Social preference, voluntary donation to climate mitigation and framing

Last registered on October 19, 2022

Pre-Trial

Trial Information

General Information

Title
Social preference, voluntary donation to climate mitigation and framing
RCT ID
AEARCTR-0009364
Initial registration date
July 22, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
July 26, 2022, 1:47 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
October 19, 2022, 5:10 PM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Region

Primary Investigator

Affiliation
Oregon State University

Other Primary Investigator(s)

Additional Trial Information

Status
In development
Start date
2022-05-23
End date
2022-11-30
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
According to the most recent ICCP report, we have to immediately and rapidly reach net zero carbon emissions to avoid many of the worst effects. Determining how to motivate the public to pursue carbon mitigation will therefore be essential for policy makers. This project studies the willingness to pay (WTP), for climate change mitigation in an experimental setting. In particular this project is interested if the way information about climate change is presented to people (framing), and people’s preferences on the distribution of income (social preferences), impact their WTP. Further, this project also studies if an individual’s social preferences changes the impact of framing.

The framing used in this experiment mirrors one of the real world characteristics of climate change. Developed nations pollute the most per capita, but developing nations are the most exposed to climate change. Thus, some participants will receive framing emphasizing the risks from climate change to themselves directly, whereas others will receive language emphasizing the risk to those in developing nations. It is hypothesized that the former framing may be more motivating to self-regarding individuals, whereas individuals who are more strongly concerned with other’s welfare will be more motivated by the latter framing. Results from this experiment will become part of the literature seeking to inform active and ongoing policy discussions. For example, this project’s findings will be of use to policy makers trying to word a referendum. Environmental activists engaged in outreach such as deep conversation may also find the results helpful in tailoring their messages. Additionally, this project will help serve as a rare experimental check on what are typically hypothetical prices for carbon mitigation in WTP surveys.
External Link(s)

Registration Citation

Citation
Thompson, Charlie. 2022. "Social preference, voluntary donation to climate mitigation and framing." AEA RCT Registry. October 19. https://doi.org/10.1257/rct.9364-2.0
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
This experiment will have two main components. It will measure individual's social preferences (via a dictator game). It will also present participants with an option to contribute towards carbon mitigation. Just prior to the option to contribute, present different framing languages describing climate change. The effects of social preferences, framing, and interactions will be studied.
Intervention Start Date
2022-05-23
Intervention End Date
2022-11-30

Primary Outcomes

Primary Outcomes (end points)
Contributions towards purchase of cap and trade permits, both in absolute and percentage terms.
Primary Outcomes (explanation)
The effect on contributions with respect to both treatment and underlying individual and demographic
characteristics are examined with the following hypotheses:

Hypothesis 1: Only those who believe in climate change will donate.
As shown in Berger (2021), those who do not believe in climate change donate zero to mitigation. This is because, regardless of their social preferences, they consider the public good to be a waste of money. Belief in climate change will be collected as a secondary outcome (below)

Hypothesis 2: On average more altruistic types will donate more.
As those who do not believe in climate will donate zero, the difference in the averages will be driven by variation among those who believe in climate change between social preference types. As discussed above, the altruistic types may feel compelled to donate both from selfish framing and from altruistic framing.

Hypothesis 3: either framing will increase donations over the neutral language.
The neutral language does suggest an increased risk from climate events such as weather, however it intentionally does not specify a party who will bear the costs of those changes. In such situations individuals often face “compassion fade” avoiding taking any action rather than contributing towards a large abstract problem (Markowitz et al., 2013). By specifying a harmed party, either the self or another person, this may reduce abstraction and therefore the effects of compassion fade.

Hypothesis 4: On average, altruistic types will respond to altruistic language (relative to neutral
language).
Those who are of a more altruistic type may experience a stronger sense of warm glow, from contributing in a way that provides benefits to someone less well off. As such, highlighting this aspect of climate mitigation will prompt them to donate more.

Hypothesis 5: On average, selfish types will donate more only when faced with selfish language
(relative to neutral language).
Those who are more self-focused will likewise be driven to donate more when the damages that they will suffer will be put into context for their own lives. They will likely be more sensitive to language that impacts them directly.

Secondary Outcomes

Secondary Outcomes (end points)
Subject specific covariates:
Demographic variables
(a) gender
(b) age
(c) highest educational attainment
(d) income (bucket)
(e) if they have children
(f) zipcode
of high school (for connection to census block information as a proxy for socioeconomic
background)
(g) Three question likert scale questionnaire for belief in climate change

Task order (dictator game or contributions task) will be varied at the session level.
Secondary Outcomes (explanation)
All of the demographic characteristics have been found to impact WTP for climate mitigation. The
following are the expected relationships to WTP:
1. gender: WTP (for public goods in general) is lower among males
2. age: WTP is inversely related to age
3. education: WTP increases with education
4. income: WTP increases with income
5. children: WTP is higher among those with children (stronger bequest value)
The socioeconomic background may influence people’s development of social preferences. For example, participants who came of age in a recession may demonstrate higher inequality aversion. Based on previous research, WTP will be lower among people who do not believe in climate change
The impact of order of tasks is unknown. It will either have no effect, or the person may act more selfishly in the second task, as they feel they have moral licenses to act selfishly after. In this case they may either contribution less (if the contribution task is second) or they may act less inequality averse (if the dictator game is second). Alternatively, even though dictator game funds cannot be used towards contributions, they may create a higher endowment effect resulting in higher contributions if the dictator game is first.

Experimental Design

Experimental Design
Participants will be asked to do two tasks, with the order varied between which task is presented first. In a laboratory setting this project will elicit social preferences using a modified dictator game. It will also elicit an estimate of WTP for climate mitigation though offering participants a chance to purchase cap-and-trade permits. When eliciting these WTP estimates, there is a control language framing, and two treatment language farmings. One framing focuses on the effects of climate change in Oregon, the other focuses on its impacts on developing nations. To account for free riding, participants will be asked to guess the average contribution of their group, with the expectation that they will project a value close to their WTP as being the norm. This guess will be incentivized with money if they are right. Finally, demographic information which has been shown to impact WTP previously such as education, age, income, and parental status, will be collected. To reduce cognitive load, the whole experiment will be designed to take no more than twenty minutes, though an hour will be allocated per session to give participants time to consider their decisions, read instructions, and carry out tasks.
Experimental Design Details

Contributions towards a public good:
The main outcome of interest is the contribution towards climate mitigation. Building on Goeschl et al. (2020), the participants will be given money ($15), and the opportunity to contribute to to the purchase of CO2 cap and trade permits to be stored, thus reducing the amount of CO2 in the atmosphere. Money for distribution to CO2 mitigation will only be received after a set of short cognitive tasks to “earn” money. This is to counteract any gift effects (List, 2007). After receiving a “gift” from the experimenter, participants may feel compelled to reciprocate by helping the experimenter find positive results (Zizzo, 2010).

Contribution will be measured, both in absolute terms, and as a percentage of the total endowed money. The use of absolute contributions is typical in the handful of similar experiments which have been conducted(Diederich and Goeschl, 2014; Goeschl et al., 2020; Paetzel et al., 2018; S. Berger and Wyss, 2021). However, Laury and Taylor (2008) found that social preferences were only significantly related to percentage of endowment contributed, not the absolute amount contributed. To be thorough, both possibilities will be examined.

To help participants understand how their contribution will result in reduced carbon being emitted, the function of a cap-and-trade market will be briefly explained, as will the process of retiring permits. The amount of carbon mitigated per dollar will be related to common real world activities, such as transportation or energy use. For example, the price of one ton of CO2 on the California market is around $18. A metric ton of carbon is produced by driving 2,500 miles on average. Therefore the participants would be told “Each dollar contributed is equivalent to driving 140 fewer miles,” equivalent a reduction of 5.5% of a ton of CO2. At this time some of the participants will be provided with one of the two additional language framing treatments, selfish or altruistic. Following recommendations in List et al. (2011), since the neutral language is used as a comparison group for each of the two framing groups, it will receive 50% of the trials, with the remaining 50% of trials being split evenly between the two other framing treatments.

Modified Dictator Game:

A modified dictator game will be used to determine social preferences. Following Kerschbamer (2015), the participants will be told that they will be choosing how to distribute money between themselves and another participant. This method resembles the framework of risk aversion, in that participants are given a set of choices between an equal payoff and a unequal payoff. The payoff structure includes both advantageous and disadvantageous inequality. This method was chosen because it requires relatively few questions in this case only ten binary choices reducing decision fatigue. At the end of the experiment half of the participants will be selected at random to be dictators and half to be recipients. 8 Only one of the ten trials will be paid, also selected at random. This method of paying one of a number of trials is common in the behavioral literature, and is not likely to change participant behavior. (Charness et al., 2016).

As part of the analysis individuals will need to be sorted into social preference types. Kerschbamer (2015) provides a mechanical way to do this based on recipient responses. This will be my primary method of sorting. However, as a robustness check, this project will also engage in latent profile analysis, which seeks to probabilistic sort participants into groups (Bruhin et al., 2019). The similarity or difference with regards to sorting participants between these two methods may be examined in another paper.
Randomization Method
Randomization of participant roles (dictator or recipient) will be done by computer during the session. The randomization of treatment types (language and order of tasks) will be done by computer before the first trial.
Randomization Unit
Language treatment is to be conducted at the session level.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
12 Sessions, each with approximately 10 participants. Some may have 12 participants. More session may be needed depending on enrollment per session. While tasks and decisions are individual and so there should be no session effects, this clustering is to control for any unobserved differences that might arise between session. It is possible that maximum enrollment for each session may not be possible due to difficulty recruiting or no-shows. In that case additional trials will need to be run until the target number of participants are reached, and the number of clusters will increase accordingly.
Sample size: planned number of observations
244
Sample size (or number of clusters) by treatment arms
There are two language framing, a control framing, and randomized order of contributions or mitigation tasks. This results in 6 variations of the experiment.
At the start of data collection, each planned session will be allocated to one treatment type. However it is the underlying distribution of the of participants which matters. In the event that more sessions are needed (due to under-recruitment) new sessions will be assigned variations based on the participants still needed for a variation. If multiple types of participants are still needed assignment of the variation for a session will be randomly determined by computer. The following is a breakdown of the planned 24 sessions:

3 sessions (30 participants) will receive the dictator game first and the language focused on Oregon.
3 sessions (30 participants) will receive the contribution game first and the language focused on Oregon.
6 Sessions (60 participants) will receive the dictator game first and the control language.
6 Sessions (60 participants) will receive the contribution game first and the control language.
3 sessions (30 participants) will receive the dictator game first and the language focused on developing nations.
3 sessions (30 participants) will receive the contribution game first and the language focused on developing nations.

The spread of these treatments is designed such that there are the same number of control language treatments as the other language treatments combined (List et al 2011). In the event that there needs to be more trials due to under-enrollment during the 24 planned sessions, sessions will be added until each variation above has at least the correct number of participants. The minimum of 2 participants per session may result in some slight over allocation to a given variation.

Given that sessions were of different sizes, session fixed effects will be used in the analysis.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
To determine the necessary sample size, this project relies on the effect size from Paetzel et al. (2018). The authors tested for the effect of social nudges on WTP for climate mitigation in Europe. The SD of the WTP in the control group was 2.66. A single treatment arm with an effect size of $1, a significance level of 0.05, and a power of 0.8 requires a sample size of 56. If using the higher standard deviation from the treated group of 2.98, this results in a group size of 70. As the order of the dictator game and the contribution task will vary, that will double my sample size to between 112 and 140. There are three framing languages, resulting in interaction terms that increase the necessary number of participants. While the number of social preference groups present is not known in advance, past research suggests three to four types is a reasonable number. That number of types means a total of nine to twelve interaction terms. If the significance level is adjusted accordingly, to achieve full power, a sample of between 188 (control SD, nine interactions) and 244 (treated SD, 12 interactions) participants is required. To err on the side of caution the highest number of participants (244) was selected.
IRB

Institutional Review Boards (IRBs)

IRB Name
Human Research Protection Program and Institutional Review Board
IRB Approval Date
2022-05-02
IRB Approval Number
IRB-2022-1404

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials