In part 1, we established that Oster is likely mistaken as to the mechanism behind FAS/D, and concluded there's no theoretical basis on which to assume there is a threshold effect for problematic drinking in pregnancy, particularly in the first trimester. We also established that risk and BAC were both dependent on weight, which absolute recommendations like "1 drink" don't account for.
In this part, we're going to address the claim that it's reassuring that European nations have both more liberal policies towards drinking in pregnancy and have less FAS/D, citing the so-called "American Paradox," a paper written in 1998 by Ernest L. Abel.
Oster writes:
Think about Europe. Much of the continent is much more permissive about light drinking during pregnancy. Heavy drinking is frowned upon everywhere, but some places in Europe have recommendations suggesting that a few drinks a week are fine. An occasional glass of wine or beer is much more common there. Yet there is no evidence of more fetal alcohol syndrome in Europe; if anything, rates are higher in the United States. If having a couple of glasses of wine a week lowered IQ, we would see big differences between the United States and Europe.*[1]
* As fetal alcohol syndrome is typically a result of binge drinking, it is possible that it could be due to the United States having more inequality in drinking—a lot of people not drinking at all and a few engaging in binge drinking—as opposed to other countries where most people drink moderately.
There are a lot of problems with this besides, namely:
It's ecological data which is pretty much the worst possible type of data to draw conclusions from. Ecological data is vulnerable to ecological fallacy amongst other statistical problems.
Some European countries used to have more liberal guidance. But by the time Oster was writing, this statement was essentially false; nearly every European country recommended abstinence in pregnancy.
We don't actually know if true rates of FAS/D are higher in or lower in individual countries in Europe, let alone Europe as a whole, because there are no universal screening programmes for it in any country, nor many good national prevalence estimates (Oster also mentions average country IQ, which is an even more conflated measure than FAS prevalence rates.)
Europe is very heterogeneous in terms of actual drinking prevalence during pregnancy, with some countries drinking more, and some less than the US.
For instance, France is commonly thought by Americans to be liberal about drinking in pregnancy, but France issued abstinence-only advice in 2002, eleven years before Oster published her book. Women in France today report drinking very little in pregnancy, not even light drinking; a survey in 2021 reported 1% of pregnant women having more than 1 drink a month, with 7% having had reported any use. Though the surveys aren't directly comparable due to different methods, surveys of alcohol use in the US report ~15% of women reporting any drinking, and ~4% reporting binge drinking.
The UK, where I live, was one of the very last European countries to change their guidance, and did so after the book came out in 2016. Today (2026) there are a few European countries that don't specifically recommend abstinence in pregnancy, but these countries simply have no guidance published on it at all. All the countries that highlight drinking in pregnancy recommend abstinence.
The ecological fallacy
The canonical example of the ecological fallacy has to do with literacy and immigration. In 1950, statistician William S. Robinson found that the higher percentage of immigrants there were in a state, the lower the illiteracy rate was. This implied that immigrants were more likely to be able to read English than native born Americans!
This is, of course, as Robinson writes, nonsense. When you look at individual level data, immigrants were more likely to be illiterate than native born Americans; it was that immigrants were more likely to settle in big cities in wealthier states. Those wealthier states had lower levels of illiteracy, but a higher proportion of immigrants. Adding in more immigrants just didn't have the power to reverse the direction of the effect. If you look at individual level data, rather than across the state, the correlation is reversed.
Oster clearly is familiar with the ecological fallacy. In her footnotes she gives one plausible example demonstrating this, that the US might have a different distribution of binge drinking, with more binge drinkers than in European countries. But this precise example, of course, supports her argument that light drinking is fine.
This pattern also happens to not be correct, as near as I can tell. I didn't check every country, but i.e. British women are more likely to binge drink than American women in pregnancy, not just drink in general. And French women today both binge in pregnancy less and also do less light drinking in pregnancy too. Per capita, France today consumes more alcohol than the US. But this isn't a paradox, it's simple ecological fallacy. Pregnant women drink less, and non-pregnant women and men drink more. But even though it's not true today, maybe it used to be true in France 30 years ago; let's fine out.
Prevalence of drinking and FAS: today and 30 years ago
It simply isn’t the case that countries with more liberal drinking rules (historically or today) have less FAS, at least that we know of. In reality we just don’t have enough data on it. The very first ever national estimate of prevalence of FAS in France was published in 2021, and only included hospital discharge data, which underestimates prevalence. I wasn't able to find national prevalence estimates in any of the other European countries I looked at (not exhaustive). I found two papers published after Oster that tried to compare country level prevalence; one used models based on alcohol use numbers, not actual diagnosed cases. I did find one slightly more modern paper that tried to (published 2017) with some some real prevalence data, which concluded, "The WHO European Region had the highest prevalence (19.8 per 1000 population; 95% CI, 14.1-28.0 per 1000 population)." However that data still mostly uses estimates.
The reason Oster ended up using a 15 year old paper is plausibly in part because ecological data is widely understood to be inferior to individual level data, so it's a fruitless research direction. And the real prevalence data which might be useful for other reasons that's available just wasn't, and today still isn't, good enough to do this analysis.
So what's wrong with this historical data? The 1998 paper Oster cites references conclusions of Abel's own analysis done in a book. The 1998 paper didn't report these data, but I tracked down the actual data tables that were used to argue this and reanalysed them to make it easier to understand the original analysis. Feeding Claude these data tables, it becomes clear the analysis is very naïve from the outset in a way that should have also been obvious in 1998 and does not support the idea of a paradox.
For one, it didn'teven show clear evidence Americans reported drinking less in pregnancy than the rest of the world, even in the early 90s and late 80s. Studies in various American cities run the gamut, with studies in Detroit reporting more drinking in pregnancy than any European city, and Philadelphia, which had less drinking in pregnancy than any European city. If you naively average across studies all with different methods, the confidence intervals overlap substantially. There is only one study from France, Roubaix, and it's middle of the pack at 48.5%. Roubaix was specifically chosen for this study because of high alcohol consumption per capita relative to other cities in France. So we don't know what drinking in France looked like then in general because it's almost certainly not representative. It's possible but would be surprising if a city cherry picked for a higher per capita rate was.
Claude was also asked to read table 11.4 on the relationship between heavy drinking in pregnancy with FAS rate into a CSV and plot them, and do a simple linear regression. Claude in this case reached the same conclusion as the book; no relationship. So Oster's repeating of the hypothesis in the paper of being more heavy drinkers in the US than in "Europe" to explain the "paradox" is contradicted by his own book (the discrepancy might be that the book was under review but not published by the time the paper was. However in itself it's very flawed, because the data aren't independent, caveat emptor.)
In the context of the American paradox, Table 11.4 indicates that the twenty-fold difference in incidence between the United States and other countries cannot be attributed to differences in the number of heavy drinkers at different sites. For example, the rate of heavy drinking in Cleveland, a city with one of the highest rates of FAS, was 1.7%; whereas Boston, with a rate ranging from 2.7% to 13.0%, had only one case of FAS; Australia [Scotland] [2], with a rate of 0.4% to 13.0%, and with an even higher rate of 35.6%, reported no cases.
Looking at the data, I think it's kind of strange the take home story of this data via a game of telephone ended up being was that Europe had less FAS even to begin with.
A striking majority of the studies reported zero cases of FAS, including most of those that took place in the US, with the exception of a few studies of specific cities in the US, France, and Sweden. The studies with no cases were generally small and did not actively screen for FAS, and we usually studying something else (i.e. prevalence wasn't the primary outcome). The sites in the US and France that reported cases were larger, and locations were cherry-picked by looking for places with high alcohol consumption, and the methods used in these studies were active case ascertainment rather than passive case ascertainment. They were simply not intended to estimate country wide prevalence, but to validate screening methods. They were used for prevalence after the fact. Nevertheless two Sokol papers with the highest rates were still not statistically distinguishable from nearly all of the European studies due to overlapping confidence intervals. It took pooling to do that.
Table 11.1 notes race and SES/class for each study, but those weren't used in the analysis. The data is analysed raw, with little attempt to statistically control for methodology, SES, or race, or region. Instead, Abel simply pooled all the cases together.
If you look at the pooled data in table 11.1, from which Abel concludes the US has 20 fold more FAS than Europe, Europe has 4 cases out of 18,766 births and the US has 91 cases of FAS out of 47,097 births. This is statistically significant. However, 87 of the 35,156 births in the US come from a single author (Robert J. Sokol) and three papers, which performed targeted screening infants of single hospitals in Detroit (1 study) and Cleveland (2 cities). 57 of 14,707 were from Detroit. Screening studies in low SES areas are valuable, but pooling the data naively like this implies 32% of the US as a whole is like Detroit and 45% is like Cleveland.
If you based anything else in the US like this off primarily these two cities- for instance, illiteracy rate - you would overestimate the prevalence of illiteracy as well, because illiteracy rates are lower in other areas of the country than they are in major cities, and Detroit and Cleveland in particular.
I'm quite confident overall difference in pooled cases between US and the rest of the world in Abel (1998) almost certainly comes down to different study methodology and the lack of representiveness of the populations used. (The 2017 study at least excluded passive case ascertainment and only included screened infants.) I would argue the central difference is the fact that Sokol lived in the US. In an alternate universe where Sokol ran the same three experiments in areas of the UK or Eastern Europe with heavy alcohol consumption instead, the pooled result would likely be flipped. The 2017 study that found higher rates in Europe than the US included additional data from Croatia, Italy, and France that was published after Abel but before Oster. It also only included screening studies, so a subset of the Abel data. To produce a US rate of FAS composed nearly entirely of data from Detroit and Cleveland but excluding similarly high drinking populations from Europe was never going to produce an accurate comparison.
I wasn't able to get all the papers cited, but for the ones I was able to check, confidence intervals between the continents overlap once you pool papers by whether they screened infants or not even if you don't bother controlling for any of the other factors like race and SES. This isn't even taking into account the screening methods were different; some only counted "complete" FAS (with facial characteristics) and some used much looser criteria, such as counting it as FASD where IQ and behavioural abnormalities were documented in children of women who reported drinking, despite an absence of distinct facial characteristics.
Coming back to the present, a study of Greater Manchester in the UK was published in 2021, and it found that indeed we did have FASD cases, and that in the UK prevalence was likely to be much higher than the actual diagnosis rate. But the study still couldn't provide a national prevalence estimate and there are currently no national prevalence estimates published for the UK that I could find, nor very many other European countries I looked into this for.
Studies that do active case ascertainment suggests the true prevalence of FAS/D is substantially higher than the diagnosis rate. There are good reasons for this; a diagnosis is highly stigmatising causes mental distress for the parents. There's also no treatment for it. Especially in public health systems where cost-benefit calculations are made for screening programmes, it's no surprise those programmes don't exist for FAS/D.
At the present time, this lack of data makes it virtually impossible to compare prevalence rates against countries in any reasonable way. And because ecological evidence is bad evidence to begin with, there's not much reason to invest in it. Oster states this lack of evidence as being reassuring; but could only plausibly be the case if prevalence had actually been studied and found no evidence of elevated risk. This lack of data is fundamentally neutral.
On public health advice
Though the recommendations all have now converged on no alcohol use, alcohol use itself in pregnancy in Europe still varies substantially. One multinational review in 2017 found alcohol use in pregnancy was highest in the UK (~41%) and lowest in Sweden (~9%) which is a large spread. (US is around ~15%). (I have found still other papers with wildly different numbers; and it changes over time, which makes a true comparison quite difficult.)
This entire argument seems to be a that old chestnut “Europe is better!” I live on the continent, technically, and there are many things I like about "Europe."
But the approach to children's public health, in the UK at least, I've found fairly lackadaisical in general. For instance, in the UK we don't screen children for lead poisoning or iron deficiency. We're also sluggish about adding new vaccines. It took 26 years after the US introduced the Hep B vaccine for the UK to start a universal vaccination programme, and 31 years after the US to introduce universal chickenpox vaccination. Supplementation of flour with vitamin B9 to reduce neural tube defects starts this year, 2026; the US started their programme in 1998, 28 years before. Recommendations on alcohol use in pregnancy lagging 35 years behind the US is generally in keeping with this, rather than evidence of the UK knowing something the US does not.
A fair argument is that the CDC and NHS's current recommendations don't really matter anyway because no one pays attention to them. This might certainly be the case. A study found recommendations to reduce alcohol consumption in the UK overall had only a temporary effect. But the very low alcohol consumption in pregnancy in France could mean that their 2002 switch to abstinence only worked. Or it could also just mean that France had a low level of consumption in pregnancy before the recommendations got stronger and labelling was introduced, since we don't have good data on that. Regardless, it's still interesting as to whether the recommendations are correct, too, even if it official advice is irrelevant.
The bottom line
The argument that Europe is more permissive about light drinking and has less FAS/D is false because:
The difference in FAS/D rate noted in the 1998 paper is almost certainly uninformative and comes down entirely to 1) over-representation of Sokol's papers in particular cities in the US rate, and 2) not controlling for methodologies across papers and over. No cross-country analysis between the US and any single European country has ever been done based on actual prevalence estimates. Newer studies estimate Europe has a higher FAS/D rate than other continents, but mostly these estimates are due to drinking rates and not real prevalence data.
Even if there were comparable estimates of prevalence, it doesn't solve the ecological data issue. Ecological data is notoriously difficult to interpret. The presence or absence of a relationship is probably meaningless at the outset.
Light drinking in pregnancy is not currently more permissible in Europe than in the US. The consensus in European countries had converged on abstinence in pregnancy largely before the 2014 publication. In 1998 it would have been accurate to say some European countries like France and Italy were more liberal than the US on drinking in pregnancy in their official recommendations, but by 2013 that was largely no longer the case, with the exception of the UK.
Note
I initially hesitated about writing this post because it's attacking the weakest part of Oster's argument, which even she admits is weak. I have now published it because beginning the post I have since seen people repeat this argument in the wild, and it bears debunking on its own merit irrespective of Oster's repetition of it. The best arguments for light drinking will come from individual level data, which we'll address later. No one should ever be using this argument at all, because none of the premises hold up. It's very likely Europe has more FAS, because we have more heavy drinking in pregnancy than the US, especially in parts of the UK and Eastern Europe.
AI use
Every word of this post was written by me. Claude was used for data extraction and analysis, and fact checking the final post for any errors; all edits were made manually. The original data tables, code, and output have all been uploaded to github.
In part 1, we established that Oster is likely mistaken as to the mechanism behind FAS/D, and concluded there's no theoretical basis on which to assume there is a threshold effect for problematic drinking in pregnancy, particularly in the first trimester. We also established that risk and BAC were both dependent on weight, which absolute recommendations like "1 drink" don't account for.
In this part, we're going to address the claim that it's reassuring that European nations have both more liberal policies towards drinking in pregnancy and have less FAS/D, citing the so-called "American Paradox," a paper written in 1998 by Ernest L. Abel.
Oster writes:
There are a lot of problems with this besides, namely:
For instance, France is commonly thought by Americans to be liberal about drinking in pregnancy, but France issued abstinence-only advice in 2002, eleven years before Oster published her book. Women in France today report drinking very little in pregnancy, not even light drinking; a survey in 2021 reported 1% of pregnant women having more than 1 drink a month, with 7% having had reported any use. Though the surveys aren't directly comparable due to different methods, surveys of alcohol use in the US report ~15% of women reporting any drinking, and ~4% reporting binge drinking.
The UK, where I live, was one of the very last European countries to change their guidance, and did so after the book came out in 2016. Today (2026) there are a few European countries that don't specifically recommend abstinence in pregnancy, but these countries simply have no guidance published on it at all. All the countries that highlight drinking in pregnancy recommend abstinence.
The ecological fallacy
The canonical example of the ecological fallacy has to do with literacy and immigration. In 1950, statistician William S. Robinson found that the higher percentage of immigrants there were in a state, the lower the illiteracy rate was. This implied that immigrants were more likely to be able to read English than native born Americans!
This is, of course, as Robinson writes, nonsense. When you look at individual level data, immigrants were more likely to be illiterate than native born Americans; it was that immigrants were more likely to settle in big cities in wealthier states. Those wealthier states had lower levels of illiteracy, but a higher proportion of immigrants. Adding in more immigrants just didn't have the power to reverse the direction of the effect. If you look at individual level data, rather than across the state, the correlation is reversed.
Oster clearly is familiar with the ecological fallacy. In her footnotes she gives one plausible example demonstrating this, that the US might have a different distribution of binge drinking, with more binge drinkers than in European countries. But this precise example, of course, supports her argument that light drinking is fine.
This pattern also happens to not be correct, as near as I can tell. I didn't check every country, but i.e. British women are more likely to binge drink than American women in pregnancy, not just drink in general. And French women today both binge in pregnancy less and also do less light drinking in pregnancy too. Per capita, France today consumes more alcohol than the US. But this isn't a paradox, it's simple ecological fallacy. Pregnant women drink less, and non-pregnant women and men drink more. But even though it's not true today, maybe it used to be true in France 30 years ago; let's fine out.
Prevalence of drinking and FAS: today and 30 years ago
It simply isn’t the case that countries with more liberal drinking rules (historically or today) have less FAS, at least that we know of. In reality we just don’t have enough data on it. The very first ever national estimate of prevalence of FAS in France was published in 2021, and only included hospital discharge data, which underestimates prevalence. I wasn't able to find national prevalence estimates in any of the other European countries I looked at (not exhaustive). I found two papers published after Oster that tried to compare country level prevalence; one used models based on alcohol use numbers, not actual diagnosed cases. I did find one slightly more modern paper that tried to (published 2017) with some some real prevalence data, which concluded, "The WHO European Region had the highest prevalence (19.8 per 1000 population; 95% CI, 14.1-28.0 per 1000 population)." However that data still mostly uses estimates.
The reason Oster ended up using a 15 year old paper is plausibly in part because ecological data is widely understood to be inferior to individual level data, so it's a fruitless research direction. And the real prevalence data which might be useful for other reasons that's available just wasn't, and today still isn't, good enough to do this analysis.
So what's wrong with this historical data? The 1998 paper Oster cites references conclusions of Abel's own analysis done in a book. The 1998 paper didn't report these data, but I tracked down the actual data tables that were used to argue this and reanalysed them to make it easier to understand the original analysis. Feeding Claude these data tables, it becomes clear the analysis is very naïve from the outset in a way that should have also been obvious in 1998 and does not support the idea of a paradox.
For one, it didn't even show clear evidence Americans reported drinking less in pregnancy than the rest of the world, even in the early 90s and late 80s. Studies in various American cities run the gamut, with studies in Detroit reporting more drinking in pregnancy than any European city, and Philadelphia, which had less drinking in pregnancy than any European city. If you naively average across studies all with different methods, the confidence intervals overlap substantially. There is only one study from France, Roubaix, and it's middle of the pack at 48.5%. Roubaix was specifically chosen for this study because of high alcohol consumption per capita relative to other cities in France. So we don't know what drinking in France looked like then in general because it's almost certainly not representative. It's possible but would be surprising if a city cherry picked for a higher per capita rate was.
Claude was also asked to read table 11.4 on the relationship between heavy drinking in pregnancy with FAS rate into a CSV and plot them, and do a simple linear regression. Claude in this case reached the same conclusion as the book; no relationship. So Oster's repeating of the hypothesis in the paper of being more heavy drinkers in the US than in "Europe" to explain the "paradox" is contradicted by his own book (the discrepancy might be that the book was under review but not published by the time the paper was. However in itself it's very flawed, because the data aren't independent, caveat emptor.)
Looking at the data, I think it's kind of strange the take home story of this data via a game of telephone ended up being was that Europe had less FAS even to begin with.
A striking majority of the studies reported zero cases of FAS, including most of those that took place in the US, with the exception of a few studies of specific cities in the US, France, and Sweden. The studies with no cases were generally small and did not actively screen for FAS, and we usually studying something else (i.e. prevalence wasn't the primary outcome). The sites in the US and France that reported cases were larger, and locations were cherry-picked by looking for places with high alcohol consumption, and the methods used in these studies were active case ascertainment rather than passive case ascertainment. They were simply not intended to estimate country wide prevalence, but to validate screening methods. They were used for prevalence after the fact. Nevertheless two Sokol papers with the highest rates were still not statistically distinguishable from nearly all of the European studies due to overlapping confidence intervals. It took pooling to do that.
Table 11.1 notes race and SES/class for each study, but those weren't used in the analysis. The data is analysed raw, with little attempt to statistically control for methodology, SES, or race, or region. Instead, Abel simply pooled all the cases together.
If you look at the pooled data in table 11.1, from which Abel concludes the US has 20 fold more FAS than Europe, Europe has 4 cases out of 18,766 births and the US has 91 cases of FAS out of 47,097 births. This is statistically significant. However, 87 of the 35,156 births in the US come from a single author (Robert J. Sokol) and three papers, which performed targeted screening infants of single hospitals in Detroit (1 study) and Cleveland (2 cities). 57 of 14,707 were from Detroit. Screening studies in low SES areas are valuable, but pooling the data naively like this implies 32% of the US as a whole is like Detroit and 45% is like Cleveland.
If you based anything else in the US like this off primarily these two cities- for instance, illiteracy rate - you would overestimate the prevalence of illiteracy as well, because illiteracy rates are lower in other areas of the country than they are in major cities, and Detroit and Cleveland in particular.
I'm quite confident overall difference in pooled cases between US and the rest of the world in Abel (1998) almost certainly comes down to different study methodology and the lack of representiveness of the populations used. (The 2017 study at least excluded passive case ascertainment and only included screened infants.) I would argue the central difference is the fact that Sokol lived in the US. In an alternate universe where Sokol ran the same three experiments in areas of the UK or Eastern Europe with heavy alcohol consumption instead, the pooled result would likely be flipped. The 2017 study that found higher rates in Europe than the US included additional data from Croatia, Italy, and France that was published after Abel but before Oster. It also only included screening studies, so a subset of the Abel data. To produce a US rate of FAS composed nearly entirely of data from Detroit and Cleveland but excluding similarly high drinking populations from Europe was never going to produce an accurate comparison.
I wasn't able to get all the papers cited, but for the ones I was able to check, confidence intervals between the continents overlap once you pool papers by whether they screened infants or not even if you don't bother controlling for any of the other factors like race and SES. This isn't even taking into account the screening methods were different; some only counted "complete" FAS (with facial characteristics) and some used much looser criteria, such as counting it as FASD where IQ and behavioural abnormalities were documented in children of women who reported drinking, despite an absence of distinct facial characteristics.
Coming back to the present, a study of Greater Manchester in the UK was published in 2021, and it found that indeed we did have FASD cases, and that in the UK prevalence was likely to be much higher than the actual diagnosis rate. But the study still couldn't provide a national prevalence estimate and there are currently no national prevalence estimates published for the UK that I could find, nor very many other European countries I looked into this for.
Studies that do active case ascertainment suggests the true prevalence of FAS/D is substantially higher than the diagnosis rate. There are good reasons for this; a diagnosis is highly stigmatising causes mental distress for the parents. There's also no treatment for it. Especially in public health systems where cost-benefit calculations are made for screening programmes, it's no surprise those programmes don't exist for FAS/D.
At the present time, this lack of data makes it virtually impossible to compare prevalence rates against countries in any reasonable way. And because ecological evidence is bad evidence to begin with, there's not much reason to invest in it. Oster states this lack of evidence as being reassuring; but could only plausibly be the case if prevalence had actually been studied and found no evidence of elevated risk. This lack of data is fundamentally neutral.
On public health advice
Though the recommendations all have now converged on no alcohol use, alcohol use itself in pregnancy in Europe still varies substantially. One multinational review in 2017 found alcohol use in pregnancy was highest in the UK (~41%) and lowest in Sweden (~9%) which is a large spread. (US is around ~15%). (I have found still other papers with wildly different numbers; and it changes over time, which makes a true comparison quite difficult.)
This entire argument seems to be a that old chestnut “Europe is better!” I live on the continent, technically, and there are many things I like about "Europe."
But the approach to children's public health, in the UK at least, I've found fairly lackadaisical in general. For instance, in the UK we don't screen children for lead poisoning or iron deficiency. We're also sluggish about adding new vaccines. It took 26 years after the US introduced the Hep B vaccine for the UK to start a universal vaccination programme, and 31 years after the US to introduce universal chickenpox vaccination. Supplementation of flour with vitamin B9 to reduce neural tube defects starts this year, 2026; the US started their programme in 1998, 28 years before. Recommendations on alcohol use in pregnancy lagging 35 years behind the US is generally in keeping with this, rather than evidence of the UK knowing something the US does not.
A fair argument is that the CDC and NHS's current recommendations don't really matter anyway because no one pays attention to them. This might certainly be the case. A study found recommendations to reduce alcohol consumption in the UK overall had only a temporary effect. But the very low alcohol consumption in pregnancy in France could mean that their 2002 switch to abstinence only worked. Or it could also just mean that France had a low level of consumption in pregnancy before the recommendations got stronger and labelling was introduced, since we don't have good data on that. Regardless, it's still interesting as to whether the recommendations are correct, too, even if it official advice is irrelevant.
The bottom line
The argument that Europe is more permissive about light drinking and has less FAS/D is false because:
Note
I initially hesitated about writing this post because it's attacking the weakest part of Oster's argument, which even she admits is weak. I have now published it because beginning the post I have since seen people repeat this argument in the wild, and it bears debunking on its own merit irrespective of Oster's repetition of it. The best arguments for light drinking will come from individual level data, which we'll address later. No one should ever be using this argument at all, because none of the premises hold up. It's very likely Europe has more FAS, because we have more heavy drinking in pregnancy than the US, especially in parts of the UK and Eastern Europe.
AI use
Every word of this post was written by me. Claude was used for data extraction and analysis, and fact checking the final post for any errors; all edits were made manually. The original data tables, code, and output have all been uploaded to github.
https://academic.oup.com/alcalc/article-abstract/33/3/195/173477
Here he writes "Australia", but that's an error. The data is from women attending three antenatal clinics in Edinburgh, Scotland, UK. https://onlinelibrary.wiley.com/doi/10.1111/j.1360-0443.1984.tb00264.x
https://obgyn.onlinelibrary.wiley.com/doi/10.1111/1471-0528.12246