Great question!

There are currently three or four areas I'm really looking out for at the moment:

1. The Demographic and Health Surveys, which were collecting various types of data from many lower- and middle-income countries on population sizes, maternal and child health, urban living, income, water & sanitation, etc. and were terminated earlier this year. It costs around $40M per year. I think that's modest for such a large program, but also because it feeds into a very wide range of datasets used by researchers on many topics. Some of the data it collects are used as covariates or denominators (e.g. population size) for many metrics; and it's really important if we want to track progress and make estimates of the cost-effectiveness of other global health programs in those countries.
2. The 'urgent and vetted' list by the Project Resource Optimization team includes several likely highly impactful and cost-effective programs. They started with all the projects whose USAID funding was cut and whittled it down to what are likely to be the most impactful global health and humanitarian programs that were cut. I donated to one of them (installing clean water pipes in Nigeria), hoping it would help get it across the threshold to be funded, and it looks like it has since then!
3. The various clinical trials running abroad whose funding were cut, especially for new vaccines, treatments and testing methods for tuberculosis and HIV. I think there's a possibility this might be reversed or it could be restored in some way. But knowing there are some potentially promising treatments in the pipeline that could increase efficacy or reduce costs in the future that have suddenly lost funding for the final stages of testing is quite depressing!

I think the biggest ones are being able to put studies into context and convey accurately how much confidence people should have in them. 

I often see people hype up early-stage drug candidates as breakthroughs without conveying that they'll go through much more testing to confirm their efficacy & safety. 

I think that's a shame because it gives people a misleading picture of what's actually likely to work, might lead to disappointment later on & pessimism in science more broadly, and also means hearing about the wrong things - there often are great breakthroughs that have demonstrably worked in late-stage research or been verified by other teams, but people don't hear about them. It's strange as well because I think it's those later stage findings that are most useful to policymakers or people using them in their own lives, rather than early studies in a handful of mice.

Advice is much harder! I often talk to my friends and family about things I'm working on and take notes on things they find difficult to understand, or common questions they have, and I try to incorporate those in what I'm writing. I'd also suggest keeping up with researchers or methods in the field people are writing about - I don't think people need to do a PhD in statistics or methods in order to write about science, but it helps a lot to understand some of the common types of study designs and their limitations, and have an understanding of the replication crisis and think about potential red flags when reading research papers.

Great question! I think there are two ways to answer this — one is whether we can predict completed cohort fertility rates in advance, and the other is whether we should rely on other metrics instead.

On predicting cohort fertility rates:

To some degree, I think we can. One way is with tempo-adjusted fertility rates, where you look at the average age of mothers giving birth in a given year and how that has changed over time, then adjust the total fertility rates accordingly. I’m slightly skeptical of this though, because the average age isn’t always that informative: you could see important changes at the tails of the age distribution, which affect the final rate in ways that an average wouldn't capture. (There are also variations on this metric, of course.)

Another approach is to use cohort fertility rates up to a certain age (say, up to age 40) and then, based on assumptions about childbearing decisions and success rates after that, estimate what the final rate will be by the end of childbearing years. But this gets harder to project accurately, especially as new reproductive technologies make births more common at older ages and shift patterns in ways we haven’t seen before. Sometimes it's not a monotonic delay - after the baby boom, for example, the US actually saw a widening of ages at childbirth, after the baby boom contracted that period.

These would be helpful if you're interested in whether younger people are having more or fewer children over their lifetimes than older generations did.

But another way to look at it is, what question are people actually asking?

If someone just wants to know how many births happened in a year, they don’t care about how many will happen in the future, then looking at the actual number of births makes sense. This is useful for understanding the shape of the population pyramid, for example.

I've noticed that birth rates often trip people up. They’ve declined a lot over the 20th century partly because they use the whole population as the denominator — so as a country’s population ages, you’d expect the birth rate to fall, even if young adults are having just as many children as before. That’s why the total fertility rate (TFR) is generally more useful for this purpose: they use the population of women of childbearing age as the denominator, not the total population.

But if you’re specifically interested in the ratio of births to the whole population, then the crude birth rate is actually the right statistic to use.

Since you mentioned issues about the workforce, that often involves other factors also coming into play, like the size and productivity of the working-age population and how that's changing. I'd be asking questions like Are women spending more time working? As people are living longer and healthier, are fewer people retiring early due to disability? Are people working more productively? How many dependants does each person actually have? As infectious diseases & mortality rates have fallen, are people spending less time out of work caring for their sick children, parents or relatives? etc. So I think looking at births alone could give a misleading picture of the demographic challenges that countries might face in the future.

Thanks so much!

Great questions. Just for context, the Demographic and Health Surveys were primarily funded by USAID but that funding was terminated by the Trump administration in February. It involved surveys roughly every five years per country, for over 90 low- and middle-income countries worldwide since it started in 1985.

I think there are two parts to this: one is about the surveys that had started and were meant to be completed this year or next (that was in ~23 countries) and the other is about the longer-term future of the program.

I'm hopeful that the ongoing surveys can be funded and completed. It would be a real shame and quite wasteful if they weren't, especially for those that had already started collecting data (or in some cases actually finished this). But I'm not certain, and there's always a risk something else happens in the meantime, so hopefully continuing to write about it makes a difference.

In the longer term, I'm less sure. Having an independent organisation collect that data in a standard way across all those countries is really very valuable, and it's across many topics (population sizes, maternal and child health, urban living, income, water & sanitation, etc.) and quite widely used by many researchers.

On how to do it more efficiently, I'm not the best person to ask but some rough thoughts. The surveys involve multiple 'modules', and one way might be to scale down some of the modules to be collected from a smaller, randomly selected subset of the overall population. Or some subnational surveys could be scaled down if national level data is considered enough; or maybe some modules could be shortened to focus on the key questions. But these all come with trade-offs and it really depends on what the priorities are.

I've often thought that innovation and policy were underrated.

With innovation I'm often thinking: rather than choosing from existing tools that are already cost-effective, can we make new tools that are cheaper, more effective, or easier to scale up? And if that's the case, we should be investing more in them.

With global health policy as well, market-shaping, regulation, and other incentives often seem like they can make a much larger impact than traditional EA targets. I think about this less often these days though, after knowing more EAs. Maybe I just didn't know people who were already thinking about these issues? Or they've shifted focus?

To me, the big thing that feels missing is a clear sense of what, concretely, individuals can do about global health and foreign aid. There’s so much (often depressing) news right now, and I think journalism tends to cover global health and foreign aid as distant problems that feel hard to relate to. From my experience, what motivates people is the sense of connection: knowing that their actions can actually make a real difference to people. I think it helps to frame this more as a shared societal problem — things that anyone could face if they were in that situation — and to be clear about the fact that people really can help, and exactly how they can do that. (Maybe I should shout out my recent blogpost on why I decided to sign up to the 10% pledge and where I donated to.) But that feels like it's missing to me.

Very tough question... perhaps the toughest one here. Pistachio, honeycomb, and dulce de leche are probably my favourites, but maybe there are other exotic flavours out there I haven't tried yet. (I did try 'beer flavoured gelato' once and it was terrible.)

Answering Seb's question: 'Is AI for science underrated or overrated?'

Can I cheat by saying both?

Say we’re thinking about AI and protein structure prediction for drug discovery, for example. I’m quite excited about what that could make possible - it could help narrow down potential drug targets, improve our understanding of protein structure and function, and also give us a better sense of which drugs might fit particular protein structures. Protein design is also really exciting, including improving proteins or enzymes that are used as drugs or in industry.

But I’m often thinking about the bottlenecks too, and I feel like those are underrated right now. That includes the inputs — like the datasets that need to be collected to train models, who will actually gather them and how they'll do it — and the outputs of AI-driven research. For example, validating predictions in clinical trials is still one of the biggest bottlenecks in medical research.

Some parts of that process could be sped up with AI, but actually running experiments, recruiting participants, securing science funding, navigating policy and regulation, and coordinating how scientists work with other people; all of that broader system that influences science will still be there and still often inefficient, I think. (This isn't to say that AI can't influence lab science or other parts of the process, though; it's already used in improving DNA sequencing and microscopy, for example.)

But I liked Owlposting’s recent blog post on why AI didn’t replace pathologists, which makes a good case in point.

Re: Bad Data Takes's question:

This takes me back! I guess I've always enjoyed writing and explaining things I was learning about, but I started writing freelance pieces during my masters degree and PhD, including writing about the Covid response in the UK. 

Roughly then, Sam Bowman and I had the idea to start a magazine together for long-form writing on science and economics, which turned into Works in Progress magazine. 

For one of our early issues, I reached out to Hannah Ritchie at Our World in Data to ask if she was interested in writing a piece for the magazine (How we fixed the ozone layer), since I'd been a fan for a while, and we had also met briefly the previous year. She replied sayjng she was interested in writing for us and, surprisingly, also asked if I was interested in working with them at OWID, which I didn't expect at all. I remember having the vague thought before that that it would be a very cool place to work but thought they already had enough great people working there and I hadn't considered writing as a potential career until that point.

I was still doing my PhD at this point though, so could only work with them part time on the side - until a few years ago, when I graduated and decided to take their offer to work there full time.

Thanks for the question!

In a way, it's quite to really understand and imagine - my life would probably be very different if that was the case, and I doubt that I'd have the same education, qualifications, and sometimes also skills if that was the case. 

But I'd probably still be interested in something quantitative - maybe I'd apply to work at Data For India (which feels like a very close alternative to OWID), or try to work as an economist, data journalist or researcher in some way.

Maybe the most helpful thing to do, though, would be to explain some of the concrete problems people face in India to a wider audience. I think there's often a lack of data-driven journalism from lower- and middle-income countries that actually gets communicated to a global audience, and explaining them clearly seems valuable. (I tried to do this a little with this blogpost on snakebites, but there are likely so many other areas that could be written about.)