U.S. EIA’s International Energy Outlook 2023

Available Downloads

This transcript is from a CSIS event hosted on October 11, 2023. Watch the full video here. 

Joseph Majkut: I’m Joseph Majkut. I’m the director of the Energy Security and Climate Change Program here at CSIS. And I can’t tell you how excited I am for today’s event.

This is an event that CSIS hosts on a biannual basis and it’s one of our – one of the things we get most excited about at a staff level. This is a benchmark product not just for the United States, but for the world as we try to consider and plan for our energy future. And I have to think that, like, the task that our colleagues at the EIA are approaching gets harder and harder every year as we think about the pace of prospective energy transition. As we think about the challenges that accompany meeting our energy needs, documents like this provide guidance. They provide insight. And in particular one of the things we’re going to hear today is that they are not meant to be forecasts; they’re meant to be stories. They’re meant to be scenarios that help us understand and plan for various outcomes that could come about. And I’m really looking forward to our discussion.

Before we get into the formal program, for those of you in the room I do have to say if anything happens, if there is an emergency and an alarm goes off, exit signs are well-lit. Please follow the instructions of a CSIS staff member. Our rally point is around the back of the building in a counterclockwise direction when you leave the front. And I am obligated to buy you all ice cream in the event we have to evacuate. (Laughter.) So thank you very much for joining us today.

For colleagues online, you are welcome as well. If you’re watching live, there should be a question box on the event page that will allow you to interact with me and Director DeCarolis later in the – in the program.

The brief run of show. We’re going to see some of the key outputs of this new IEO this morning from Director DeCarolis and his colleague Angelina LaRose. Then Joe and I are going to have a few minutes to talk about how policymakers, how industry people, how civil society and the interested public should look at the IEO and where it fits into the broader, like constellation of energy forecasts and planning documents that we have available to us.

Thank you again for joining us. And, Director DeCarolis, if you’d like to come up, I think we give you the floor – the podium to you, sir.

Joseph Decarolis: All right. Good morning, everyone. It’s an honor to be able to kick off our presentation of the International Energy Outlook for 2023.

And before I get started, I just wanted to give a huge thanks to the EIA staff. There’s just an enormous amount of effort that goes into doing the analysis and delivering the report. So I just want to thank them. And I also want to thank Joseph and the CSIS staff for hosting this event.

I want to talk a little bit about what EIA does. So we’re the Energy Information Administration and we’re the statistical and analytical agency within the Department of Energy. Among other things, that means that we’re vested with the unique authority to collect energy data from U.S. industry, and that data we publish. It’s used by a lot of different stakeholders. But one of the things we do is we use that data internally to inform our model projections. And so we view our data as a critical source of information within the United States.

The other thing is that by law all of our products are independent of approval by any other officer or member of the U.S. government, and we really prize that independence. It’s very important to us. This particular product, the International Energy Outlook, explores long-term trends in energy supply and demand around the world.

  1. So what’s new in this year’s report? Well, if you read our Annual Energy Outlook this year, you’ll see a similar look and feel. You’re going to see improvements in the narrative. So I have a background in modeling myself, and I think it’s really important that analysts and modelers be able to explain the model results and translate them into a real-world context. And the only way to do that is to focus on the – on the narrative. So there’s that.

You’re going to see technical notes that will appear as light-blue boxes within the – within the narrative. And that’s an opportunity to take a deeper technical dive on particular subjects for readers who might be interested. So in this IEO we’ve got three of those technical notes. There’s one focused on our electric-vehicle projections, our representation of storage in the electricity sector, and then our representation of global refineries.

We also emphasize the range of results around the cases that we – that we model; again, similar to the AEO. In this IEO, we’re also introducing new cases that focus on examining the capital costs associated with zero-carbon technologies.

And then we’ve also made several improvements to the modeling work itself. We have new analysis regions, so we have 16 different regions, and we think that they offer better geographic alignment. We have a new – a brand-new oil and natural gas supply module that’s now been incorporated into our modeling framework, so you’re going to be seeing the results from that. We’ve increased the temporal resolution in our electricity model so we now model electricity supply and demand across 288 separate time slices, and that’s really important to model wind, solar, and battery dispatch. And then, finally, we’ve made several assumptions about the impacts of Russia’s full-scale invasion of Ukraine, and so you’ll see a whole appendix where we go sector by sector and talk about what assumptions we’ve made and how it’s informed our analysis.

I want to talk a little bit about the cases that we model in the IEO. The first thing I want to point out is that we only model current laws and regulations, and we freeze our consideration of current laws and regulations as of March of 2023, OK? And this is looking at policy around the world. For the United States in particular, the results that we present are from our Annual Energy Outlook, so the results will be the same. And those policy assumptions were frozen in November of 2022, OK? And these assumptions about policy carry through all of the cases that we model.

  1. Getting on to the specific cases, we have our reference case. In our reference case, we’re assuming that global annual GDP grows at 2.6 percent annually. We assume that the price of Brent crude starts at $102 a barrel, it varies a bit, and then it ends up very close to $102 a barrel again in 2050. We also assume that there are reductions in the cost of zero-carbon technologies – and I should mention those technologies are wind, solar, batteries, and nuclear – and they can achieve cost reductions up to 20 percent in the reference case. And the way we handle this is through technological learning. So every time the capacity of one of those technologies doubles, there’s a reduction – there’s a reduction in cost.

The second set of cases focus on economic growth. So in the low case, we assume that the average annual GDP growth rate is 1.8 percent. In the high case, it’s 3.4 percent. And I will say that these macro cases have a big impact on the results, because when you have high macro growth that leads to high end-use demand, and that in turn requires higher supplies of energy. And then the converse is true: When you have lower macro growth, it means lower end-use demand, and therefore less energy supply is required.

Next, we have our oil-price trajectories, right? We have a low and a high. These are formulated outside of the model framework and they become an input. So in the low case, we assume that Brent crude spot prices reach $48 a barrel in 2050; and in the high case, they reach $187 a barrel in 2050.

And then, finally, again, the zero-carbon technology cost assumptions. This is new to this IEO. In the low case, we look at what the cost was that was achieved for each of those technologies in 2050 and then we – in the reference case – and then we look at a trajectory that takes us 40 percent below that cost. And in the high case, we just assume constant cost through the projection horizon.

These are the highlights from this year’s report. The first is that increasing population and income offset the effects of declining energy and carbon intensity on emissions, and I’m going to be providing you more details in a moment on that. The second is that the shift to renewables to meet growing electricity demand is driven by regional resources, technology costs, and policy. And finally, energy security concerns hasten a transition from fossil fuels in some countries, although they drive increased fossil-fuel consumption in others. I’m going to give, you know, some high-level view of that, and then Angelina’s going to add some additional detail.

There are some things that I want you to keep in mind. It looks like – OK. There’s some things I want you to keep in mind as you – as you look at these – at these results. When we model these cases, we take a deliberately restrictive approach in the IEO. The way that I would characterize these cases is that they’re plausible but sober, OK? And here’s – here are some of the assumption we make.

As I said, we only account for current policies, and we’re looking at policies that are legally enforceable, right? So we don’t model aspirations or targets unless there’s a legally enforceable policy that backs it up. Second, we look at evolutionary rates of technological change that are based on recent history. And then, third, we don’t consider sweeping changes and consumer preferences or major geopolitical events that could produce durable change and, you know, shift the trajectory of the system.

Now, it’s certainly possible that those things can happen, right? It’s entirely possible we could get new policies. There could be unforeseen geopolitical events or technology breakthroughs that we don’t model. And that’s exactly why you should not think of the IEO as a forecast. We’re trying to do something different here. What we’re doing is providing a set of policy-neutral baselines that focus on the current trajectory of the energy system, and I think that that actually provides a really useful point of reference for decision-makers against which they can judge future actions and developments.

  1. So I’m – this brings us to our first set of results. Across most of the cases, we find that energy-related CO2 emissions continue to rise through 2050 under current laws. So, again, we have a regional model, 16 regions. I’m showing the results aggregated up to the – to the global scale. Just to orient you, the line represents the reference case, and then the gray bands around the reference case show the full range across all of the cases that we – that we model. So, moving left to right, we have global gross domestic product, which is – which is – you can tell is growing very rapidly, OK? In the middle, we have primary energy usage; again, you can see an upward trend. And then, finally, to the right we have energy-related CO2 emissions. In each of these three panels, the bounds on those gray bands are due to the high and low macro cases, so they’re setting the bounds on each of these.

One thing that’s interesting is, as you look from left to right, you notice that the slope of the cone keeps decreasing, right? So GDP’s growing the fastest, then primary energy, and finally energy-related CO2 has the least slope.

So to get some more insight, we can further break this – those three panels down into four panels. And some of you might immediately recognize this as the terms in the Kaya identity. So, again, moving from left to right, we have global population. You notice there’s no gray band there because we assume one – (mic feedback) – oh, sorry – one population projection that carries through all of the cases that we’re – that we’re modeling. Next, we have GDP per capita, which you can take to be a rough measure of per capita annual income. Next, we have energy intensity, which is a measure of energy per dollar GDP. And then, finally, carbon intensity, which is the amount of carbon emissions you get per unit of energy. The uncertainty bounds in the two middle panels, again, are being set by the high and low macro cases. And for carbon intensity, not surprisingly, it’s being set by the zero-carbon technology cost cases.

If you take the product of the first three terms, you get total global energy consumption. If you take the product of the four terms, which is the way the Kaya identity is designed, you get total global CO2 emissions.

And you can see very clearly that in the first two panels – population and GDP per capita – they’re increasing at a pretty fast rate. And this makes sense. There’s more people, and as they gain wealth they tend to demand more energy-intensive goods and services. But you can see that at the same time we’re getting – we’re able to use less energy per dollar of value in the economy, so energy intensity is clearly declining. And then, for every unit of energy we consume, we’re emitting less carbon emissions. And that brings us to the insight on the title, which is that the upward pressures of population and GDP tend to outweigh the downward pressures that we’re seeing from reduced energy and carbon intensity over time.

Again, we can disaggregate even further, right? So this is just looking at GDP growth rates by region in the model. Again, there’s 16 regions. We have the high-income countries on the left. We have the low-income countries on the right. And I’ll just point out that India has the highest average GDP growth rate, but it varies quite a lot by region. So when I present the results at the global scale, just remember that there’s all of this detail under the surface.

We can also look at total energy and break that – disaggregate that a little bit. So here we’re looking at fossil fuels versus non-fossil fuels, and what we find is that increasing demand and current policies drive steady growth in fossil energy but even faster growth in non-fossil sources. So, again, you can see the reference case as a line. The bands represent the range across the cases.

With fossil, which is the black and gray, it starts at 505 quads in 2022 and then it grows anywhere from 1 to 40 percent, depending on the case. And then, with non-fossil – so this would include all the renewables and nuclear – we start at 133 quads, but that grows by anywhere from 70 to 125 percent by 2050. So wind and solar in particular are growing at a remarkably fast rate, and again, you can see that by this increased slope associated with the blue line.

  1. And then we can, again, break it down even a little bit further here, OK? So we’re looking at fossil versus non-fossil, but we’ve now broken that down to specific fuel types. So you see the stacked bar for 2022. The black outline represents all the fossil sources. The gray outline represents all of the non-fossil resources. So we’ve got 2022 all the way to the left, and all the other bars represents a snapshot across the cases in 2050. So you can see how each of these fuel types is changing over time.

Generally, what we find is that fossil fuels hold onto their share through time, but that renewables really pick up, and most of that renewable development is taking place in the electricity sector. So we find that if you look at the electricity sector, renewables plus nuclear combined represent roughly 55 to 65 percent of global electricity supply in 2050 across – again, across the cases.

Now, obviously, there’s a lot going on here. At the regional level, there are regional patterns that we’re not showing here. And that’s going to depend on prevailing policy, trade patterns, and then the cost of local resources.

  1. With that, I’m going to go back to the highlights and turn it over to Angelina LaRose, who’s our assistant administrator.

Angelina LaRose: Good morning. I’m going to keep on with the trend of breaking things down and go into more detail surrounding our three highlights. I do encourage you all to take a deeper dive into our website following this event. We have a lot of data and analysis we’re going to be posting related to the IEO.

But for now, I’m going to just walk you through some of the interesting findings of our main highlights, starting with the first one about increasing population and income driving the growth in consumption and energy-related emissions despite declining energy and carbon intensities.

So what this slide is showing is energy consumption by sector. As Joe oriented to you all, the solid line represents our reference case and the range – the area, the light-colored area, represents the range of projections from our side cases.

So end-use energy consumption grows across all sectors through 2050 and across all – (off mic). You can see on this slide the industrial sector, shown in green – which includes manufacturing, refining, and other sub-sectors – has the largest share of energy consumption and also has the largest share of growth through 2050. The industrial sector has the widest range of consumption across cases due to the broad range of industrial growth output assumptions across our cases and sectoral sensitivity to these macroeconomic drivers. The highest growth we see in the industrial sector comes from our high economic growth case, reaching a growth rate of 1.7 percent per year. But even in our low economic growth case, we see increasing industrial energy consumption.

So this slide is very similar to the one we just saw, but it’s just looking at liquids consumption across cases and sectors. So, similar to what we saw in the last chart with total energy across all fuels, liquid consumption also continues to rise through 2050, and the fastest growth in liquids consumption comes from our industrial sector. Petroleum is used as a feedstock in industries like chemical production and – as well as in diesel fuel and construction and agricultural equipment. So although transportation still maintains its largest share of liquids consumption, its growth is tempered by efficiency gains and the shift towards electric vehicles. However, overall there’s growth in liquids consumption driven by the growth in industrial and transportation.

So I’m going to take a closer look at the industrial consumption. This slide is looking – focusing on industrial consumption in China and India in particular. So across the globe, while we’re projecting increasing energy consumption in the industrial sector, we’re seeing declining energy intensity. At a regional level, there are two prevailing drivers of energy consumption in the industrial sector. These are industrial gross output, which is a measure of economic activity, and energy efficiency or energy intensity.

So we see different trajectories in China and India. As you can see on the left, in China some cases see industrial-sector consumption declining or leveling off, compared to India where we have growth across all the cases. In China, slowing industrial gross output after 2024, combined with significant energy-efficiency gains, slows and decreases industrial energy consumption in some cases. In particular, China is projected to significantly increase its production of recycled steel, which is significantly less energy intensive. In India, the growth in industrial gross output, which more than triples in most cases and even quintuples in the high economic growth case, is the main driver of that sector’s increase in India. India specifically sees growth in energy-intensive industries, so primarily metals, chemicals, and nonmetallic minerals.

So I’m shifting to the transportation sector, and there is a lot to unpack on this slide. (Laughs.) So I want to do my best to walk you through some of the main highlights. On the left is a graph of passenger travel demand in select regions. And on the right is a graph of passenger travel demand by mode, indexed to 2025. So, again, that graph shows an index to 2025, when we project travel demand to approximately return to pre-pandemic levels.

So, looking on the graph at the left, you can see that the overall travel demand is growing. Travel demand is highly sensitive to changes in disposable income per capita, as well as employment. And there are several interesting findings at the regional level. Much of this growth in travel demand is concentrated in India, where both disposable income and employment is growing significantly over the projection period. Regions with slower income growth, with – and with lower absolute income per capita like – such as Africa, continues to see growth in travel demand because of increases in employment, but to a less degree. In regions with both slower growth in income and employment, such as Western Europe and Japan, per capital travel demand growth is limited. Globally, travel demand per capita growth across all cases nearly doubles in the high economic growth case. So, combined with growing population, this is a strong upward pressure on energy consumption.

So, now turning to the graph to the right, travel demand for less-efficient modes of transportation, particularly light-duty vehicles and air travel, grows in regions as incomes increase. So rising income in several regions enables travelers to shift from inexpensive but more efficient modes of transportation, like two- or three-wheelers or buses, to more convenient but less-efficient modes, like light-duty vehicles. You can see this particularly in China, India, and other Asia-Pacific region, and you can see that in blue on that graph on the right where we have faster growth in LDVs – light-duty vehicles – as well as air, which is in the first two panels; and slower growth in buses and two- and three-wheelers, which you can see in the third and the fourth graph.

Aircraft travel, which is highly sensitive to changes in income, is noticeably increasing across all regions. In regions with slower income growth, such as Africa you can see in the yellow as well as the “other Americas” region – part of the green – use of two- or three-wheelers persistently grows compared with aircraft and LDV travel. Efficiency improvements within each of these modes/technologies offset a significant portion of energy consumption from this travel demand growth as well as from this shifting towards less-efficient modes of transportation.

  1. My next slide. (Laughter.) Imagine it. (Laughs.) Four panels. OK, here we go. See, it’s just as you all imagined, I’m sure. (Laughter.)

So this slide is looking at the share of electric vehicles in select regions, breaking out battery electric and plug-in hybrid electric vehicles. So the aggregate increase in light-duty vehicle travel leads the on-road fleet of light-duty vehicles to grow from 1.4 billion in 2022 to more than 2 billion vehicles by 2050. Within light-duty vehicles, we see a technological shift from internal combustion engines to electric vehicles. Globally, EVs will account for 29 to 54 percent of new vehicle sales by 2050 across cases in our projection. And the continued increase in EV adoption leads to a peak in the global fleet of internal combustion engine light-duty vehicles between 2027 and 2033 in all cases.

Looking at the EV adoption, battery electric vehicles – which you can see in blue – cannibalize plug-in hybrid electric vehicles – in yellow – particularly early in the projection period. And that kind of makes sense, as plug-in hybrids are – make more sense in the interim, but as battery costs drop out battery electrics are much more competitive. China and Europe, the adoption of electric vehicles is largely policy-driven, whereas in India and Japan it is largely based on economics – hence, the wider range of results.

So, like transportation, the building sector shows a lot of interesting stories at the regional level. So this slide is showing residential and commercial delivered energy consumption per capita in India. India, with its significant GDP growth and population increase, exemplifies the relationship between energy use, income, and service-sector growth. Disposable income and an expanding service sector support significant increases in overall buildings energy use, which almost triples by 2050 relative to 2022 across cases. Electrification on the buildings stock grows electricity consumption, increasing more than any other energy source in the residential and commercial sector.

Which leads us to our second highlight, which relates to the shift to renewables in meeting global energy demand. So this slide is showing the change in installed electricity capacity in 2050 compared to 2022 levels across cases. So to meet increased global electricity demand, installed power capacity increases, reaching a total of one-and-a-half to two times what it was in 2022 by 2050. Across cases, the 4,600 to 9,200 gigawatts of generating capacity installed by 2050 is predominantly solar, wind, or storage, which you can see in the yellow, green, and purple shares of the bars. From 2022 to 2050, zero-carbon technologies make up between 81 to 95 percent of new global generating capacity installed across cases.

So what this slide is showing is electricity generation by fuel. So, as expected given the capacity builds I just showed you in the previous slide, solar and wind show the highest levels of generation growth. However, while the last slide showed changes in capacity from 2022 levels, it’s important to note that existing coal and natural gas power plants continue to operate. In 2022, coal, natural gas, and liquid fuels combined constituted more than half of the world’s electricity generation capacity. By 2050, the share from these fuels for power generation decreases to 27 to 38 percent of the world’s generating capacity across cases, so it’s still a noticeable share. So while coal is flat or declining – it declines in most of the cases we modeled – natural gas is flat or rising. So the percentage growth is nothing compared to what we’re seeing in solar and wind, but both gas and coal remain a stable part of the generation mix.

That brings us to our last highlight, and that’s related to the role that energy security has in the transition from fossil fuels. So this slide is showing electricity capacity in select regions broken out by zero-carbon technology – so renewables and nuclear – and by fossil-based technologies. Across a large majority of regions, zero-carbon technologies increase through 2050 under current policies. There is regional variation, however, in the timing of that growth. While Western Europe and India’s growth – which you can see in the first two panels – in zero-carbon technology capacity is projected to accompany a flat to declining change in fossil fuels, China and Africa – shown in the third and fourth panels – see growth in zero-carbon and fossil-based technologies.

In Western Europe, energy-security considerations favoring the use of locally available resources such as wind and solar increase installation and planned builds for these technologies earlier in the – plus batteries – earlier in the projection period.

In India, rapid growth in zero-carbon technology is seen after 2030, heavily influenced by assumptions of economic growth.

In China, coal-fired generation makes up about 62 percent of the electricity generation mix in 2022 and decreases by less than 10 percent throughout the projection period in all cases, except the low zero-carbon technology case and the low economic growth case. Local resources are key in the Asia-Pacific region overall, where electricity demand growth is the most rapid and local coal is both cheap and abundant.

And so while Africa sees significant increases in installations of zero-carbon technologies over the projection, it’s accompanied by an increasing share of fossil fuels as the region takes advantage of locally available fossil-fuel resources and the absence of any uniform policy.

One fossil fuel in particular, natural gas, is an important part of the electricity generation mix in several regions of the world in our projection through 2050. The power sector and its continuing need for natural gas is a key component of the global trade dynamics, which is seen on the next slide.

So this slide is showing net natural gas trade. Above the zero axis represents natural gas imports. Below the zero axis is net natural gas exports. So notable on this slide is the range of imports from Asia-Pacific and the range of net exports out of the Middle East. These differences are largely driven by macroeconomic assumptions. Total global demand for natural gas differs significantly between the reference case and the high economic growth case. By 2050, global natural gas demand reaches close to 200 Tcf in the reference case and grows to over 240 Tcf in the high economic growth case, so about a 20 percent difference.

Most of the demand occurs in China, where consumption rises across all sector(s), particularly the electric power sectors in the later years. India also drives significant natural gas import growth in the Asia-Pacific region overall because of the growth in its industrial sector.

So, as regions run out of cost-efficient resources, they will import natural gas from regions that have the cheapest resources. So, as Joe mentioned earlier in this remarks, the IEO uses our most recent Annual Energy Outlook for the U.S. projections. So the U.S., which is clearly part of the North America grouping in this graph, supply of natural gas on this chart is what we’ve published in the AEO in March. And so although North America is the second-largest source of supply on a net basis from all the regions, there’s only a limited growth in supply between the reference case and the high economic growth case, as published in our Annual Energy Outlook. So given this lack of significant growth from the United States in the macro case and the limited growth from Russia, the Middle East’s role as a natural gas supplier increases significantly in the high economic growth case.

So that leaves – I’m going to leave us with our three highlights. I just want to thank you all for your attention, thank CSIS again for hosting this event, and thank all our analysts and modelers for all their work in making this the best IEO yet. Thank you. (Applause.)

Dr. Majkut: As we get settled, please let me thank Angelina for a really helpful description of the fundamental pieces.

You know, I said at the beginning of this that this document tells us stories and that the task ahead of the EIA is so hard. And that alighted the really fundamental and quantitative work that goes into modeling all these different sectors of the economy, all these different regions.

Joe, thank you so much for coming today. Colleagues online or colleagues here in the room, if you want to use the event page to offer a question I’m happy to kind of incorporate those on a rolling basis.

But I want to start with your early message. This is not a projection – wait, this is not a forecast.

Dr. DeCarolis: Forecast. Forecast.

Dr. Majkut: Right, but it’s a set of projections. And in particular, EIA has always had a practice of saying the base-case scenario is those laws that are on the books today –

Dr. DeCarolis: Yes.

Dr. Majkut: – which is separate from the ambitions of the – of governments around the world related to economic growth, you know, the kind of – like, the composition of the vehicle fleet, or greenhouse gas emissions. When you – when I kind of look at the results, I say, oh, man, I mean, this is kind of an energy addition and not an energy transition scenario. Is that the right key takeaway? And how do you guide or recommend global public-policymakers look at the scenarios that you’ve – that you’ve – or interpret the outcome of the scenarios that the EIA has crafted? Let me put it that way.

Dr. DeCarolis: Sure. Yeah. Great question.

I think it always comes back to the context. You have to – any time you’re looking at a modeling exercise or an outlook, it’s very important to understand, you know, what were the assumptions that went into it. And we’re taking a very particular approach here where we’re assuming a strict interpretation of current policy. Again, we’re assuming evolutionary rates of technological change, no major surprises. And so that informs the projections.

So the way that I – if I were to make an analogy, if the global energy system is a car, we’re basically looking at what happens when you shifted into cruise control. Where do we end up? Now, we do explore some of the key sensitivities in the models through different cases, but looking at it as a whole that’s kind of the perspective that we take.

Dr. Majkut: Over the weekend I was reading Emerson’s lecture on the duty of the American scholar.

Dr. DeCarolis: (Laughs.)

Dr. Majkut: He’s got this great line in there that it’s the duty of the scholar to cheer, to raise, and to guide men by showing them facts amidst appearances. Which facts are coming out of this IEO that you think are important for, like, the general public and policymakers to understand as we look at the energy future?

Dr. DeCarolis: I think the biggest thing is, you know, we all bring – when we think about energy, we all bring our perspectives, right, our own life experience when we’re interpreting the news or we see what’s going on. And I think with an exercise like this, it’s really informed by data. So you have to step – you know, it’s a very careful exercise to look across the world and to see what’s happening. And sometimes, I think, in developed countries we forget about how much development is actually taking place.

So I think for me one of the key takeaways is, you know, as governments explore low-carbon futures, just know that there’s this backdrop of continued development. And that’s putting pressure on energy demand, and that’s – that informs our projection. So we’re seeing, you know, sort of steady or increased use of fossil fuels, but also rapid growth of renewable energy. So I think the clean-energy story is taking place in the electricity sector. It’s largely being driven by solar and wind.

Dr. Majkut: So then there are other groups that look at global forecasts. I don’t want to get into, like, a modeling debate, but help us understand, you know, they’re – we’re sort of – even if you take your results, which one of the things, I think, I kind of took when I – from this presentation as well as our conversations in advance of today’s event, you know, the sort of – the character of the energy system going forward is changing rapidly in your results as well; it’s just that there is still this growing demand around the world for energy services. Fossil fuels meet those in a cost-effective way. So help us understand or help the audience understand, you know, if another modeling exercise, another set of projections says, well, fossil-fuel use is going to peak this decade, it’s going to peak next decade, you know, what differences are those – what differences is that – is that showing us in how scenarios are constructed or how modeling is being done?

Dr. DeCarolis: I think one thing to keep in mind is that you should expect to see big differences across outlooks. I can’t emphasize enough how much uncertainty there is. And I think there’s always a lot of uncertainty when you produce an outlook like this, but I think we’re entering a period of even higher uncertainty. So the expectation shouldn’t be that you look across outlooks and see the same thing. I think when you see differences, that’s actually healthy.

I would – as a modeler myself, I’d be concerned if they all converged to the same number. I’ve spent a lot of time over the last year going back and looking. We have a long history. We’ve done a lot of projections. Being a modeler requires humility. And so I think if all the outlooks are saying the same thing, we actually – we have a problem. So I think diversity in what you see across the outlooks is healthy.

Just to give you an idea of what drives some of those differences, right, so even if you’re looking at another outlook and, oh, there’s a scenario that looks kind of similar to EIA, you have to look at how was policy interpreted, right? Again, I go back to what we did: strict interpretation of existing policy that’s legally enforceable. What do you assume – what do they assume about economic growth? What do they assume about technology innovation, so what do the cost pathways for renewables and other technologies look like? What do you assume about technologies that might be on the horizon that are highly uncertain? And also consumer preferences, right? We’re largely doing economic modeling here, but there’s a big piece of this – even with existing policy, we’re incentivizing consumers and nobody knows exactly how they’re going to respond.

Dr. Majkut: So I want to touch on uncertainty in a couple ways. There’s, like, a hard version of the question that I want to ask –

Dr. DeCarolis: (Laughs.)

Dr. Majkut: – right, that, like, when I look at the various projections, right, and you sort of – you give us this envelope, a shaded envelope in the graphical display, I look at that and I go, I don’t know if there’s enough uncertainty here, right? I don’t – I don’t know if that’s a wide enough band, in particular on the downside of greenhouse gas emissions, you know, just, like, as an external person who reads a lot about this stuff.

So maybe we can dive into an example, right? So EVs are a key one. There’s, like, rapid growth in – 40 percent of the Chinese auto market is now EVs at the point of new sales. The U.S. is like 5 to 7 percent depending on the season. Some people will tell you that at somewhere around 10 percent of penetration you hit this tipping point and then consumer preferences will shift rapidly toward EVs. If that happens, you might see a lot less fossil-fuel demand for transportation, and then, like, your lines on greenhouse gas emissions or global fossil-fuel consumption might have to go down a lot.

Dr. DeCarolis: Yes.

Dr. Majkut: So, like, how does that – how does this structure model that kind of – those dynamics?

Dr. DeCarolis: OK. So one thing I want to just acknowledge upfront is we do show those uncertainty bands; that is not the full scope of the uncertainty, right? There’s this wide universe of things that can happen. We’re very clear about how we approach it and what assumptions we make across the cases.

Dr. Majkut: So just as –

Dr. DeCarolis: So we’re kind of taking – we’re taking a narrow path through this pretty wide universe. So those –

Dr. Majkut: Would it be appropriate to say, like, just as you’re not making a forecast, that is not a probability distribution?

Dr. DeCarolis: Correct.

Dr. Majkut: OK.

Dr. DeCarolis: Absolutely, right? The bands are wider than that, and it depends on what you test. So if I was trying to interpret our outlook in the context of others, think about looking at the projections all on the same chart and then you start to – because different outlooks make different assumptions, they use different models, so you’re exploring different parts of the future. And I think that’s – you know, that’s a healthy exercise.

Getting to – let me get to EVs, though, and I can – we can talk a little bit –

Dr. Majkut: Yeah, I want to talk about it because we talked about in our pre-call –

Dr. DeCarolis: Yes.

Dr. Majkut: – and I actually learned a lot, and I want these people to learn that stuff too. (Laughter.)

Dr. DeCarolis: Sure. Sure. So I want to give a little bit of perspective, so I’m going to – I’m going to cite some of the numbers Angelina already gave. But I think they’re – I think they’re really important to understand.

I mean, first of all, passenger travel demand is going to increase by something like 65 (percent) to over a hundred percent by 2050 across the cases.

Dr. Majkut: Pure economic growth.

Dr. DeCarolis: Yep.

Dr. Majkut: OK.

Dr. DeCarolis: And that’s because there’s more people, but also – but also they want to travel more miles, right, as they gain wealth.

Angelina also mentioned we’re at – right now we’re at about 1.4 billion light-duty vehicles on the road. That’s going to grow to about 2 billion by 2050 in most of our cases.

You have to remember that with electric vehicles it’s not just pure economics; there’s also consumer preferences. And so we have some representation of that, you know, in our model. Policy interpretation also matters a lot in terms of what you project going into the future from where we are right now. And as Angelina said, what we’re finding is that, you know, EVs are projected to make up about 30 to 55 percent of global sales by 2050, but there’s a lot of uncertainty. So is it possible that things can unfold in a way where we get – we end up with higher projections? Absolutely.

I’ll give you one example. We assume, again, evolutionary rates of technology innovation. So we look very carefully at the prevailing cost of batteries. We look at how the cost of those lithium-ion batteries has declined over time and we project that into the future. If there’s a new battery chemistry, some new breakthrough, you know, all bets are off, right? Things can change.

Dr. Majkut: And when you think about sort of – and here you’re doing something difficult because you’re modeling consumer choices, right? And we just – like, I don’t think we really know particularly well how to do that.

Dr. DeCarolis: Yeah.

Dr. Majkut: So what are – like, what dynamics go into trying to assess, like, how do consumers weigh relative cost curves, performance, social trends, et cetera?

Dr. DeCarolis: So there is – in our National Energy Modeling System, which we use to produce our Annual Energy Outlook, we have more of a – consumer-choice models. We have a simplified version in the model that we use to produce the International Energy Outlook, so it’s not a full consumer choice model but it’s – but it’s also not just about cost.

So the four factors that go in – and this is in one of the technical notes – you’ve got the cost of the vehicle, there’s the cost to drive it, but then there’s also model availability, right? Like, how many EV models are out there relative to internal combustion engines? And also, fuel availability. So as people see the availability of more models and they see that there’s more infrastructure available to charge or refuel, they’re more apt to make that purchase. And this is – you know, this is a standard methodology to look at vehicle uptake.

Dr. Majkut: Yeah. I don’t know, I always wonder because I just don’t think of a car dealership as a place to make, like, a rational decision. (Laughter.)

Let’s maybe shift to, like – I want to come – like, I think it’s good to talk about the technical sophistication, but we’re running short on time and a lot of people want to know about your call on peak fossil production. But –

Dr. DeCarolis: (Laughs.)

Dr. Majkut: But maybe, like, let’s – let me put that in a different context, right? So you have two energy price cases with respect to oil prices, high and low. I would say that, like, I didn’t see, like, big differences in the results, but maybe that’s at the top line. Can you help us understand what we learn from those two cases with respect to the shape of global oil markets, the role of the U.S. as a growing producer? I mean, the kind of insights that EIA can help us derive I think are really important for the sort of policy conversation that we have here in Washington.

Dr. DeCarolis: Sure. So, I mean, as far as oil demand goes, what we see is – so we’re roughly – total liquids right now is around 99 million barrels per day. What we see is by 2050 that – again, across the cases we model it’s something like, very roughly, a 10 to 40 percent increase in demand for liquids.

The U.S. piece of that – again, remember that the U.S. results are fixed to what we had in the Annual Energy Outlook. But what we found there is that we remain a net exporter of petroleum across all of the cases, but there’s a pretty wide variation. So it’s anywhere from half-a-million barrels per day all the way up to 9 (million barrels per day).

Dr. Majkut: Yeah. OK.

Let me – let me ask you about the future of these exercises, right? So over the last year, I’ve had the real honor of hosting a lot of people who look at – put together different forecast products – oh, excuse me, different projection products. Some of them will daresay call it a forecast, right? You know, how do we evolve these tools? I know, you know, from your academic background, you’ve worked on modeling exercises a lot, and I’m interested in talking with you a bit about how we – how are you thinking about evolving these tools to inform the public-policy conversation and better understand the dynamics. So, like, you know, what do – what do you think about these critical minerals challenges that are really important for the kind of uptake of solar and wind that you see in this, in your modeling? How do you think about making these results more transparent and more available to the public or to policymakers so we can get a better understanding of what these models and what these scenarios can teach us? You know, I’d love to just have – you know, hear your open thoughts on how we keep innovating to make sure we’re meeting informational needs of the moment?

Dr. DeCarolis: That’s a great, huge question.

Dr. Majkut: Yeah.

Dr. DeCarolis: How much time do we – (laughs) – to do that.

Dr. Majkut: You have eight minutes, sir. (Laughter.)

Dr. DeCarolis: So I guess a few thoughts. The energy system is in a period of pretty rapid change, and that creates a challenge as energy modelers. It means we need to make sure that our models stay up to date with what we see – what we see happening.

Very quickly, I will say that with the modeling exercises that we do at EIA we try to be as transparent as possible. So I do want to say that upfront. Like, if you’re – if you want to dig into what we’ve done, you can read the narrative, you can download the data tables you can read all of the documentation for all of the modules there, you can request the source code, and we’re planning to make the source code publicly available under an open-source license. There’s still a lot of logistics to work through, but we’re trying to be as transparent as we possibly can. OK, so there’s that.

We’re in the process of retooling our models. So I had announced earlier that we’re taking a break from the Annual Energy Outlook in 2024 so that we can add some of the low-carbon pathways that we need to be able to model, so we’re going to be doing that. But on a parallel track – separate effort – we’re also building a next-generation model, OK? And I think the key is, by talking with internal and external stakeholders, we really need modeling to be nimble and flexible, right? We need to be able to look at a wide range of scenarios, and the considerations that we now need to bring to the model are increasing.

So you mentioned critical minerals, right? So that’s something that we’re looking at. So I can go into critical minerals a little bit more, or if you want to pivot, that’s –

Dr. Majkut: I mean, like, what is – what are you learning?

Dr. DeCarolis: So I would say the quick answer is with critical minerals we’re currently working with the U.S. Geologic(al) Survey. They’ve done a lot of work looking at supply chain risk from critical minerals, so we’re collaborating with them. We view ourselves as the experts in producing these energy projections. They understand the supply chains better. So we’re working collaboratively to figure out how we can build critical mineral supply into our models. I would say that, like, trying to, you know, build those supply chains into the modeling framework would be part of this effort to build a new next-generation model.

Dr. Majkut: And do we have enough information – I mean, you – the EIA is the information keeper for oil and gas production in the U.S. You’re a statistical agency, so you get to demand it.

Dr. DeCarolis: Right.

Dr. Majkut: We don’t have an analog on the mineral side. Do we have enough information to do, like, a meaningful kind of analysis in the way that EIA is used to doing on the – on the oil and gas side?

Dr. DeCarolis: I think that’s something we have to evaluate. I mean, my sense is that you always work with the data that you – that you have, and I think if it helps to highlight where there could be particularly important risks to consider then that’s valuable. But I think we have to look at it more carefully as we build –

Dr. Majkut: And then this is a question that comes from online that actually is really interesting. The question is about learning, right?

Dr. DeCarolis: Yeah.

Dr. Majkut: How do you – how do you – when you embark on the IEO every year or you’re embarking on this taking a pause on the AEO to kind of renew the tools that we have, how do you make sure you’re learning from past successes and past failures?

Dr. DeCarolis: Yeah, that’s a good question. I mean, we are in contact with stakeholders. We often have workshops with our next-generation modeling effort. We’re actually about to have a series of engagements with external stakeholders so we can learn from them.

We also do model benchmarking work. So with the Annual Energy Outlook, we have a retrospective report where we go back and look and see how we did. And we’re actually in the process of trying to expand that work ourselves. I’ve personally, you know, tried to go back and look across past projections and see what did we get right, what did we get wrong, and how can that, you know, inform our projections moving forward.

Dr. Majkut: Do you see that as being a process that can get formalized, right? I mean, there’s lots of fancy modeling tools that allow us to really take in quantitative information or change projections based on – based on what we learn about over time. Do you think that stuff can be formalized?

Dr. DeCarolis: Yes. I mean, absolutely. I mean, one of the things we did in the Annual Energy Outlook was we had uncertainty cones. So you can come up with different projections by changing input assumptions in your model, but what the cones do is they look at the difference between what was in the reference case and what actually happened and said, well, you know, here are the errors, you get a distribution, and then you can use that to derive a cone that says, you know, you may or may not believe our reference case, but if you look at past differences this is where things could head.

Dr. Majkut: So –

Dr. DeCarolis: So that’s just one example, but there’s a lot of different ways that you can – you know, you can formalize that – the consideration of that.

Dr. Majkut: And then, like, as you think about putting these scenarios together, as you think about the – like, the numerous attempts that you have to make at parameter estimation – and I don’t know whether I buy this idea that we’re all living in a simulation or not, but posit we do for a moment, and I give you access to the source code, and you get to learn one thing with, like, absolute precision and accuracy to inform making better projections for our energy future. What would it be?

Dr. DeCarolis: I get access to the real-life –

Dr. Majkut: Yeah, yeah. (Laughter.) Yeah, we give you access to the real-life source code.

Dr. DeCarolis: Is it “The Matrix”?

Dr. Majkut: Exactly.

Dr. DeCarolis: I think the biggest challenge we face is – it’s the human behavioral piece of it, right? We can characterize cost data, and that’s uncertain, but I think we don’t have a good sense for what – you know, as we offer novel technologies, will consumers take them up? How do personal preferences influence, you know, political movements and, you know, like, how does – you know, how does all of that aggregate up in a way that informs whether we’re going to – would we end up with new policy? Do consumers prefer these technologies? I think that’s a really huge challenge that we – that we struggle with as modelers. I mean, we do the very best we can with the data we have, but we don’t have a good understanding of how human beings are going to respond to external events, to new technologies, and so forth.

Dr. Majkut: I mean, I kind of agree. One thing that we’ve been thinking a lot recently here is, like, there’s all this tech innovation that’s probably going to come from the results of the bipartisan infrastructure law, from the IRA. How quickly does that diffuse around the world is an economics question, it’s a public-policy question, it’s a social question.

Dr. DeCarolis: Right, right.

Dr. Majkut: And there’s just a huge veil of uncertainty around that question.

Dr. DeCarolis: Absolutely. And I think that just brings me back to the earlier point that – (laughs) – that’s why it’s important to – we shouldn’t expect to see outlooks all, you know, sort of conforming to the same – the same results because there is all this uncertainty. We all make different assumptions, and so you get this array of outcomes. And then we all get surprised by certain things, right? (Laughs.)

Dr. Majkut: Right. Oh, man, yeah.

  1. We’re running up against our time window. I just want to say, like, we are very grateful you came to talk about this today. I think it’s – like, it’s an interesting, in some quarters provocative set of results that you’ve found. I think it’s going to really inform the conversation, not just here in Washington but around the world.

So thank you to you, Joe. Thank you, Angelina. And also, those who are joining online don’t know, but many of the EIA’s subject-matter experts and analysts have come today for what is, unfortunately, the lamest launch party that I can possibly imagine. (Laughter.) But I – you know, please allow me to extend our appreciation for your hard work and your dedication to public service. Thank you all. (Applause.) And I wish you well, everybody joining us here in person as well as online. Colleagues, this is Joseph Majkut, signing off, and we hope to see you next time here at CSIS.