Yes, we do consider the benefits of alcohol, including that many people enjoy it.
James Snowden put together a short document discussing this when he made the largest current Open Philanthropy alcohol grant in 2021 (the grant was recommended by GiveWell but funded by Open Philanthropy; any extension / renewal will sit within Open Philanthropy). At the time GiveWell / James applied a 10% reduction to the (implicitly net) burden of alcohol harm on this basis.
The dismissal of consumer surplus-based ways to value alcohol consumption is puzzling. The main justification is that “it seems likely to us that consumers are behaving irrationally”, but this is an overly broad statement. What fraction of alcohol consumption is irrational? If you believe that 30% of consumers are consuming irrationally high amounts, you could easily exclude the 30% heaviest drinkers and estimate consumer surplus for the remainder. In general, you can choose a population where you believe people are consuming more out of enjoyment than addiction.
This would require a bit of primary investigation, but you could use Nielsen scanner data with alcohol prices and consumption to a) drop the people who drink the most, b) estimate consumer surplus on the remainder. I’m pretty sure Nielsen has similar data in some LMICs if you want a more representative population. My prior is that you would arrive at substantially more than a 10% downward adjustment.
(Speaking for myself; the 10% estimate comes from work I did at GiveWell but others at Open Phil and GiveWell may disagree with me)
I agree we shouldn’t dismiss consumer surplus entirely, and in retrospect would soften some of the wording in that doc – I think the irrationality point is important but not totalizing. The Nielsen idea is interesting and I’d like to think about it more. I think internalities are less bimodally distributed between people than your model, which muddies the waters, but I wonder if an analysis like that could still be informative.
Fwiw the program we funded is primarily focused on taxation, which is a nice mechanism to balance a recognition of externalities / internalities with a general prior towards personal choice. I’d estimate higher than 10% if that wasn’t the case. A focus on tax means the reduced consumption will be from the drinks for which people had the lowest willingness to pay, limiting lost consumer surplus.[1] It also results in increased tax revenue, so could be considered as trading off against alternative ways of raising tax revenue with their own deadweight loss in consumer surplus.
TBC, I recognize the inherent fragility / subjectivity of the 10% estimate and I suspect different people would come to quite different conclusions about what input to use, so I’d be excited to see more efforts to estimate this considering the broad sweep of evidence.
Of the two studies I could find on consumer surplus, the one which attempted to estimate consumer surplus from a marginal increase in price (rather than for typical consumption) estimates a loss of €58 million in consumer surplus, compared to a €700m improvement in “health, productivity, and non-financial welfare losses” (Anderson and Baumberg 2010, pg 34), implying an offsetting impact of ~8%. (Though I think there are a bunch of ways in which that study isn’t analogous to the models we use, including a higher estimate of non-health impacts, so difficult to know what to make of it).
This also raises a separate worry about the extent to which taxation affects heavy drinkers, where the marginal harm is likely highest, which we tried to account for separately in the effect size estimate.
Fair points, I agree that taxation has a lower bar. The bimodal point was illustrative, you could take some other individual characteristics as proxies for the extent of internalities (e.g. education) and weight people by that when estimating.
Thanks for this. I’m not sure I follow the claim that if you assume that alcohol taxation merely shifts the tax burden, there aren’t strong reasons to think the deadweight loss will be greater from alcohol taxation vs other forms of taxation. The subjective wellbeing study found that drinking increases people’s wellbeing by almost as much as spending time with friends. It seems unlikely to me that if the tax were instead eg on income that the benefits of the income would be as large as this. Intuitively, this seems off.
On your botec on the benefits of alcohol, a lot rides on you assuming that a death from alcohol accounts for 40 units of value (I’m assuming this means life years lost, but not sure). But in the sheet you suggest that most deaths would be among older people. If you revise this figure to 10 years of life lost (which seems much more plausible to me), on the median case, the value wiped out by reduced booze enjoyment is 67% on the median case and 134% on the pessimistic (or optimistic, depending on your view) case. If you reduce the years of life lost to 5 years, then on the median case, the 134% of the value is wiped out. i.e. it seems like on some more plausible assumptions, the policy is net negative.
On the other hand, none of this considers hangovers.
Having said that, this paper suggests that most the median age of death would be around 30-40, not among older people.
**
Ah, I was looking at the spreadsheet and I think a unit of value is different to a year of life lost.
Another thought—you measure the effects of alcohol on subjective wellbeing as a fraction of someone’s waking hours. This seems right from a subjective wellbeing perspective. But is that also the way you think about the value lost by a death? By consistency, you would also need to implicitly downweight the disvalue a death by a third for the time people spend asleep. Or do you already do that in your moral weights?
>Another thought—you measure the effects of alcohol on subjective wellbeing as a fraction of someone’s waking hours. This seems right from a subjective wellbeing perspective. But is that also the way you think about the value lost by a death? By consistency, you would also need to implicitly downweight the disvalue a death by a third for the time people spend asleep. Or do you already do that in your moral weights?
Oh that’s interesting. It’s been a while now since I did this, but I think I was implicitly doing that with this calc
Implicitly, yes. Though don’t use that exact formulation. money vs daly comparison is based on reported preference not swb. Daly vs swb comparison implicitly writes off time spent asleep where I assumed 1 daly = difference between 40->100 on swb scale.
If didn’t exclude sleep in botec, would make alcohol look worse as happiness bump from alcohol would be for lower % of time. (Set row 21 to 24)
>I’m not sure I follow the claim that if you assume that alcohol taxation merely shifts the tax burden, there aren’t strong reasons to think the deadweight loss will be greater from alcohol taxation vs other forms of taxation. The subjective wellbeing study found that drinking increases people’s wellbeing by almost as much as spending time with friends. It seems unlikely to me that if the tax were instead eg on income that the benefits of the income would be as large as this. Intuitively, this seems off.
Interesting. That doesn’t seem off to me. If I’m understanding correctly, the implication of your view is that people would generally be better off if they consumed more alcohol and less of other goods on the margin. Is that right?
To put it another way: increasing taxes on alcohol has two effects on consumer surplus: (i) deadweight loss (ii) a transfer from consumers to the government. I think (ii) is probably positive. Almost all taxes involve some amount of deadweight loss, but we do them anyway because we think public goods and redistribution are worth it.
TBC, I’m not claiming that higher excise taxes on alcohol relative to other goods merely shifts the tax burden. If we assume perfect rationality (which I believe would be mistaken), having unequal marginal taxes between goods does result in some additional deadweight loss. But it is a counterveiling factor.
>On your botec on the benefits of alcohol, a lot rides on you assuming that a death from alcohol accounts for 40 units of value
A unit of value (in GiveWell’s terms) is equivalent to doubling consumption for a person for a year. A DALY is 2.3 units of value. So you want to be dividing your estimates by 2.3.
The Global Burden of Disease estimates ~30 YLLs and ~10 YLDs per death (I didn’t include YLDs in the BOTEC and I underestimated YLLs which makes it conservative, though I also didn’t discount the GBD estimates for imperfect evidence quality and black market consumption not addressable through policy which makes it optimistic. I’d guess these ~cancel).
Edit: didn’t see your second comment when writing this where you saw this
>On the other hand, none of this considers hangovers.
Yeah, although interestingly (IIRC) the Baumberg study didn’t find any effect on SWB the day after drinking (though I’m skeptical—maybe people didn’t feel like inputting how sad they were on their phone when they were hungover!)
Yes, we do consider the benefits of alcohol, including that many people enjoy it.
James Snowden put together a short document discussing this when he made the largest current Open Philanthropy alcohol grant in 2021 (the grant was recommended by GiveWell but funded by Open Philanthropy; any extension / renewal will sit within Open Philanthropy). At the time GiveWell / James applied a 10% reduction to the (implicitly net) burden of alcohol harm on this basis.
I’m reviewing this issue in greater detail now.
The dismissal of consumer surplus-based ways to value alcohol consumption is puzzling. The main justification is that “it seems likely to us that consumers are behaving irrationally”, but this is an overly broad statement. What fraction of alcohol consumption is irrational? If you believe that 30% of consumers are consuming irrationally high amounts, you could easily exclude the 30% heaviest drinkers and estimate consumer surplus for the remainder. In general, you can choose a population where you believe people are consuming more out of enjoyment than addiction.
This would require a bit of primary investigation, but you could use Nielsen scanner data with alcohol prices and consumption to a) drop the people who drink the most, b) estimate consumer surplus on the remainder. I’m pretty sure Nielsen has similar data in some LMICs if you want a more representative population. My prior is that you would arrive at substantially more than a 10% downward adjustment.
Thanks for the thoughts Kartik!
(Speaking for myself; the 10% estimate comes from work I did at GiveWell but others at Open Phil and GiveWell may disagree with me)
I agree we shouldn’t dismiss consumer surplus entirely, and in retrospect would soften some of the wording in that doc – I think the irrationality point is important but not totalizing. The Nielsen idea is interesting and I’d like to think about it more. I think internalities are less bimodally distributed between people than your model, which muddies the waters, but I wonder if an analysis like that could still be informative.
Fwiw the program we funded is primarily focused on taxation, which is a nice mechanism to balance a recognition of externalities / internalities with a general prior towards personal choice. I’d estimate higher than 10% if that wasn’t the case. A focus on tax means the reduced consumption will be from the drinks for which people had the lowest willingness to pay, limiting lost consumer surplus.[1] It also results in increased tax revenue, so could be considered as trading off against alternative ways of raising tax revenue with their own deadweight loss in consumer surplus.
TBC, I recognize the inherent fragility / subjectivity of the 10% estimate and I suspect different people would come to quite different conclusions about what input to use, so I’d be excited to see more efforts to estimate this considering the broad sweep of evidence.
Of the two studies I could find on consumer surplus, the one which attempted to estimate consumer surplus from a marginal increase in price (rather than for typical consumption) estimates a loss of €58 million in consumer surplus, compared to a €700m improvement in “health, productivity, and non-financial welfare losses” (Anderson and Baumberg 2010, pg 34), implying an offsetting impact of ~8%. (Though I think there are a bunch of ways in which that study isn’t analogous to the models we use, including a higher estimate of non-health impacts, so difficult to know what to make of it).
This also raises a separate worry about the extent to which taxation affects heavy drinkers, where the marginal harm is likely highest, which we tried to account for separately in the effect size estimate.
Fair points, I agree that taxation has a lower bar. The bimodal point was illustrative, you could take some other individual characteristics as proxies for the extent of internalities (e.g. education) and weight people by that when estimating.
Thanks for this. I’m not sure I follow the claim that if you assume that alcohol taxation merely shifts the tax burden, there aren’t strong reasons to think the deadweight loss will be greater from alcohol taxation vs other forms of taxation. The subjective wellbeing study found that drinking increases people’s wellbeing by almost as much as spending time with friends. It seems unlikely to me that if the tax were instead eg on income that the benefits of the income would be as large as this. Intuitively, this seems off.
On your botec on the benefits of alcohol, a lot rides on you assuming that a death from alcohol accounts for 40 units of value (I’m assuming this means life years lost, but not sure). But in the sheet you suggest that most deaths would be among older people. If you revise this figure to 10 years of life lost (which seems much more plausible to me), on the median case, the value wiped out by reduced booze enjoyment is 67% on the median case and 134% on the pessimistic (or optimistic, depending on your view) case. If you reduce the years of life lost to 5 years, then on the median case, the 134% of the value is wiped out. i.e. it seems like on some more plausible assumptions, the policy is net negative.
On the other hand, none of this considers hangovers.
Having said that, this paper suggests that most the median age of death would be around 30-40, not among older people.
**
Ah, I was looking at the spreadsheet and I think a unit of value is different to a year of life lost.
Another thought—you measure the effects of alcohol on subjective wellbeing as a fraction of someone’s waking hours. This seems right from a subjective wellbeing perspective. But is that also the way you think about the value lost by a death? By consistency, you would also need to implicitly downweight the disvalue a death by a third for the time people spend asleep. Or do you already do that in your moral weights?
>Another thought—you measure the effects of alcohol on subjective wellbeing as a fraction of someone’s waking hours. This seems right from a subjective wellbeing perspective. But is that also the way you think about the value lost by a death? By consistency, you would also need to implicitly downweight the disvalue a death by a third for the time people spend asleep. Or do you already do that in your moral weights?
Oh that’s interesting. It’s been a while now since I did this, but I think I was implicitly doing that with this calc
Yeah wrt to your botec, I wasn’t sure whether you were implicitly writing off time spent asleep.
I suppose you would also have to do the same for measuring the effects of money on wellbeing. Do you do that?
Implicitly, yes. Though don’t use that exact formulation. money vs daly comparison is based on reported preference not swb. Daly vs swb comparison implicitly writes off time spent asleep where I assumed 1 daly = difference between 40->100 on swb scale.
If didn’t exclude sleep in botec, would make alcohol look worse as happiness bump from alcohol would be for lower % of time. (Set row 21 to 24)
Yeah I’m not making the ‘you’ve underestimated net benefits of alcohol’ point, just trying to think through your assumptions
Interesting. That doesn’t seem off to me. If I’m understanding correctly, the implication of your view is that people would generally be better off if they consumed more alcohol and less of other goods on the margin. Is that right?
To put it another way: increasing taxes on alcohol has two effects on consumer surplus: (i) deadweight loss (ii) a transfer from consumers to the government. I think (ii) is probably positive. Almost all taxes involve some amount of deadweight loss, but we do them anyway because we think public goods and redistribution are worth it.
TBC, I’m not claiming that higher excise taxes on alcohol relative to other goods merely shifts the tax burden. If we assume perfect rationality (which I believe would be mistaken), having unequal marginal taxes between goods does result in some additional deadweight loss. But it is a counterveiling factor.
A unit of value (in GiveWell’s terms) is equivalent to doubling consumption for a person for a year. A DALY is 2.3 units of value. So you want to be dividing your estimates by 2.3.The Global Burden of Disease estimates ~30 YLLs and ~10 YLDs per death(I didn’t include YLDs in the BOTEC and I underestimated YLLs which makes it conservative, though I also didn’t discount the GBD estimates for imperfect evidence quality and black market consumption not addressable through policy which makes it optimistic. I’d guess these ~cancel).Edit: didn’t see your second comment when writing this where you saw this
Yeah, although interestingly (IIRC) the Baumberg study didn’t find any effect on SWB the day after drinking (though I’m skeptical—maybe people didn’t feel like inputting how sad they were on their phone when they were hungover!)
Hangovers are a myth
The way people downvote jokes on this forum lol. Really discourages it, I feel we need a decent chunk more humor here!
(this was a joke)