As I explain later, I think this comparison is a bit silly, but sympathise with the overall sentiment: Europe is not doing a good job of protecting people from heat. Things don’t have to be this bad; it’s a choice.

But, unfortunately, the chart has several issues.

The heat death numbers for Europe are modelled as “excess deaths” during summer months; this attempts to capture people dying from a range of health conditions — cardiovascular disease, stroke, and others — earlier than they would have in more optimal temperatures. This is a common way researchers measure heat deaths (or cold, for that matter). I’ve done a deep dive on these methods previously.

But the US number isn’t based on this type of modelling; it’s based on heat deaths recorded on death certificates. If you used death certificate figures for Europe, they’d be far lower.

It also uses European Union figures for gun deaths, but a broader definition of Europe for heat deaths.1

I thought I’d have a quick go at making a better version. These numbers are still not perfect, but hopefully a closer comparison.

Here are some of the changes:

• I’ve used more comparable excess heat death figures for the US (details in the footnote), so we’re not relying on death certificate records.2 This is the annual average between 2000 and 2020 and for adults aged 25 to 84 (it’s not a perfect comparison, but the best I could do with available data).3

• I’ve taken the average of 2022 to 2024 for Europe, so we’re not catching an “irregular” year and comparing it to a more average year in the US.4

• Created one version that consistently uses EU-27 figures for both guns and heat.

• A second version that uses the broader European definition in the heat deaths paper, sourced in the original chart: the 26 EU countries (minus Malta), plus the UK, Switzerland, Bosnia and Herzegovina, Montenegro, Norway, and Serbia (so 32 in total).5

• Shown the breakdown of gun deaths by homicide, suicide, and other causes (which is small but includes unintentional injury and deaths by law enforcement).6

Here’s the comparison for countries in the European Union only.

And now using the broader “Europe” definition.

Does it change the conclusions much? Not really. But hopefully it provides a slightly more accurate comparison. And I think the breakdown of gun deaths provides more useful context.

What about per capita?

As always in a comment section, people asked: “What if we adjust for population?”

Here’s the same data adjusted for rates per 100,000 people in the total population7

It does actually change the order a bit. Gun deaths in the US are now slightly larger than European heat death rates. But to me, it doesn’t matter much for the conclusion.

As I mentioned, I found the initial framing of guns and heat as a contest to be a bit silly — they’re not substitutes. Europe is doing a bad job of protecting itself from life-threatening heat. That the US has disproportionately high rates of gun deaths compared to other rich countries is also bad. The two are not mutually exclusive.

But it does highlight something important about status quo bias. Both places take the status quo as a given: a high-mortality situation in one domain that they’d never accept in another.

Subscribe now

Much of these discussions use the United Nations’ population projections as the starting point.2 But these, like any demographic projection, rely on assumptions for how fertility rates, life expectancy, and migration rates will change in the future. And by future, I mean the next 70 years.

It often takes a bit of work to understand what the UN assumes in its models. And then it’s difficult for anyone who’s not a demographer to work out how things might look different if reality turns out different.

What if fertility rates fall faster or slower? What if there is no rebound in rates, like many scenarios assume? What if we see unprecedented improvements in healthcare and life expectancy, or net migration rises, or countries close their borders?

It can be fun and informative to play around with what the future might look like. At Our World in Data, we have built a population simulation tool that lets you do this (kudos to Daniel Bachler and Sophia Mersmann).

It’s a relatively simple model (but stacks up pretty well when we tested it against UN data). For any country, you can change three variables: fertility rates, life expectancy and net migration at three points of time in the future. You can then see how total population numbers change, alongside the population pyramid.

Try the population tool

I gave the tool a go, asking the question: “What would it take to stop South Korea’s population from shrinking?” The article is here.

TLDR: to achieve it through life expectancy gains or migration would require really unprecedented change (by that, I mean average life expectancy increasing rapidly to 130 years).3 Neither will happen to that extent.

Fertility rates would need to rapidly increase from 0.7 children to 2.1 per woman, and stay there for the rest of the century. Perhaps more plausible than the other options, but still extremely unlikely.

Hopefully, you find the tool useful in exploring more questions like this.

South Korea article

The problem is that until recently, the latest figures from organisations such as the International Energy Agency were for 2024. It’s hard to have a serious conversation about AI's energy use with data from before AI really took off.

But a few weeks ago, the IEA published its latest report on Energy and AI, with estimates for 2025, and updates of its future projections.1

I want to run through the high-level numbers here.

I’ll make one caveat here so I don’t have to repeat myself throughout: things become increasingly uncertain once we get into medium-term projections. I am not here to present these projections as ground truth — other analysts produce outlooks that are quite different — but to reflect what the latest IEA report says in an understandable way.

How much of global electricity is used for data centres and AI?

Last year, around 1.5% of the world’s electricity was used to power data centres.

Now data centres are not just AI: they’re facilities that contain the servers and IT infrastructure behind all of our digital services. That’s everything from Substack and the rest of the internet, to Netflix streaming, Google Maps, online banking, and messaging friends.

AI data centres are dedicated facilities for running AI models. The distinction between the two is not always that clean-cut, but the chart below shows the breakdown.

Non-AI data centres consumed more than twice as much electricity as AI-focused ones.

By my estimates, AI consumed around 0.5% of the world’s electricity in 2025.

Electricity is just a subset of total energy use, so data centres consume less than 0.5% of final energy, and AI less than 0.2%. That’s useful to keep in mind: the AI debate is an electricity one, not an energy one.

I’ve also included the IEA’s base-case projection of data centre demand in 2030 (I’ll look at this in more detail later).

Most of the growth in data centre demand will come from AI-focused facilities. In the IEA’s base scenario, data centres grow to 3% of global electricity in 2030. AI centres then use about the same amount as non-AI ones.

Data centre demand is very much US-centric

1.5% of the world’s electricity doesn’t seem like much. But in some places, that share is far higher.

In the chart, you can see the share of electricity used for data centres in different regions. 5% of the US’s electricity is used for data centres. For AI-focused data centres specifically, we’re probably talking around 2%.

But in fact, this demand is even more locally concentrated. Beneath Europe’s 1.6% figure, we have Ireland, where data centres account for more than 20% of its electricity consumption. Beneath the 5% US figure, there are a number of states where data centres make up more than 10% of demand, and in states such as Virginia, it’s more than a quarter.2

This is really the key challenge with growing data centre demand: it’s geographically concentrated, meaning the entire world’s demand is being served by a small number of electricity grids. You’d also see bottlenecks and local grid issues if you were trying to serve all of the world’s demand for microwaves, kettles, or washing machines from a handful of places.

How have the IEA’s projections changed?

What’s the IEA now saying about future demand?

It publishes four scenarios: its base case, a lift-off case where AI grows much faster, and two cases with lower power demand, either due to even more rapid efficiency improvements or lower-than-expected demand.

You can see these four scenarios in the chart below. Projections to 2030 are already pretty uncertain, but the IEA stresses that those out to 2035 are even more so (what it calls “explorations”).

There aren’t huge differences between the four scenarios in 2030. In all scenarios, global power demand for data centres basically doubles from 2025 levels.

In the base case, global electricity demand reaches 945 TWh. The high and low scenarios are within around 100 TWh of this. The reason these scenarios vary so little is that much of this supply is already “locked in”. It takes several years to build a data centre, there are near-term bottlenecks in supply chains and chips, and there are local bottlenecks in how quickly new power supplies can be built.

After 2030, the IEA’s projections diverge as some of these nearer-term barriers become unlocked.

Data centres could be a modest driver of electricity growth globally, but a large one in the US

Data centres are not the only driver of new electricity demand. Countries will need more electricity for transport, heating, and industry. Low- and middle-income countries are still seeing rising electricity demand as people move out of poverty and raise their standards of living.

How much of the power demand growth in the next five years could come from data centres?

The IEA projects around one-tenth (9%) of it.

In the US, it could be as much as half. That’s for two reasons: most data centres are being built in the US, so most of the growth in global demand is really being centralised there; and growth in electricity demand for other uses is slower than elsewhere. The US is not seeing the electric vehicle boom that other countries are; almost everyone has air conditioning, and it’s not a country adding huge amounts of new industrial capacity.

What’s the energy footprint of individual queries?

I’ve written several articles on the footprint of individual LLM queries.

A key takeaway from the numbers was that asking a chatbot a question — which is what most people were using AI for in their day-to-day lives — consumes very little energy.

Tech companies have not been very transparent about the energy use of their AI models (and I think they should be), but the numbers seemed to converge around 0.3 watt-hours (Wh) per typical text query. To put this into context, asking ChatGPT or Gemini 10 simple questions is equivalent to about 10 seconds of microwaving or mere seconds of showering.

This power consumption varies by the size of the text query, as this chart from Epoch AI shows.

What does the IEA say about this in its latest report?

It quotes two things. First, the company-reported figures I discussed previously: 0.24 Wh per median text query using Google Gemini (published by Google) and around 0.34 Wh per average ChatGPT query (mentioned by Sam Altman).3

Second, AI Energy Score benchmarks for different levels of tasks. These figures only measured the energy used by the GPUs (basically, the chips); on top of this, energy is needed for networks, cooling, lighting etc. In hyperscale data centres, GPUs can consume 50% to 60% of a facility’s total energy. So the total energy use per query could be as much as double this.

Both these measures are included in the chart below.

The AI Energy Score benchmarks are in decent agreement with the company-reported figures. A medium text query uses very little energy: 0.05 Wh for GPUs, which could be around 0.1 Wh in total. A large text query consumes around 0.3 Wh on GPUs (up to 0.6 Wh in total).

Depending on the length of the text query, asking an LLM a question is somewhere in this 0.1 to 0.6 Wh range. Company-reported figures and benchmark scores are all in this ballpark.

As we’d expect, more complex tasks use more energy. Once we get into the realm of agentic, reasoning, or agentic-with-reasoning tasks, we enter the range of several to tens of watt-hours.

Agents are AI tools — Claude Code is one example — which can plan, iterate and call on tools autonomously. In the IEA methodology, this involves 4 to 6 sequential calls and responses, so you can see how this is inherently more complex than a simple text query.

What do these numbers mean for individual footprints?

Many people are using AI for medium- to long-form text queries, such as asking a quick question or requesting a short fact-check or correction. Their energy use is very small, even if they’re asking tens or hundreds of questions a day. A hundred questions have a footprint of around 30 Wh. That’s roughly the amount of electricity the average American consumes in just over a minute (or for the average European, every two and a half minutes).4

The footprint of someone who uses agents heavily is not so negligible.

Let’s say they do 4 agentic queries per hour (how many you can do in an hour is limited by the fact that complex tasks can take 15 minutes or more to complete). And they do this for 6 hours a day. That’s 24 per day. We’ll assume that the total electricity use per query is actually 100 Wh (50 Wh multiplied by two).

They’ll consume 2,400 Wh (or 2.4 kWh). That’s like running a tumble dryer for one cycle, or driving an electric car eight miles. It’s around 7% of the average American’s electricity use (but a much smaller share of total energy use).

It’s not blowing up their footprint, but it’s not nothing either.

One follow-up question is the energy use of image and video generation. Unfortunately, I don’t know, and this report doesn’t include numbers on this.

But it leads to one of the most interesting — and unanswered — points in the report.

The “missing” energy puzzle

The IEA report includes some useful napkin math.

Let’s assume a text query to a chatbot consumes 1 Wh of electricity (this would be a fairly large text query, but it’s a nice round number).

If the world made 10 billion queries a day — similar to the number of Google searches, and four times the number of queries that ChatGPT reports — then we’d be consuming 10 GWh of power a day, or 3.65 TWh per year.5

But AI-focused data centres consumed 155 TWh in 2025. So, all of the world’s text queries accounted for around 2% of this electricity consumption. Where did the other 98% go?

That’s the puzzle, and we don’t know because companies don’t release breakdowns of exactly where their compute and power are going.

Some of it will be for training the models. But even some of the largest models consume less than a terawatt-hour during training (or at least that’s what’s reported).6

Another explanation is image and video generation. Video generation in particular is more energy-intensive per task, but without good data on per-video energy costs or volumes, it’s hard to know how much they account for.

Perhaps the most plausible explanation is that the largest energy consumer is the more diffuse deployment of AI across the internet: AI-generated summaries in Google Search results, AI-driven algorithms on social media, content moderation, advertising, and translation.

Enterprise and industrial use of AI is also increasing, ranging from AI models in Microsoft Office, Teams, Gemini, and Salesforce to AI in drug discovery, weather forecasting, and financial modelling.

The most visible form of AI use is asking a chatbot a question. But in reality, many are using it all the time, in subtle (and even involuntary) ways. The energy consumption of these interactions is likely far higher than that of the average person’s chatbot requests.

To be clear, this isn't about whether the total AI figure is right; the IEA's estimate is not adding up queries, but from a bottom-up model based on server shipments and their power draw. The puzzle is that we know how much electricity AI infrastructure uses overall, but companies don't disclose what they’re actually being used for.

Subscribe now

Last week I was in London for the Unwin Award Ceremony. It was held at the Royal Institution. Special as it’s where they’d host the infamous Science Christmas Lectures I watched as a kid on television.

I was one of six authors shortlisted for “non-fiction authors in the earlier stages of their careers as authors whose work is considered to have made a significant contribution to the world.”

It was an honour to even be shortlisted, but even more so to win it. Thanks to the judges and everyone at the Publisher’s Association that made the prize possible. And of course, to my editorial team for helping me publish some award-worthy books.

The award comes with a £10,000 prize. I’m donating all of this to the Against Malaria Foundation. More than 600,000 people die from malaria every year, most of them children. Interventions against malaria are among the most cost-effective ways to save a life, which is why the money is going there.

£10,000 will protect over 5,000 people for three years, and will save approximately three lives.

In other news, for the entire month of May you’ll be able to buy my first book, Not the End of the World, for 99p on Kindle. Enjoy!

Not the End of the World

The focus of the post is always one of a few things. First, surprise that India has almost fully electrified its rail networks, putting most of the rest of the world to shame. Second, the fact that basically none of America’s railways are electrified. Or three, the mediocre performance of countries like the UK especially compared to their European neighbours.

But there are also a few things about that statistics that initially seem confusing. As someone that spends a lot of time on trains in the UK (I am actually writing this on one), the fact that just over one-third of the lines are electrified seems odd to me. I barely remember being on one that wasn’t. Japan, who we hold up as the pinnacle of modern, high-speed rail seems like it should do better.

So I took a bit of time to figure out if these numbers are actually believable and why some of them seem to be at-odds with what many people will experience day-to-day.

First, are the numbers in these viral infographics and tweets correct?

Pretty much. I did my own data-gathering exercise across a selection of countries, drawing on national data sources and databases like Eurostat. The results are in the chart below, and they tally pretty well with the original figures.

If you want to explore data for other countries in Europe, here is a good resource.

What’s important to note about these numbers, though, is that they’re purely based on the length of the track that’s electrified. So if a country has 1000 kilometres of railways lines, and 500 kilometres are electrified, then their electrification rate is 50%. How busy these lines are is irrelevant. Not a single train could run on the non-electric part, and the rate would still be 50%.

This goes a long way toward explaining why frequent train travellers’ experiences of trains is at odds with some of these stats. Here’s a very helpful map of the rail network in Britain. The lines in red, yellow and green are electrified. The lines in grey are not, and run on diesel.

The very busy transport corridors — across the South-East of England, the central belt of Scotland, and long links between the two are electric. When I’m travelling up and down between Edinburgh and London, I’m always on an electric line, regardless whether it passes through major cities on the East or West coast.

For those living and travelling outside of this central, high-traffic zone, the experience is quite different. The trains run far less often, and they’re usually running on diesel.

What happens if we weight by how much traffic actually runs on these lines?

I’ve tried to do that in the chart below.1 This tells us what share of passenger train kilometres were electrified. Now, this takes account of the fact that trains on some routes are running back and forward a lot (see the footnote for an example of the calculation).2

This metric doesn’t capture how many people are on the train. One route could be rammed while the other has two people per carriage, and they’d be treated the same. If you were to adjust for this, the numbers would change even more.

What we see if that the share for a lot of countries is higher than the earlier metric covering just the distance of the railway track. The UK moves from just over one-third to almost two-thirds. Japan moves up to over 90%.

It makes little difference to US and Canada: their rail networks are just not electrified, and there’s no metric that’s going to tell a different story.

Another big factor in these statistics is freight: the transport of goods and materials.

I’ve done the same adjustment for freight, in the chart below. This time it’s adjusted for tonne-kilometres.

Here, the UK does particularly badly: almost no freight is transported on electrified tracks. That didn’t surprise me much. What did was how high this was for other countries. My assumption would have been that most passenger links are on electrified lines, but the majority of freight still travelled by diesel. But the latter is not true across many countries, ranging from the Netherlands, Spain and France in Europe, to Japan, India and China elsewhere.

The US has basically no freight running on electric lines. It built a large diesel-powered network before the arrival of electrified overhead lines, and in fairness, it would probably take a lot of work to replace them over such large distances.

Why should we care that we electrify more rail networks?

The answer for many readers of this newsletter is obvious: to reduce carbon emissions. There’s no easy way to decarbonise diesel, but we do know how to produce low-carbon electricity.

If this is what we’re interested in, how many journeys are electric versus diesel matters more how much of track is electrified. In the same way that how many coal plants you have matters much less than how often they’re running.

Even then, it’s hard to get excited about rail electrification for the direct emissions benefits. Globally, rail accounts for just 0.2% of CO₂ emissions.3 We’re building out large infrastructure projects to skim off a tenth of a percent of emissions.

Making rail travel cleaner isn’t going to make much difference. But having people switch from the car or plane to the train, could. That’s the strongest climate case for electrifying rail: making the experience better and faster so that people are more incentivised to switch.

Electric trains tend to be quieter and smoother. They can accelerate faster, and even standard ones reach higher top speeds than diesel ones. All of these can add up to more frequent, faster and pleasant journeys. And to even consider high-speed rail you need to go electric.

There are benefits beyond climate, too. Electric trains are much more energy-efficient. They tend to cost less to maintain and run. They improve air quality, which is particularly important in dense cities. Faster services can keep more remote places better-connected.

Again, what progress metric matters most depends on your ranking of these benefits. If it’s about keeping all areas of a country well-connected, maybe it is about the length of track that’s electrified. If it’s about costs, the focus would be on most high-traffic areas.

Britain’s progress on electrification is slow

So, how is Britain doing on its journey to electrifying its rail network?

Not great.

Below you can see how much electrified rail was completed each year since the 1990s. Progress has been very stop-start, without a consistent buildout. In some years, we’ve upgraded basically no track at all.

Let’s put these numbers in context.

In an article for Britain Remade, Ben Hopkinson estimates that for the UK to reach its electrification targets by 2050, we’d need to electrify around 13,000 km of track over 27 years. Or, around 500 km per year.

Just one year in three decades actually passed that target. In most years, we’re doing less than 200 kilometres. In 2022, we managed a whopping 2 kilometres.

Other countries have not been as stop-start. In that same article, Ben compares Britain’s efforts to that of Germany. It has been electrifying at a very consistent pace for decades.

This likely makes the whole programme cheaper. Ben estimates that electrifying Britain’s rail network will cost approximately £31 billion. If it were able to match the costs of Germany or Denmark, it would cost half as much.

Subscribe now

Before then, I had what I’d guess is the common understanding of how cancers form and spread. The cells in our body are continually growing, dividing, repairing and dying. Cancers start when mutations escape the usual process of repair or die, forming clusters of abnormal cells. These then divide and replicate, forming a tumour and invading nearby tissue.

Most of these mutations are random, so developing cancer is often down to bad luck. And then there are risk factors and lifestyle choices that increase the odds of getting various cancers. We all know the common ones: smoking, obesity, skin damage.

While I knew that certain viruses increase the risk of some cancers: human papillomavirus (HPV) for cervical cancer or hepatitis B and C for liver cancer, I underestimated how strong that link was. I also hadn’t quite internalised that many other cancers are caused by infections. Around 13% of cancers are, and for some types, infections are the largest cause.

This finally clicked for me a few years ago when talking to my former colleague, Saloni Dattani (who is an expert on all things health, and I’ve learned a lot from). You should follow her Substack, and the publication she co-founded, Works in Progress.

Many cancers are caused by infections, and we often know how to prevent them

The fact that many cancers are caused by infections is good news: humans have a decent track record of finding ways to beat infections. At least, we have a better idea of how to do that than just stopping random, unlucky cell mutations.

The most dramatic example of progress is stomach cancer. Most cases are caused by infection from Helicobacter pylori, which causes long-term damage and inflammation in the stomach and can eventually lead to cancer.

Improvements in hygiene and food safety standards have reduced the prevalence of H pylori across many countries, leading to a dramatic fall in stomach cancer death rates in recent decades.1

Globally, stomach cancer death rates have fallen by 60%. In some countries, the change has been even more dramatic. In Japan, stomach cancer was once the largest cancer killer, but death rates have fallen by more than 70%, and that’s no longer the case.

This decline in stomach cancer is not that well-known. The fact that it was caused by an infection is key to our success in tackling it. That is also true for cervical cancer, which we can and are making progress on.

The HPV vaccine offers extremely high protection against cervical cancer

I was among the first cohort of schoolgirls in Scotland to get the HPV vaccine. Class by class, we were marched down to join a queue to get “a jab that protected against cancer”.

I don’t think I quite internalised that this single vaccine could essentially stop me from getting an extremely common type of cancer later in life. Perhaps this was also because the vaccine’s effectiveness in large populations was not as well established as it is now. “This will reduce the risk of cervical cancer” is how it was framed, but by how much was underappreciated.

Nearly all cases of cervical cancer are caused by the human papillomavirus (HPV) infection.2 Many types of HPV are spread through sexual contact. An effective HPV vaccine, therefore, stops someone from becoming infected (and dramatically reduces the risk that they later develop cervical cancer as a result).

The fact that it typically spreads through sexual contact means that HPV vaccines need to be given to children before they reach puberty to be most effective. They’re typically recommended for girls aged 9 to 14 years. Vaccinating older adolescents also reduces the risk, but it is less effective than vaccinating at younger ages.

There is now a lot of evidence that these vaccines work. One thing that makes these studies difficult is that the risk of cervical cancer (like most cancers) tends to increase steeply with age. The cohorts of girls/women who were vaccinated before adolescence are still relatively young, so we can’t say from population data that girls vaccinated early are much less likely to develop cervical cancer in their 50s or 60s.

But we can say a lot about the prevalence of HPV infections and the development of cancerous cells among younger women.

Cochrane Reviews — often seen as the gold standard of research reviews — consistently find positive outcomes. A review of randomised control trials found that the HPV vaccine significantly reduces pre-cancerous outcomes.3 The same is true for reviews of population-level studies, which found that:

“There are now long‐term outcome data from different countries and from different study designs that consistently report a reduction in the development of high‐grade CIN and cervical cancer in females vaccinated against HPV in early adolescence. Data show that there is greater benefit to vaccinating younger adolescents prior to becoming sexually active.”4

This reduction in risk of developing cervical cancer was at least 80%.

We can see this reduced risk clearly in studies from countries with national immunisation programs. Below is a chart that my former colleague, Saloni Dattani, made of cervical cancer rates among cohorts of young women in England.

On the x-axis, we have the women’s age. On the y-axis, the rates of new cervical cancer cases. As you can see from the blue and green lines (which represent different cohorts of women), cervical cancer risk increases with age. The blue line shows the new cases among women who were not offered the HPV vaccine. The green line shows those who were offered it at age 16 to 18. Girls who were vaccinated, even in their older teenage years, had a lower risk than those who were not vaccinated at all.

But you can then see the large reductions in risk among those vaccinated at younger ages. In purple are those vaccinated between the ages of 14 and 16. In red, those aged 12 to 14. These lines don’t extend as far along the x-axis, because these cohorts of women had not yet reached age 28 or 30 at the time of follow-up. But you can see that the line is consistently shallower and lowest among women vaccinated in their preteen years.

Similar evidence exists among populations of women in other countries with high vaccination rates. Australia is another stand-out example. It was the first country to launch a nationwide, publicly-funded vaccination program, and could be the first to eliminate cervical cancer completely.5

Combined with national smear test screening programs, both the UK and Australia have seen a dramatic reduction in death rates.

While most vaccine programs have been rolled out across high-income countries, there is huge potential for them in low- to middle-income countries, too.

A new study looking at the success of trials among young girls in South Africa is especially promising.6 It found a similar 80% reduction in HPV infections among girls who were vaccinated (achieving similar results with a two-dose program as those seen in three-dose programs in rich countries).

Importantly, this was also seen among girls who were living with HIV. This matters because having HIV can increase the risk of a HPV infection developing into cervical cancer. The high-burden of HIV is one of several reasons why death rates from cervical cancer are much higher in South Africa than in the UK.

Hundreds of thousands of women die from cervical cancer every year; we now know how to stop this for future generations

The fact that we have an incredibly effective vaccine against cervical cancer is huge.

Globally, there are between 600,000 and one million new cervical cancer cases each year, and around 370,000 women die from it.

What’s crucial is the rollout of vaccination programs across the world. This is happening. In the two maps below, you can see the change in the number of countries with HPV vaccines in their national vaccination programs.

Nonetheless, only one-in-five fifteen-year-old girls today received the HPV vaccine schedule in their early adolescent years. There is still a long way to go.

The exciting thing about these programs is that they can scale incredibly quickly, even in places with fairly limited health resources. Just look at Tanzania, another country with particularly high rates of cervical cancer due to HIV. Within a few years, it went from a minority of girls being vaccinated to almost universal coverage.

More efforts like this could see tens of millions of lives saved in the coming decades, and an end to one of the most common cancer types.

Subscribe now

Understandably so: the United States is the world’s largest aid donor in pure dollar terms (even though it’s far down the list when compared to the size of the economy). As I wrote previously, this aid was probably saving the lives of several million people each year through its health programs alone, so taking it away likely costs many lives, too.1

But some other countries have followed suit and started to cut their aid budgets as well. These have received very little attention by comparison.

One of those — and the one that has announced the steepest cut — is my own country, the United Kingdom. The coverage of it has been so sparse that you’d be forgiven for completely missing it.

I wanted to take a look at how these cuts fit within the UK’s historical contributions. It’s only when you visualise this data that the depth of these cuts becomes clear.

In 2024, the UK’s foreign aid budget was equivalent to 0.5% of its Gross National Income (GNI). This was down a bit from the 0.7% of GNI given throughout most of the 2010s, which was actually the UN target for wealthy countries to contribute.

Last year, the Labour government announced it would cut this to 0.3% by 2027, with reductions in stages over the next few years. In the chart below, you can see this in the context of the aid budget over the past 60 years.

The UK has not seen a contribution as low as 0.3% this century; the last year it was below this level was in 1999. In my lifetime, this has happened only five times (all in the late 1990s).

You can also see how low these levels are compared to the UN target.

We might argue that, since the UK economy has grown over that time, we’re still giving more in absolute terms to foreign aid.

Below, I’ve made a chart of the budget in US dollars (this is adjusted for inflation).2

In 2027, the budget may dip below $10 billion. It’s true that in dollar terms, the budget will be higher than it was in the 1960s through the 1990s. But it will be at its lowest level in decades (last matched in 2007).

What’s perhaps most damning about these cuts is the political party carrying them out.

Below is the same chart of foreign aid as a share of GNI, coloured by the party in government. Foreign aid contributions did increase a lot in the first decade of the 2000s under Labour (the left-leaning party). But the only years where the UK met the 0.7% of GNI target were under a Conservative government (which is probably not what people would expect).

The current and announced cuts are happening under Labour. In its 2024 manifesto, the party did pledge to increase aid from 0.5% to 0.7%. Granted, it was caveated in a way that made it easy to break:

“Labour is committed to restoring development spending at the level of 0.7 per cent of gross national income as soon as fiscal circumstances allow.”

I guess the argument is that fiscal circumstances don’t allow — and won’t — while they’re in office.

Just for completeness, here’s the coloured line chart, but showing aid given in absolute terms.

Less and less foreign aid is being spent overseas

The amount of money being spent on overseas development is not just falling due to budget cuts, but because aid budgets are being spent differently, too.

In recent years, around a third of the UK’s budget has been spent at home. You can see how this has increased over time in the chart. In 2010, it was just 3%.

A number of other countries have seen similar trends, so the UK is not an outlier. You can read more about this on Our World in Data.

Domestic spending includes administrative costs, scholarships, and development awareness programs, but the large increase has been in hosting refugees within the UK. The invasion of Ukraine had a big impact on this, which is why you see the bump in 2022 in particular.

Now, there’s nothing wrong with spending money this way: one way to help international development is to send money overseas, and another is to house and look after people fleeing desperate situations.

But the reality is that less and less money has been sent internationally in the last few years.

In the late 2010s, the overseas budget was around $20 billion. It’s now about half that.

As these aid cuts go even deeper, we would expect overseas development funding to fall even more. Below are rough projections that assume the share of aid spent domestically declines gradually to around 19% in 2027, which we might expect, as the cost of hosting refugees falls.3

Based on this, while the total aid budget in 2027 might be 0.3% of GNI, the amount spent internationally will be just 0.24% of GNI.

The UK is falling down the generosity rankings

How would these cuts affect the UK’s contributions relative to other high-income countries?

In the chart, I’ve shown the shares of GNI given as foreign aid across a selection of countries [these are not all top donors] in 2024. I’ve added where the UK would sit in terms of its 2027 contribution.

It was previously far behind the top top donors, but on a par with some of its Western European neighbours. In the next few years, it’s set to slip to become one of the least generous. Remember that when adjusted for money sent overseas, its contribution could be as low as 0.24%.

The UK is not the only country on this list pledging to cut back. France, Sweden and Switzerland are also set to reduce their budgets. And who knows what the US has planned. But the UK stands out as the non-US donor that has announced the deepest cuts.

Some countries are moving in the opposite direction: Japan and South Korea, for example, have increased their budgets significantly over the last decade despite starting from a low base.

On a more personal note

As a Brit passionate about international development, these cuts are disappointing, and if I’m honest, embarrassing. Sure, not every pound spent on aid is amazingly effective, but many programs are genuinely life-saving for millions, and life-bettering for many more. This is particularly true for health interventions. A relatively small amount of money — just tenths of a per cent of the national economy of rich countries — can go extremely far in some of the poorest countries in the world.

It’s easy to feel a bit helpless at times like this. But beyond advocating for the restoration of aid budgets, there are other things we can do as individuals to cover a small amount of the loss.

I give quite a bit of money to cost-effective charities. I’ve taken the Giving What We Can pledge and donate 10% of my income each month to recommended programs (mostly in global health).

GiveWell is usually my go-to source for figuring out where these donations can do the most good. I usually donate honorariums from book festivals and other events to charities like the Against Malaria Foundation. My partner knows that a donation is the perfect birthday gift.

Small donations from individuals can’t and won’t cover the tens of billions being cut from international development spending, but it’s something. And saving one or two lives is surely better than doing nothing.

Subscribe now

I’m not going to go into the mechanics of contrails — what they are, how they’re formed, and how they drive global warming — here, because I covered this in much more detail in another Substack. If you’re new to this topic, I’d suggest you read that one first for context on why this matters:

But the core of my argument there is that contrails — which are around 2% of the world’s effective radiative forcing — could be reduced quickly and cheaply.1 We will need to find low-carbon alternatives to jet fuel, but that will take time, and it will be expensive. While the aviation industry works on that, it could substantially cut its climate impact at a relatively low cost.

By “low cost”, I mean it could be a few dollars per tonne of CO₂e and per passenger.

While this all sounds promising, a lot of this thinking has been based on modelling: how many flights could avoid the atmospheric conditions that create contrails, whether that would actually reduce contrail formation, and how much extra fuel would be burned in the process.

A few years ago, Google partnered with American Airlines to run a real-life “proof of concept” showing that AI combined with satellite imaging technologies could actually help pilots reduce contrails. They did this across 70 test flights, reducing contrail formation by 54%.

This showed that it was technologically feasible, but on a limited scale.

Google have just published the results of a larger trial, involving thousands of flights. I think these results tell us something interesting about the feasibility and the hurdles to getting this to work at an industry scale.

The trial involved around 2,400 American Airlines eastbound transatlantic flights; those flying from the US to Europe. These flights were deemed to have a substantial impact on warming through the contrails they’d generate on their standard route.2

Around half were in the control group and were not given any flight plans to help them avoid conditions that would generate contrails.

For the other half, plane dispatchers had two options: the standard flight plan and an alternative contrail-avoidance one. They could then make their own decision on which one to allocate and hand over to the pilot. This was completely voluntary; there was no obligation or incentive to choose the avoidance option.

The headline result is that across these 1,232 flights eligible for contrail avoidance, contrails were reduced by 12% compared to the control group. It’s something, I guess, but not that impressive.

This first result might suggest that AI and satellite technologies are unable to model contrails, or that there is some flaw in the science that says they can be avoided.

That’s not what was going on here. The issue was more about human behaviour and incentives than about the technology. Most of the flights that could have avoided contrail conditions didn’t even try.

Of the 1,232 flights with contrail-avoidance plans, dispatchers used them only 15% of the time (rather than the standard plan). The avoidance plan was only successfully flown in just 60% of those flights.

But for the planes that did follow the avoidance plan, contrails were reduced by almost two-thirds.3 The technology works when it’s executed.

The problem was that fewer than one-in-ten eligible flights actually took that route. Why?

There are two steps to consider: whether the dispatcher chose the route, and then whether the pilot managed to execute it. The biggest bottleneck was the dispatcher’s decision; most flight crews didn’t even receive a contrail-avoidance plan.

There are several reasons the dispatcher might not choose it. The first one is simply busyness: I can imagine aircraft dispatch being a stressful job at times, with lots of spinning plates. If you’re stressed, and someone asks whether the next flight should follow the standard route or a less familiar one, you’re probably going to choose the former. That’s especially true given that participation was totally voluntary.

To avoid contrails, planes would often have to climb and descend mid-flight. This is safe to do, but not standard, so you can see why dispatchers might opt for the simpler, more typical route.

Their choices were also limited by the maps and information they received. They only got a top-down map view, rather than a vertical profile. This meant they couldn’t really understand why up-and-down changes were needed to optimise the route. Without complete information, they were less trusting and confident in the decision.

There are also reasons why contrail-avoidance plans were only flown successfully 60% of the time. First, the pilots didn’t receive complete information in the cockpit; without properly understanding the rationale for the request and why they were being asked to follow a specific, non-standard route, they were more likely to deviate slightly. Even relatively small deviations could mean they failed to avoid the conditions that create contrails.

In some cases, the dispatcher had made manual adjustments to a plan, which directed pilots back into the contrail-formation zone. Again, because they didn’t have a vertical profile of the routes, they didn’t know that these adjustments put the planes back into the avoidance zone.

These are not insurmountable barriers.

The fact that this program was entirely voluntary clearly makes a difference. If optimisation for contrail avoidance were mandatory or even incentivised, you wouldn’t have such low uptake rates. It was also very new to dispatchers, making them understandably resistant; the more common this became, the more comfortable they’d be in choosing alternative routes.

The authors of the study think that a better user interface, especially with a vertical profile of the routes, would have made a big difference. And further down the line, a more automated workflow where the contrail-optimised route became the default (or standard) one would also change the uptake rate dramatically.4

This study was an insightful test of the difference between technical feasibility and operational reality. Yes, the technology can improve, but it will be the operational barriers — and the regulatory or financial support needed to overcome them — that will determine whether the aviation industry can beat its contrails problem.

Subscribe now

I wanted to answer the basic question: how many hours could the average person run an air conditioner for with their daily electricity “budget”?1

You can see the results below. Forget hours — in many of the world’s energy-poor countries, people only consume enough electricity to use an AC for mere minutes.

The point was not that air conditioners are terrible energy-guzzlers (I did the same comparison for an electric fan, and the results were also pretty depressing). It’s that billions of people in the world still live in deep energy poverty.

How might this change over the next 20 to 30 years?

The International Energy Agency’s scenarios to 2050 get a lot of attention, but most of this focuses on how much CO₂ they generate, and how far we are from net-zero. Here I want to look at what they suggest about energy poverty, with a focus on electricity.

In the chart below you can see how much electricity different countries or regions generate per person today.2 These regional and country selections come from the IEA; I realise that they use a broad brush and talking about “Africa” or “Asia Pacific” as some single entity hides a lot of variation.

As you can see, the average across Africa is around five times lower than the global average, tens times lower than the European Union, and twenty times lower than the US. Keep in mind that this data is not just about households: it also includes electricity used in industry and commercial services, so having extremely low levels of consumption not only matters for individuals at home, but reflects a lack of infrastructure and economic engine.

How does the IEA suggest this will change by 2050?

In blue, you can see the projected growth in electricity consumption per person in the IEA’s “Current Policies Scenario”. This tries to reflect where countries are heading under energy system policies that are already in place. It takes factors such as population and economic growth, and changes in the costs of technologies into account.

The gap between Africa and the rest of the world does not shrink. It gets bigger. Per person, it would consume around six times less than the global average. The gap between Africa and other energy-poor regions will grow too. By 2050, Africa still wouldn't have reached the consumption level that India has today, which is already a very low baseline (the average person in India could not power just one hour of AC).

The outlook for Africa is no different in the “Stated Policies Scenario”, which aims to reflect what will happen if countries deliver on their stated policies. I’ve made a small interactive grapher showing these different scenarios and charts, so you can explore the data for yourself.3

The main reason why energy poverty rates are still so high in these scenarios is poor economic growth. It’s one of the key inputs into the scenario, and is based on projections from the IMF, and later, modelled growth based on factors such as labour supply and total factor productivity.

These projections for Africa do not look good. My colleague, Max Roser, recently wrote about the implications of these projections for extreme poverty. The number of people living on less than $3 per day in Sub-Saharan Africa is expected to grow over the coming decades. That’s because many of the world’s poorest are currently living in countries stagnant — or even shrinking — economies. If this is what monetary poverty looks like, it has obvious spillovers to energy poverty.

Many will argue that these projections are too pessimistic in a world with increasingly cheap solar and storage. Cheap power not only offers people more electricity on a lower income, but at a macro-level, can unlock industrial output and services, and stimulate faster economic growth than expected. I think that’s plausible, but how much it will change these final figures I don’t have a good sense of.

Affordable power will make a big difference, but I also want to emphasise how crucial economic growth is. It’s popular to say that “GDP is not all that matters”, which is true, but for the poorest countries it really does matter a lot. If these countries do not grow, billions will still be living in deep energy poverty decades from now. Unable to power a basic fridge to keep food fresh, or find relief from the gruelling heat with an hour of air conditioning.

Even if the world were to get anywhere close to zero emissions by 2050, this would be a huge failure. A low-carbon world where billions get left behind is not one that we should be happy with.

Subscribe now

The chart is one that I know all too well: it comes from Our World in Data (based on data from Ember).

It’s not the first time I’ve seen people share a version of this chart, arguing that China’s huge growth in electricity generation means it should (or will) dominate global industrial output.

The replies are always the same: China has a much larger population, and when we adjust for that, the US generates far more. That’s certainly true. Here is the same comparison in per capita terms. China has overtaken the European Union in the last few years, but the US is still far ahead.

What’s also important here is the difference in the trend: electricity use in the US and EU has been flat for decades, while China is very much on a steep upward trajectory.

Now, we can argue whether it’s total or per capita generation that matters.

If it’s about the living standards of individuals, per capita makes more sense. A country could use a lot of electricity but still have most people living in energy poverty if it has a large population. If it’s about industrial capacity or dominance, total electricity is more revealing: your ability to dominate global solar panel or battery markets, robotics or other manufacturing sectors, is about how much stuff you can build and sell, not how much you can do per person.

Nonetheless, let’s say these per capita comparisons do matter. America’s dominance is not as stark as these electricity-per-person figures first suggest. That’s because the US and China use electricity for very different things. The US uses most of its electricity in households and commercial services; China uses most of it for industry.

Here is the comparison of industrial electricity use per capita. The main source of this data is the International Energy Agency (IEA).1 You will notice that the US is shown twice: one line using the IEA data, and another using the US Energy Information Administration (EIA) for comparison. I’ll explain why I’ve done this in the footnote.2

Per person, China consumes slightly more industrial electricity than the US, and again, its trajectory looks very different. Both use more than the European Union.

The big per capita gap between the US and China comes from the residential sector. Per person, Americans use more than four times as much as the Chinese at home.

The difference in electricity allocation is really stark when we look at all of these sectors together.3

China’s huge growth in electricity demand is really about industrial capacity.4 It not only uses far more than the US in absolute terms, but it also matches it on a per capita basis.

This story is hidden in the raw per capita comparisons that everyone shares as a rebuttal to Musk’s (and others’) argument at the top.

Subscribe now

2

If you plot the raw IEA series for “Industry” for the US, there is a very large drop in electricity use in the early 2000s. While directionally correct, the magnitude of this drop seemed implausible to me.

After looking at the other sectoral data, this drop happens alongside a substantial rise in electricity use in the “Other end use” category. I think there is a reclassification of some industrial electricity in the US in the IEA dataset around this time, which leads to some odd results.

For the IEA series shown in the chart, I’ve therefore combined the “Industry” and “Other” category for the US. To check that this wasn’t unreasonable, I cross-compared it to the US EIA’s dataset, and it gives a decent (but not perfect) match.

My IEA series for the US might contain small amounts of electricity that are not industrial, but without more granular breakdowns, I couldn’t separate out the industrial from non-industrial use in the “Other” category. I still think that for the sake of this comparison, the estimates seem reasonable.

3

Here is the breakdown from the International Energy Agency:

The motivation for this was simple: it’s extremely hard to get a sense of magnitude for how things compare, and I know people struggle with this a lot. That affects what people prioritise, and is a problem for public discussion. We get countless headlines focused on the energy saved from “deleting emails” or streaming Netflix, and not enough focused on swapping a boiler for a heat pump.

I asked for feedback on how I could improve it, and I got a lot of responses. Thank you to everyone who engaged, commented, and emailed.

I’ve spent some time over the last few weeks addressing many of these comments. For various reasons, I couldn’t implement all the suggestions (some would make things too complex, which goes against the point of the tool), but I did make quite a number of improvements.

In the tool, I’ve included a “Change log” which details the changes that I’ve made along the way.

Perhaps the largest and most visible one is that I’ve tried to provide a rough sense of energy costs for a selection of countries. That is often more meaningful to people than the number of watt-hours they’re consuming.

I’m still very open to more feedback (especially if you think something is wrong!), but I won’t be as active in changing things as I was this time around.

I hope it’s useful for you in some way. Feel free to use it and share.

Does that use a lot of energy?

Energy comparison tool

First, thank you very much to everyone who played around with the Energy Comparison Tool that I published yesterday, and wrote to me with feedback. I received many emails and comments with great suggestions.1 I feel lucky to have an audience that is so engaged in trying to build something useful.

As promised, I read them all and I’m working on improvements. I’ll add a change/update log to the tool, so you can follow what’s changed. I’ll share more on this in the next few weeks.

There was also a lot of enthusiasm for a carbon emissions equivalent. It will take quite a bit of work, but it’s now on my list.

My new book — Clearing the Air — is published in North America today

It seems like it’s been ages since it came out in the UK (and some other international markets), but I’m excited for it to hit the shelves in the US and Canada.

You can find it in all good book outlets, and a selection of places here on MIT Press’s website.

Order 'Clearing the Air' in the US

I’ve been shortlisted for the Unwin Award

I’m very pleased to have been one of six authors shortlisted for the Unwin Award. It was launched last year by The Publisher’s Association, and nominations are based on “recognising non-fiction authors in the earlier stages of their careers as authors whose work is considered to have made a significant contribution to the world.”

It’s quite an honour to be recognised in this way, and to be among such fantastic company on the shortlist. Thank you very much to the judges for the nomination.

One area where this is extremely apparent is in energy use and carbon emissions. It’s hard to know what matters a lot, and what very little.

A lot of my work with data is about getting a sense of whether something is big or small, effective or not. Is that a big number?

To help, I’ve built an interactive tool that lets you compare the energy use of different products and activities: from lighting and cooking, to heating and driving. This is focused on “household” energy use: what people consume at home or while commuting.

Does that use a lot of energy?

In the video below, I’ve shown a quick example of what you can do with it. Select the products or activities you want to compare, and adjust the basic inputs, which are usually the number of hours or minutes of usage, or miles driven.

To build this, I developed a dataset of energy consumption based on typical usage. As I stress in the tool, these are approximations. They will not reflect your life down to the watt-hour level. I have assumed an LED lightbulb is 10 watts; yours might be 8.5.

Getting caught up in these small differences is actually against what I’m trying to do here. I’m trying to give people (and myself) a sense of what’s small and big. What order of magnitude are we talking about? Is driving 10 miles equal to 10, 100, or 1000 hours of lighting?

If this tool looks useful and is making a difference, I might build a similar one for carbon emissions (it will allow me to include things like food and other purchases).

Do you have feedback?

This is still a work-in-progress, so if you have feedback to share, I’d really appreciate it. There are now more than 70,000 of you following this newsletter, so I can’t promise a reply to everyone, but I do promise to read your messages and use them to make improvements.

Below the tool, I’ve written a methodology document detailing my assumptions for each product. It was a long list, took quite a bit of work, and there might be a few mistakes in there.

Feedback that is useful is along the lines of:

• Your assumption for the energy rating or power consumption of X is really off

• This component/configuration is broken

• This tool would be much more useful if it had X functionality

• It would be great to add [some product/activity not listed]

Comments that are less helpful:

• Those focused on small differences between my numbers and your actual usage. As I said, I’m going for typical approximations here; it won’t reflect precise reality for everyone

Things that make the tool a lot more complex. The simplicity here is important and deliberate, so people can make relatively quick comparisons. I don’t want people to have to put in the specific model of television, dishwasher, or car; or give precise details about the fabric of their home.

Try it out

In it, he touches on one question I haven’t seen a particularly good answer to: Do meat and dairy subsidies make them much cheaper?

Bruce thinks the answer is no. It reduces the price a little, but not much. This section in the book was fairly brief, so I thought I’d try to dig in a bit more to see what I could find.

This post is fairly long and dense to get into the details, so here’s the TLDR: Removing subsidies would raise meat prices a little, but only by cents to tens of cents per kilogram. Across the EU, US, and global modelling, removing direct subsidies would likely raise retail meat prices by well under 5%, and often closer to 1% or 2%. This drop would not be enough to change the economics relative to meat substitutes. I should also say that I found it extremely difficult to find a clear answer on this (for reasons I’ll explain), so if you’re hoping for a definitive dollars-per-kilogram figure, apologies in advance.

Let’s start with the common argument we’re here to test. As I’ve written about before, meat substitutes tend to be more expensive than their equivalent meat product. Here’s the chart I made a few years ago.

One response to this — and, in fact, the title of my article — is that meat substitutes are too expensive, and need to get cheaper. Another is the opposite: that meat is too cheap. Or, artificially cheap.

That’s where subsidies come in. It’s true that many countries subsidise their meat and dairy industries (as they do for some other agricultural products). This can make animal products cheaper than they would be in a “free” or “fair” market. Consumers might not be paying the true price, but someone is indirectly paying through taxes.

The solution (if you’re in favour of accelerating a transition to more plant-based diets) seems simple then: take the subsidies away and make it a level playing field. Or, better yet, reallocate them to alternative products that have a lower environmental impact.

But how large are these subsidies, and do they really make meat and dairy cheaper?

Direct (explicit) versus indirect (implicit) subsidies

Before we get into the numbers, we need to be clear about definitions and what we’re measuring here.

It’s extremely common to hear large figures quoted like “meat and dairy subsidies are $X billion dollars”. We hear similar statements about fossil fuels, the common one being “Fossil fuels received $7 trillion in subsidies”.

This does not always mean what people think it does. Sometimes people imagine that fossil fuel companies are literally being handed $7 trillion a year. Or meat and dairy farmers get that amount in handouts.

These figures often include several components. There are direct (or explicit) subsidies, which are, in a simplified way, handouts. That either goes to the producers of these products or is used as tax breaks or price cuts for consumers. Then we have indirect (or implicit) subsidies. There are basically negative externalities, such as damage from air pollution and climate change. To be clear: these are real and genuine costs (although we can debate what size those costs are) borne out across society. But they are not handouts.

This chart is from an article I wrote last year, looking at how much subsidies fossil fuels receive. As you can see, explicit subsidies are around $1.3 trillion. The estimated cost of externalities is much larger: $5.7 trillion.

Being clear on this difference is important. I think some people, when they hear “$7 trillion in subsidies”, imagine there is some $7 trillion money pot sitting somewhere that could either be taken away or reallocated to something else. In reality, the pot is “only” around $1 to $1.3 trillion. The other $6 trillion is costs that are dispersed across society, and harder to collect.

It also matters for understanding which levers you would use to reduce these subsidies. Again, you cannot pay for the externalities by pulling subsidies (there is nothing there to pull). The answer would be a carbon tax: charging producers or consumers more to cover the costs of the damage.

What’s true for fossil fuels is also true for meat and dairy. They do create externalities in terms of greenhouse gas emissions, air pollution, land use, and other environmental damages, so there are indirect subsidies (again, we can debate the price). The lever to pull there if you wanted to factor in the “true” cost of these products would be something like a meat tax (or carbon tax if you just wanted to account for climate damages).1

That is not what I’m going to be looking at here. I’m only looking at the direct subsidies farmers and food producers receive. These are the subsidies that a government could, in practice, remove or reallocate, if it wanted to. These are the ones that people would argue “distort” the price difference between meat and meat substitutes.

I can’t cover every country here, so I’ll focus on the US and the European Union (EU) markets, which are often the focus of government subsidies.

How much do subsidies reduce the cost of meat and dairy in the EU and US?

It would be really nice to give a concrete “subsidies reduce beef prices by $X per kilogram” number. Unfortunately, that doesn’t really exist.

That’s because not all subsidies are passed throughout to consumer prices. They can be given as “decoupled payments” to farmers, which are subsidies that are not linked to how much they produce. These are often given to maintain good agricultural land and are therefore reflected in land prices or rents. Since they’re not tied to any marginal unit (X amount per kilogram of meat), these decoupled tend to have almost no impact on shelf prices.

Farmers also receive other payments, though, often in the form of animal feed subsidies or insurance. When crop subsidies are allocated to livestock via animal feed, they account for a large share of the support embodied in animal products. These can reduce consumer prices, but not by as much as the raw numbers suggest. That’s because not all of these subsidies are passed through the production chain. Often, they’re there to keep farmers in business.

Let’s look at some examples.

One of the key recent studies in this area looked at EU subsidies.2 Its Common Agricultural Policy budget — basically payments to farmers — was around €56 billion. It suggested that 80% of this went towards animal products using the authors’ specific accounting approach (so, €45 billion).

Note that this 80% is somewhat disputed as it was based on data from 2013. Another study in this area suggested this was actually more like 50%.3

But let’s go with €45 billion. That’s quite a lot of money. But how much is it per kilogram of meat or milk?

The authors did their own calculations that are far more detailed, but let’s do our own quick sense-check first.

The EU produced 41 million tonnes of meat (of all types) and 140 million tonnes of milk (raw equivalents, before conversion into dairy products) in 2013. So, around 180 million tonnes in total. Divide our €45 billion by this figure, and we get around €0.25 per kilogram of product.

Now, this is very simplified for several reasons: we’ve used raw production rather than retail weights and are assuming subsidies are split evenly across all meat and dairy, which isn’t the case. But it gives us a sense of the magnitude of how these large subsidies break down per kilogram.

In the paper, the authors estimate €1.42 per kilogram of beef, €0.15 for chicken, and €0.28 for pigmeat.4 So our very crude calculation wasn’t that far off!

How much of a difference would this make to their price?

At my local supermarket, a relatively cheap but reasonable quality chicken breast costs around £7 (also €7) per kilogram. A 15-cent discount would cut the cost by around 2%. For bacon, a 28-cent discount might cut the cost by 4%. Beef costs about €14 per kilogram. A €1.42 discount is 10%.

It wouldn’t close the price gap to meat substitutes, though. In my previous analysis, I found the gap for products like chicken was often €3 or €4, not €0.15. The same for bacon or sausages.

But the crucial point is that these are the absolute maximum price impact these subsidies have. It’s probably quite a bit less because only a fraction filters through to consumer prices. Much of it goes towards farmer land rents, inputs or incomes. In reality, that 2% discount might be less than 1%, and the 10% more like a few per cent.

Are things any different in the US? No, if anything, the impact is smaller.

A modelling study suggested that phasing out tariffs and subsidy policies in the US would lower the farm price of beef cattle by about 3.3%, while pig, chicken, and milk prices would fall by far less than 1%.5

Finally, we can broaden this analysis to global markets, drawing on a paper by Springmann and Freund.6 They estimated that, globally, agricultural support measures totalled $233 billion in 2017. Again, not small. Around one-fifth of this was for meat production, and 10% for milk.

They modelled the impact of removing subsidies completely; commodity prices of most meat and dairy products were expected to increase by around 1% to 2%. The largest increase was a 3.5% rise for beef and lamb in OECD countries. Again, not all of that would be passed through to consumer prices.

It’s frustrating to not be able to pinpoint a specific “consumer prices would rise by $ per kilogram” from the research. But my takeaway from this is that, at most, prices would increase by 1% to 2%. We’re talking about cents, or at most, tens of cents of a difference. And this would not come close to closing the price gap with meat substitutes.

Wouldn’t lots of farmers go out of business without subsidies, and that would drive prices up?

One reason subsidies have less impact on consumer prices than we’d expect is that many subsidies are absorbed by farmers as land rents or income support.

But there is, of course, the argument that if these subsidies were removed, some farmers would go out of business, countries would produce less meat and dairy, and prices would increase.

There is some truth to this.

Around three-quarters of all farms (not just meat and dairy ones) would be expected to lose some income if EU subsidies were removed. How many of those would struggle to stay in farming?

I couldn’t find much recent data on this, but surveys from France around a decade ago found that around one-fifth said they would exit farming if all subsidies ended.

Among livestock farmers, smaller-scale beef and dairy farmers would probably take the largest hit to income and margins. Larger, more “efficient” producers of chicken, pigmeat and eggs, less so. In fact, some might be able to absorb some of the production losses elsewhere and see their incomes increase.

Overall, it seems reasonable to assume that production might drop a bit, but not so dramatically that prices surge. Again, we’d be talking about something in the range of a few percent, perhaps up to around 10% for beef.

Another data point to rely on here is New Zealand’s historical experience. It removed all agricultural subsidies in 1984, with many predicting huge drops in production, rural poverty, and farmer exits. But the impacts were far smaller than expected; most farmers found ways to adapt and build sustainable and profitable businesses without financial support. That is only to say that the real-world impact of removing subsidies would likely be lower than pre-policy surveys or expectations, because these often underestimate adaptation and innovation from farmers.

Once again, meat substitutes need to get cheaper

Here are the conclusions I came away with.

First, meat and dairy subsidies are pretty large in aggregate. Tens of billions of dollars every year. You could argue that reallocating these to meat substitutes — relatively immature technologies — could make a big difference to their development.

Second, when we break this down per kilogram, it’s not much. Removing subsidies wouldn’t change prices much. We’re probably talking a maximum of a few percent for most products, maybe slightly more for beef.

Third, this difference falls far short of closing the gap with meat substitutes.

I stick by my conclusion from my previous article: meat substitutes need to get cheaper. I also think this is important if we want to reduce the environmental impacts of food while ensuring people can afford a nutritious diet (and continue to support farmers who put food on our plates). A solution that relies on making food more expensive sacrifices the latter more than I’d want.

In this sense, it’s similar to where I stand on energy. Simply making fossil fuels more expensive without having affordable alternatives just makes energy more expensive, which hits the poorest hardest. That’s why the huge drop in the price of renewables and storage is so crucial to building a more sustainable energy system. You could say the same about food.

It’s at this point that some will argue we already have cheap plant-based protein in the form of beans, peas, and other legumes. That’s true, but people are not really switching to them. Across much of the world, the opposite is happening: people are switching from these plant-based staples to meat and dairy as their incomes rise. That gets to the point of Bruce Friedrich’s book: we have to find a way to meet that demand and want in a way that is better for the environment and animals alike.

Subscribe now

Speaking of books, my second one — Clearing the Air — comes out in North America next week (February 17th). Many of you have been asking about it since it came out in the UK. You can now find it here.

According to the International Diabetes Federation, it’s Pakistan.

Take a look at the map below, which plots the prevalence of diabetes among 20 to 79-year-olds.

Now, this data is what we call “age-standardised”. The risk of diabetes, like many diseases, increases with age. So if we were to map the raw (or crude) rates of diabetes, it would strongly reflect how old populations are.

To understand changes in prevalence and risk beyond aging, we use age-standardised metrics that hold the age structure of the population constant over time and across countries. It imagines that the age distribution of every country is the same.

“Diabetes” here includes both type I and type II, but the vast majority of cases are type II. And it doesn’t reflect diabetes diagnoses across the population. Why this matters will become clear later. They are estimates of how common it is based on blood glucose surveys among smaller, representative populations.

Countries like Pakistan, Saudi Arabia, Egypt, Malaysia and a number of Pacific island states have very high rates of diabetes. Pakistan’s share is more than double that of the United States and four to five times that of most European countries.

Why?

Diabetes is a disease where the body cannot properly regulate blood sugar levels. It fails to do so because the body becomes resistant to insulin, the hormone that allows cells to take up glucose. Obesity is the largest modifiable risk factor for type II diabetes because excess fat — in particular, visceral fat — increases insulin resistance.1

Below, you can see the relationship between body mass index (BMI) and the relative risk of developing diabetes across studies involving millions of individuals.2 It’s not a perfectly straight line (for some reasons I’ll come on to), but the relationship is strong.

While obesity was mostly a Western (or rich country) problem 20 to 30 years ago, it isn’t anymore.

A growing share of populations across Asia, Africa, South America, and the Middle East are overweight. This often happens while many are still underweight and undernourished (the double, or even triple, burden of malnutrition). In some countries, this rise in obesity has been rapid, far faster than the historical rise in rich countries.

So that is the first explanation: obesity has been rising quickly in countries like Pakistan, India and Malaysia. But that can’t be the full explanation: their obesity rates are growing, but they’re still lower than the US or the UK. How can diabetes rates be two, three, or even four times higher?

The risk of developing type II diabetes if you’re overweight depends on other factors. One is ethnicity.

Let’s plot the ratio of the prevalence of diabetes to the prevalence of being overweight. This helps us isolate differences in vulnerability or susceptibility to developing diabetes, independent of obesity rates.

A higher ratio means someone would be more likely to develop diabetes in that country if they’re overweight. On this map, Asia (particularly South Asia), North Africa and Sub-Saharan Africa are flashing red. Based on these ratios, we might say that someone in South Asia is three or even four times more likely to develop diabetes if they’re overweight than someone with a European genetic background.

That is perhaps oversimplistic, but it’s broadly in line with what we find from the medical research. People from South Asia, in particular, are more likely to develop diabetes at a lower BMI than Europeans or North Americans. This is reflected in many national dietary guidelines. The UK’s National Health Service, for example, uses different BMI scales depending on a person’s ethnicity. As they explain in their calculator:

“The calculator will also ask for information on your ethnic background. This is because people from an Asian, Black African, African-Caribbean or Middle Eastern ethnic background have a higher chance of developing health problems at a lower BMI. When you enter information on your ethnic background, the calculator will give you more accurate advice about your BMI result.”

When I select “White” as my ethnic background, it tells me that a healthy BMI range is between 18.5 and 24.9. When I select “Asian or Asian British”, my healthy range is narrower: just 18.5 to 22.9.

The UK NHS basically says that, if you’re white, you’re overweight if you have a BMI of 25 or more. But if you’re from an Asian background, it’s 23.

There are a couple of (at least partial) explanations for why this might be the case. First, for a given BMI, some ethnicities tend to have higher overall body fat percentages. Second, where that fat is stored matters. South Asian populations, for example, tend to have more visceral (or abdominal) fat, which accumulates around key organs and is particularly bad for insulin resistance. Even if Europeans had the same amount of fat, more of it is stored subcutaneously — that is, the pinchable fat on someone’s stomach, thighs, and arms — which is less metabolically harmful.

As a sidenote, this distinction is one reason why men are at a higher risk of chronic diseases like heart disease, stroke, and diabetes (and don’t live as long). They’re more likely to accumulate visceral fat. It’s not uncommon to see men with relatively skinny arms and legs have a protruding belly.3 Women hold much more subcutaneous fat, which is actually much better metabolically.

Exactly why there is this genetic lottery when it comes to fat distribution and chronic health risks is not clear to me. While there are a range of hypotheses in the scientific literature, I couldn’t find a clear answer.

These differences matter for how we understand the shifting burden of diabetes (and other metabolic diseases). The geography of diabetes is changing faster than many health systems are prepared for. In countries where diabetes risk rises at lower levels of body fat, even modest increases in overweight can translate into large increases in disease. This is happening in countries where getting a diagnosis is difficult, and getting treatment is even more so.

That means most of the global burden is invisible. The longer these cases go untreated, the harder they are to treat and reverse.

Subscribe now

First, we ran a small thought experiment. The world currently uses around 32 million hectares of land for biofuels, roughly the size of Poland or Germany. Let’s imagine we wanted to keep using that land for energy production, but switch to another source. How much electricity could we generate if we put solar panels on it?

The numbers were quite staggering. So staggering in fact, that I doubted myself. I ran the calculations many times, convinced I’d accidentally added a zero somewhere. I asked Pablo to also come up with an estimate, without telling him how I got to my numbers. As it turns out, we took slightly different approaches, but landed somewhere similar. We wrote up all of our assumptions and methodology if you’re interested.

If we put solar panels on those 32 million hectares of biofuel land, we could generate around 32,000 terawatt-hours (TWh) of electricity. Incidentally, that’s the same amount of electricity as the world consumes in a year.

So we could keep the biofuels, which amount to around 1,400 TWh of energy, and meet around 4% of global transport demand. Or we could use it for solar and produce enough electricity to meet the world’s current electricity demand.

Of course, there are other things to consider when it comes to solar: energy storage, where these panels are built compared to where the electricity is needed. Our point is not that we should cover all of this land in solar panels, or that it would easily power the world on its own.

But this small thought experiment does make one thing clear: there are many, many more land-efficient ways to produce energy than growing biofuels.

We then ran a second thought experiment. Let’s say we specifically wanted to use that biofuel land to decarbonise transport. That is, after all, what most liquid biofuels are used for. How much of the world’s car and truck fleet could we power with solar on that land if they were all electrified?

The answer is all of them, and then some.

We estimated that based on current global mileage, an electrified car and truck fleet would consume around 7,000 TWh of electricity. Less than a quarter of how much of solar could be produced on biofuel land.

Again, we can use biofuels to meet 4% of transport demand, or we could use that land to power all road transport.

These comparisons all boil down to the efficiency benefits of solar over crops. Crops convert less than 1% of sunlight into biomass; solar panels convert around 20%. Add the efficiency gains of electrified cars over liquid fuels, and the efficiency gap widens further.

I’ve often wondered why biofuel land use doesn’t get the same public attention as solar or wind power. I think part of the reason is that many people don’t really think about agriculture as “land use”. Humans are encroaching on natural landscapes when they put metal and concrete down. That’s why sprawling cities, towns, and roads are seen as the big “human land use problem,” even though they only cover around 1% to 2% of habitable land. Agriculture makes up around 50%. But because it’s plants growing, it seems like a more natural part of the landscape.

The same applies to renewable energy technologies. Biofuels are just plants. Solar panels are big chunks of metal, concrete, and minerals.

You can read our full article on this here.

How much of global aviation could be powered by biofuels?

In another article, we looked at the numbers on aviation.

Here’s a brief summary of the main points.

1. Currently, less than 1% of the world’s biofuels are used for aviation. Almost all of it goes to road transport.

2. Those biojet fuels account for far less than 1% of global aviation demand.

3. If we took all of the world’s biofuels and allocated them all to aviation, it would meet a maximum of one-third of current demand (and probably quite a bit less once we take co-products and conversion losses into account). This comparison is in the chart below.

4. Waste cooking oils and fats get a lot of attention from aviation companies. If the world were able to collect all waste oils and fats, we estimate it could meet up to 4% of global flight demand. Not nothing, but waste oils won’t decarbonise the world’s planes.

You can read our full article on this here.

Again, none of the above is a specific policy proposal. We wanted to get a sense of perspective on the relative land use of different decarbonisation options. That can then be one input into the decisions that policymakers, investors, and others make about where to go next. If we were surprised by the numbers, I’m sure many others will be too.

Subscribe now

“The media has told you that the bees are dying. But look at this chart: there are more bees than ever!”

They’ll then show a chart like this:

Or similar data showing the global production of honey and beeswax:

Now, there’s nothing wrong with this data. It is published by the Food and Agriculture Organisation of the United Nations. You’ll find similar data from the US Department of Agriculture (USDA).

But the interpretation of this data — and what it means for bee populations more broadly — is sometimes off.

What this data shows is the number of honeybees (or, in this case, beehives, which are often used as a proxy for individual bees) that are farmed to produce honey and beeswax.1 They are, essentially, livestock.

It’s true that managed honeybee populations are doing well overall.2 Hive numbers have risen over time, and global honey production has increased too.

[Note that this is not true everywhere. There are fewer beehives in the United States than there were 50 years ago. I’ve plotted this data across selected regions here].

Yet every few years, there are headlines about honeybees in crisis somewhere in the world — sometimes even warning of a world without any bees. These concerns are often real. Honeybees do struggle: beekeepers can lose a large share of their colonies each year. But somehow the system muddles through. Colonies are replaced, production continues, and the headlines become fodder for people trying to dismiss concerns about insect decline.

But crucially, “bees” doesn’t just mean “honeybees”. A chart of honeybee populations — which is basically just one species, Apis mellifera — tells us nothing about how the thousands of wild bee species are doing.

It would be like showing this chart of chickens (which are mostly factory-farmed) and saying, “People say that bird populations are struggling. But there are more chickens than ever!”

Or showing the number of farmed cows and concluding that global mammals are flourishing.3

If all we care about is food production, these metrics are useful. We’ve become extremely good at raising livestock at scale: producing feed, managing disease, and keeping populations productive.4

But again, the debates about the state of honey bee populations are completely disconnected from what’s happening to bumble and wild bee species. Just as conversations about cows or pigs are disconnected from ones about elephants or chimpanzees.

Why should we care about wild bees?

A couple of reasons. First, the intrinsic one: if you care about protecting a wide range of species on the planet, you might care about maintaining more than one species of bee. Second, the practical one: wild bees often pollinate different plants (and some crops). Wild bees can also be more effective pollinators. For many crops and wild plants, they deliver more pollination per visit than honeybees.

Wild bee populations are not doing so well. Global data is much poorer than the data we have on managed honeybees. But from the available evidence, I think it’s reasonable to conclude that they’re in decline. Not “insectageddon” level declines (again, overblown narratives on this make it easier for people to discredit valid concerns about biodiversity loss). But many species have seen a shrinking of range, population reductions, and are creeping up the rungs of extinction risk.

A range of local and regional studies find negative trends.5 And while global studies are rare, one major analysis using records in the Global Biodiversity Information Facility found that the number of wild bee species recorded each year has fallen by around 25% since the 1990s — even as the total number of records increased.6

Crucially, one of the threats to wild bee species is the presence of honeybees themselves.7 Since the latter are effectively managed populations, they can be introduced to an ecosystem in much higher densities. These honeybees introduce competition for nectar and pollen, and can also spread pathogens and parasites via shared flowers and hive movements.8 In many places, the honeybee is a non-native species competing with native ones for resources.

So, the chart of honeybees doing well might actually mean the opposite for other bee species.

To be clear: this isn’t about “good bees” and “bad bees”. Honeybees and wild bees both matter. But when you see claims about flourishing bee populations or rallying cries to “Save the bees”, it’s worth pausing and questioning what bees they’re talking about.

Subscribe now

Back in 2015, YouGov asked more than 18,000 adults exactly this question: whether the world was getting better, worse, or neither. Those who thought the world was getting better were in the firm minority. Below, you can see the results across various countries. Overall, rich countries were more pessimistic. This makes sense, as those in low- to middle-income countries have likely seen significant improvements in living standards over the past few decades.

Only a few percent of people in France, Australia, Germany, or Great Britain thought that the world was improving.

If people were pessimistic about the state of the world in 2015, I can only imagine that they are even more so today.1 The post-pandemic world seems gloomier than the pre-2020 one.

People also tend to think that things in their own country are going in the wrong direction. Again, this is most pronounced in richer countries; those at lower incomes tend to be less pessimistic following recent periods of strong economic growth and improved living standards.

But despite their global and national pessimism, most people tend to be optimistic about their own lives. Most say their lives are improving, and believe they will continue to do so.

This perception gap is what I want to explore in this post. This is not an article about whether people are “right” or “wrong” to be optimistic or pessimistic. It’s about the disconnect between how we perceive our own lives and everything happening around us.

Many of us are individually optimistic, but collectively pessimistic

I’m far from the first to highlight this gap. It’s a well-discussed phenomenon: people are often individually or locally optimistic, but collectively pessimistic.2 We have lots of data and research on this on Our World in Data.

Three years ago, Derek Thompson published an article in The Atlantic titled ‘Everything Is Terrible, but I’m Fine’ on this exact theme. It was primarily focused on the United States, specifically the US economy. Survey data from the Federal Reserve consistently show that people tend to be relatively positive about their own financial well-being, less positive about the local economy, and even less so about the national economy. I’ve updated this chart to 2024, which you can see below.3

But we see this mindset in many different areas.

First, on aggregate, very recent survey data from Ipsos Mori show that people are more likely to say that last year was bad for their country than to say it was bad for them and their family. In the chart below, I’ve plotted results from a 30-country survey. On the y-axis, we have the share of people who said 2025 was a “bad year for me and my family”. On the x-axis, the share that said it was a “bad year for my country”.

The blue line represents where these responses would be even; for example, where 60% of people said it was a bad year for them, and 60% said it was bad for their country. Every country, with the exception of Singapore, lies below this line, which means they were more likely to say it was a bad year for the country than they were to say it was bad for them personally.

For example, 85% of the French said it was a bad year for France, but only 45% said it was bad for them. The French were particularly negative compared to other countries. Sixty per cent of the Dutch said it was a bad year for the Netherlands, but only half as many said it was bad for them.

That same survey also suggests that people are also more likely to be personally optimistic about the future than they are to think that the country will start to feel more optimistic. 58% of Brits are optimistic that 2026 will be a better year for them, but only 32% think that Brits overall will start to feel more optimistic about the country’s long-term future. Again, this pattern was consistent across the countries in the sample.

People consistently think that people in their country are less happy than they actually are. On the x-axis of the chart below, we have the share of people in each country who said they were “very happy” or “rather happy”. In most surveyed countries, it was the majority: often over 80% or 90%. On the y-axis, we have the average guess that people made about what share they thought would say they were very or rather happy. These guesses were far lower than reality. In South Korea, for example, people thought that only 25% of their fellow citizens were happy. In reality, 90% were. Since all the countries are below the orange dotted line, people tend to underestimate others’ happiness everywhere.

As I’ve written about extensively, people consistently underestimate how many others in their country “believe” in climate change or support action on it.4 Again, we often think we’re in the minority of those who care about a given issue, but we’re not. We’re simply too pessimistic about others.

In the US, 49% of people say that crime at a national level is extremely or very serious. Just 12% say that crime in their local area is. As Gallup put it: “Americans have consistently seen crime in their local areas as less serious than crime in the country at large.”

Finally, people are more likely to see problems as collective, not personal: they say an issue affects the country, but not them. As Tom Calver showed in The Times, on almost every issue, Brits were pretty shy about saying it was important to them personally.

32% thought that immigration was an important issue for the country. But just 4% said that it was actually important to them personally. Very few say a lack of faith in politics is one of the most important issues affecting them personally, but a much larger share sees it as one of the biggest issues facing Britain. Crime, education, poverty, and the economy were bigger issues for the country than for them.

Of course, it’s not wrong that some issues affect others, but not us personally. It’s not incoherent for me to say that I’m not directly affected by poverty myself, but can see that it’s a problem for others in my country, or the world as a whole.

But the fact that some of these individual-collective gaps are so large, and almost always consistent (with the exception of inflation), is another nod to the “I’m Okay, But You’re Not Okay” mentality.

Why are we more optimistic about our own lives than other people’s?

There is a range of plausible theories as to why this gap exists.

One is that people are simply too optimistic about their own lives. In other words, they suffer from “optimism bias”. This would partly explain why they are more likely to be positive about their own future; they would overestimate the likelihood of good things happening tomorrow, next month, or next year. It’s less effective at explaining why this gap exists for single-snapshot perceptions, such as rating people’s happiness or their support for action on climate change.

Most theories tend to focus less on why we have a personal optimism bias and instead on why we have a pessimistic social one.

Here, there are a few related ones that make some sense to me.

Theory 1: Our information diet

Imagine a series of nested circles, with you at the centre. Each circle represents one zone removed from us: so, our partner, our family, our circle of friends, our local area, our country, the world.

Think about the “completeness” of information and experiences we have at each level.

We have a pretty complete picture of our own lives. Or, at least, we have a complete story we tell ourselves. We are aware of our history, our challenges, and how we have overcome them. We can see the positives in our lives, the victories. We know about our work, relationships, hobbies, and overall standards of living.

But as we move outwards from there, we have fewer inputs to rely on. We know a great deal about our partner and close family members (although some of those details are private).

We know a bit less about our local area, although we know what it’s like to live there, what the health services and schools are like, the place is clean and looked after, and we have a sense of how safe we feel walking the streets.

At a national level, we rely on small tidbits from people who live in other parts of the country or stories from the national news. What makes it into the national news? Usually bad stuff. The latest horrible murder, political scandal, or disaster. The exceptional, not the everyday, stuff. It’s hard to gauge the lives of tens or hundreds of millions of people, even in our own country, from these drops of stories and headlines alone.

And at a global level, we rely on very little information. Tiny slivers. If we’re basing it on the news, most of what we hear about other countries is negative: it’s a war, a disaster, a threat from one world leader to another. The rest is in darkness. How many of us can really say much about life in Kazakhstan, Peru, China, or [insert almost any country here]?

Most of what we’re hearing about the world as a whole is negative. That’s because bad things happen quickly (and that makes them headline-worthy), whereas good things happen slowly. We’re not hearing about reductions in child mortality, improvements in healthcare or schooling from the other side of the world. Of course, one way to overcome this is to look at statistics, but most people do not.

When someone is asked to reflect on their own life, they draw on a wide range of experiences. When people are asked about “the world”, they recall a very small number of recent events (the availability bias). Those events are likely to be negative, hence they give a pessimistic answer.

Theory 2: Agency and control

The same circle framework can be applied to our sense of agency.

Agency is often crucial to how people feel about the state of things: if they feel like they have some sense that they can do something to change it, they tend to feel more positive that it can be changed.

We have the most agency about our own life outcomes (or, again, perhaps the illusion that we have a lot of agency…this is not the place for a discussion about free will). If things are going poorly, it’s within our control to bounce back (and we often have previous experiences to show that we can).

People often think they have some agency over outcomes in our local area, but much less at a national level. In a recent Ipsos survey in the UK, 72% disagreed with the statement “I can influence decisions affecting the UK as a whole”. Only 56% disagreed that they “can influence decisions affecting my local area”.

At a global level, people feel absolutely helpless. They feel like they have no control over what happens to the world at all.

If we think everything is falling apart, we’re less motivated to improve it

Why does this collective pessimism matter? Again, I’ll try my best not to interrogate whether people are correct in their optimism or pessimism.5 But I do believe that deep-seated, blanket fatalism (which is stronger than pessimism) about our countries, systems and world is unhelpful if we want to improve things.

If we think our institutions can’t be saved, we become susceptible to grand promises from those outside of those institutions. We become less cooperative and trusting of others. Social cohesion breaks down.

If we think our health systems are corrupt and crumbling, then we become sceptical of public health advice. Some might question whether they should vaccinate their kids. If we think foreign aid is wasted money that improves nothing, it’s much easier politically for it to be taken away (with a tragic cost for the millions of lives it saves each year). We become less generous in our own personal giving.

If we think no one else cares about climate change, we stay quiet about it. We don’t expect or advocate for more action.

When we think everything is falling apart — and that’s the common narrative — we struggle to prioritise the areas that need the most attention. Failure to acknowledge where things are going well means we can’t learn lessons about what works and what doesn’t.

It comes back to agency. If we think that nothing can be done to improve things, we’re unlikely to try. This is one reason why I try to emphasise that there are things that each of us can do to make the world a better place. We don’t have to just sit on our hands. Without a sense of agency, we can become cynical and fatalistic that anything can change.

Subscribe now

The United States is the clear standout. It spends a lot more on healthcare than its peers, yet appears to receive little in return in terms of life expectancy. Americans have a lower life expectancy than other rich countries, despite paying far more.

Recently, I saw the economist Tyler Cowen spark a discussion on whether the US looks bad on life expectancy at birth because of far higher death rates at young ages. It has higher rates of young people dying from homicides, suicides, drug overdoses, and car crashes, and this really pulls down its “at birth” numbers.

The argument follows that once Americans reach older ages, their life expectancy is very good and comparable to that of other wealthy countries.

I’ve seen others make similar claims, so I wanted to figure out if it was true.

To do this, I used the latest estimates of life expectancy from the Organisation for Economic Co-operation and Development (OECD).2 This means we’re mostly comparing the US to other high-income countries.

The key takeaway is that young deaths are an important part of the US’s disadvantage, but not all of it. Americans who live to 65 still have lower life expectancy than those elsewhere. The differences narrow by age 80, but this is at least partly due to the selection bias of who survives.3

The US ranks extremely low on life expectancy at birth

Let’s start with a well-established point: the United States has a particularly low life expectancy at birth, both in terms of its level of income and its level of health spending.

Life expectancy at birth does not tell us what the most common or “modal” age of death is. It tells us the average lifespan a newborn would be expected to live if mortality rates at each age stayed the same throughout its lifetime. If the average life expectancy is 75 years, this does not mean a 50-year-old should expect the most probable outcome is that they die in 25 years.

In the chart below, you can see male life expectancy at birth in 2023 across a range of OECD countries, with the US in blue. It’s around 76 years. That’s less than both the mean of 78.2 years and the median of 79.4 years in the dataset. It’s about 6 years lower than the leaders.

Women in the US rank even lower relative to their peers (although they still live longer than American men). At 81 years, the US is close to the bottom of the list. Again, they’re around 6 years behind the leaders, and below the mean and median, which are both 84.

America does have particularly high rates of death among younger adults

A big part of Americans’ lower life expectancy can be explained by the fact that they are much more likely to die from causes such as car crashes, violence, suicide, or drug overdoses. In the chart below, I’ve plotted death rates from these causes among 15- to 49-year-olds in 2023.4 Each dot represents a high- or upper-middle-income country.

The US is in red. There are a few other countries that have comparable or even higher rates. In this dataset, the few that do are Mexico and Costa Rica.

The gap between the US and most other countries is big. Death rates from drug use are almost 10 times higher. Road injuries, three times higher. Violence/homicide is more than five times higher.

Another thing to keep in mind is that estimates of infant mortality really impact life expectancy outcomes. One issue is that infant mortality rates are sensitive to what countries consider to be a “live birth”. This can vary. In some cases, an extremely premature baby that takes just one breath, then sadly passes, is counted as a live birth, and therefore included in mortality metrics. In others, they would not be counted at all.

I previously wrote about this question of how infant mortality is counted for Our World in Data, if you want to read more. I found that the US does have a definition of a live birth that is more stringent (in this case, harsher for the US’s final statistics) than many other countries.5 But even when we adjust for that (which I show in the article), its rates are still high compared to other high-income countries. So its infant mortality counts probably play some role in its low life expectancy at birth numbers, but this is not the key explanation.

Life expectancy at age 65 is still not great for Americans

But what about those who make it to retirement age? Are they just as likely to live a long life as their other rich country peers?

The gap between countries at older ages is smaller. For men, the US remains firmly in the bottom half of the rankings, comparable to countries such as Germany, Finland, and Denmark. It’s two years behind leaders such as Switzerland, Australia, and France.

American women still lag quite far behind, though. The US is in the bottom third (granted, countries like the Netherlands, the UK, and Germany are too). It’s around 3 to 4 years behind leaders like Japan and Spain.

Another way to test whether it’s young deaths that make the US an outlier is to plot life expectancy at birth against life expectancy at age 65. I’ve done this first for men. Overall, we observe a strong correlation: countries with higher life expectancy at birth also have more years of life remaining at age 65.

For men, the US is somewhat of an outlier (hence why there’s a bit of distance between it and the trend line). But it’s not particularly dramatic.

The standout is Mexico. It has a particularly stark problem with very young deaths, which pulls down its life expectancy at birth figures relative to those for 65-year-olds.

For women, the US shows a fairly small deviation from the trend, with slightly higher expectancies at older ages for its life expectancy at birth figures.

Americans still spend a lot on healthcare for poor returns at older ages

The point that Americans spend far more on healthcare without significant gains in health outcomes is again clear when we plot the first chart in this article, showing how much countries spend on healthcare, but replace life expectancy at birth with expectancy at 65 years. This data on healthcare spending is adjusted for differences in living costs across countries and covers all financing schemes (both public and private).

The US is still a clear outlier. Around $12,000 is spent per American per year. Most other rich countries spend around half that. Some even less.

Here’s that same plot, but with female life expectancy at age 65. There, the returns on investment look even worse.

To be clear: healthcare alone doesn’t determine longevity — social conditions, lifestyle, and prices matter too — but the US is still a clear outlier on spending relative to outcomes.

80-year-old Americans can expect to live as long as octogenarians from other rich countries

So far, I’ve taken 65 years as the definition of “older adults”. Or rather, used it as a threshold for removing younger deaths from the equation.

Some may argue that 65 is not old at all, and that healthcare spending matters most for those far older than this.

The OECD doesn’t have life expectancy data for age groups above 65, but the Human Mortality Database does have a metric for individuals at 80. I’ve plotted this for the selection of countries we previously looked at.

The differences across most countries are small. On this metric, the United States does fine. Not at the top of the pack, but higher in the rankings than it was for younger metrics. Americans aged 80 can expect to live as long as their peers in other wealthy countries. Granted, it’s more expensive to achieve the same outcome, and this convergence partly reflects selection: in countries with higher mortality earlier in life, those who reach 80 are likely to be a more robust group.

Early deaths are a major part of the story, but they don’t fully explain the US gap. Even among people who reach 65, Americans have fewer years left than their peers, meaning the US disadvantage persists into older adulthood. This is particularly true for women. The gap narrows by age 80, at least partly because of selection.

Subscribe now

Over the weekend, the United States bombed Venezuela, and captured its president Nicolás Maduro. There has been a lot of speculation about the legality, true motive and implications going forward.

Oil has been a central part of the discussion. I wanted to get a quick overview of what the global picture looks like. So here are five(ish) simple charts that give some context on the history of oil in Venezuela, and why the United States — which is, by far, the world’s largest producer itself — would care so much.1

1) Venezuela has the world’s largest oil reserves

While we often think about the Middle East when it comes to large oil stocks, it’s Venezuela that has the largest proved reserves in the world.

The chart below shows the ten countries with the largest proved oil reserves. These are deposits that are deemed economically feasible to extract under current market conditions. This number can change as new reserves are found, or become economic.

Venezuela has about as much oil as Iran and Iraq combined. Its current proved reserves are more than four times larger than the USA’s.2

This data comes from the Energy Institute’s Statistical Review of World Energy.

Update: A few people have asked whether these reserve statistics are, in some way, politically manipulated to make Venezuela look far larger than it actually is. This is a valid point, so let’s take a quick look at why there might be concerns about this, and what the implications are.

The charts above — and actually many international datasets on oil reserves — rely on this definition of proved reserves. Again, these are quantities expected to be recoverable from known reservoirs with reasonable certainty under existing economic and operating conditions.

That’s one step up from probable reserves, where there is less certainty/confidence in the technological or economic feasibility.

The argument is that Venezuela’s reported proved reserves were revised upward— using optimistic economic and technical assumptions — because reporting large reserves has a political advantage. This is not an unreasonable assumption, especially when we look at a time-series of how Venezuela’s reported proved reserves have changed over time. This is data from the Energy Institute again.3

The big change happened from around 2007 to 2012, when parts of the Orinoco Belt were classified as proved reserves. The Orinoco Belt does have very large oil deposits. But most of it is “heavy” or even “extra heavy” oil (more on this later) that is often expensive to extract and process. The question, then, is whether this classification is too optimistic, and those reserves can’t really be recovered economically.

Of course, whether it’s economically feasible depends on the oil price. It might be profitable when oil prices are high, but a money-sink when they’re low.

Some analysts suggest far less of Venezuela’s reported reserves would qualify as proved under more conservative price and cost assumptions — closer to around 100 billion barrels. The other two-thirds is then less certain, and perhaps only recoverable when prices are high.

100 billion barrels is still a lot of oil! More than the US’s current proved reserves. But it would relegate Venezuela from being top to being around 7th.

2) Venezuela produces relatively little oil

Despite having a lot of oil, Venezuela doesn’t produce much compared to its own reserves and other large producers.

Here are those same countries (not necessarily the top ten producers, although the US is the world’s largest). Venezuela is bottom of the list. It produces less than 5% of the US’s output, despite having far more available.

3) Venezuela has an extremely large reserves-to-production ratio

Combining these two metrics gives us the “reserves-to-production ratio” (R/P). It tells us how many years a country’s proved reserves would last if production stayed constant and no new reserves became economically viable. Those are big ifs.

Because reserves and production can change a lot, this is not a prediction of when a country will “run out of oil”.

Since Venezuela has huge proved reserves and produces comparatively little oil, its R/P ratio is very large. In 2020 — a low point for oil production — the Energy Institute had it at over 1500 years. Note that this is based on the 300 billion barrels of proved reserves figure. If we think that’s too optimistic, and only 100 billion barrels are economically viable, it would be 500 years (still a lot!).

This figure is also very sensitive to production: if output were higher, the ratio would fall proportionally (for example, doubling production would halve the R/P ratio). So don’t focus too much on the exact number; the point is that it is big.

That contrasts sharply with the 11 years for the United States. Again, this does not mean the US will run out of oil in 11 years: its proved reserves fluctuate a lot as prices, technology, and investment change.

The contrast is stark: Venezuela has a massive proved reserves base that is not being translated into production, while the US is very active in production and continually replenishes its proved reserves through new drilling and development.

4) Venezuela produces much less oil than it did in the past

Venezuela’s oil production hasn’t always been so low. It was more than three times higher than it is today in the early 1970s, the late 1990s, and the early 2000s. You can see this in the chart.

Venezuela’s output comes from its state-owned oil company’s (PDVSA) operated fields and joint ventures with foreign partners (such as Chevron). In these joint ventures, PDVSA typically owns at least 60%.

Production dropped off rapidly post-2014, when oil prices were very low (which hurt a lot because oil was a large part of the economy). Chronic underinvestment in infrastructure and mismanagement throttled supply. Economic sanctions (including from the US) and limited access to foreign investment throttled export markets and financial support.

Nicolás Maduro, Venezuela’s now-captured President, first came to power in 2013. Most of this recent decline in oil output has largely occurred during his presidency.

Just for perspective, this is the difference in oil production between Venezuela and the US. While output in Venezuela has declined in the last decade, the US has been at or close to record highs in recent years.

5) Much of US crude oil production is light, while many refineries still rely on heavier crude oil

All oil products are not equal. You can get very different products, and in particular, different densities, ranging from “heavy” oil (which is very viscous) to “light” oil (which is much less so).

Oil extraction in the US has become increasingly “light”, particularly since the shale revolution (which produces light products). The issue is that many of its refineries are designed to process heavy oil, and it would be expensive to completely retrofit them.

That means the US still imports a lot of oil (mostly heavy oil), despite producing vast amounts domestically. It imports heavy crude oil from countries like Canada and Mexico, and exports lighter crude and refined products. You can see this in the chart below: it produces vast amounts of petroleum domestically, but still exports and imports a lot. This wouldn’t make sense if these were the same products, but they’re quite different.

Where does Venezuela come in? Well, a lot of its oil reserves are the “heavy” kind (or when it comes to the Orinoco Belt, “extra heavy” kind), which the US doesn’t produce a lot of domestically. Venezuela is one potential source of heavy barrels compatible with US refineries.

In summary

• Venezuela has the largest proved oil reserves in the world. It has far more than the United States.

• A lot of Venezuela’s reported proved reserves are very heavy oils that can be expensive to extract (if they’re economically viable at all). So the exact figure is sensitive to assumptions about the economics of extraction, and the oil price. These are not cheap oil deposits to exploit.

• Today, it produces very little. Far less than the United States.

• That means it has a large amount of proved reserves that are not being translated into production.

• Historically, it produced a lot more, and in principle, could do much more again. However, the timeline for this is uncertain; it could take up to a decade for production to return to previous levels, even with substantial investment.

• Venezuela has exactly the type of oil (“heavy oil”) that the US lacks and needs. So its interest is not only about the quantity, but about the type.

Subscribe now