Promises and expectations

Florian Ederer, Alexander StremitzerGames 2013

TL;DR

This paper provides the first experimental evidence that people keep promises primarily because they want to avoid disappointing the expectations of others, especially when those expectations are directly tied to a promise. This "expectation-based" explanation for promise-keeping is supported by a novel experimental design that isolates the effect of expectations from other factors like commitment and reputation.

What Problem Does This Paper Address?

Why do people keep their promises, even when there's no legal enforcement or reputational cost? This question has puzzled economists, psychologists, and philosophers for decades. While it's clear that promises significantly boost cooperation, the underlying psychological mechanism remains debated. Two main theories exist: the "commitment-based" theory, suggesting people have an intrinsic preference for keeping their word, and the "expectation-based" theory, arguing that people are motivated by the guilt of disappointing others' expectations. Previous experiments struggled to disentangle these, often finding support for commitment-based explanations. This paper aims to definitively test the expectation-based theory by creating a scenario where expectations are high due to a promise, but the ability to fulfill that promise varies.

Methods

The authors designed a controlled laboratory experiment using a modified dictator game. In this game, a "dictator" (Player A) can send a message to a "recipient" (Player B). Player B then decides whether to "opt out" of the game, in which case both players receive a small payoff ($2). If Player B stays in, a "random device" determines whether the dictator is able to fulfill a promise. This device is either "reliable" (5/6 chance of being able to perform) or "unreliable" (1/6 chance). Crucially, the dictator knows whether they can perform, but the recipient does not until later. The experiment manipulates the reliability of this device, thereby exogenously varying the recipient's expectations and the dictator's beliefs about those expectations, while keeping the "commitment" (the promise itself) constant. This allows them to isolate the impact of expectations.

Datasets

Not discussed in the paper. The study uses data generated from their controlled laboratory experiment with 280 UCLA undergraduate students.

Key Insights and Ideas

  • Promises Matter, But How? The study confirms that promises significantly increase cooperative behavior, but crucially, it pinpoints why: the desire to meet the expectations they create.
  • Lexicographic Promise Keeping: The authors propose a "lexicographic" theory of promise keeping. This means expectations only matter if they are supported by a direct promise. If there's no promise, even if expectations are high, they don't influence behavior.
  • Expectations are Key, Not Just Commitment: By varying the likelihood of being able to fulfill a promise (through the reliable/unreliable device), the researchers could change expectations without altering the promise itself. This manipulation led to significant changes in behavior, supporting the expectation-based theory.
  • Guilt Aversion as a Driver: The model incorporates guilt aversion, where individuals feel bad if they fall short of others' expectations. This guilt is amplified when expectations are high due to a promise.
  • The Power of a Direct Link: The findings suggest that the direct promissory link between parties is essential for expectations to influence behavior. In situations where this link is absent (e.g., third-party promises), expectations may not play the same role.

Key Results

  • Higher Performance with Promises and Reliability: When dictators made a promise, they chose higher performance rates (contributing more to the recipient) when the random device was reliable (5/6 chance of performance) compared to when it was unreliable (1/6 chance). This difference was statistically significant.
  • No Effect Without Promises: When dictators sent no message or only "empty talk" (no explicit promise), the reliability of the random device had no significant impact on their contribution decisions. This strongly supports the lexicographic theory.
  • Recipient Expectations Rise with Promises: Recipients' first-order beliefs (expectations about the dictator's action) were significantly higher when a promise was made, and even higher when the device was reliable.
  • Dictator's Second-Order Beliefs Adapt: Dictators' beliefs about the recipients' expectations (second-order beliefs) also adjusted based on the device's reliability, and these beliefs significantly influenced their decisions when a promise was present.
  • Opt-Out Rates Reflect Promises: Recipients were significantly less likely to opt out of the game when they received a promise, indicating they expected a better outcome by staying in.

Limitations and Future Work

  • Generalizability of Guilt Aversion: While the model incorporates guilt aversion, the paper acknowledges that other factors like social preferences could also drive behavior. The distribution of guilt aversion parameters suggests a significant portion of subjects are not driven by guilt.
  • Experimental Setting: The study was conducted in a controlled laboratory setting with student participants. Future work could explore whether these findings hold in real-world scenarios with different populations and stakes.
  • Future Research Directions: The authors suggest further investigation into the "lexicographic" nature of promise keeping and how it interacts with commitment and reputation in more complex settings. They also propose exploring the distribution of altruism and guilt aversion parameters in more detail.

Summary updated

BibTeX
@inproceedings{ederer2013promises,
  author = {Florian Ederer and Alexander Stremitzer},
  title = {{Promises and expectations}},
  booktitle = {Games},
  year = {2013},
}
Figure