reviewed; the assumptions are hotly debated. We decided on our first cut to just use the EIA’s assumptions, with the exception of the treatment of employment because we thought that was a pretty glaring shortfall given current underemployment. But we didn’t want to make selective changes to cost assumptions in the code.
In subsequent analysis we’re going to do that. There are big capacity additions in our analysis from nuclear and CCS. But there are uncertainties about both of those, in many ways greater than the uncertainties surrounding efficiency and renewables. They will have a big impact on carbon prices, electricity prices, rates of deployment, and overall employment.
Q. If it turns out nuclear and CCS are more expensive and deployment is more limited than reflected in the current analysis, what’s the effect?
A. You see more renewables, a higher carbon price, and an increase in international offsets.
Q. You said you’re going to do another round of modeling, suggesting ways the bill could be improved. Can you give us any kind of preview of that?
A. We’re still in the midst of all of that, so I don’t want to speak out of school too much. The one thing I’d give a nod to is: If employment creation is a policy priority, there are a number of things the drafters of legislation can do that will significantly improve the jobs outcomes relative to what we see with the current draft. Most important is how allowance revenue is used to offset the impact to affected consumers and affected businesses.
In the early years, the primary mechanism for offsetting the impact to affected consumers is free allocation to LDCs [local distribution companies], which is supposed to be used for the benefit of ratepayers. Assuming that it is in fact used for the benefit of ratepayers, that prevents some improvement in energy efficiency that you would otherwise see. Consumers aren’t seeing higher prices, which increases the overall economic cost of the program.
If your focus is on employment, remember, a significant amount of allowance revenue is going for commercial and industrial sales from LDCs. If you use that allowance revenue for a rebate of the payroll tax, you create a significant incentive for employment that would offset the impact of higher energy prices — through wages, not through a check to consumers.
Those are some of the tradeoffs that we’re going to explore. We’re going to look at ways the bill could be more successful in addressing energy-security concerns and discuss some of the issues surrounding the long-term environmental integrity of the legislation given where the price collar is.
Q. I hope you make the point that there’s much more efficiency to be had.
A. Right. What is really shocking to me, if you look at our results, is how little overall energy demand is reduced. It’s reduced 5 percent below business as usual by 2030. That’s because the barriers to energy efficiency are non-economic barriers, so adding another 5 percent to electricity prices is not going to make the average household run out and buy a more efficient air conditioner. What that means is, there’s economic loss there — you’re forcing the power sector to make investments that have 10-year payoffs instead of households making investments that have three-year payoffs, because of those market barriers. So yeah, we will certainly focus quite a bit on that in the forthcoming work.
Q. Thanks to the influence of the Cantwell-Collins cap-and-dividend system, more allowance revenue is being directly rebated in the Senate bill than in the House bill. For the most part that money got taken from the weakest lobbies, generally renewable energy and energy efficiency. What’s the macroeconomic effect of that?
A. I don’t have the answer for you now, but that’s something we’re looking at. I think the core point is probably right. As a good economist, I should be ecstatic about something like the CLEAR Act. It puts a simple price on carbon and rebates it to consumers.
The problem is, the energy space if rife with market failures, so an economically optimal policy needs to do two things in addition to pricing carbon. It needs to take care of efficiency-market failures, which are at the left hand of the cost curve — the stuff we’re not taking advantage of even though it’s profitable today. And it needs to deal with R&D for the stuff at the right hand of the cost curve, so that when carbon prices get up to $80, $100 a ton, we have mitigation options like CCS or advanced vehicles that the private sector is not necessarily going to be able to deliver.
Whether or not the American Power Act gets the balance right is a separate question, but I think it’s important that any piece of [climate] legislation have those complementary measures in them.
Q. But that increases the page length!
A. Our litmus test for legislation that will fundamentally alter the behavior of the U.S. energy sector — a $2.2 trillion part of the U.S. economy — for the next 40 years is whether or not it can be kept under 100 pages?
Q. It gets lost in the debate that in the House bill, ACES, the bottom 20 percent of the income scale came out ahead.
A. The problem is that our frame is, how low can we get household expenditures? If households — the middle 50 percent of U.S. households, not the lowest deciles — are not willing to pay an extra $100 to $200 a year in higher energy prices on top of the already $5,000 to $7,000 a year they pay, then we just can’t get this thing done. Trying to minimize that number to zero is a losing strategy. Paying $100 to $200 extra a year now is going to allow us to make investments that prevent us from paying $600 to $1,200 three decades from now.
Q. EPA is analyzing the bill now. Any chance they’ve been in contact with you? Any indication your approach might be making headway?
A. EPA can’t use our approach — a full employment model by definition can’t really account for the impact of an increase in investment during a recession.
Q. That’s just dumb, though. Is it not just dumb?
A. EPA would probably say that 200,000 jobs in a 135 million employee labor force is noise, which is not the politically astute answer but is the economically correct answer. Their model is there to look at 50-year trends. It’s not supposed to be a near-term macroeconomic model.
I think CBO will look at this. I don’t know whether they will do it quantitatively or just qualitatively, but I think they will explore changes in investment. In the past they’ve always assumed no change in GDP or employment when they model the bill. But they recently put out a literature review of studies that look at the employment impacts of pricing carbon, which suggest they are ready to wade into this debate.
EIA has a work item on this, on how to better capture changes in investment in the power sector. It’s coming up with a methodology for doing it that’s robust … it’s taken them a little bit of time. But they’re looking at this approach as well.
Q. One more nerdy question?
A. Dave, I learned what I know about CBO scoring from you, so the nerdier the better.
Q. Under traditional full-employment models, if government takes money out of one part of the economy and put it back in another part, it reduces economic efficiency and incurs transaction costs. According to these models, government literally can’t act without creating a drag on the economy.
A. Always. By design.
Q. One way of getting around that is your way, which is by taking into account recession. Is anybody out there challenging it more directly, by arguing that in some circumstances it is possible to determine that the economy is not deploying resources efficiently and correct that with tax and regulatory policies?
A. There’s four areas. First is what the energy-efficiency literature focuses on, a whole category of market failures. The second is regulatory reform. There’s some work on RES’s [renewable energy standards] in the West that show net-worth enhancement because the RES forced reform of utility monopolies in a way that was welfare enhancing. The third is infrastructure investment.
The fourth one is R&D. There’s a lot of literature on the welfare-enhancing benefits of R&D, because over the long term, you can only grow your economy in three ways: you pile on capital, you pile on labor, you change technology. Nothing else you can do. Those first two variables are somewhat fixed, barring immigration policy changes. So it comes down to technological change. That’s the hardest to quantify but potentially most powerful economic impact of pricing carbon and all the complementary polices that come along with it: Does it force innovation? It’s a really tough thing to model.
If you can’t improve the model, then you need to be very upfront about its shortcomings. That was the biggest weakness of the CBO report on employment. I understand the weaknesses of our existing models in capturing some of these employment effects, but people who don’t work with models don’t understand how frail they are. They think they are definitive answers on what the economy will look like in 20 years. They think models are science, when they are tools for evaluating possible scenarios, and flawed tools at best.
Stronger, more upfront statements about what models can do and can’t do should be on all of these reports right up at the top. The press will grab what the press will grab, but a clear acknowledgement of the shortcomings of models would be useful.