I’ve become interested in the literature on the measurement of social value orientation (SVO) because SVO is of increasing interest for behavioral economists. What puzzled me about the SVO tests is reciprocity. Assume you want to measure SVO for two players, A and B, using a test that relies on decomposed games like Liebrand’s ring test or the slider measure by Murphy, Ackermann and Handgraaf. Take as an example the dictator game in which both players choose between two self-other-allocations, x=(2,0) and y=(1,1). Assume that A chooses y. There are two reasons: (1) a concern for B’s payoff, and (2) a preference for reciprocity. (2) is a possible motivation behind A’s choice if A expects B to choose y and wants to reward B’s kindness by being kind. But this implies that SVO is sensitive to players’ expectations.
In the literature, I’ve found no discussion of reciprocity motives in relation to SVO tests, hence my question is: Has anyone seen a discussion of this issue, or does anyone know why the SVO literature remains silent about this issue?
Very interesting question, and worth to investigate!
I do know that SVO can be used as a dependent variable as well (I've published about it used in this way, even) as a trait-like (independent) variable - using a continuous scale.
Also, the underlying reasons why people are prosocial might very well be reciprocity expectations, but these simply cannot be measured in the current version(s) of the SVO. Perhaps the reason why someone is prosocial isn't as interesting as the fact that he/she is?
Karlijn, Shafig, thanks for your answers. Let's assume for the moment that SVO is a stable trait. Here is some evidence in support of my claim:
Rigdon and Levine (http://ideas.repec.org/p/pra/mprapa/19372.html) present empirical evidence from two treatments of the so-called Secret Santa Game. The game consists of two stages without any feedback between stages, and participants are divided into two groups, A and B.
First consider the control treatment. In the first stage group A participants are in the role of a dictator and decide about a transfer, which is given to a randomly chosen participant from group B. In stage two, participants from group B are in the role of the dictator and decide about a transfer, which is given to a randomly chosen participant from group A. Hence, in the baseline treatment participants have homegrown expectations about the transfer they will receive.
In the other treatment the transfer form group B participants is not decided by the participants but determined by a random mechanism. This is common knowledge, so that all group A participants expect to receive the same transfer (in expectations).
The results reveal significant differences between treatments regarding the behavior of participants from group A. Clearly, the difference stems from differences in participants’ expectations, which are either homegrown or controlled for through the use of the random mechanism.
Current methods for measuring SVO are similar to the treatment with homegrown expectations, which would imply that SVO measures some combination of social preferences and expectations. Now is this an interesting issue or just a small methodological point?
There might be a way to formally test this. I recall when I was collecting my data, I initially told participants that the 'other' person is also doing the very same task (SVO) for them. Interviewing the subjects and looking at data directly confirmed what you suggested: the SVO had become a kind of a second order PD game. I had to change the instructions, and then the responses became more variable and in accordance with the literature. Maybe you could test this formally by creating different instructions.
My paper is here: http://www.ncbi.nlm.nih.gov/pubmed/25224998 - I used an updated version of the ring measure.
Thanks George, I will take a look at the paper.
Actually, I'm currently running an experiment to test this. It consists of the following four treatments:
1. Certainty: Player 1 chooses an allocation that determines player 1 and player 2's payoffs.
2. Uncertainty: Players 1 and 2 simultaneously choose allocations. Then, a coin toss determines whether payoffs are determined by the allocation chosen by player 1 or 2.
3. Decomposed: Players 1 and 2 simultaneously choose allocations. Then, payoffs are determined by the sum of the allocations.
4. Matrix: Same as Decomposed but the game is presented in matrix form, i.e., participants see the recomposed matrix-game.
A participant's SVO is measured in each of the four treatments (within-subject design). After a first look at the data we see two features:
- homogeneity at the aggregate level / the SVO distributions are not significantly different across treatments
- heterogeneity at the individual level / a participant's SVOs differ across treatments.
I wonder if you actually did the experimen - if you have this published I would be happy to read it
@George: Here is the link to the working paper: http://www.uni-marburg.de/fb02/makro/forschung/magkspapers/paper_2016/06-2016_greiff.pdf
- Badges
- Science topic
- Similar topics
- Psychology
- Social Psychology