Thanks for the interesting references! In particular, you’ve inspired me to look into Wikileaks (which I hadn’t heard of before aside from the recent news).
Concerning microfinance; I myself currently know almost nothing about the topic; my knowledge comes almost exclusively from following the GiveWell blog. See
It would be great if you looked at and critiqued some of their work. I think you might enjoy doing so. According to their page about their process
We believe that “Where should I give?” is one of the hardest questions there is, and there’s no single field of knowledge or expertise that can take it on completely. Instead, we seek to put our thoughts in public as a starting point, and get as many perspectives from the outside as we can. If you have thoughts on our analysis, please don’t hesitate to contact us.
My own experience has been that they take this statement to heart.
They compare average microfinance with best health interventions.
Valid comparison would be either average microfinance with average health intervention, or marginal microfinance with marginal health intervention. “Best” is never a valid measure, let alone comparing best something with average something else.
Compare this situation:
Donors pay $1bln to fund amazingly efficient vaccination program
Government pays $10bln for random not too efficient health services
With:
Government pays $1bln to fund amazingly efficient vaccination program, and $9bln for random not too efficient health services
Donors pay $1bln for not too efficient health services
Donors will surely feel a lot more awesome in first scenario than in the second, but there’s no difference between them at all. And unless the government is too stupid or too evil to fund this amazingly efficient vaccination program, if donors pulled out the result would be in either case:
Government pays $1bln to fund amazingly efficient vaccination program, and $9bln for random not too efficient health services
Of course this was based on optimistic assumption that configuration wouldn’t be:
Donors pay $1bln to fund amazingly efficient vaccination program
Government pays $9bln for random not too efficient health services, and $1bln for bombs to bomb neighbouring country
In which case donations would have massive negative utility. That’s fungibility of money. Analysis of net effect of donations isn’t impossible, but they’re not doing it, so they should stick with broad-based averages as the second best thing.
On the other hand what can have a lot of value is researching relative effectiveness of various interventions when we don’t know yet which works better. This is true not just in case of poorest countries, but even more so for huge welfare programs ran in rich countries with nearly no research.
Now this might be wrong, but impression I’m getting is that microfinance attracts a lot more research than other types of interventions, which tend to be rather hostile even to the idea of randomly assigning people to control and intervention group as a rule.
From quite glance GiveWell seems to be using totally wrong margins.
They compare average microfinance with best health interventions.
Yeah, I’m not really sure what the intended purpose of the linked post was. I would guess they were trying to say something like “donating to one of GiveWell’s top recommended health charities seem to be a better bet than donating to a random microfinance charity” but I agree the implied comparison of health as a sector with microfinance as a sector is misleading.
“Best” is never a valid measure
GiveWell is focused on finding the best charities for casual donors rather than assessing the merits of entire charitable causes. Note that they recommend Small Enterprise Foundation as an outstanding microfinance charity.
Donors will surely feel a lot more awesome in first scenario than in the second, but there’s no difference between them at all.
On the other hand what can have a lot of value is researching relative effectiveness of various interventions when we don’t know yet which works better. This is true not just in case of poorest countries, but even more so for huge welfare programs ran in rich countries with nearly no research.
In which case donations would have massive negative utility.
My (vague) impression is that developing world countries with militaristic governments which devote a lot of financial resources to military spending often do so despite the fact that their poorer citizens have unmet basic needs. To the extent that this is true it points in the direction of supplying health interventions being unlikely to displace government money in the direction of military spending.
On the other hand what can have a lot of value is researching relative effectiveness of various interventions when we don’t know yet which works better. This is true not just in case of poorest countries, but even more so for huge welfare programs ran in rich countries with nearly no research.
I have a similar impression here.
Now this might be wrong, but impression I’m getting is that microfinance attracts a lot more research than other types of interventions, which tend to be rather hostile even to the idea of randomly assigning people to control and intervention group as a rule.
This could be; I know almost nothing about the topic.
GiveWell is focused on finding the best charities for casual donors rather than assessing the merits of entire charitable causes. Note that they recommend Small Enterprise Foundation as an outstanding microfinance charity.
I understand this point of view, but if Vilalge Reach is clearly so much better than an average health charity (as GiveWell seems to be certain of), shouldn’t they just get other health charities to reallocate a small portion of their vast funds to Village Reach?
How large a violation of EMH are we willing to accept here with how little evidence? Now EMH fails in many contexts for many reasons, but this really begs for some explanation. Why should we trust GiveWell if even other health charities seem not to? Even if they’re not certain, even modest level of agreement should result in transfer of funds a lot larger than what Village Reach currently gets.
Unfortunately there are only two possibilities in equillibrium here:
Other health charities drastically disagree with GiveWell.
Other health charities agree with GiveWell, some funds get reallocated, Village Reach gets very high coverage, marginal utility of health dollars falls down to mid tier charities.
EMH says GiveWell should only be trusted if we can observe ongoing large shifts of funding towards charities it promotes. It will lose informational value eventually but donations accelerate this shift towards more efficient charities.
Do we see this happening (in which case go ahead and donate to Village Reach), or is everyone ignoring GiveWell (in which case the crowd might have a point, and don’t blindly trust GiveWell).
My (vague) impression is that developing world countries with militaristic governments which devote a lot of financial resources to military spending often do so despite the fact that their poorer citizens have unmet basic needs. To the extent that this is true it points in the direction of supplying health interventions being unlikely to displace government money in the direction of military spending.
There was never really any country that could entirely disregard population’s needs.
My model—you need some level of spending X to keep country’s economy from collapsing and population from revolting. Everything more than that goes to military. If foreign donors give you Y for that, then X-Y of your spending will be enough to keep country’s economy from collapsing and population from revolting, leaving X more for military hardware. Even if Y>X, they’ll figure a way to embezzle excess funds.
This model is too extreme, but so is naive assumption of no offsets.
I understand this point of view, but if Vilalge Reach is clearly so much better than an average health charity (as GiveWell seems to be certain of), shouldn’t they just get other health charities to reallocate a small portion of their vast funds to Village Reach? [...]
VillageReach stands out for transparency, monitoring and evaluation and focus on a cost-effective program. Where this places VillageReach relative to other health charities in impact per dollar depends in some measure on what one’s default assumption is about a charity’s effectiveness in the absence of information. The GiveWell staff have a skeptical default assumption; the plausibility of which can be questioned. Note however that there are charities that that focus on the cause of clean water which is the DCP report lists as being something like 500 times less cost-effective than increased immunization; this pushes in the direction of adopting a skeptical default assumption.
The idea of getting other health charities to reallocate money to VillageReach is an interesting one but I suspect it’s infeasible for political reasons (employees of a generic health charity are motivated to keep the money donated to the charity within the organization as sending it elsewhere might entail layoffs, etc.).
There’s no profit motive attached to providing health services more efficiently so the usual hypotheses of the EMH are not in place. Again recall my earlier comment about donors not paying attention.
In view of this it seems that essentially the only factor that pushes in the direction of EMH in the non-profit world is altruism; but altruism in humans is limited and easily crowded out by tribalism and by (wishful thinking)/(need for self-image preservation).
My observation (e.g. as a student in public high school some time ago) is that there’s a tendency for badly inefficient policies to persist in the non-profit sector because the people who are perpetuating them find it uncomfortable to admit that the policies that they’ve been adhering to are defunct and correspondingly delude themselves into believing that they’re just fine as they are.
Do we see this happening (in which case go ahead and donate to Village Reach), or is everyone ignoring GiveWell (in which case the crowd might have a point, and don’t blindly trust GiveWell).
GiveWell has only been around since 2007 and only has four employees. My impression is that it’s presently little known within the philanthropic sector. Last year they leveraged a million dollars which is only 1⁄300000 of the total U.S. charitable contributions in 2009.
They have not made a systematic attempt to publicize their findings beyond their website up until now (no advertising, etc); so far preferring to focus on making their research and research process more robust with a view toward building a more solid product of broader interest.
In interest of building credibility, GiveWell has been asking outside volunteers to vet their research.
Potential donors seldom reject GiveWell’s recommendations on account of doubting the credibility of their analysis; much more common reasons for rejection are:
Squeamishness about mixing rational analysis with charity on account of seeing rational analysis as “cold and calculation” and charity as “warm and fuzzy.” This is not at all common on LW (!) but fairly common in broader society.
Lack of interest in the causes that GiveWell has covered so far.
Prior commitment to particular charities that they’ve become attached to.
My model [...]
I follow, but your model seems so extreme as to be a self-characiture! :-)
This model is too extreme, but so is naive assumption of no offsets.
Thanks for the interesting references! In particular, you’ve inspired me to look into Wikileaks (which I hadn’t heard of before aside from the recent news).
Concerning microfinance; I myself currently know almost nothing about the topic; my knowledge comes almost exclusively from following the GiveWell blog. See
Where We Stand On Microfinance (From a year ago, possibly dated.)
Microfinance charity
It would be great if you looked at and critiqued some of their work. I think you might enjoy doing so. According to their page about their process
My own experience has been that they take this statement to heart.
It might be my inner contrarian speaking.
From quite glance GiveWell seems to be using totally wrong margins.
They compare average microfinance with best health interventions.
Valid comparison would be either average microfinance with average health intervention, or marginal microfinance with marginal health intervention. “Best” is never a valid measure, let alone comparing best something with average something else.
Compare this situation:
Donors pay $1bln to fund amazingly efficient vaccination program
Government pays $10bln for random not too efficient health services
With:
Government pays $1bln to fund amazingly efficient vaccination program, and $9bln for random not too efficient health services
Donors pay $1bln for not too efficient health services
Donors will surely feel a lot more awesome in first scenario than in the second, but there’s no difference between them at all. And unless the government is too stupid or too evil to fund this amazingly efficient vaccination program, if donors pulled out the result would be in either case:
Government pays $1bln to fund amazingly efficient vaccination program, and $9bln for random not too efficient health services
Of course this was based on optimistic assumption that configuration wouldn’t be:
Donors pay $1bln to fund amazingly efficient vaccination program
Government pays $9bln for random not too efficient health services, and $1bln for bombs to bomb neighbouring country
In which case donations would have massive negative utility. That’s fungibility of money. Analysis of net effect of donations isn’t impossible, but they’re not doing it, so they should stick with broad-based averages as the second best thing.
On the other hand what can have a lot of value is researching relative effectiveness of various interventions when we don’t know yet which works better. This is true not just in case of poorest countries, but even more so for huge welfare programs ran in rich countries with nearly no research.
Now this might be wrong, but impression I’m getting is that microfinance attracts a lot more research than other types of interventions, which tend to be rather hostile even to the idea of randomly assigning people to control and intervention group as a rule.
Yeah, I’m not really sure what the intended purpose of the linked post was. I would guess they were trying to say something like “donating to one of GiveWell’s top recommended health charities seem to be a better bet than donating to a random microfinance charity” but I agree the implied comparison of health as a sector with microfinance as a sector is misleading.
GiveWell is focused on finding the best charities for casual donors rather than assessing the merits of entire charitable causes. Note that they recommend Small Enterprise Foundation as an outstanding microfinance charity.
Yes, this is true. See negative and offsetting Impacts.
My (vague) impression is that developing world countries with militaristic governments which devote a lot of financial resources to military spending often do so despite the fact that their poorer citizens have unmet basic needs. To the extent that this is true it points in the direction of supplying health interventions being unlikely to displace government money in the direction of military spending.
I have a similar impression here.
This could be; I know almost nothing about the topic.
I understand this point of view, but if Vilalge Reach is clearly so much better than an average health charity (as GiveWell seems to be certain of), shouldn’t they just get other health charities to reallocate a small portion of their vast funds to Village Reach?
How large a violation of EMH are we willing to accept here with how little evidence? Now EMH fails in many contexts for many reasons, but this really begs for some explanation. Why should we trust GiveWell if even other health charities seem not to? Even if they’re not certain, even modest level of agreement should result in transfer of funds a lot larger than what Village Reach currently gets.
Unfortunately there are only two possibilities in equillibrium here:
Other health charities drastically disagree with GiveWell.
Other health charities agree with GiveWell, some funds get reallocated, Village Reach gets very high coverage, marginal utility of health dollars falls down to mid tier charities.
EMH says GiveWell should only be trusted if we can observe ongoing large shifts of funding towards charities it promotes. It will lose informational value eventually but donations accelerate this shift towards more efficient charities.
Do we see this happening (in which case go ahead and donate to Village Reach), or is everyone ignoring GiveWell (in which case the crowd might have a point, and don’t blindly trust GiveWell).
There was never really any country that could entirely disregard population’s needs.
My model—you need some level of spending X to keep country’s economy from collapsing and population from revolting. Everything more than that goes to military. If foreign donors give you Y for that, then X-Y of your spending will be enough to keep country’s economy from collapsing and population from revolting, leaving X more for military hardware. Even if Y>X, they’ll figure a way to embezzle excess funds.
This model is too extreme, but so is naive assumption of no offsets.
VillageReach stands out for transparency, monitoring and evaluation and focus on a cost-effective program. Where this places VillageReach relative to other health charities in impact per dollar depends in some measure on what one’s default assumption is about a charity’s effectiveness in the absence of information. The GiveWell staff have a skeptical default assumption; the plausibility of which can be questioned. Note however that there are charities that that focus on the cause of clean water which is the DCP report lists as being something like 500 times less cost-effective than increased immunization; this pushes in the direction of adopting a skeptical default assumption.
The idea of getting other health charities to reallocate money to VillageReach is an interesting one but I suspect it’s infeasible for political reasons (employees of a generic health charity are motivated to keep the money donated to the charity within the organization as sending it elsewhere might entail layoffs, etc.).
There’s no profit motive attached to providing health services more efficiently so the usual hypotheses of the EMH are not in place. Again recall my earlier comment about donors not paying attention.
In view of this it seems that essentially the only factor that pushes in the direction of EMH in the non-profit world is altruism; but altruism in humans is limited and easily crowded out by tribalism and by (wishful thinking)/(need for self-image preservation).
My observation (e.g. as a student in public high school some time ago) is that there’s a tendency for badly inefficient policies to persist in the non-profit sector because the people who are perpetuating them find it uncomfortable to admit that the policies that they’ve been adhering to are defunct and correspondingly delude themselves into believing that they’re just fine as they are.
GiveWell has only been around since 2007 and only has four employees. My impression is that it’s presently little known within the philanthropic sector. Last year they leveraged a million dollars which is only 1⁄300000 of the total U.S. charitable contributions in 2009.
They have not made a systematic attempt to publicize their findings beyond their website up until now (no advertising, etc); so far preferring to focus on making their research and research process more robust with a view toward building a more solid product of broader interest.
In interest of building credibility, GiveWell has been asking outside volunteers to vet their research.
Potential donors seldom reject GiveWell’s recommendations on account of doubting the credibility of their analysis; much more common reasons for rejection are:
Squeamishness about mixing rational analysis with charity on account of seeing rational analysis as “cold and calculation” and charity as “warm and fuzzy.” This is not at all common on LW (!) but fairly common in broader society.
Lack of interest in the causes that GiveWell has covered so far.
Prior commitment to particular charities that they’ve become attached to.
I follow, but your model seems so extreme as to be a self-characiture! :-)
Agree.
What’s the best evidence against it, or quick test that would be able to at least tell maximally naive model apart from maximally cynical model?
Do we really know totally nothing about that?