To give or not to give. Explaining earnings choices in an online labour market experiment.

How to cite

Kovacevic, Sonja; Knutsen, Tora. 2021. "To give or not to give. Explaining earnings choices in an online labour market experiment.." AEA RCT Registry. AEARCTR-0008345. https://doi.org/10.1257/rct.8345-1.0

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

Sonja Kovacevic* , Tora Knutsen

* Primary investigator

Registry ID
AEARCTR-0008345
DOI
https://doi.org/10.1257/rct.8345-1.0
First registered
Last updated
Status
In development
Trial dates
2021-10-12 – 2021-10-25
Countries
Private
JEL codes
C91; D63; H23
Registry record
http://www.socialscienceregistry.org/trials/8345

Abstract

We run an experimental study on the online labour market Prolific. Allowing for one-sided earnings choices after joint production in pairs, we investigate drivers of giving behaviour by varying the level of hierarchy and connectedness stakeholders experience in the experiment. To measure the effect of hierarchy for giving, we compare choices under boss-assisstant to co-worker vignettes. We also compare co-worker vs. neutral vignettes to measure the effect of connectedness on giving.

Intervention

We invite participants on the online platform Prolific to participate in a study. In this study, participants are working on an effort task, get matched into pairs and some participant get randomly selected to split earnings within pairs.
Within the group of (randomly chosen) deciders, we randomly vary how we describe the person they are matched with. In the hierarchy treatment the decider is referred to as the boss and the other participant as the assistant. In the connected treatment, both participants are called co-workers. In the neutral treatment, the decider learns about the other participant only after completing the task and the instructions do not contain words like "match" or "pair" that could generate connectedness perceptions.
We study behaviour differences across these treatment conditions.

Experimental design

We conduct our experiment on the online platform Prolific, where we randomize one of the participants to decide about earnings in that pair. Each pair receives a bonus and the decider can take the whole bonus for him/herself or allocate a portion or the whole bonus to the receiver. The receiver cannot review or change this decision in any way, so that strategic considerations should not play a role for the decider.

The treatment consists of three different vignettes, randomly shown to the deciders. The first vignette addresses the decider as the boss and names the other player the assistant, the second vignette refers to both players as co-workers, while the third vignette minimizes references to the other player to reduce connectedness. After receiving this information regarding their respective roles, participants work 5 minutes on a task. This task consists of counting the number of 0's. On completing the tasks, deciders learn their relative performance in the pair and make their decision about the bonus.

Design details

Randomization method
We randomize participants by their incoming player number in Prolific to the different treatments. This way, both participants entering early or late will be balanced across the different treatment conditions.
Randomization unit
Individual participants on Prolific.
Clusters
4214
Observations
4214 This is the maximum number of observations we can realize. This number may go down, if we have dropouts in our study. Dropouts will not get paid so that we can in principle recruit new participants to replace them. However, partners of dropped out participants will receive their participation fee (but not the bonus), so that the number of observations will decrease in dropouts, but at a rate of less than 1. Additionally, if inflow of participants to Prolific should be very slow, we may have to increase the participation fee which would also reduce the number of observations. In both cases, we will reduce observations symmetrically across treatments, so that each of our three treatment arms for the deciders are of the same size and that we maintain equal numbers of deciders and non-deciders. This is also a necessity given that people are matched in pairs and only one participant per pair can decide.
Arms
- 100 participants in the dictator game benchmark for external validity - 2057 will be non-deciders who only serve to be decided on - 2057 will be deciders, with 685 in each of the three decider treatments: --> 685 bosses --> 685 co-workers --> 685 neutral (non-connected) deciders.
Primary outcomes
Share of bonus taken by deciders. This is a percentage of the 2£ bonus available in each pair of participants.
Secondary outcomes
Messages from the deciders in the boss and co-worker treatments and explanations of earnings choices for the deciders (free text).

Cited by

No citation data available for this registration yet.

Keywords