Social drinking is widespread in many areas of the world, and there is at least anecdotal evidence that participation in moderate consumption within a social context can be bene cial for individuals.1 It is argued in this paper that social drug consumption, in particular alcohol drinking, can be used directly to partially reveal a player’s personality, in particular his trustworthiness, and, in a separating equilibrium, can serve as a credible signal. A drug is thought of as a technology which can be used to voluntarily give up some degree of control over one’s action and words; if this consumption takes place within a social context both the consumption itself and the drug-induced behavior is observable.
While the revelation of truth a er the consumption of alcohol is proverbial in many cultures, it su ces for the idea of this paper that the observation of alcohol-induced behavior can be used to better estimate a person’s personality type, or that alcohol as a “social lubricant” heightens social interaction.2 If alcohol can be used to give others better information about one’s personality, the social consumption of alcohol can bene t those who would like to honestly reveal their type. For example, if players play a game of trust, a trustworthy person might su er a subjective disutility from exploiting trust and therefore choose to reward trust. If his trustworthiness is common knowledge, this player would be trusted by others if this leads to a higher payo for both. If his trustworthiness is not observable, other players might not be willing to take the risk of trusting him. e trustworthy player, therefore, has an incentive to employ alcohol in order to credibly reveal information about himself. A non-trustworthy player may not be willing to imitate the behavior of the trustworthy players.3
In general, many di erent drugs are potential signals. If intoxication is associated with a loss of productivity, e.g. a reduction in the pie which is to be split up, there is a natural trade o between the gains due to revelation of trustworthiness and the decrease in productivity. If the drug is highly addictive, it would also be used out of the social context, a use which is only unproductive.
Besides being useful to obtain a better estimate of someone’s type, a suitable drug should therefore i) be not addictive for moderate consumption, ii) have only short term e ects, iii) allow a gradual degree of loss of control which can be easily observed and reciprocated by others. Alcohol satis es these requirements.
If it is e cient for a society to employ a universal, exclusive technology, then it is e cient to coordinate on one drug. Laws or social norms can then be used to enforce the use of only one technology that is well suited for the purpose. is can help to explain why alcohol is legal in many countries, whereas other drugs which are just as harmful or even less harmful from a medically point of view are illegal.4
Furthermore, competing suppliers of signaling technologies such as religions have incentives to discourage drug use. is can help to explain why many religions recommend or require abstinence.
From a modelling point of view, the approach in this paper is also novel: We model the strategic improvement of the informational content of a noisy exogenous signal (similar to the one used in Frank 1987).
e rest of the paper is organized as follows. Section 2 gives a simple example of a game of trust in which trustworthy players employ a drug to reveal their type. Section 3 introduces a more detailed model with a noisy exogenous signal. Section 4 analyzes the model when the noisy signal is normally distributed. It is shown that both separating and pooling equilibria can arise. In a separating equilibrium, only the trustworthy types consume the drug. e pooling equilibrium di ers from usual pooling equilibria considered in the literature on asymmetric information: Both types of players consume the same amount of the drug, but this behavior is nonetheless informative. Long run evolutionary stable equilibria of type distributions are considered. Section 5 discusses the results of the model and possible extensions. Section 6 concludes.