The GiveWell Blog

The case for cash

Our choice to name GiveDirectly as our #2 charity has drawn some surprise and criticism. GiveDirectly seeks to deliver 90c directly into the hands of the very poor (no strings attached) for every $1 of total organizational expenses. There are many people who consider this intervention “unproven” (since there is not research linking cash transfers directly to the sort of health impacts associated with other top charities’ programs) and even dangerous (with the idea being that the people receiving the transfers are unlikely to spend them well).

We believe that cash transfers face a lower burden of proof than other charitable interventions, yet have been studied more than any other non-health intervention we’re aware of, with results supporting the idea that they have net positive impact; that GiveDirectly fits the description of a “charity with documented positive impact”; that the magnitude of this impact is certainly open to debate, but appears reasonably high and could be competitive with the most cost-effective interventions; and that much of the intuitive resistance we see to the idea of unconditional cash transfers may be driven by misleading analogies between the developing world and the developed world.

This post will

  • Lay out the basic case for the appeal of cash transfers.
  • Address what we see as misleading analogies between the developing world and the developed world, analogies that lead to what we see as excessive pessimism about the impact of cash transfers.

The basic case for cash transfers
Improving consumption, in the short and long run

For several years, our definition of “evidence of impact” has included evidence that “wealth is being transferred to low-income people.” Documentation of positive impact is usually (necessarily) documentation of proxies for improved quality of life, and of such proxies, we have long found “increased wealth/income/consumption” to be one of the stronger ones. (Other strong proxies include improved nutritional status as indicated by height/weight-related measures, and reduced incidence/prevalence of symptomatic diseases such as malaria and diarrhea. A more direct measure of long-term impact is reduced mortality, but we do not wish to consider interventions based solely on their impacts on mortality, as different people have very different intuitions about how this impact should be valued relative to life improvement.) Improvements in wealth/income/consumption are a common goal of health programs, job training programs, and other programs; we would consider such improvements to be evidence of positive impact in evaluations of such programs, and it seems consistent and appropriate to consider them evidence of positive impact in evaluations of cash transfer programs as well.

It may seem near-tautological to say that cash transfers improve recipients’ wealth/income/consumption. We recognize this, and this is why we sometimes describe cash transfers as carrying a “lower burden of proof” than other interventions. If a charity can establish that it is placing significant wealth in the hands of low-income people (not necessarily an easy thing to establish), we believe this is essentially tantamount to evidence of positive impact. (That isn’t to say that it proves the charity’s impact is highly positive, exclusively positive, or even net-positive; but we believe that it establishes a case for such impact comparable to the best cases we’ve seen for other interventions.)

However, it isn’t the case that this is the only argument for cash transfers. Cash transfers also happen to be the most extensively studied non-health intervention we know of. In a large number of high-quality studies, researchers have looked to see whether cash transfers have indeed increased consumption, what sorts of consumption they’ve increased, and whether common concerns about them are supported by evidence. The consistent picture that emerges from these studies is that cash transfers generally do increase consumption, particularly on food, and that evidence to support common concerns has not emerged despite being looked for. (More at our writeup on cash transfers.)

As discussed previously, there is a smaller set of studies implying that people get significant return on investment from cash transfers, even over the long run; the case for longer-term impacts of cash transfers is broadly comparable to the case for longer-term life improvement impacts of our other top charities’ health interventions, and the cost-effectiveness according to our best estimates is in the same ballpark as well.

Leveraged impacts

One common objection to cash transfers is along the lines of “They probably do help some, but can’t we do better?” We think the answer is that we likely can do better, but it isn’t obvious. For one thing, it’s important to keep in mind the limitations of giving as a casual donor. While a major philanthropist may be able to take big risks with big upside, a donor without special expertise – looking to translate dollars directly into improved lives – has limited options. (For what it’s worth, we feel similarly about LLIN distribution and deworming: we think it’s likely that a major philanthropist could do better than paying directly for delivery of these interventions.)

More importantly, cash transfers can have leveraged and transformative impacts. The most direct evidence of this is pair of long-term studies finding annual rates of return on invested funds in the range of 35-75%, over ~5-year time frames, which we have stated constitutes evidence that is roughly as good as the evidence for long-term impact for deworming.

Many people see claimed returns in this range and find them implausible, as though earning such returns would require great ingenuity and/or risk-taking. For relatively wealthy people, this is an accurate perception; but for people with extraordinarily low levels of wealth, the returns to a little bit more – especially delivered as the sort of lump-sum payment that helps them get around the challenges of volatile and uncertain incomes described in Portfolios of the Poor – could be significant. The right analogy is not to an investment vehicle that can provide returns on arbitrary amounts of savings; rather, it’s to the kinds of returns we can realize every day by being able to spend money up front rather than piecemeal. (For example, if I buy a $5 water bottle that allows me to save $1 per week on bottled water, that’s an annual return of over 1000%. I have enough flexibility in my finances that I would never think twice about an expenditure like this, but for someone on an extremely low income, the situation is different.)

More specifically,

  • One common reported use of GiveDirectly’s transfers is purchasing livestock – something that is often portrayed as having lasting, leveraged, transformative impacts. (Though the purchases of livestock by cash transfer recipients have important differences with the livestock distributed by livestock-specific programs – namely that recipients choose which livestock to purchase, and can purchase something else instead if they’d prefer.)
  • Another common reported use of GiveDirectly’s transfers is purchasing a metal roof, to replace a roof made of mud and thatch. We asked about this sort of purchase on our site visit; we were told that mud/thatch roofs take repeated applications of money and time to repair, and can leak to the point of compelling people to move their families and belongings into others’ homes when it rains heavily (or when their roofs are not in good shape). It’s not hard to imagine that a metal roof could make a major and lasting difference to a family (and GiveDirectly has stated to us – though we have not yet examined the details – that the effective “rate of return” on a metal roof is about 17% annually, which would be very high). Unlike with livestock, I’ve never heard of a charity giving out metal roofs; this is arguably an illustration of the idea that recipients have ideas for improving their own lives that haven’t occurred (or can’t be sold) to donors.
  • There are many more possibilities for how recipients could leverage cash transfers into lasting, high-return impacts. One person on our site visit stated that he had used the money to purchase a motorcycle instead of having to rent it by the day for his job; this sort of story is common among microfinance recipients and in Portfolios of the Poor (the most credible attempt we’ve seen to understand how very low-income people in the developing world manage their finances). Indeed, the high repayment rates and high interest rates associated with microfinance give some indication that it is common (though not universal) for low-income people in the developing world to have fairly simple opportunities to earn high returns on cash, and the evidence we have seen on long-term returns from cash provides further evidence.

Two more perspectives from which cash transfers look attractive:

  • One way to think of cash transfers is as “giving low-income people their choice of intervention [via purchase] to improve their own lives.” There are many plausible reasons to think that their choices may be inferior to the choices of more educated aid professionals, and not all such reasons rely on the idea that aid professionals are better informed. (Alexander covered several of these in a post earlier this year, defending his preference for bednet distribution over cash transfers.) But there are also plausible reasons to think that recipients’ choices may be superior to those of aid professionals. For our part, we’d guess that cash transfers are more beneficial than many, but not all, charitable interventions, and our charity rankings and recommended allocations are consistent with this position.
  • Another argument for cash transfers is that they are the best intervention to support given “maximum skepticism.” If I put no credence whatsoever in expert analysis and academic studies, unconditional cash transfers would be my intervention of choice. Since I do put some credence in such analysis and studies (particularly the strongest ones), I don’t go all the way to this position; however, I think that such analysis and studies have more weaknesses than most people recognize, and I don’t find the position of extreme skepticism absurd or indefensible.

Important differences between the developed- and developing-world poor
Some people object to GiveDirectly based on the idea that “cash transfers have been tried and haven’t worked” in the U.S. They cite examples such as government welfare programs and studies of lottery winners.

We believe that it is very misleading to analogize the poor in the developing world to the poor in the U.S. This is primarily because


  • Because the U.S. poor generally have access to basic necessities, there is generally little low-hanging fruit in terms of easily purchased items that can materially improve one’s standing. This is (we believe) precisely why there are such better giving opportunities in the developing world.
  • When we think of the U.S. poor, we generally think of people who have access to basic necessities but may face more daunting challenges, such as living in high-crime neighborhoods, being unable to hold what we consider desirable jobs, having substance abuse or other mental health issues, etc. Overcoming these obstacles and becoming “middle-class” would take either a great deal of money (reasonably well spent) or fundamental changes in educational status, behavior, environment, etc. By contrast, the developing-world poor generally lack the ability to afford very basic things; purchasing these things could make them much better off, while still at a lower standard of living than the U.S. poor.
  • We also think the nature of “temptation to spend money unproductively” is very different in the two settings. While the developing-world poor certainly have opportunities to spend money on gambling, alcohol, cigarettes, etc., we haven’t (on our site visits) observed the same level of opportunities to waste large amounts at once that we see here. Furthermore, it’s quite plausible to us that the greatest temptation for a clinically malnourished individual would be to legitimately improve his or her diet (e.g., by eating more protein) or address some other pressing basic need (such as a leaky roof).The last few of the above points are speculative, and shouldn’t be taken as an attempt on our part to demonstrate that developing-world cash transfers are well spent. Rather, they should be taken as illustrations of why it isn’t safe to extrapolate from the U.S. to the developing world. (Instead, one should focus on the studies that have been done of cash transfer programs.)

    Another point worth noting is that even if cash transfers have had disappointing results in the U.S., it isn’t clear that any other anti-poverty intervention has had better results. (More at our discussion of U.S. equality of opportunity.) One way of putting our view is that (a) in the U.S., poverty is often too complex to be solved by money, and therefore donors have trouble helping significantly whether they are making cash transfers or funding other interventions; (b) in the developing world, poverty often includes a lack of very basic, very helpful necessities that can be easily purchased, and therefore donations can do a great deal of good in the form of cash transfers or other interventions.

    Bottom line
    We don’t take the position that cash transfers are the best intervention out there. But we think they are a highly promising intervention, and that many of the concerns we see raised about them are unwarranted and/or exaggerated. In terms of both the intuitive case and the evidential case, we think cash transfers are on the very short list of the most promising interventions we’ve seen for individual donors.



  • Scott Weathers on December 26, 2012 at 4:07 pm said:

    In the first paragraph of the case for cash-transfers, you write that GiveWell uses proxies to measure quality of life, like increased wealth/income/consumption and nutritional measures. What exactly are these proxies for? In other words, what are we using to represent quality of life? Does this refer to self-reported well being, or something else entirely?


  • Ted Gudmundsen on December 28, 2012 at 4:32 am said:

    Team GiveWell,

    I’m a huge fan of your work, and I found this post to be insightful and thoughtful, as always. Thank you for your work.

    I worry with direct cash transfers about creating strange incentives, and the effect of such transfers on the fabric of society. $1000 is a lot of money in a village where most people are living on ~$1 a day; adjusted to the US, it’s like giving some family in a poor neighborhood ~$100,000. That’s enough to make everyone else in the village think–“wow, how do I get selected next time?” At best, they’ll regard this as a random and unrepeatable event. At worst, I think it will incentivize people to suck up to rich outsiders and hope for a windfall.

    Even in the best case, it suggests that getting ahead is not a matter of hard work and good decisions, but luck and forces outside of one’s control. My experience in Zambian villages is that the wealthiest, most powerful folks in the village are usually the best educated or at least the cleverest, savviest families. That’s not always true, and it’s certainly not “fair”, but it does provide a powerful incentive to all families to invest in the future, whether that means buying fertilizer or sending kids to school. But now we turn that dynamic upside down. What message will the other kids in the village take from this? Some of the wealthiest kids in the village school aren’t from the best educated or most responsible family, but are some random brats whose parents smooth talked some NGO people.

    For context, I’m not a Republican, I don’t believe that increasing the marginal tax rate stifles hard work in the US by any appreciable amount, and I don’t think that the free market “knows” best. But my experiences in sub-Saharan Africa suggest that cultural dynamics are complicated, that the effects of even simple interventions can be complex and unintended, and that it’s very difficult to figure for an outsider to collect good data on those effects. I don’t know or even suspect that this intervention will create these negative externalities, but I worry about it. I would be hesitant, personally, to donate to GiveDirectly without at least some exploration of this topic.

    Kind regards,

    Ted Gudmundsen

  • John Hunter on December 28, 2012 at 5:33 am said:

    I think the direct cash donations are a great option to evaluate. I am concerned with giving cash directly to those I think there is a likelihood won’t use it well – say a random homeless person on the street. Others may disagree, that is my concern.

    I gave cash directly in the past to people I knew (in developing countries) through friends who suffered twists of fait. The idea of giving cash to “screened” recipients I really like (say a priest, or someone, selects).

    I think it will be interesting to see if this attempt is successful. I think it will be. I am sure there will be plenty of instances of “waste” but eliminating all “waste” (bad spending or whatever people want to call waste) wouldn’t be my goal. My goal would be doing the most good compared to other options.

  • Ted,

    Great post, but a big reason why GiveWell is supporting GiveDirectly is exactly because they want to support learning more about the intervention.

  • BTW, I’m upping my GD allocation at the expense of SCI by 10 percentage points in large part due to these write ups.

  • I found Givewell through researching Kiva (decided against donating through them, so thank you.)

    Suggestion: make your logo a “pinnable” image. (You might also add a “Pin It” button to your posts.)

    I wanted to Pin your site and/or one or more of your pages on a Pinterest board for later reference but you don’t have any pinnable images, so I had to upload an image and copy/paste the URL. Not user-friendly.

    If you made this change, I think it would increase traffic to your site (assuming that’s something you would like to do.)

    Here’s a blog post about it for you:

    Looks like you’re doing good work. Thank you.

  • Ted, thanks for the comment. While I think that there is merit to the concern you raised, broadly speaking, I also think it’s worth noting the following points:

    • GiveDirectly’s process is designed to be objective and transparent: grants are given based on the materials people’s houses are made out of. We have examined the process in depth and find it to be well designed. Therefore, I think the concern about “smooth talking NGO people” is overstated.
    • GiveDirectly’s transfers are estimated to be the equivalent of about a year’s income for the average (extraordinarily poor) family receiving them. Therefore, I think that comparing them to $100,000 transfers to poor Americans is unwarranted. If one wanted to make this analogy, the right figure for Americans would be substantially lower (the poorest Americans make substantially less than $100,000 per household per year); more importantly, however, we think this analogy is problematic for reasons discussed in the post.
    • We don’t believe that GiveDirectly’s transfers are large enough to turn the poorest families in a village into the wealthiest, especially on a long-term basis. (Information about incomes of different villagers is available here.) If this did take place, it would probably have to be through the mechanism of very high rates of return on cash (not merely saving the cash that is received), which would bode well for the overall impact of the transfers.
    • I agree with your characterization of the risks in terms of incentives, but I also think that people in the developing world already often seek to win over outsiders who can give them money. To the extent that GiveDirectly’s program increases their incentive to do so, I don’t think it’s necessarily a major or important change in behavior.
    • We agree that it is easy for aid to have unintended consequences. But we don’t think this issue is confined to cash transfers; we think it applies to all aid. Any program that benefits some people and not others risks disrupting social dynamics and distorting incentives, and it is very difficult for a program to be designed in a way that helps everyone equally. We’d like to see other charitable interventions approached with the same skepticism that many approach cash transfers.

    Further discussion of GiveDirectly’s potential negative impacts is available in our review.

  • Sharon Lipinski (@SharonGives) on January 2, 2013 at 6:43 am said:

    Thanks for this follow up post. Before we donated to GiveDirectly, my giving circle ( talked extensively about the “can we do better?” question.

    Given the lack of information and options sometimes available to poor people and the human tendency to discount future pain for current pleasure, are aid organizations better at deciding what to provide?

    We decided that the answer was sometimes yes and sometimes no. Sometimes aid organizations are better equipped to identify and provide the best aid.

    But at what cost?

    There are two costs to consider. First, the administrative costs associated with aid organizations identifying and providing specific kinds of aid are large. GiveDirectly is providing 90% of their donations directly into the hands of the poor. Other aid organizations have to pay for studies, salaries, transportation, etc. Even at their most efficient, a much smaller percentage of donations will go directly towards the recipient (this is why we don’t use admin costs to evaluate the effectiveness of a charity).

    Second, what is the psychological cost to the participant? When GiveDirectly provides an unconditional cash transfer to a family, they are trusting and respecting that family’s ability and right to make their own decisions. I don’t know how to measure or quantify this cost, but I do believe that providing people with the means and confidence to be the agents of their own lives pays ongoing benefits.

    Thanks for continuing this great discussion!

  • Ted Gudmundsen on January 2, 2013 at 11:46 pm said:


    Thanks for your comments. I appreciate the way you and the other members of staff consistently engage with feedback from the community. Let me respond to several of your points:

    * Regarding the numbers, after more careful consideration, I agree that the intervention is smaller in magnitude than I had estimated. I said $100,000 in USD, but I think $22,000 is a better number. I had been comparing average US income with median Kenyan income, and I should have compared average to average or median to median. I agree that thinking about median household income makes more sense, and I agree that describing this intervention as 121% of annual income seems reasonable. That means that an equivalent US intervention, if we assume that we’re taking families in the 20th income percentile, is about 120% x $18,000 = $22,000. ( I agree with all of your comments about why this intervention is very different from a similar intervention in the US. My point in translating to US dollars was only that we’re talking about a lot of money, enough to turn heads.

    *Let me also retract my assertion that this amount of cash will cause poor kids to become “the wealthiest kids in the village school”. I think it’s probably more accurate to say that this intervention allows a family in the bottom income quintile to consume as if they’re in the second or third income quintile for a year. (50% of rural families are below $1/person/day. I’m guessing at average household size.) In other words, it’s enough for new school uniforms, but not enough egregiously conspicuous consumption. As you point out, it’s a one-time transfer, and therefore probably not even enough to move a family to the middle class permanently. I was thinking about the transfer as an individual income multiplier ($1 day –> $4 day for a year!), but it makes more sense to think of it as a household income multiplier, which is less dramatic.

    *Concerning the external impression of GiveDirectly, I understand that GiveDirectly has an objective, well-characterized formula. I like their formula. I understand that it’s difficult to estimate wealth and income, and in my experience, looking at a family’s house is one of the better measures of prosperity. However, despite GiveDirectly’s best efforts, it may be difficult to convey the rule-based nature of the program to the local population. In my experience, in the US, nearly all programs, charitable or otherwise, are designed in a rule-based, impersonal way designed to strongly curtail the agency of the executing employee. You can’t negotiate the price of an orange with the cashier at the grocery store. People in the US trying to get government assistance are accustomed to this—they try to learn the rules and work through them. In Zambia [I haven’t worked in Kenya, so Zambia is the best I can do], I found that many activities were highly personal. So, even when you explained that a particular program had such-and-such rules, people would still make appeals based on their personal worthiness, not based on those rules. I think there’s a cultural difference, here. Cultural differences are difficult to prove and can be easily abused by outsiders, so I invite you to take this observation with circumspection. Still, I think this cultural difference potentially changes the way people will perceive the program—and therefore the way in which it could skew their incentives and world view.

    *On your final point, let me just say that I agree. It’s brave of you to go out on a limb for cash transfers, and I agree with nearly all of the points you made in favor of this kind of intervention. It IS hard for a casual donor to properly analyze complicated programs, and I agree that simple programs like this one are appealing if one is skeptical of the implementation details of most NGOs. A 90% pass through rate to poor families is amazingly good. Like Sharon, I like the respect that this intervention shows poor families. I believe you yourself have written that we should start from a position of skepticism for all non-cash interventions, since the market is amazingly good at filling people’s needs, even those of the very poor. You wrote in this post that cash transfers may not be the absolute most efficient intervention, but it’s certainly a good one, and I agree with that. I think the externalities on incentives should be kept in mind, but that hesitation doesn’t negate the many positive aspects of this program. I appreciate your research and thinking about this topic.

  • Holden on January 3, 2013 at 1:58 pm said:

    Ted, thanks again for the thoughtful comments. I agree with your points. We certainly agree that there are legitimate concerns around GiveDirectly’s work (though we also think there are legitimate concerns around the work of other charities, including our other top charities).

  • Allan Benamer on January 8, 2013 at 8:27 pm said:

    As an immigrant myself, I can tell you that direct cash transfer works for most immigrant families. There’s a huge amount of remittances for migrant families in the US and that’s basically how my country of origin, the Philippines receives its hard currency. Sure, it distorts the national economy but it does work. Take a look at its macroeconomic effects in this 2008 blog posting at the World Bank:

  • Charlie McKelvey on February 20, 2013 at 1:39 pm said:

    Great article. A 90% pass through rate is exceptional. I believe this is the biggest concern that donors have with any charity – how efficient is the charity run and just how much actually goes towards the mission of helping people. This is the first responsibility of any charity and why some smaller charities are gaining popularity.

    Best Regards and God Bless,

    Charlie McKelvey
    Bethlehem PA

  • Madeline Anglin on April 6, 2013 at 2:23 am said:

    Please don’t ignore the growing numbers, of adults and children, here in the US, that are having a harder & harder time buying food, and keeping a roof over our children’s heads.I just realized, what we’ve been eating, most definitely has my children vitamin deficient!!!! GOD HELP US ALL !!!

  • I read this post a few weeks ago and found myself agreeing with tha main points, and quite happy that such a simple act could make a difference. However, I just listened to the Freakonomics podcast called “Would a Big Bucket of Cash Really Change Your Life?” (found here ) and the evidence suggests that giving a chunk of money (or property in the case of the podcast) makes very little difference in human capital (A measure of the economic value of an employee’s skill set).

    Here are some quotes from the transcript of the podcast

    “Hoyt Bleakley and Joseph Ferrie, found a fascinating data set from a fascinating moment in history ? a big land lottery in Georgia in 1832. They realized they could use this data, along with U.S. Census data, to follow families over time, comparing lottery winners to losers, to see how this shock of sudden wealth affected those families”

    “So, we see a really huge change in the wealth of the individuals, but we don?t see any difference in human capital. We don?t see that the children are going to school more.”

    “school attendance rates are pretty much the same, the literacy rates are pretty much the same”

    “Whether their father won the lottery or lost the lottery their occupation looks the same. The grandchildren aren?t going to school more, the grandchildren aren?t more literate.”

    Possibly the difference in the podcast is that a poor family won a large amount of wealth, whereas in these cash transfer models it seems we are helping the extreme poor obtain some true basics (a tin roof as an example) which could have a more positive effect?

    Any thoughts?

  • Alexander on October 9, 2013 at 4:08 pm said:

    Josh – Thanks for the comment. We’ve seen the study you’re discussing, and thought it was interesting. However, we do not believe that it should affect our view of GiveDirectly much, for a couple reasons:

    • we haven’t modeled intergenerational effects of cash transfers, and don’t think that improving human capital over multiple generations is central to the case for cash transfers.
    • we doubt the relevance of results from these land lotteries from the U.S. in 1832 to the expected returns of contemporary cash transfer programs in low-income countries.

Comments are closed.