Ok, so the title of the post gives the paper more pizzazz than it really has, but hey? Today, a little discussion about Nava Ashraf, Iris Bohnet and Nikita Piankov's (2006) paper 'Decomposing Trust and Trustworthiness'.

The main reason that it is of interest to me is that it used a sample of students from South Africa in Cape Town (which somehow was spelt Capetown in the paper, ATROCIOUS editing!), along with students from Boston, US and Moscow, Russia.

The next point of interest is that, unlike many economic experiments, it assesses within-subject variation across two different games, the dictator game and the trust or 'investment' game.

The authors had 4 main hypotheses that they wished to test:

1. Trust is only based on expectations of trustworthiness.
2. Trust is only based on unconditional kindness.
   
3. Trustworthiness is only based on reciprocity.
4. Trustworthiness is only based on unconditional kindness.
   

The results from the paper for the dictator game and trust game do not differ significantly from conventional results as reported by Camerer (2003) on a survey of experimental results. Pertinently, non-white South Africans in the sample were significantly less trusting than white South Africans using an index of trust that included 'trust'-probing questions such as “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?”

With respect to their hypotheses, they find that, "trustors derive satisfaction from trusting independent of amounts expected back and unconditional kindness" (201). This dismisses hypotheses 1 and 2. With respect to hypotheses 3 and 4, they find that "[In] contrast to trust, unconditional kindness accounts for most of the variance explained while reciprocity plays a comparatively small role" (202-204) but, "Overall, we reject both our hypotheses" (202) they do this because their results seem to explain approximately 20% of variation, which they do not feel is sufficient to 'prove' their final hypothesis.

This doesn't give us much insight, however their one conclusion is quite interesting, in response to the question posed by Camerer “What game do people think they are playing?”, they argue that:

Our results suggest that many people may play a different game than researchers thought they were playing when confronted with the “investment game.” Only 36 percent of the 159 trustors who decided to send any money in our “investment game” expected to make money (204).

Moreover, one of their final points is that:

Our design allowed us to solve one of the important trust puzzles, namely that people trust even though hardly anyone makes money by doing so. We found that generally, people are aware of this. They trust even though they know it does not pay monetarily. They enjoy the act of trusting and being kind to others, even to anonymous strangers (204).

This paragraph makes sense to me, up to the point when they say 'enjoy'. In a world where people get windfall cash (as that is basically how experimental money often is thought of, see John List, 2007) people might behave as if they enjoy giving away money to others or as if they 'enjoy' trusting others, but I remain unconvinced that they actually derive 'joy' from such acts in the real world.

Which makes me speculate. What if you could run experiments, and then have a 'beggar' set up outside the experiment (not actually a beggar but an actor, or something) and you gave the individuals the money and then you observed, after the experiment how many individuals chose to give to the 'beggar'. Apart from the 'deceiving the experimental subjects' problem, this might be interesting to observe and to see whether individuals behave consistent with their experimental generosity and 'enjoyment' of giving. It is this correspondence of field observation and experimental economics that requires propagation and increased funding in order for us to validate within subject variation and validity of our laboratory results.

Reference
Nava Ashraf, Iris Bohnet, Nikita Piankov (2006). Decomposing trust and trustworthiness Experimental Economics, 9 (3), 193-208 DOI: 10.1007/s10683-006-9122-4