I highly recommend Leif Wenar’s essay “Poverty Is No Pond” – especially to those not yet familiar with, but interested in, the empirical complexities involved in giving to overseas poverty-fighting charities.  Wenar’s main aim in his essay is to criticize Peter Singer’s 2009 book The Life You Can Save for (i) being overly optimistic about the quality of information available on the effects of giving to various charities, and (ii) failing to emphasize that every charitable donation also comes with some risk of harming people living in extreme poverty.  I’ll only briefly address (i), and then turn to and focus primarily on (ii).

(i)  Wenar raises a number of very good points about the difficulties of assessing the effectiveness of government and NGO (non-governmental organization) aid directed toward people living in extreme poverty.  He does not confine himself to a single type of aid, e.g. development or humanitarian aid, but calls our attention to important challenges facing a wide range of aid categories.  I do think, however, that it is reasonable to be optimistic about the quality of information available on many health interventions, particularly the ones studied in detail by the Disease Control Priorities Project and the World Health Organization.  Wenar would probably also be pleased by how far GiveWell has developed in the five years since he wrote his essay, and by the convergence between them and Giving What We Can on the tremendous cost-effectiveness of deworming and providing malaria nets (both charity evaluators very highly rank Against Malaria Foundation, Schistosomiasis Control Initiative, and Deworm the World).  Wenar is right that it is difficult to know how much it costs to save a life by donating to, e.g. Against Malaria Foundation, because there is uncertainty about what would have happened had the donation not been made.  But charity evaluators like those mentioned above take counterfactuals and uncertainty into account and would agree that we can at best provide rough ranges rather than exact figures, for how much it costs to save a life.  Wenar writes:

“There are presently few paths or incentives for NGOs to translate academic studies into changed behavior.  External audits on aid NGOs cover only the basics of financial probity, without touching on the results of the NGOs’ projects.”  (p. 20)

But this doesn’t accurately reflect, for example, the in-depth process that GiveWell uses in evaluating charities.  Moreover, now there is a new breed of NGOs that are very focused on transparency and accountability; in addition to Against Malaria Foundation and Deworm the World, consider GiveDirectly and Possible (there are still several others).  Because of effective altruist organizations like GiveWell, Giving What We Can, and The Life You Can Save (Singer’s book spawned an organization), these NGOs do have an incentive to continue down this path, and others to follow their lead.  (As one example, endorsements from these charity evaluators feature prominently on the Against Malaria Foundation website.)

(ii)  I find Wenar’s criticism that Singer’s framework does not take account of risk of harming others to be the most philosophically interesting aspect of his essay.  I’ll briefly remind the reader of Singer’s general framework, and then present and discuss Wenar’s criticism.  In his 1972 article, Singer offered the following principle:

“If it is in our power to prevent something bad from happening, without thereby sacrificing anything morally significant, we ought, morally, to do it.”  (p. 231)

He then gave the famous pond example, eliciting a powerful and widely shared intuition that plausibly supports the above principle:

“[I]f I am walking past a shallow pond and see a child drowning in it, I ought to wade in and pull the child out.  This will mean getting my clothes muddy, but this is insignificant, while the death of the child would presumably be a very bad thing.”  (p. 231)

Finally, Singer noted that poverty-related death and suffering are bad things, and that we can prevent them from happening without sacrificing anything morally significant by donating to effective aid agencies.  The conclusion is that we ought, morally, to donate to effective aid agencies.

Wenar’s criticism of this Singer framework is that (Wenar claims) it doesn’t take into account the fact that our donations invariably come with a risk of harm to others, particularly those living in extreme poverty.  He writes:

“Singer’s principle does not capture the correct factual relationship between affluent and poor individuals.  So this principle does not capture the moral relationship between them, and it cannot ground a call to action.  You are not in a situation analogous to knowing that you can save a child drowning in a shallow pond.  Closer is this:  if you hand cash to a stranger he may – along with other strangers hired by other people – try to save some children who have fallen into a lake.  Yet it looks like these strangers can only get to the lake by pushing through a crowded rave on a pier with no railings.”  (pp. 25-6)

Wenar thinks Singer’s principle doesn’t apply to our situation, because our situation involves risk of harm to others, and (Wenar claims) Singer’s principle,

“…draws attention to only one kind of negative effect of our contributing to aid:  that we ourselves may ‘sacrifice’ – that we ourselves may have less disposable money or time.”  (p. 24)

But (as Wenar himself acknowledges in footnote 46 of his essay) Singer intends for “sacrifice” to be understood more broadly than this, such that his principle would not imply that you ought to prevent bad when this would mean violating a deontological constraint (e.g. against harming, stealing, lying, promise-breaking, and so on).  I think it’s clear that this has been Singer’s intention in all of his writings making use of this principle.  Still, Wenar is right that Singer did not explicitly draw attention to risk of harm to people living in extreme poverty as a possible clash with deontological constraints.  And Wenar is right that in the pond example there’s no mention of risk of harm, whereas we donors live in the real, risk-filled world.

In response to Wenar, one could point out, as Nick Beckstead does, that:

“While it is plausible that some aid involves [deontologically prohibited] harms (e.g., aid to regions that are overrun by corrupt warlords), other cases seem less worrisome, for example, providing technical assistance to a deworming program in India.  Admittedly, providing such assistance may divert employment from the private sector, produce bad incentives, or partly replace government effort, and hence weighing these losses against the gains produced by the program might require some cost-benefit analysis and thinking about how to aggregate harms and benefits.  However, implementing the program does not seem to involve any tragic violations of rights or have any other worrisome deontological features.”  (p. 416)

So in cases of giving to charities that would impose only these kinds of “economic harms” for the sake of big health benefits, Singer’s principle would indeed ground a call to action – that is, if imposing such harms for the sake of providing such benefits isn’t in fact prohibited by a plausible constraint against harming.

I am sympathetic to Beckstead’s style of response, and would like to say a bit more to extend it.  In particular, notice that the size of the risk of any significant harm of giving to a deworming charity – or for that matter any of GiveWell’s top-ranked charities – seems fairly small.  The following position seems to me sensible:  one would be in violation of no plausible deontological constraint if one imposed a fairly small risk of harm on some in order to (with a high degree of certainty) substantially benefit others.

Some might disagree.  Imagine an absolutist view that implied that, as long as there is some risk that your act will do harm, it’s impermissible.  Given that we’re stuck in the real, risk-filled world, no matter what we do there is some risk we’ll do harm, and thus this view implies that no matter what we do, we act impermissibly.  However, I acted permissibly just now when I blinked my eyes, so this view is false.  (I apologize if through some Butterfly Effect my blinking will in fact cause an avalanche, destroying a town in the Alps…)

The more moderate position above that I said seems sensible contains a number of vaguely specified quantities (in italics).  How must these fields be filled in such that no plausible deontological constraint is violated?  I am not sure how much precision can, in principle, be offered here.

But I do think I can offer something useful.  Consider a parent who wants to give her 9-year-old son a fun weekend at Disneyland.  She also has an infant daughter, who is too young to get any enjoyment out of the trip.  The parent drives both children for six hours on the freeway to get to Disneyland, knowing that there is a risk they will be killed in a car accident.  Nonetheless, most of us would say that the parent did not violate any plausible deontological constraint; this seems to be in part because the risk of harm she imposed on her children is fairly small.

I would wager that the risk (of economic harm) to those living in extreme poverty from you donating $50 to Deworm the World is also fairly small, probably comparable to the risk of death associated with six hours of freeway driving.  And if it’s permissible for a parent to impose this fairly small risk of death for a weekend of fun, then presumably it’s permissible for a donor to impose a comparably small risk of economic harm to achieve major health benefits for many children ($50 would be enough to deworm 100 children).

A worry is that the bigger the donation you give, the greater the risk.  Very large donations may no longer yield fairly small risks of harm, but sizeable ones.  One point to keep in mind here is that while a very large donation may impose more risk of harm, it would bring about correspondingly more benefit.  Still, it is unclear how plausible it is that if the amount of risk imposed in donating $50 is made permissible by deworming 100 children, then the amount of risk imposed in donating $5,000 is made permissible by deworming 10,000 children.  For one thing, there may well be a non-linear relationship between donation size and risk size.

But suppose you added up all the annual donations to a particular deworming charity.  How much risk (of economic harm) is imposed, in aggregate, by all these donations?  I welcome any corrections on this score, but – based on what I know about the activities and results of deworming charities over the past several years – my guess is that not much risk of economic harm is imposed even by a whole year of donations.  I can add to this that there are documented economic benefits of deworming (it’s easier for people to go to work and school when they aren’t infected).  Maybe the greatest risk of a mass donation to a deworming charity is decreased government effort.  But it turns out that in many places where deworming projects take place it is unlikely governments would have provided funds if NGOs had not.

One philosophically interesting question is what to make of cases – unlike deworming – where very large donations do carry with them sizable risks of harm.  Then we will have to determine whether the proportionately greater good done by the large donation makes it permissible to impose the large risk.  It’s possible to hold the view that, once the size of the risk (of some particular harm) is sufficiently large, it would be a violation of the deontological constraint against doing harm to impose this risk even if the good brought about were proportionately greater.  That is, maybe R is the risk threshold, and while it’s permissible to bring about .25R risk for the sake of saving 1,000 lives, it’s impermissible to bring about R risk for the sake of saving 4,000 lives.  This view has the implication that, while it would be permissible for me to make a donation every three months to a charity, each which imposes .25R risk and saves 1,000 lives and which would taken together over the course of a year impose R risk and save 4,000 lives, it would be impermissible for me to lump these donations together, making a single annual donation.  I don’t intend for this to be a convincing objection to the view, I merely want to flag a further philosophically interesting complication.

Wenar has done us a service by clearly and carefully presenting a number of important considerations that are relevant to donating to overseas aid agencies that aim to help people in extreme poverty.  However, I agree with Beckstead that Singer’s principle applies to the risk-filled world we donors find ourselves in; and the most highly-ranked charities are also ones that are pretty innocuous from the standpoint of deontological constraints.  If you are skeptical of this, but still want to help, you could give to a reputable organization that rigorously studies extreme poverty interventions, such as IPA or J-PAL.  There’s almost surely no harm in that.

(I am grateful to Peter Singer, Nick Beckstead, and Tom Ash for helpful feedback.)