Let the artefactual field experiments begin!

The first project I was involved in here was an evaluation of Dutch funding to Congolese NGOs. We want to know if the NGOs did stuff that improved the lives of Congolese people (Like massages? No, like help with farming or schooling). To find out, we carried out a survey of about 1,000 households in about 100 villages in Uvira Territory and Fizi Territory in the Province of South Kivu (of which Bukavu is the capital). That required recruiting 35 research assistants. We put up signs at l’Université Catholique de Bukavu (UCB), sent out emails, and called people in our database. Altogether we collected about 120 CVs. We invited the most promising 60 to take a tablet-based general intelligence test with questions like, “If you add every third number between 1 and 50, is the result odd or even?” The top 40 from the test were invited to two weeks of in-class training. Here’s the group on Day 1 in a classroom at UCB’s new campus:

And here’s a view from the back of the class:

We went over the tablet-based questionnaire again and again:

And went through practice scenarios:

The training was so long because the survey included behavioral games a.k.a artefactual field experiments (AFEs) (by those who don’t play no games), which are tools for measuring trust, risk preferences, time preferences, and other characteristics for which self-reports may be unreliable. This survey had two AFEs. In an AFE to measure trust, the interviewee was given six laminated cards with a bushel of corn on one side and 500 FC (500 Congolese Francs is about 55 cents) on the other side. She was told that they could donate as many or as few as they want to a randomly selected member of their village, but they wouldn’t find out who, and neither would the recipient ever know their identity. NOW HERE’S THE KICKER: each bushel of corn that the interviewee donates becomes THREE bushels of corn (it’s, like, an investment). And THEN, after the corn multiplies, the anonymous receiver can send as much or as little back to the anonymous sender. Get it? The wealth-maximizing choices would be for the interviewee to send all six, which become 18 (18*500FC = 9000 FC = $10 USD), and then for the receiver to send back nine. That requires trust on the part of the interviewee. 

In the AFE to measure risk preferences, the interviewee was asked to choose one of six lotteries. Each lottery had two outcomes, each with a probability of 0.5. The interviewee would literally put her hand into a bag with a black ball and a white ball and pull one out. In the least risky lottery, both the white and black ball yield 4000FC (~$4.5 USD). In the most risky lottery, the white ball yields 1400FC (~$1.5 USD) and the black ball yields 8200 FC (~$9 USD). So the expected value of the riskiest lottery is 1400*.5 + 8200*.5 = 4800 FC. Thus, if an interviewee chooses the least risky lottery, she is giving up 800FC (almost $1 USD) in exchange for less uncertainty about the outcome. To see how intra-household bargaining influences risk preferences, we asked husbands and wives to play separately and then together.
In a country where the average daily income is roughly $1USD, these AFEs involve significant decisions, so, the theory goes, they should reveal interviewees’ true preferences. I was a little weirded out at first by the idea of asking subsistence farmers to play games with money, but two factors help assuage my concerns: first, it’s impossible to lose money. Every interviewee receives 1000FC in addition to the payout of one randomly selected AFE, and none of the AFEs involve losses (you can only win less). Interviewees could make up to 10,000FC (~$11USD). Second, village life has been described as boring by more than one villager, and it’s possible that these bizarre AFEs would be kinda fun. It’s also possible, though, that the stress of making these choices involving relatively large amounts of money outweighed the benefits. In my next post I’ll write about how it all turned out.