The Last Dictator Game? Dominance, Reactivity, and the Methodological Artefact in Experimental Economics

Abstract

The Dictator Game (DG), one of the best-known designs in experimental social science, has been extensively criticized, and declared by some to be defunct, on the grounds that its results are the product of a research artefact. Critics of the DG argue that the behaviour observed in the game (where agents frequently decide to share their money with strangers) is not the outcome of genuine pro-social preferences but must, instead, be interpreted as a response to the cues given by the experimental design, where these cues signal that the game is about ‘sharing’ (i.e. about altruism or pro-social behaviour). Despite this criticism, the DG continues to be extensively used, and some have defended its validity as an instrument capable of measuring the role of social pressure and social norms against economic motivations. This article examines the assumptions implicit in the claim that the DG results are artefactual and spells out the conditions under which the game can be used to test hypotheses about pro-social behaviour. To conclude we show how the DG debate helps to illuminate the pitfalls embedded in the abandonment of the dominance principle, and argue that this is crucial to the expanding agenda of experimental economics.

Acknowledgements

The research for this article was supported by grants FFI2011-28835 and FFI2014-57258-P of the Spanish Ministry of Science. I am grateful for the valuable insights supplied during the refereeing process.

Notes

[1] For an early review on the effects of anonymity and original results on identifiability of recipients, see, for example, Small and Loewenstein (2003).

[2] It should be noted that Bardsley offers two other potential readings for his findings (linked, respectively, to ‘range effects’ and to stochastic choice models), though the one associated with Hawthorne effects seems to prevail in the discussion.

[3] More often, and honouring the origins of the term, which can be traced back to the study of productivity arrangements in industrial organizations at the beginning of the twentieth century, the Hawthorne effect is used to refer to the instances in which subjects modify their behaviour to ‘look good’ in the eyes of researchers that are studying them. Adair (1984) offers a clarifying article in which he dissects the original studies that led to the coinage of the term Hawthorne effect: a series of studies taking place during the 1920s in the factory Hawthorne Works, in the vicinity of Chicago. Perhaps the best known among these are the so called Illumination Studies, in which it was observed that both the treatment and the control groups of workers increased their performance during an intervention in which the treatment consisted in gradually decreasing the illumination of the premises.

[4] See, for example, Adair (1984) and Chiesa and Hobbs (2008), who warn against some of the confusions that come from the multiple meanings associated to the term, ‘Hawthorne effect’.

[5] This notion of altruism simpliciter is akin to Andreoni’s pure altruism. Andreoni (1990) distinguishes between pure and impure forms of altruism to accommodate phenomena such as warm glow and interdependence of preferences.

[6] For an extended discussion of the precepts, see Guala (2005).