Elizabeth Hoffman, Kevin McCabe, and Vernon L. Smith in "Social Distance and Other-Regarding Behavior in Dictator Games," American Economic Review, 3, 1996, pp. 653-660, use instructional and procedural manipulation to examine the effects of social norms. They find that the degree of social distance - separation of subjects from the experimenter - plays an important role in determining and understanding the behavior in the experiments with salient fairness considerations.

The experimental design implements a dictator game (in particular, the ultimatum game dictator control version). In the experiment, there are two players. Player 1 (the dictator) makes an offer of how to split $10. Player 2 has no decision to make and the proposal of Player 1 is implemented to give the final payoffs to each of the players. The Nash equilibrium is to offer zero, just like in the ultimatum game. However, player 1's behavior is free of strategic considerations. Therefore, any positive amount offered in the dictator version of the game is a measure of altruism or another type of other-regarding preferences. In the ultimatum game, the modal offers are around a half of the pie, whereas in the dictator game they are significantly lower (see Forsythe et. al [1994]). This difference is due to strategic concerns and not due to fairness.

Hoffman et al. hypothesize that the concept of social distance between the dictator and the paired player as well as other's who know the dictator's decision (in this case experimenters) matters. They define the social distance as: "... as the degree of reciprocity that the subjects believe exists within a social interaction. By design, economic or material action is unidirectional in the dictator game." (pp. 654) In the actual experiment they vary the distance by changing the procedures and wording in instructions that affect the dictator's social isolation and anonymity.

The authors note that, "In laboratory experiments we cannot assume that subjects behave as if the world is completely defined by the experimenter. Past experience is important in so far as beliefs are based on experience. The future is important in so far as people are accustomed to operate in an environment in which there is ongoing social interaction, and in so far as subjects may be concerned about the extent to which their decisions have post-experimental consequences, or that others may judge them by their decisions. In short, subjects bring their ongoing repeated game experience and reputations from the world into the laboratory, and the instructional language, especially in single-play sensitive experiments like the dictator game, can subtly suggest more or less isolation from that interactive experience. It is well documented in the experimental literature that the "framing" of a decision can influence expectations by associating a subject's decision with past experience." (pp. 655)

Treatments:

There are two rooms, A and B, into which the subjects are randomly assigned. People in room A are the dictators. In the description of treatments, sentences labeled (i)-(iv) represent conditions that are modified in the corresponding instructions.

• Double Blind 1 (DB1)
  (i) One subject in room A is chosen to be a monitor and is paid $10. The experimenter leads the instructions aloud to verify that all subjects have the same instructions.
  (ii) The instructions inform the 15 subjects in room A that there are 14 envelopes. 12 envelopes contain 10 one dollar bills and 10 blank slips of paper and 2 envelopes contain only 20 blank slips of paper. This prevents the situation that the experimenters would know what every subject did had all of them taken all the dollar bills and filled the envelopes with blank sheets only. Subjects are asked one by one to take a random envelope and bring it to the back of the room in a carrel to maintain a privacy.
  (iii) The subject opens the envelope in private and decides how many dollar bills to keep for himself and how many to leave for a person from room B. All bills taken are replaced by blank sheet of papers to preserve the same thickness of envelopes.
  (iv) After making the decision, the subject seals the envelope and returns it to a box by the exit. This is done for all the subjects. The experimenters and the monitor then take the box to outside of room B. Subjects are called one by one, in their presence a random envelope is chosen and opened. The monitor records the contents preserving anonymity. The content of the envelope is then given to the subject who leaves the experiment. This is done for all subjects. At the end the monitor is paid.
  
  
• Double Blind 2 (DB2)
  The second treatment omits (i) the paid monitor and (ii) the two blank envelopes. This causes that complete anonymity is not guaranteed but is very likely provided at least one person leaves some money in the envelope. Because of weakened social isolaton, the offers are expected to increase.
  
  
• Single Blind 1 (SB1)
  SB1 differs from DB2 only by modifying (iv) so that the experimenters, but not the other subjects, learn the identities of each decisionmaker by opening the envelopes at the experimenter's desk and recording the decisions, in isolation from the other subjects. Because of further decrease in social distance, the offers are expected to increase comparing to DB2.
  
  
• Single Blind 2 (SB2)
  SB 2 is identical to SB1 except for (iii). The envelope now does not contain money but a decision form filled out by a subject in private. The returned envelope is opened by the experimenter at the front of the room, data recorded, and subject is paid according to the made decision. Then the enveloped is stuffed with money (or not) based on the decision of the subject and sealed. The subject then drops it to a box and exits the experiment. As Hoffman et al. note, "Since SB2 created a direct transaction between the subject and the experimenter (in order to get paid), social distance is narrowed, and we predict that offers will increase relative to SB1." (pp. 657)

Research Hypothesis:

H_R: F(DB1) > F(DB2) > F(SB1) > F(SB2)
where F(.) is the population distribution of offers made by the dictators for each of the above treatements.

As predicted by the research hypothesis, the offer distributions decrease as social isolation and anonymity conditions are weakened from DB1 to SB2, although the individual pair-wise treatments are not generally significantly different. The authors interpret the data as generally supportive of the assumption of self-regarding preferences but place three caveats in this assumption. First, in DB1 a few subjects sent considerable amount of money to their paired players, suggesting the presence of some other-regarding preferences. Second, there is a possibility of socialization - the subjects were all undergraduates from the same university. Third, the double blind procedure may not result in more self-regarding offer distributions than in ultimatum games what would occur only if the players' 1 expectations that offers will be rejected dominated the effects of complete anonymity.