The Effect of Commitment Types in Repeated Games with Long-lived and Short-lived Players

How to cite

Knutsen, Magnus Våge; Heggedal, Tom-Reiel. 2025. "The Effect of Commitment Types in Repeated Games with Long-lived and Short-lived Players." AEA RCT Registry. AEARCTR-0015903. https://doi.org/10.1257/rct.15903-2.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Magnus Våge Knutsen* , Tom-Reiel Heggedal

* Primary investigator

Registry ID
AEARCTR-0015903
DOI
https://doi.org/10.1257/rct.15903-2.1
First registered
Last updated
Status
Completed
Trial dates
2025-05-04 – 2025-07-31
Countries
Norway ()
Registry record
http://www.socialscienceregistry.org/trials/15903

Abstract

Repeated games with a long-lived player facing a sequence of short-lived players has, sins these
games were first studied in Fudenberg and Levine (1989), been the focus of much theoretical work. In more recent years these models have received much attention due to their application to trade on online platforms such as Ebay and Amazon (which is becoming increasingly important). Like the more studied case of two long-lived players, the case with one long-lived player also suffers from
multiplicity. Thus, while it is clear from a theoretical perspective that repeated interactions can
help long-lived and short-lived players coordinate on mutually beneficial outcomes, it is unclear
how this will come about.

We set out to experimentally study strategy choices in repeated games with long-lived and short
lived players by running a series of lab experiments. Our perspective is on the role of commitment
types in this framework. That is, to what extent the presence of commitment types help coordi-
nation towards beneficial outcomes. From a theoretical perspective, commitment types will help
long-lived and short-lived players to coordinate on beneficial outcomes by excluding equilibria in
which cooperation never takes pace.

We implement the model of reputational concerns introduced by Fudenberg and Levine (1989,
1992), which has become a workhorse model in the theoretical work on situations with reputational
concerns. In the model, a long-lived player interacts with an infinite sequence of short-lived players.
The long-lived player has a commitment types and short-lived players observe the complete history
of play. Despite the vast theoretical literature, the experimental evidence on infinitely repeated
games building on the paradigm of Fudenberg and Levine (1989, 1992) is scarce. The environments
with a long-lived player facing a sequence of short lived players has been studied in experiments
before, this has been in setting with finite horizons. Thus, to our knowledge we are the first to take
the model of Fudenberg and Levine (1989) to the lab.

Intervention

The aim of the experiment is to empirically investigate the effect of commitment types in a repeated game where the long-lived player faces a trade-off between short-run and long run gains, and where the efficient outcome is an equilibrium. However, as it is only one of many, it requires that players coordinate. We hypothesis that the presence of a commit types will facilitate this coordination.

Our plan is to run two different treatments: One without commitment types, and one with commitment types. Commitment types are induced by way of computerized long-lived players who always chooses a specific action.

In addition we run a control treatment where both players are short lived. This treatment will save as a empirical benchmark.

Experimental design

See attached document.

Design details

Randomization method
Randomization done in office by a computer. Sessions are randomised into one of two treatments. The control treatment is without commitment types, and our treatment is with commitment types. Commitment types are induced by way of computerized long-lived players who always chooses a specific action.
Randomization unit
The unit of randomization is session.
Clusters
Six matching block per treatment (two treatments). There are eight participants per matching block in the treatment without induced commitment types, and seven in the treatment with induced commitment types (one participant is replaced by a computerized long-lived player.
Observations
90 (6*8 + 6*7)
Arms
There are 48 participants in the treatment without induced commitment types, and 42 in the treatment with induced commitment types.
Primary outcomes
Our focus is on four main outcome measures: The payoffs of the long-lived player, the frequency at which the long-lived player chooses the actions that most benefits the short-lived players, the frequency at which the short-lived players chose the action the most benefits the long-lived player, and the frequency at which the long and short-lived player are able to coordinate on the efficient outcome.

Cited by

No citation data available for this registration yet.

Keywords