Wednesday, July 18, 2012

Fun Things With Data Entry.

As you know from the previous post I'm entering data. Part of this data is information about the lab-in-the-field experiments I am conducting in Eastern Congo. Normally for such experiments the players are explicitly selected so that they are not acquainted with each other, or this is assumed, or the players play the games anonymously. Researchers do this so as to single out a particular mechanism that impacts a player's strategic behavior. Now of course the real-life strategic behavior of individuals is very much influenced by the context and whom one is interacting with. What is important are often things different than only 'observables'; people make decisions towards others based on things such as frequency of interaction and shared experiences. "We are member of the same church", "He is family", et cetera. In contrast to most lab-in-the-field experiments, our games in Eastern Congo take the player's natural context -- his underlying, latent network so to say -- at heart. To do so we take polaroid pictures of each of the 18 players and have them make strategic decisions toward each of the other 17 players. From the information that this gives we can construct trust and benevolence networks (see here for more information). We then ask to each of the 18 players questions about each of the other 17 players, such as "Are you family of this person?", "Are you friends with this person?", etc. This provides us the latent-network information -- information that is not observable. From combining these two we can thus learn what is important for levels of trust and benevolence. Maybe being member of the same church is important for levels of trust. Maybe having been displaced together is crucially important for levels of benevolence. Et cetera.


Eustache at work: showing the picture of
one of the players, to another player.


So what is important for people to contribute to others? I have not yet combined the network information with the trust and benevolence networks for two reasons. First, I want to pre-register the hypotheses before looking at the data so that I cannot cherry-pick my results. Together with collegue Neelan we are about to finish a paper in which we present our hypothesis and present results based on fake data. We expect to register this soon. Second, I have only entered data for 2 villages up to now. However, some interesting things I already would like to share via this blogpost. After each game we do an in-depth debriefing with each player to understand why the players played as they did. I've now filled out only two villages but this already gives a potential 2*18*17 = 612 reasons.


So what are reasons for people to contribute to each other? [Do note that you should take this with a grain of salt because this is what people say; it is thus not necessarily the truth. People for example might be unwilling to say "I don't contribute to him because he/she is from a different ethnic group". Indeed this is exactly the reason in the first place why we conduct lab-in-the-field experiments, and the data from the network questions will shed light on this in due time. However, this might already be very informative.] Common reasons not to give are "He/she is arrogant", "He/she is hypocritical", "Not from my generation", etc. Often-mentioned reasons to give include "he gives me advice", "we are both migrants", etc. Now of course the great thing of just asking why a person contributes is that one also receives more exotic reasons for (non)contribution. Hereby a small selection (I translated this from French), note that I added whatever is in square brackets:
  • "Once he paid the school fees for my son."
  • "When I'm alone at home, she will collect firewood for me."
  • "He always buys drinks (alchohol), but without giving anything to me."
  • "We were caught by the Interhamwe [a rebel group in Eastern Congo] but he freed me."
  • "She assists my child by giving food in my absence."
  • "He is a miner and therefore his family is not social."
  • "He reminds me of my father who died. And I therefore like him."
  • "My daughter married his son."
  • "He gave me food when I came from the forest and was famished."
  • "If my house would catch fire he would help."
  • "He put me in prison."
  • "I need the points to buy a pen."
  • "He is ANR [The Congolese security service]."
  • "He tells bad stories about me to the other villagers."
  • "He never pays his debts. He owes me a lot of money."
  • "He is a sorcerer, and I am sure he placed a spell on me."
  • "He does't salut me when we pass on the street"
As expected, it is likely that latent (unobservable, but measurable) networks are going to be very important to understand the trust and benevolence networks in Congolese villages. Also, this reminded me of a great reply from one of our enumerators for the evaluation I'm conducting with Macartan and Raul in Eastern Congo. The question was "Where was the person in 2006?". Below is the raw input by the enumerators into their PDA. We make use of codes: "-8" for example means "not applicable" and therefore appears very often. However, some enumerators are more complete and write things such as "-8, was not born yet". Two replies are specifically great. I highlighted them in yellow below. Translated in English the person in 2006 was "Still in the belly of his/her mother". Brilliant! (h/t Raul).
 ************************************************
  ************************************************
tab qf019_location_2006_1

   QF019_LOCATION_2006_1 |      Freq.     Percent        Cum.
-------------------------+-----------------------------------
    n'etaitpas encore ne |          1        0.00        0.00
                      -8 |          2        0.00        0.01
                      +8 |          1        0.00        0.01
                     --8 |          2        0.00        0.01
                      -0 |          1        0.00        0.02
                      -5 |          1        0.00        0.02
                      -7 |          1        0.00        0.02
                      -8 |      3,508        7.89        7.91
       -8   PAS  VIVANTE |          1        0.00        7.92
     -8  PAS  ENCORE  NE |          1        0.00        7.92
                     -8. |          2        0.00        7.92
                      -9 |         10        0.02        7.95
                     .-8 |         30        0.07        8.01
                    ..-8 |          1        0.00        8.02
                       0 |         31        0.07        8.09
                    1720 |          1        0.00        8.09
                      50 |          4        0.01        8.10
                    7010 |         45        0.10        8.20
                    7020 |         10        0.02        8.22
                    7030 |          1        0.00        8.22
                    7052 |          3        0.01        8.23
                    7053 |          3        0.01        8.24
                    7054 |          4        0.01        8.25
                    7065 |          1        0.00        8.25
                    7071 |          6        0.01        8.26
                   70710 |          1        0.00        8.26
                    7073 |          9        0.02        8.28
                    7074 |          8        0.02        8.30
                    7075 |          1        0.00        8.30
                    7076 |          8        0.02        8.32
                    7090 |         10        0.02        8.34
                    7094 |          4        0.01        8.35
                       8 |          2        0.00        8.36
                     A-8 |          2        0.00        8.36
                      A0 |          3        0.01        8.37
                      A1 |     37,165       83.64       92.01
                    A1-8 |          1        0.00       92.01
                     A1. |          8        0.02       92.03
                   A1... |          1        0.00       92.03
                     A1… |          1        0.00       92.03
                      A2 |      1,962        4.42       96.45
                      A3 |        401        0.90       97.35
                      A4 |         94        0.21       97.56
                      A5 |         41        0.09       97.66
                      A6 |        740        1.67       99.32
                      A7 |         55        0.12       99.44
Dans  leventre  desamere |          1        0.00       99.45
      N'etait encore nee |          1        0.00       99.45
  N'etait pas  encore ne |          1        0.00       99.45
   N'etait pas encore ne |          5        0.01       99.46
  N'etait pas encore nee |          1        0.00       99.46
   N'etait pas encore né |          1        0.00       99.47
       N'était encore né |          1        0.00       99.47
   N'était pas encore né |          3        0.01       99.48
                     NON |          4        0.01       99.48
                       O |          1        0.00       99.49
     PAS  NE  EN  2007   |          1        0.00       99.49
           PAS ENCORE NE |          2        0.00       99.49
                   Pa ne |          2        0.00       99.50
                 Pas  ne |          1        0.00       99.50
              Pas encore |          1        0.00       99.50
           Pas encore ne |          5        0.01       99.51
                  Pas ne |        159        0.36       99.87
                  Pas né |          1        0.00       99.87
                   Pasne |          1        0.00       99.88
                  Ventre |          1        0.00       99.88
                      _8 |          1        0.00       99.88
                  pas ne |         49        0.11       99.99
                  pas né |          1        0.00       99.99
                   pasne |          3        0.01      100.00
-------------------------+-----------------------------------
                   Total |     44,435      100.00
  ************************************************
  ************************************************

No comments:

Post a Comment