A journalist visits GiveDirectly villages in Kenya

In February, Jacob Kushner, a journalist living in Kenya, contacted us. We have long been interested in seeing more substantive coverage of philanthropy, so we were excited to talk to him.

As a pilot project, Mr. Kushner decided to visit villages in which GiveDirectly had distributed some of its earliest cash transfers. We spoke with Mr. Kushner several times to offer thoughts and feedback, but we encouraged him to write about whatever he found (positive or negative about GiveDirectly). We also put him in touch with GiveDirectly to confirm that staff there were amenable to this project.

Mr. Kushner completed his trip in April, and his full article follows. He also shared his full interview notes with us which we’ve posted here.

We’ve summarized what we took away from his article here. Carolina Toth, Manager, People and Partnerships at GiveDirectly responds here.

When giving out cash to the poor, what happens when some are left behind?
A closer look at whether GiveDirectly’s cash transfers stoke community tension in Western Kenya

By Jacob Kushner

For several years now, the charity GiveDirectly has experimented with different ways of deciding who among Western Kenya’s rural poor should receive cash transfers. It’s an important consideration, because $1,000 means a lot to the families that receive it—and it can mean a lot of disappointment to the families that don’t. Last month I traveled to Western Kenya to speak with both lots, and I found that the discrepancy did not go unnoticed in their communities. To date, GiveDirectly has undergone five different transfer programs in Siaya over the past three years, with different metrics for selecting recipients. I interviewed recipients from three of those cohorts: • The Google Cohort (approximately 850 ‘thatch-roof only’ recipients whose transfers were completed in October 2013) • The 200k Cohort (approximately 200 ‘thatch-roof only’ recipients whose transfers were completed in January 2013) • The 2M cohort (approximately 2,000 recipients divided into ‘thatch-roof only’ villages and ‘saturation’ villages (in which nearly everyone is eligible) who have received one major transfer and will receive the second and final one in July 2014). In a follow-up to a randomized controlled trial, GiveDirectly asked residents if they’d heard any complaints about GiveDirectly in their community. Sixty-four percent of respondents in Siaya County answered “yes,” as did 48 percent of those in the “Google” cohort (in Rareida it was 28 percent). Fewer than 6 percent of respondents in all four groups said shouting or angry arguments had ensued because of the transfers, and fewer than 4 percent said they’d experienced crime, theft or violence or felt threatened as a result. Virtually no one said they’d argued with family members over how to spend the money, and no more than 7 percent in any group said their village elder had approached them asking for money. Carolina Toth, Kenya Field Director for GiveDirectly, explained the results of a series of informal community group meetings in which GiveDirectly led residents in a discussion of who should be eligible for transfers. Sixty-two percent of respondents in thatch-only villages said they’d heard complaints relating to ineligible households, compared with 46 percent in saturation villages. Thirty percent of those in thatch-only villages said they’d heard complaints about different criteria being used across different villages, compared with only 4 percent in saturation villages. GiveDirectly concluded that the strongest takeaway from the discussions is that poorer ‘thatched’ households are more deserving but also that certain households that have mabati or permanent houses are deserving of the transfers as well. When asked about their own villages, residents preferred the saturation method. When asked about other villages, they preferred thatch-only. No one thought it would be “bad” if cash were given to some wealthier households. Because recipients in saturation villages have yet to receive their second transfer (due in July), it’s too early to draw definite conclusions. But this and other previous reports leave several question unanswered: To the extent that community tension may result in the wake of cash disbursements, how does that tension actually unfold? Who are the parties and what are some examples? Most importantly, what do non-recipients in those communities think about the fairness of the selection process? Do they feel stigmatized for not having received the money, and how does their perception of whether animosity resulted from the cash transfers compare with those of the recipients’ themselves? In April I made a reporting trip to Siaya County to interview recipient and non-recipients in the communities where GiveDirectly has made those disbursements. Over three days I interviewed 15 people, asking whether they were happy with GiveDirectly’s selection process and whether any tension arose in their communities as a result of it. I interviewed some recipients from each of the three cohorts and also interviewed recipients and in both the ‘saturation’ and ‘thatch’ divisions of the 2M cohort. I interviewed four non-recipients, at least one in each of the three cohorts. My interviews seemed to reflect many of the conclusions of the RCT and subsequent follow up interviews and meetings. No one reported intra-family arguments about how to spend the money or being coerced by a spouse or family member to spend it in a particular way. Only one recipient said he’d originally disagreed with his spouse but that they eventually came to a mutual agreement. No one reported theft or that their own money had gone to waste in any way. But 12 of the 15 respondents did indicate that some amount of tension had fostered in their community as a factor of some people having received money while others did not. By far the most tangible conflict mentioned to me occurred in the 200k cohort in the village of Koga. There, the village elder did not receive a cash transfer. He was, however, consulted by GiveDirectly staff to assist in a tour of the boundaries of the village so GiveDirectly could identify eligible households, for which he was given a small token payment as compensation for his time. But in the words of one recipient there, “there was a scandal.” The elder “had conspired (to enlist) some households that were outside the area and had better houses, with the understanding that they would give him some money.” GiveDirectly staff say the elder seems to have directed residents who lived in tin-roof houses to “squat” in vacant thatch roofed houses in order to receive the money. Subsequently, the assistant chief, with the support of the other village council members, dismissed the elder from his position. When I spoke with the elder, he confirmed that he had misrepresented certain households in the village so they would be enrolled in the program. He justified that decision saying, “I was the village elder and I was working for the (entire) community.” He said tension resulted when the initial disbursements were made and some families, including his own, were left out. “I felt degraded by my community members. They were laughing at me that I didn’t receive any help even though I was the leader of the community. I was so humiliated.” He said the incident led him to ‘resign’ after more than 35 years of serving as an elder in Koga (he is 62 years old). The second most tangible takeaway was the resentment and frustration expressed by the four non-recipients I interviewed. One woman in a “saturation” village was visibly angry as she described how she was not selected because the living room in her tin roof house is cemented, even though her other rooms are not. Another Koga man said he was cheated out of a transfer: ￼￼“The time the GiveDirectly team was working in the village, they came to my home ￼but at that time I was grazing cattle outside the compound and I saw them in my ￼sister-in-law’s house. I was curious. But due to how relations within households go ￼sour, my sister told the GiveDirectly team that I had left and was never around.” ￼Despite an appeal he said he made to GiveDirectly field staff, this man did not ￼receive a transfer. He says his economic situation is similar to that of the other recipients: ￼“I live in a house like this—(a) grass thatch house. I have children in school and I struggle to pay their fees. Some of my children for lack of funds have to be supported by my relatives in other areas, in Nairobi. I have only two cattle.” GiveDirectly staff pointed out that “targeting” is a universal problem in development aid. Other methods used to select recipients—such as letting communities vote on who should receive, or requiring people to go to some lengths to prove they are indeed quite economically poor-off—have major drawbacks: Cronyism, and excessive bureaucracy and burdens, respectively. As an alternative, GiveDirectly employs another common method that uses easy-to-observe characteristics such as roof style to judge how wealthy or poor a household is. According to GiveDirectly’s own research, less than 5 percent of people in the 2M cohort villages complained, legitimately or otherwise, of being unfairly excluded. (￼In comparison, a recent study ￼￼of the Kenya Hunger Safety Net Program found an exclusion error rate of 46 ￼percent). ￼The man in Koga who says he was unfairly excluded also expressed sympathy for ￼the Koga village elder. “I would not be happy with what has happened to him, ￼because the feeling he has now at losing his job is the same feeling I have at not ￼getting the money. I feel bad for him because I am also going through some pain.” ￼The man also aired some critiques as to how some people in the community spent their money. ￼“I saw some beneficiaries, the way they misbehaved when they got the money, and ￼that made me feel it is important that recipients receive training on how to spend it. For example there are people who wasted it on drinking sprees, and others bought ￼items that they didn’t understand how they would maintain. For example, one bought a motorbike and used it for a few months, but now it is unused and has not ￼really helped him.” ￼Indeed, several interviewees mentioned the need for training to accompany the transfer process. GiveDirectly currently does not provide training or advise ￼recipients as to how they should spend their money. GiveDirectly does, however, provide a brochure that lists different possible categories of expenditure such as home construction, business, and farming. GiveDirectly is considering experiments ￼in which brochures also list the average returns that previous beneficiaries earned on each category of investment. After completing the interviews, I asked Carolina Toth, the GiveDirectly field director, what she made of it all. I asked Toth what she thought about the village elder scandal in Koga—that a man who had served as elder for 35 years lost that position not because he violated a community custom, but simply a rule imposed by GiveDirectly. “The village elder more often than not is one of the richer members of the community,” Toth said. As to his “previous feelings of entitlement to benefit from whatever is happening … I don’t think that’s an expectation we want to uphold.” Toth and I also discussed the consequences for individuals who are excluded in a community where most residents receive the cash. “It’s definitely a psychological event in their live,” Toth said. “But we know from the (randomized controlled trial) that there are huge spillover effects to the people who didn’t receive.” When I asked Toth about the man who says he missed out on the transfer because his sister-in-law misinformed the GiveDirectly staff that he was not living in the village, Toth said it’s certainly true that some people get left out by mistake. But she said such cases are rare. As to the woman with the cemented living room who didn’t receive cash even though the rest of her home is not yet cemented, Toth said the GiveDirectly field staff can only make decisions based upon what they see—and that the distinction between a cemented house and a non-cemented house is not always entirely clear under such circumstances. The vast majority of people who aren’t selected, said Toth, are skipped because they come from a marginally higher socioeconomic standing to whom the money would be less useful. “What is the value of$250 given to a family that’s richer? Wouldn’t that be more valuable in the hands of people who are really poor?” Toth asked. “We have a mission of giving to the extreme poor, so by excluding some people who are not in the extreme poor, you are able to reach more extreme poor.”

Ultimately, the question any cash transfer implementer must decide is, “Is the possibility that community tension may result from a non-universal disbursement so great or concerning that transfers should be made to all residents in a village despite the opportunity cost that fewer, even poorer people in other villages will not receive any cash?”

Thus far GiveDirectly has answered that question in the negative. With certain exceptions (such as allowing communities to nominate a pre-determined number of otherwise unqualified people for the disbursements) and with increased nuance (by considering more advanced criteria than simply thatch versus tin roofs and indoor plastering), GiveDirectly intends to continue excluding those residents who do not qualify as the poorest of the poor.

Jacob Kushner is a journalist based in Nairobi. He reports on foreign aid and investment in Africa, human rights and the extractives sector.

• Elie on May 28, 2014 at 12:18 pm said:

We’re excited that we found a journalist to independently visit GiveDirectly on the ground. To date, the information we have on GiveDirectly has come from (a) GiveDirectly staff, (b) a randomized controlled trial co-authored by a former GiveDirectly board member, and (c) our own site visit to GiveDirectly in Kenya. Mr. Kushner’s visit represents an independent perspective on GiveDirectly’s program, and a chance to identify issues we couldn’t have otherwise.

Mr. Kushner findings are broadly in line with what we expected based on the information we previously had, and we think this fact should bolster confidence in GiveDirectly.

Based on information GiveDirectly has produced from internal monitoring and an RCT of its program, we expected to see what Mr. Kushner found: some evidence of tension as a result of its program. We were previously aware of the specific case involving a village elder who lost his title. These problems are concerning and worth addressing — a sentiment we believe GiveDirectly agrees with — but are not surprising or new, and we don’t think these costs come close to outweighing the benefits the program provides.

Moreover, we would not expect it to be hard to find examples of complaints were someone to interview recipients and non-recipients of virtually any aid program. It is worth keeping in mind that very few aid agencies have allowed themselves to be subject to this type of analysis, so we are grateful to support an organization like GiveDirectly that is ready to open itself up to outside criticism.

As a side note, this article raises the question of whether GiveDirectly should distribute funds to all (or most) households in a village (the saturation model) or to only those that meet certain characteristics (e.g., having a thatch roof).

This is something GiveDirectly and I (thought not necessarily all GiveWell staff) have long disagreed about. In my opinion, the evidence for conflict is strong enough that GiveDirectly should take additional steps to alleviate it. Non-recipients in non-saturation villages are wealthier but still extraordinarily poor by any measure and I’m not convinced that cash transfers will have significantly greater impacts on the poorest households than the merely very poor ones.

GiveDirectly believes that with limited resources, it should serve the poorest whom it believes will benefit most from transfers. It also has conducted surveys in its saturation vs traditional villages and has not found significant differences in complaints between the two. It also finds minimal cost savings from its saturation model. (GiveDirectly has shared a summary of these surveys with us. We have not yet carefully vetted them. We hope to publish them shortly, with GiveDirectly’s permission.)

• Carolina Toth on May 28, 2014 at 12:20 pm said:

As Jacob’s piece points out, all development programs have to decide whom to help and whom to leave out, as there are never enough resources to help everyone. When GiveDirectly began we reviewed existing research on different approaches and chose to prioritize the poorest in rural areas using thatch and mud houses as the indicator of poverty. This method performed well, but GiveWell raised the question: would it make more sense to simply give money to everyone in a given village? We thought this question was important enough that we designed an experiment to examine it. We randomly assigned 37 villages to either a “thatch” condition, in which we used our standard criteria, or a “saturation” condition in which (almost) every household received a transfer.

The data from this experiment did not present a clear case for saturation. Enrollment cost savings from saturation were small (<5%) and reductions in ineligible people trying to game the system were also very small (4.3% in thatched villages vs. 3.6% in saturation). Only about half of the indicators we use to measure tension and conflict were lower in saturation villages, and the other half were actually higher. We also conducted focus groups in six villages to better understand the results on tension. Interestingly, community members in saturated villages reported tension primarily with neighboring villages that did not receive transfers. We also presented focus groups with the same choice that Give Directly has to make as an organization: given two villages, would you rather give to the poorest in both villages, or give to everyone in one village and no one in the other? All focus groups chose to prioritize the poorest in both villages.

We think that the most reliable basis for decision-making are data like these, drawn from experimental comparisons and using samples that are large enough to be representative. Based on the data we have thus far, we plan to continue giving to the poorest. We are piloting community based methods to enroll the relatively smaller number of extreme poor who are living in iron-roofed houses, and will continue conducting experiments and relying on data to improve our methods.

Carolina Toth
Manager, People and Partnerships
GiveDirectly

• Elie on May 29, 2014 at 7:48 pm said:

Note: After publication, GiveDirectly told us that because Jacob Kushner was interested in understanding possible tension caused by its distributions, it sent to him a location where it knew that there had been trouble with the village elder. The place he went to was not randomly selected.

• Uri Katz on June 1, 2014 at 1:23 am said:

Correct me if I am wrong, but haven’t direct money transfers demonstrated that the recipients know how best to spend aid funds? If so, why not extended that to allowing the communities themselves to choose who receives funds, or maybe to use part of the money for a community project (e.g. paving a road). Surely, this opens the door to corruption and violence, but in much the same way, so do our tax dollars, and still we think our system of government is the best of evils.
Wouldn’t creating stronger communities sereve recipients better in the long run? Isn’t the gift of self reliance and responsibility ultimately greater than that of money?
I would guess that this will not necesarily reduce conflict and some people will still think the division was unfair, but it adds value in other respects.
In any case, what makes GD’s staff more qaulified to make these decisions?

• James on June 2, 2014 at 1:42 pm said:

Uri,
If the people in the community decided that they wanted to use the funds for a community project they are perfectly free to do so. GD only decides who receives money, not what they spend it on.

• Colin Rust on June 2, 2014 at 2:15 pm said:

Uri, if the community determines who receives the funds, then it seems very possible that the money would go disproportionately to the most powerful and wealthy members of the community (the opposite of the result of the current process).

The community project idea is interesting to think about. I disagree with James: because of free-rider and other problems, that seems unlikely to happen without a mechanism to make it realistically possible. Maybe in villages where money is being distributed universally, they could have a vote as to whether they want to distribute the money, or use it for a specific collective purpose or a mix. But it is easy to imagine this getting contentious in a variety of ways.

• Alexander on June 2, 2014 at 8:56 pm said:

Uri – I agree with the thrust of James and Colin’s responses. While an organization could try to use funds to facilitate community decision-making, doing so might increase the likelihood of resources being redirected to the benefit of the community’s most powerful members, and would also likely be significantly more costly than GiveDirectly’s current targeting efforts. I agree that it might be an interesting experiment to have a community vote about how the GiveDirectly transfers are spent, with both individual transfers and some community-wide projects as options, though concerns about manipulation by the most powerful members of the community would again arise.

• Carl Shulman on June 2, 2014 at 9:05 pm said:

“When asked about their own villages, residents preferred the saturation method. When asked about other villages, they preferred thatch-only. No one thought it would be “bad” if cash were given to some wealthier households.”

Any thoughts on how seriously to take own-village vs other-village views?

• Elie on June 2, 2014 at 9:21 pm said:

Carl – not really. We haven’t looked closely at this data yet.

• Uri Katz on June 3, 2014 at 12:36 am said:

Alexander, Colin & James – You seem to have missed the main point I was making. Why do we assume that our own governments (from the municipal to the federal levels, not to mentions organizations like the E.U. and the U.N.) are the best way to make collective decisions about spending money, but in a small village in Africa this is prima facie a bad idea, that will enhance inequality?
There could be a cultural difference that I am not accounting for (like India’s caste system), or something else I am missing, but a reason other than a gut feeling should be provided for ruling this idea out, especially in light of the evidence that GD is causing tension and conflict. Isn’t it prejudiced to say collective decisions work fine for “us”, but “they” could not handle it?
The reason I raise this with relation to GD is that their novel idea is to trust that people know best what is good for them, even the poorest people on the planet. Also, they seem concerned with the question of whether to give money to all residents of a village or to a select few, but there are other options.

• Tomer Chachamu on June 3, 2014 at 3:22 am said:

Uri, does collective decision making work for us? There’s a lot of disillusionment and disapproval of the current system and turnout is less than 50%. Besides that, I think that in a small village, social considerations will dominate.

Say that giving only to the poorest is preferable, will the poor people vote for that knowing that they are causing their friendly, richer neighbours to go without? It would only increase tension past GiveWell’s current levels.

You might want to look at microsavings though, where everybody contributes to a communal pot and then collectively approves loans for village members.

• Meenal Mamdani on June 3, 2014 at 5:08 pm said:

Would it be possible, in the saturation method, to vary the amount of funds based on the poverty level? This way every one gets something, so no one feels excluded.

• Michael Kramer on June 6, 2014 at 12:00 pm said:

This suggests to me that the target communities recognize that thatch-only is fairer but that it is also more disruptive to the status quo. Assume that what I recommend to other villages does not affect me personally. Then I tell the interviewer that other villages should do the best thing which is to reduce inequality. What I tell the interviewer about my village does affect me personally. Reduction of inequality in my village means overturning long-standing social relations of power. This is sure to increase social tension.

In saturation, everyone is made better off and no one is made worse off, in absolute terms. If we rank people on a wealth ladder, it is also true that no one changes rank. In relative terms, some people are made worse off. Assume that prior to the giving everyone $1,000, the wealthiest person was worth$100 and the poorest was worth $1. Thus the wealthiest person was one hundred times wealthier than the poorest. After the$1,000 transfer the wealthiest person is only worth 9.9% more than the poorest.

With thatch-only some people are made better off and no one is made worse off in absolute terms. But now some people can be made worse off both in positional (rank) terms and in relative terms. Thus thatch-only is more disruptive than saturation.

In the interviews it was not made clear how universal the sentiment was that thatch-only is good for other villages and saturation is good for my village. If I were to guess, I would think that those in the highest wealth quartile would be more likely to agree with that sentiment and those in the lowest wealth quartile would be less likely. But I could be wrong. The wealthier rulers might have more social connections with which to make life difficult for those with new money.

• reader on June 21, 2014 at 11:05 pm said:

Let’s devote more resources (time, money, print space) to this unanticipated effect, i.e. envy among humans.
Duh.

—————–
Project designimplementation (intenionally one word) is directly and unevitably responsible for either stoking or assuaging envy.

Assuage through INCLUSION of all households.

Continue the \$1k to poorest households.
INCLUDE all others by giving some money PLUS a formal certificate to be used in a lottery of some number of drawings for some bigger money.

Certificate identifies the household,
and informs that funds will not be distributed at lottery event.

Important: Donor organization conducts lottery.
Important: A brief, non-celebratory business event.

• reader on June 21, 2014 at 11:07 pm said:

typo: inevitably