Originaltitel: Przepiorka, W. (2009): Reputation and Signals of Trustworthiness in Social Interactions. ETH Zürich: ETH-Diss.-No: 18649.
Titel der Replikation: Wittgrebe, Franziska: Signaling im Labor: Eine Replikation von Przepiorkas Experiment zu Signaling in Vertrauensspielen mit asymmetrischer Information
Zentrales Ergebnis: Die Resultate der Originalstudie konnten größtenteils repliziert werden.
Abstract Original: This dissertation contributes to the research on trust and trustworthiness. Based on a rational choice approach, it explores theoretically and empirically to what extent reputation and costly signals provide the right incentives for long-term cooperative behavior in trust relations. The first chapter gives an overview of the research on trust and trustworthiness, reputation, and signaling theory. While questions concerning the determinants and consequences of social trust are usually researched in surveys and macro-comparative studies, experimental approaches are used to determine the motives behind actors’ decisions in trust situations. Actors are heterogeneous in their preferences and in the constraints that shape their incentives in social interaction. This fact introduces uncertainty in actors’ decisions to form trusting relationships. Reputation and costly signals are two social mechanisms that help actors to overcome this uncertainty. Chapter 2 devises two game theoretic models comprising the trust game with mincomplete information. The idea behind the models is that trustees are heterogeneous as to their time preferences and only trustees with a sufficiently long time horizon have an incentive to behave cooperatively. Long-term trustees can afford to invest in costly signals or reputation building and thus can be distinguished from short-term types without cooperative incentives. From the two models hypotheses are derived which are tested empirically in the subsequent chapters. In a series of three laboratory experiments hypotheses derived from the signaling trust game devised in chapter 2 are tested. The first experiment (chapter 5) was conducted in Zurich (Switzerland) and Nizhny Novgorod (Russia). Results from the two countries were not significantly different. In line with theoretical predictions, long-term trustees are more cooperative and invest higher amounts in costly signals, and these higher investments increase trusters’ propensities to trust. Contrary to theoretical predictions, trusters in the experimental conditions with investment opportunity did not exhibit a higher level of trust than in the control condition.
Link: https://www.research-collection.ethz.ch/bitstream/handle/20.500.11850/22145/1/eth-798-01.pdf
Abstract Replikationsstudie: In the context of my master’s thesis I replicated a laboratory experiment which Wojtek Przepiorka (ETH Zürich) conducted for his dissertation. The aim of this work was to show that Signaling is a solution for trust games with incomplete information in reality, or at least in the lab. Therefore I accomplished 8 experimental sessions, each with 10 students, which played the trust game with incomplete information either in the role of the truster or the trustee. Trustees were either “short-term” types (playing only one game per interaction) or “long-term” types (playing on average 3 games per interaction). Because of the payoff structure and the repeated interactions, “long-term” trustees preferred to honor trust while “short- term” trustees had an incentive to abuse trust. In the 6 treatment sessions trustees had the opportunity to send a costly signal to the truster, in order to reveal their interest in the interaction. There was no such possibility in the control sessions. Differently from the original study, I conducted 3 sessions with smaller outcomes (original outcomes divided by 3), assuming that subjects could cope better with small numbers and consequently would behave more rationally.
The results are widely the same as in the original experiment. It can be shown that “long-term” trustees send higher signals than “short-term” trustees. It also appears that the probability to be trusted is higher for the sender the higher the signal was he chose to send. So Signaling seems to work for the sender as well as the receiver of the signal. Nevertheless, it cannot be proved that trusters in the control sessions trusted less than in the treatment sessions, which in theory they should. Apparently, there seems to exist a basic level of trust that cannot be explained by the theory. Additionally, there is no effect of smaller outcomes on rational behavior in the game.