Guest Post by Paul Niehaus
GiveDirectly got started when some grad school friends and I decided we wanted to give our money – mostly hypothetical future money, at that point – to the poor.
Why start a new NGO? To our surprise, we couldn’t find an existing one that would do the job for us. And when we opened up GiveDirectly to the public in 2011 we began to realize why. In the nonprofit world giving money to the poor is heretical. Some called us “peculiar;” many pointed out that the poor are known substance abusers; one funder even said that we ourselves “must be smoking crack.”
GiveDirectly has nevertheless grown in two years from nothing to the #2 rated US nonprofit, according to GiveWell. With glowing press coverage and promising results from another study by Blattman, Fiala and Martinez on the effects of similar cash transfers (large grants to groups of young unemployed individuals in Northern Uganda), the narrative has shifted; instead of smoking crack, people now ask if I’m drinking Kool-Aid. Are cash transfers the future of international development? Should we replace everything with cash?
Of course, not. Public goods are still public goods, externalities are still externalities. Direct transfers are unlikely to solve large-scale collective action problems like finding a malaria vaccine or establishing effective governance in post-conflict states. But the role of direct transfers can and should expand as last-mile payments technology improves. We see GiveDirectly driving this change and serving the development community in three main ways.
First, GiveDirectly provides a benchmark for international charitable giving. Keep in mind: this is an industry where “efficiency” is measured using accounting concepts – “overhead,” “fundraising,” and “program expenses” – that are legally meaningless. “Program expenses” is the most egregious as many “programs” are simply sub-grants to other organizations, whose cost structures remain hidden. It’s the industry of gift catalogs that let people in Manhattan choose what assets people in Africa get. (While we’re at it, why not have African farmers pick our stock portfolios?) Only 3% of donors even claim to base such choices on effectiveness research, probably in part because “useful information about what different charities do and whether it works isn't publicly available anywhere.”[1] Worse, charities that do conduct evaluations can easily bury results they don’t like. A friend recently showed me RCT evidence that one much-touted nonprofit has net negative impact; sadly, the results are embargoed and will never see the light of day.
Compare all of this to direct transfers. They generate simple, transparent cost structures: in GiveDirectly’s case, we’ve been spending about 5% of donated dollars targeting poor households, 3% on transfer fees, and putting the remaining 92% into the hands of recipients. Recipients then get to buy whatever they want. Their choices will not be perfect, but the accumulated evidence from dozens of studies says that they will be good on average and lead to sustained improvements. Forthcoming evidence from IPA’s independent evaluation of GiveDirectly (which was pre-announced before it ran) is in line with these earlier results. On net I view the evidence as strong enough that the burden of proof rests with nonprofit entrepreneurs to show that they can do better. I think of it like investing: there surely are fund managers out there that are worth their management fees, but I want to be convinced before I move my money out of index funds. I see GiveDirectly as the index fund of charitable giving.
Second, GiveDirectly provides a laboratory for product design. For development professionals the question these days is less whether cash transfers are useful than how they are most useful. What do big lump-sum transfers do as opposed to streams of smaller payments? What are the impacts of giving to men vs. women? What would a cash transfer optimized for maternal health look like? The list goes on.
GiveDirectly should accelerate the pace at which we learn about questions like these. We provide a relatively frictionless platform for implementation, and we “play nicely” with researchers. We’ve already begun doing this in pilot collaborations with the Nike Foundation (looking at impacts on young women) and ideas42 (examining behavioral elements of framing). I’m excited about projects like these; I see them as doubly beneficial as they both generate new knowledge about outcomes and also let us field-test new operational techniques (e.g. targeting young women).
Third, GiveDirectly provides a technology-driven management platform. This is crucial for institutional funders who already use cash transfers on a massive scale – around 1B people in the developing world, according to DFID – but also grapple with the corruption and logistical challenges that come with that scale. I’ve seen both promise and pitfalls illustrated in a recent project with Karthik Muralidharan and Sandip Sukhtankar, where we measure the impact of improved payments technology in Andhra Pradesh. On the bright side, the tech substantially increased incomes and participants love it. On the downside, only 35% are able to use it despite 2 years of hell-bent effort by the (arguably) best-run state in India. That’s not a design problem: it’s a management problem.
This is why we put such emphasis at GiveDirectly on building scalable management systems. With help from technology-oriented investors like Google Giving we’re developing electronic data collection tools, remote sensing capabilities, crowd-sourced audits that run via Mechanical Turk, scalable call centers that have us in regular contact with over 99% of our recipients to check for problems, and so on. We get really granular data on human behavior throughout the organization – we know, for example, if you do work on Saturday and then try to claim pay for it on Monday. Many of these process improvements get evaluated experimentally – for example, we have ongoing experiments on incentive contracts to reward clear communication to beneficiaries, and another experiment examining the impact of different targeting rules on social cohesion. Insights from projects like these may not make headlines or lead to academic publications. But if we want to deliver value to the poor at the scale of Africa or India, we’re going to need them.
Paul Niehaus is an Assistant Professor at UCSD and co-founder and President at GiveDirectly. His research deals with corruption and implementation problems in developing countries.
GiveDirectly got started when some grad school friends and I decided we wanted to give our money – mostly hypothetical future money, at that point – to the poor.
Why start a new NGO? To our surprise, we couldn’t find an existing one that would do the job for us. And when we opened up GiveDirectly to the public in 2011 we began to realize why. In the nonprofit world giving money to the poor is heretical. Some called us “peculiar;” many pointed out that the poor are known substance abusers; one funder even said that we ourselves “must be smoking crack.”
GiveDirectly has nevertheless grown in two years from nothing to the #2 rated US nonprofit, according to GiveWell. With glowing press coverage and promising results from another study by Blattman, Fiala and Martinez on the effects of similar cash transfers (large grants to groups of young unemployed individuals in Northern Uganda), the narrative has shifted; instead of smoking crack, people now ask if I’m drinking Kool-Aid. Are cash transfers the future of international development? Should we replace everything with cash?
Of course, not. Public goods are still public goods, externalities are still externalities. Direct transfers are unlikely to solve large-scale collective action problems like finding a malaria vaccine or establishing effective governance in post-conflict states. But the role of direct transfers can and should expand as last-mile payments technology improves. We see GiveDirectly driving this change and serving the development community in three main ways.
First, GiveDirectly provides a benchmark for international charitable giving. Keep in mind: this is an industry where “efficiency” is measured using accounting concepts – “overhead,” “fundraising,” and “program expenses” – that are legally meaningless. “Program expenses” is the most egregious as many “programs” are simply sub-grants to other organizations, whose cost structures remain hidden. It’s the industry of gift catalogs that let people in Manhattan choose what assets people in Africa get. (While we’re at it, why not have African farmers pick our stock portfolios?) Only 3% of donors even claim to base such choices on effectiveness research, probably in part because “useful information about what different charities do and whether it works isn't publicly available anywhere.”[1] Worse, charities that do conduct evaluations can easily bury results they don’t like. A friend recently showed me RCT evidence that one much-touted nonprofit has net negative impact; sadly, the results are embargoed and will never see the light of day.
Compare all of this to direct transfers. They generate simple, transparent cost structures: in GiveDirectly’s case, we’ve been spending about 5% of donated dollars targeting poor households, 3% on transfer fees, and putting the remaining 92% into the hands of recipients. Recipients then get to buy whatever they want. Their choices will not be perfect, but the accumulated evidence from dozens of studies says that they will be good on average and lead to sustained improvements. Forthcoming evidence from IPA’s independent evaluation of GiveDirectly (which was pre-announced before it ran) is in line with these earlier results. On net I view the evidence as strong enough that the burden of proof rests with nonprofit entrepreneurs to show that they can do better. I think of it like investing: there surely are fund managers out there that are worth their management fees, but I want to be convinced before I move my money out of index funds. I see GiveDirectly as the index fund of charitable giving.
Second, GiveDirectly provides a laboratory for product design. For development professionals the question these days is less whether cash transfers are useful than how they are most useful. What do big lump-sum transfers do as opposed to streams of smaller payments? What are the impacts of giving to men vs. women? What would a cash transfer optimized for maternal health look like? The list goes on.
GiveDirectly should accelerate the pace at which we learn about questions like these. We provide a relatively frictionless platform for implementation, and we “play nicely” with researchers. We’ve already begun doing this in pilot collaborations with the Nike Foundation (looking at impacts on young women) and ideas42 (examining behavioral elements of framing). I’m excited about projects like these; I see them as doubly beneficial as they both generate new knowledge about outcomes and also let us field-test new operational techniques (e.g. targeting young women).
Third, GiveDirectly provides a technology-driven management platform. This is crucial for institutional funders who already use cash transfers on a massive scale – around 1B people in the developing world, according to DFID – but also grapple with the corruption and logistical challenges that come with that scale. I’ve seen both promise and pitfalls illustrated in a recent project with Karthik Muralidharan and Sandip Sukhtankar, where we measure the impact of improved payments technology in Andhra Pradesh. On the bright side, the tech substantially increased incomes and participants love it. On the downside, only 35% are able to use it despite 2 years of hell-bent effort by the (arguably) best-run state in India. That’s not a design problem: it’s a management problem.
This is why we put such emphasis at GiveDirectly on building scalable management systems. With help from technology-oriented investors like Google Giving we’re developing electronic data collection tools, remote sensing capabilities, crowd-sourced audits that run via Mechanical Turk, scalable call centers that have us in regular contact with over 99% of our recipients to check for problems, and so on. We get really granular data on human behavior throughout the organization – we know, for example, if you do work on Saturday and then try to claim pay for it on Monday. Many of these process improvements get evaluated experimentally – for example, we have ongoing experiments on incentive contracts to reward clear communication to beneficiaries, and another experiment examining the impact of different targeting rules on social cohesion. Insights from projects like these may not make headlines or lead to academic publications. But if we want to deliver value to the poor at the scale of Africa or India, we’re going to need them.
Paul Niehaus is an Assistant Professor at UCSD and co-founder and President at GiveDirectly. His research deals with corruption and implementation problems in developing countries.
[1] http://www.givewell.org/about/story, accessed 7 October 2013
Join the Conversation