One of the authors of the original CGD blog here.
Hi Nick,
Thanks again for engaging. I don’t think your criticism is correct and I’ll try again to explain why here in more detail.
Our blog argued that after a UK visa policy change in 2020, there was a change in the trend or growth rate in both nurse migrants to the UK and new nurse trainees. To show this we present data from both before and after the policy change in 2020. The data you present from only post-2021 can't therefore refute our argument. We’re arguing that the situation could have been worse in the absence of the policy change, with even fewer new nurses being accredited in Nigeria.
Between 2018 and 2020 there was increasing migration to the UK, before the visa change, but the rate of increase dramatically accelerated following the visa change. In our previous blog we left implicit the idea of the counterfactual - that absent the policy change in 2020, trends would have continued as they had previously.
What about other countries?
You’re right to point out our lack of data for other destination countries. If our theory is correct (and it is only really a theory), we should expect to see a spike in nurse migration to the UK after 2020, and no change in nurse migration to other destination countries. The best data on this would probably be going through each destination country's records, but as a short-cut I took a look at the OECD data on annual flows of Nigerian-trained nurses to OECD countries. This data is I’m sure flawed, but it is entirely consistent with our argument. The OECD suggests that in the most recent year for which data is available the UK is by far the largest recipient of nurse migrants from Nigeria, with a flow of 1,709 in the latest available year, followed by the United States (87), Ireland (82), Canada (36), New Zealand (24), Germany (9), and Italy (1). Furthermore, the trends fit our theory entirely. Unfortunately, there is no data for the United States after 2015, but for Ireland, Canada, New Zealand, Germany, and Italy, there is no change in annual flows after 2021, whilst there is a huge spike for the UK.
Is this really plausible? It probably shouldn’t be that surprising that the largest flows from anglophone Nigeria are to other anglophone countries, with a particularly high demand for jobs in the UK given historical links and the fact that it is geographically closer to Nigeria than the other anglophone countries and in a closer time zone. For another source, a recent survey of Nigerian nurses found that the UK was the most preferred migration destination (Badru et al 2024, Investigating the emigration intention of health care workers: A cross‐sectional study).
What about the short time lag between the visa policy change and nurse training?
Is it implausible that nurse graduate numbers would increase so quickly in response to visa opportunities, given that training takes 3-5 years? This is a fair concern. If there was a steady pipeline of trainees progressing smoothly through the 3-5 year training course and all trainees then taking the professional exam at the end of their training period, then yes it would be impossible to see such a sudden increase. If however on the other hand there are significant numbers of trainees who have been enrolled on and off for a cumulative period of 3-5 years but had not previously sat the professional exam, for instance due to a lack of available job opportunities, then the emergence of new job opportunities could easily lead to the observed rapid bump in exam candidates.
Is migration reducing the availability of health care in Nigeria?
For this to be true, you would have to assume that all trained health care workers are able to find jobs in Nigeria in healthcare, which doesn’t seem likely to be true to me. You dismiss the World Health Organisation workforce data on nurses per capita because it is noisy, and argue we should instead rely on the data from the Nursing and Midwifery Council of Nigeria, a government agency. I have bad news for you, because the WHO gets its data mostly from government agencies. WHO supplements government data where it can with more reliable census or survey data, but its unlikely to be significantly less reliable than the official government data. Official data from low- and middle-income countries is often noisy, but what does seem apparent to me from the long time period available in the WHO data is that there is no ongoing downward trend in the availability of nurses in Nigeria. An alternative data source are the Demographic and Health Surveys, which show the share of births attended by a skilled provider. This is relatively flat over time from 1990 to 2021.
Ultimately our blog was speculative - we have a clear theory that training should respond to job opportunities, which seems to be consistent with the data we presented. As we wrote in the blog, this data is not definitive and doesn't prove our argument. But the data you have presented doesn’t contradict our argument either, and neither do any of the other new data sources I have consulted, whether from the OECD, WHO, or DHS. None of this data is perfect and we might be wrong, but I don’t think you’ve made that case yet.
Thanks,
Lee
Interesting stuff. At CGD we're hoping to follow-up with some old RCTs to find the ideal evidence on this question - what actually happened to incomes of people who have experimentally induced higher early grade test scores: https://www.cgdev.org/blog/will-raising-test-scores-developing-countries-produce-more-health-wealth-and-happiness-later
I think you've missed the best existing study on this question which is Glewwe et al using longitudinal panel data to track kids through to adulthood and look observationally at the wage gains from better early test scores https://www.sciencedirect.com/science/article/abs/pii/S0167268121004947. They find 13% higher earnings from 1 SD better scores.
Let me explain my position - first, I agree with rejecting a pure time preference, and instead doing discounting based primarily on expected growth in incomes.
For me, the expectation that in 50 years the average person could easily be twice as wealthy, leads to quite heavy discounting of investment to improve their welfare vs spending to alleviate suffering from extreme poverty right now.
It's possible I haven't thought this through thoroughly, and am explaining away my lack of enthusiasm for your choice of 5 causes to the neglect of the classic Givewell/GWWC choices. Perhaps there is something to do with efficacy there - that I'm unsure of the likely impact of funding immigration advocacy, forecasting, and more research.
I think your choice of discount rate is going to fundamentally alter your investment decision, it's not just some kind of marginal technical tweak.
In practice either you discount fairly heavily, as most public projects do, and end up putting most of your money into solving short-term suffering (as I think you should), or you discount lightly, and put most of your money into possible future catastrophic risk mitigation.
I don't see how this is "computational convenience" - it's fundamental.
What do you mean by "using discount rates on a case-by-case basis as a convenience for calculation"?
I don't find your dissertation discussion very convincing (but then I'm an economist). I worry a lot more about the existing real children with glass in their feet right now (or intestinal worms or malaria or malnutrition or whatever) than the hypothetical potential children of the future who don't exist yet, and in any case when they do will live in a substantially wealthier society in which everyone has access to good quality footwear.
Kudos for trying, its important work.