Are economic experiments representative of underlying sentiments, or social preferences? This was the topic of a recent series of posts I did on articles by John List two of which were in collaboration with Steven Levitt (you can find them here, here and here). One of the main papers to which List referred was a paper by Nicholas Bardsley, then a working paper, and recently published in Experimental Economics. So that's what I'm reviewing today: Nicholas Bardsley's 'Dictator Game Giving: altruism or artefact'.

Bardsley's main thesis is that altruistic giving, and by extension much of social preference theory, are artefacts of the way in which the dictator game is traditionally constructed: the 'proposer' gets some sum of money, they choose to give some of it to their partner, the 'responder'. The giving act can either be one to one in money terms, or it can be multiplied by some factor (generally two). Hence 'giving' $1 means that the 'responder' receives $2. But, if social preference theory predicts that people should give why doesn't this occur more regularly in everyday life. As Bardsley points out:

A common concern is that people could always make anonymous donations to random strangers in everyday life, for example by mailing cash to persons sampled from the telephone directory, but few if any choose to do so. Transfers are made instead to family members, specific organisations or face-to-face to people requesting money. (123)

His analysis therefore is based on the fact that our lives involve interactions with many different individuals, some of whom we interact regularly and some we shall never meet again. What would happen, therefore, if we introduce an option that allows us to 'take' resources from others into the dictator game?

By standard theory of social preferences (inequity aversion, altruism, reciprocity), it would be argued that individuals would 'give' some positive amount to the other individual, because the benefits to the other enter into our own utility functions. Grossly simplified, we care about other people's material interests. If social preference theory is accurate, Bardsley reflects, then introducing taking should not affect giving behaviour. He argues that allowing taking can assist us in identifying our theory more accurately. Hence, in these experimental variations, as the proposer not only do I have the option to give money to the responder, but I can take money away from them at some rate - Bardsley proposes the inverse of the giving rate (i.e. 1 to 1 or 1/2 to 1).

The results of the experiments are fairly stark: when 'taking' is introduced into the action set, the number of individuals who choose to give drops dramatically. This is the case for all the constructions of the game in which the taking treatment is applied. I'll leave it to you to read the paper for specific nuances of this.

What does this mean? Is dictator game giving simply a result of say, a "Hawthorne Effect" (response to experimental demand characteristics): individuals give when giving is an option and take when taking is an option to be "good experimental subjects". Alternatively, is giving subject to a "range effect" where taking $1 could be seen as kind when you have the option to take $2 (this is still consistent with Rabin's reciprocity models). A third option includes stochastic choice models, which he duly dismisses. Bardsley believes that either Hawthorne effects, or range effects are predominantly 'to blame' and proposes more research in these areas.

Thus, for Bardsley,

[T]he results confirm that dictator game giving provides no evidence of context-free pro-social behaviour or, therefore, orthodox social preferences. (130)

Bardsley's attack is thus routed in the claim that social preferences, if existent, must be everywhere applied, i.e. if we have deep-seated inequity aversion then that should be our main influence when making choices. My interpretation is that he has misunderstood the nuances of social preferences - do 'context-free' preferences of any sort exist? In my reading, the idea is that social preferences can be looked at as 'general rules' of behaviour that can, in fact, be undermined by all kinds of things. For example, when offered enough money for themselves, people don't care as much about inequality. Incentives can alter the 'general rule'. The argument for this, as proposed in say, Bowles's recent paper (on which I blogged here) is that several factors can activate different underlying motivations. When we allow individuals to 'take' we 'frame' the problem differently. I believe strongly that introducing new actions changes the ways in which we understand problems. Actions alter the state in which we make choices.

However, my defence is also problematic. Basically, I have said that theories of social preferences are nigh on unfalsifiable: certain frames will highlight certain actions and social preferences can be activated or deactivated depending on the frame. The inclusion of certain actions in the action set acts as a frame and therefore we cannot tell whether social preferences exist (in general) or whether individuals are rational altruists (in general). I can't really think of a way to get around this in terms of the current research. Nevertheless, I think that this idea of framing is distinct from the idea of the Hawthorne effect that Bardsley identifies and needs to be kept in mind regardless - research must be dedicated to differentiating the two and trying to get our experimental results to be robust to these effects. Moreover, experimental economists and social psychologists should definitely dedicate a greater amount of research to range effects and to the ways in which nuanced ranges can affect experimental outcomes.

[On this last point, Dan Ariely has an entertaining Authors@Google talk where he briefly discusses range effect experiments on 'incidence of flossing'.]

References
Bardsley, N. (2008). Dictator game giving: altruism or artefact?. Experimental Economics, 11(2), 122-133. DOI: 10.1007/s10683-007-9172-2