Friday, January 20, 2012

Back from Kalehe 3: Network Field-experiment

In addition to the village mapping and the in-depth interviews (see the previous post), we also conduct a field-experiment. Actually this is the biggest of our three undertakings – we received a grant for this from Columbia’s Center for the Study of Development Studies to do this.

In brief, Neelan and I are interested in how Congolese villagers behave towards each other. How much trust do people have in each other? How benevolent are people towards each other? For my dissertation I am particularly interested in this for migrant populations. The big problem here of course is that if you ask for example the question “Do you like your migrant neighbor?” to an native villager, he/she is likely to say ‘yes’ – even if this is not the truth. In other words, it’s cheap to lie with a survey. One way to get around this problem is to do a field experiment – making sure that it becomes costly to lie (so that they don’t).

We therefore play two well-know types of field experiments: trust games and so-called dictator games. In brief, in a trust game a player makes the following decision: For each other player that plays I receive (for example) $1. I can keep this money, and then I have $1. However, I can also contribute this to the other player. If the other player (when he/she plays) also contributes back, my contribution is doubled and I received $2. If he/she doesn’t contribute back, I receive nothing. Thus you only contribute to another person if you trust this other person to contribute back to you as well. The second type of game (the dictator game) is much easier and doesn’t have this strategic component. You receive (for example) four quarters and you have to decide how much you want to give to the other player. If you give 1 quarter to the other player, you thus have $0.75; and the other player receives $0.25. This will get at benevolence. Interesting is that following theoretical work the best strategy in this game is of course to give nothing to the other player; so that you receive $1 each time. However field-work has shown that throughout the word people contribute between 20-40% (with some differences across cultures). I'm quite sure we'll find something similar in East Congo.

Just playing these two types of games would not be that interesting – thousands of studies have already done this. Our only contribution would be that we are crazy enough to do this in the Congo. No, we do something extra. These games are normally played specifically with people that have not had previous interaction. Often these experiments are played at universities and students are recruited via posters. But this is in many cases not how the world works! People often base their decisions regarding other people on experiences that they have had with them. For example, trusting a person depends on previous interaction - on networks between people that are already in place. So instead of recruiting subjects via posters at universities we go to places where people specifically know each other: Congolese villages. By doing this we will be able to find out what specific characteristics or what relationship between villagers are important in society.

So what do we expect to obtain? Playing these games will give us trust and benevolence networks. That is, we will obtain dots (indicating people) and lines between those dots (indicating contribution). But that's not all we'll do. Let’s say that we find clustering. That is, let’s say that 8 people all contribute to each other and another 8 are left out. The big question then is: why? What determines this? So we also obtain information about each player (ethnicity, migration status, etc.) and each player’s relationship to all the other players: Family? Work together? Same political party? Etc. Once we lay this information (what we call latent networks) over our trust and benevolence networks we will know what cleavages are important in Congolese villages.

There is so much more to tell (this not even half the story). Let’s see, four more things quickly. First, because of our interest in migrants we stratify our sample by migrant status. This is a difficult way of saying that we make sure that when selecting players we make sure that a certain amount is migrant (in our case half) and the others are natives.

Second, another interesting thing is that we take pictures of the players. A player of course has to make his/her decisions in private and working by name only would be difficult. We therefore make pictures and show the player that is playing the pictures of the other players, so that he/she can make a decision for each other player. This works well. It’s also great fun to be printing pictures of people in these Congolese villages - and having the people keep the picture afterwards.

Third, people that know the Congo and its people much better than we do strongly suggested not to play the games directly with money. So we needed something else to play with - something that is durable and looks like something you want to earn a lot of. We now play with colorful pins ("epingle" in french): each is a point in a lottery that we play at the end, so people want to earn a lot of them. Let's just say that it was fun running around Bukavu trying to find a substitute for money.

Fourth, we do a solid de-briefing with each player after the games to not only understand whether the player really understood the games, but also to learn why they contributed to some players and not to others. Yes, some people reply things such as “He is a witch-doctor, and has poisened somebody from my family”.





We play the games in several variations to learn even more. One variation in the dictator (the benevolence) game is the following: Players play the dictator game twice, once in public and once in private. The top picture shows Desire playing the game in private: it’s played in a schoolroom with only the player and the enumerator present. The bottom picture shows Eustache. It’s not too clear, but this is the public dictator game: the player sits in a room filled with the other players and several important people from the village. Needless to say we expect that all players will contribute more in the public game (all the other people are there present so you definitely want to show yourself from your good side). However, and this is because of our interest in migrants, we expect this change in increased benevolence to be bigger for migrants than for natives. The reason is that migrants are new in the villages and are not yet integrated, and we posit that one way of integrating is to be overly cooperative: overly benevolent towards other people. The migrants wants to signal that they are good and friendly people so that they can become members of the village – in the private game being benevolent is not a signal because it is played in private, but in the public game it is.



Last week we played our first game in Kabumbiro during a whole day. It was a lot of fun, and I think the villagers enjoyed it even more than Neelan and I did. We compensated the people for spending their whole day away from the fields (with an amount of money sufficient to buy some food for that evening). And at the end of day we have a lottery in which one person wins that same amount one more time, and also receives a Columbia University pen. We wrote a computer program in R that quickly calculates the game play of the people, and then the people who played better during the game have a higher chance of winning the lottery. In the picture we thank the players for their participation and explain one more time what we are going to do with the data.

Btw, there is reason we do a lottery and that we do not tell the people what total amount of points they scored. The reason is to avoid creating bad feelings among the players. Let’s take an extreme case: you play the trust game and contribute to all the other players. Then in the end you find out that you received a total of $0. In other words, you know that nobody contributed to you while you did contribute to the other ones. I wouldn't be very happy to figure out that my fellow villagers do not trust me. By doing this lottery we avoid these problems.



Villager, Neelan and villager. Houses are often tens of meters apart (if not more). Given several hundreds of houses to visit we’ll get our fair share of exercise. I haven’t touched my jumping rope yet.



These are women preparing cassix (not sure if I write it correctly). It’s a liquor made out of banana, water, sorghum. Put them together, then place it in jerry-cans and let it ferment for two days. Before placing it in a jerry-can it is more like banana-liquor. Better drink this without seeing how it is made; or more specifically, how many flies are on top of what you’ll be drinking.

No comments:

Post a Comment