Active and Passive Risk-Taking

How to cite

König-Kersting, Christian; Merkel, Anna Louisa; Lohse, Johannes. 2021. "Active and Passive Risk-Taking." AEA RCT Registry. AEARCTR-0007271. https://doi.org/10.1257/rct.7271-1.0

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

Christian König-Kersting* , Anna Louisa Merkel , Johannes Lohse

* Primary investigator

Registry ID
AEARCTR-0007271
DOI
https://doi.org/10.1257/rct.7271-1.0
First registered
Last updated
Status
Completed
Trial dates
2019-06-15 – 2019-07-29
Countries
Germany ()
JEL codes
D81, D91, C91
Secondary IDs
osf.io/3dcnj
Registry record
http://www.socialscienceregistry.org/trials/7271

Abstract

Risk-taking may depend on whether risks result from an action (active risk-taking) or from not taking action (passive risk-taking). We develop a new experimental risk-elicitation procedure, the Dynamic Lottery Adjustment Task, and employ it across two separate experiments to study the size and direction of potential mode-of-choice effects (i.e. differences in risk-taking between active and passive decision modes). While our tightly controlled lab study provides little evidence for such effects, we find substantial evidence for mode-of-choice effects when decisions are spread out over 10 days and attention costs are a key feature of the online choice environment we use.

This is an ex-post registration. The original pre-registration is available at https://osf.io/3dcnj

Intervention

Please refer to the original registration at https://osf.io/3dcnj

In this plan, we describe a lab-in-the-field experiment in which we will investigate risk-taking in situations where choices are made by commission (taking an action) or omission (remaining passive). While most of the existing evidence on risk-taking comes from studying behavior in situations, where risk-taking results from active choices, in many naturally occurring situations risks emerge from remaining passive.

To study this form of passive risk-taking we have already conducted a laboratory experiment that was designed to answer the question if individuals are more prone to take risks by commission than by omission. In this previous experiment, subjects were endowed with an initial lottery that they could adjust. Across different treatment conditions, adjustments, i.e. changes to the payoffs associated with the two possible lottery outcomes, were implemented either by taking an active choice or by remaining passive and not taking a choice. To disentangle the mode of choice from a status-quo effect, we furthermore varied whether subjects start from an initially safe lottery, such that any adjustment of the lottery implies more risk-taking, or from an initially risky lottery, such that adjustments lead to less risk-taking. In this previous experiment, we do not find that people take more risks when risks result from omission as compared to commission, while the initial lottery assignment strongly affects risk-taking.

We plan to follow up on this previous lab study with a lab-in-the field experiment. In this experiment, we wish to test, if our previous findings are an artifact of the elicitation method in the lab in which subject could make adjustment steps in a short succession of decisions or if they generalize to a more realistic setting where there is a larger time gap (approximately 24h) in between single adjustment steps. To do so subjects will be asked to make one decision per day within a predetermined time window of 4 hours. Depending on treatment condition, this decision is made by logging into an experimental software and pressing a button (active condition) or by remaining passive (passive condition) and only logging into the software to stop further adjustment steps. In total there will be 10 possible adjustment steps implemented in a succession of 10 days. Before the experiment starts subjects will receive instructions by one of the experimenters. Following the last day of the decision period, there will be a questionnaire. Subjects filling out this questionnaire will receive a bonus payment.

Our proposed experimental design is hence a 2x2 design varying (i) the mode by which choices can be implemented (active or passive) and (ii) if subjects are initially endowed with a safe or a risky lottery (safe or risky).

Experimental design

The study consists of three parts that run over a succession of 12 days: (i) the initial instructions (Day 1), (ii) the main decision task (Day 2 - Day 11) and (iii) the final questionnaire (Day 12). After Day 12 subjects will receive a payment reflecting their choices in (ii) and a bonus payment if they fill out the final questionnaire (iii).

The decision task consists of a succession of up to ten decisions made on subsequent days. On the first day of the decision task subjects are endowed with an initial lottery. This lottery has two outcomes (yellow or green) each of which is associated with a specific payoff. Each decision affects the payoff associated with each outcome. To make a decision subjects can log into an online program via their computer or smartphone. At the instructions day (i) subjects choose a window of 4 hours in which they can log into their software and a user code prevents multiple participation.
After logging into the software, subjects see the current lottery and its payoffs. Depending on the treatment condition an adjustment to these payoffs can be made by (a) clicking a button (active condition) or (b) taking no action (passive condition). If no action is taken (or subjects do not log into the software) in the active condition the current lottery is selected at the end of the decision window. In the passive condition, subjects can halt further adjustment steps by clicking on a button during an active decision window. If subjects do not log into the software or do not press the button to halt further adjustment steps another adjustment step is taken.
To disentangle the mode of choice from a status-quo effect, we furthermore varied whether subjects start from an initially safe lottery, such that any adjustment of the lottery implies more risk-taking, or from an initially risky lottery, such that adjustments lead to less risk-taking.
The size of an adjustment step is unknown to subjects. They are made aware that payoff is getting less (more) similar in the safe (risky) condition and that the lottery on the last day will be maximally risky (safe) in the safe (risky) condition. They are also told the highest and lowest possible adjustment step in each condition. This ensures that subjects do have an incentive to log into the software to find out the payoffs of the current lottery.
In the safe treatment, both lottery payoffs (green and yellow) of the initial lottery payoffs are set to 6. In the risky treatment, one payoff is 0 the other is at 15. Adjustment step sizes are structured in a way that ensures that the safe (risky) starting point lotteries always reach 0/15 (6/6) Euro after 10 adjustment steps. That is, there is a predetermined

set of possible step sizes. These range from 0.5 to 0.75 €. The order of implementation of these step sizes is randomly determined for each participant.

The questionnaires serve two purposes. On the one hand, in the initial questionnaire, we elicit some basic demographic information (gender, age, math grade, subject of studies, available income, smoking behavior, dental checkups). On the other hand, we use the final questionnaire as a measure for sample attrition. Subjects who do partake in the final questionnaire can be assumed to have not selected out of the experiment at an earlier stage. In the final questionnaire, we elicit additional demographic information and information regarding decisions in the study (SOEP Measure of risk aversion, aim to earn as much as possible in the study, the difficulty of remembering to take a decision, remember/forget to log in, number of participations in previous studies, issues with understanding the instructions, purpose of study)

Design details

Randomization method
Subjects are randomly assigned to one of four treatment conditions that determine if they (i) make choices by action or inaction and (ii) start with a safe or risky lottery. Randomization of starting lotteries is done by the computer on the individual level. Randomization of active and passive treatments is done on a session by session basis. However, participants do not know which treatment a session belongs to when signing up for the experiment.
Randomization unit
Randomization of starting lotteries is done on the individual level. Randomization of active and passive treatments is done on the session level.
Clusters
200 Participants
Observations
200 Participants
Arms
50 per treatment.
Primary outcomes
The primary variables of interests are the coefficients of relative risk aversion that we derive from the lotteries participants select to play in the experiment.

Cited by

No citation data available for this registration yet.

Keywords