Warm Glow, Scope and Parochialism: A Natural Field Experiment After the Climate Law Referendum

The intervention is a natural field experiment embedded in the environmental organization’s first fundraising campaign following the approval of a new climate law through a climate referendum. A total of approximately 3,367 individuals from the organization’s contact list, specifically potential and past donors, receive one of four versions of a donation request. The experiment follows a two by two design.

The first treatment dimension varies donation visibility. In the anonymous condition, donations are described as fully private. In the public recognition condition, donors are informed that they may choose to have their donation publicly acknowledged on the organization’s supporters page.
The second treatment dimension varies the framing of benefits. Half of the messages emphasize the local benefits of the newly approved climate law, while the other half emphasize its global climate benefits.

Randomization is conducted at the individual level and is stratified by past donation status and past engagement with the organization, ensuring balanced assignment across treatment groups. Each individual receives only one version of the fundraising message through the organization’s standard communication channels.

The intervention is designed to examine how donation visibility and local versus global benefit framing influence charitable giving after the provision of a public good.

The study is implemented as a natural field experiment embedded in an environmental organization’s first fundraising campaign following the approval of a climate law through a referendum. The target population consists of individuals on the organization’s contact list, including both potential donors and past donors. A total of 3,367 individuals are included in the randomized sample.

The experiment follows a two by two factorial design. The first treatment dimension varies donation visibility. In the anonymous condition, donations are described as fully private. In the public recognition condition, individuals are informed that they may choose to have their donation publicly acknowledged on the organization’s supporters page. Public recognition is optional and uptake is left entirely to the donor. Assignment to the public recognition condition therefore represents an offer of visibility rather than enforced disclosure.

The second treatment dimension varies the framing of benefits associated with the recently approved climate law. In the local framing condition, the donation request emphasizes benefits of the law for the jurisdiction in which the referendum took place. In the global framing condition, the donation request emphasizes global climate benefits. Given the characteristics of climate protection, the local and global framings are constructed to describe benefits of a similar nature. In particular, both framings refer to comparable mitigation and co benefit outcomes, such as reductions in carbon emissions and associated health benefits, using closely aligned language. The key difference across conditions is the geographic reference of the beneficiaries rather than the type of benefit emphasized.

Randomization is conducted at the individual level using Stata with a fixed random seed. Assignment is stratified by two pre treatment characteristics: past donation category and past engagement with the organization. Past donation is categorized into three groups: no recorded past donation, low past donation, and high past donation, based on the distribution of recorded donation amounts among past donors. Engagement is categorized into low and high engagement based on the median of a pre treatment engagement measure. Stratification ensures balanced assignment across treatment arms along both dimensions.

Each individual receives exactly one version of the donation request through the organization’s standard communication channels. There is no self selection into treatment assignment, as the subject line and initial receipt of the message are treatment independent, allowing for intention to treat estimation. All primary analyses are conducted based on initial assignment, regardless of whether individuals opt in to public recognition.

The primary analysis focuses on the effect of being offered public recognition on donation behavior. To maximize statistical power, the main visibility effect is estimated by pooling across the local and global framing conditions. Differences between local and global framing, as well as interactions between framing and visibility, are analyzed as secondary outcomes.

In addition to average treatment effects, the study examines heterogeneous effects across pre treatment characteristics. In particular, heterogeneity analyses are conducted with respect to the variables used for stratified randomization, namely past donation category and past engagement with the organization. These analyses explore whether responses to donation visibility and benefit framing differ between individuals with and without prior donation history, as well as between low and high engagement individuals.

The study also explores heterogeneity by geographic location as recorded in the organization’s contact database. Location is not used for randomization but is employed in secondary analyses to examine whether responses to local versus global benefit framing vary with geographic proximity to the jurisdiction in which the referendum took place. These analyses are exploratory in nature and are intended to assess whether place based attachment or distance from the policy context moderates donation behavior.

All heterogeneity analyses are conducted within an intention to treat framework and are interpreted as secondary or exploratory outcomes unless otherwise specified.

2026-01-01

2026-02-28

Donation amount (conditional and unconditional), Donation (dichotomous decision variable), engagement (measured by number of cliccs per donation campaign), de registration request, request for donation to be made visible.

Secondary outcomes include:

Engagement with the donation request, measured by the number of clicks on links contained in the fundraising email during the campaign period.

Uptake of public recognition, measured by an indicator equal to one if an individual assigned to the public recognition condition requests that their donation be publicly acknowledged.

Deregistration from the mailing list, measured by an indicator equal to one if an individual requests removal from the organization’s contact list following receipt of the donation request.

Donation amount conditional on opting into public recognition, analyzed descriptively among donors in the public recognition condition who choose to make their donation visible.

The study is a natural field experiment conducted during an environmental organization’s first fundraising campaign following the approval of a new climate law through a climate referendum. Approximately 3,367 individuals from the organization’s contact list are randomly assigned to one of four treatment conditions in a two by two design. These conditions vary whether donations are described as fully anonymous or whether donors are given the option of having their donation publicly acknowledged, and whether the benefits of the climate law are framed as local or global. The public information contained here reflects only the high level features of the study design.

Experimental Design (Hidden)

This study is a natural field experiment implemented during the environmental organization’s first fundraising campaign after the approval of a new Climate Law through a climate referendum. The Climate Law has already been passed and is legally binding at the time of the intervention. As a result, donations cannot influence the adoption or implementation of the law. This creates a post provision environment in which donations can be interpreted as reflecting warm glow and social image motivations rather than instrumental effects.

The sample consists of approximately 3,367 individuals in the organization’s contact database, including previous donors as well as individuals who have never donated but have interacted with the organization. The unit of randomization is the individual. Each participant receives one electronic message inviting them to donate. The header, layout, and general format of the message are identical across conditions. Only selected sentences in the message body vary across treatments. Treatment assignment is not visible to participants, and individuals do not choose which message they receive.

Treatment design

The experiment follows a two by two factorial design.

Visibility condition (Anonymous vs Public)
In the Anonymous condition, the message explains that any donation will remain fully private and that no public recognition is possible.
In the Public condition, the message explains that donors may choose to have their name displayed on a publicly accessible supporters page related to the Climate Law. This constitutes an encouragement design in which visibility is made available but remains optional.

Benefit framing (Local vs Global)
In the Local framing condition, the message emphasizes local benefits of the Climate Law for the jurisdiction in which the referendum took place, such as improved air quality, health benefits, and other local co benefits.
In the Global framing condition, the message emphasizes global climate benefits, including contributions to emissions reductions and broader climate mitigation.

This yields four treatment cells: Anonymous Local, Anonymous Global, Public Local, and Public Global.

Heterogeneity analyses

Heterogeneous treatment effects are estimated using interaction terms between treatment assignment and pre treatment characteristics. Pre specified heterogeneity analyses focus on characteristics used for stratified randomization, namely prior donation status and prior engagement with the organization. These analyses examine whether responses to donation visibility and benefit framing differ between previous donors and non donors, and between low and high engagement individuals.

Additional heterogeneity analyses explore geographic variation using location information recorded in the organization’s contact database. Location is not used for randomization and these analyses are interpreted as secondary and exploratory, assessing whether responses to local versus global framing vary with geographic proximity to the policy context.

ITT and per protocol analyses

All primary analyses follow an intention to treat framework based on initial treatment assignment. For the visibility dimension, additional per protocol and instrumental variable estimates are reported based on whether donors choose to opt in to public recognition, with the public recognition assignment used as an instrument. These estimates are clearly labeled and interpreted as subject to self selection.

Randomization is conducted at the individual level. Each of the 3,367 individuals in the study sample is assigned to one of four treatment conditions in a two-by-two factorial design. Assignment is implemented in Stata using a reproducible random-number seed to ensure full replicability.

Randomization is stratified on two pre-treatment characteristics recorded in the organization’s administrative data:

Past donation status: whether the individual has ever made a prior donation to the organization;

Past engagement: whether the individual has previously interacted with the organization’s communications (e.g., opened or clicked prior messages, or participated in previous actions).

Stratification on these variables ensures balanced assignment across groups that differ in historical involvement with the organization.

Participants do not choose their treatment assignment, and the assignment is not visible to them at any stage. All individuals receive a fundraising message with an identical header, format, and general structure. Only selected lines in the body of the message vary according to the experimental condition. This design eliminates self-selection into treatment and supports unbiased estimation of intent-to-treat effects.

Within each stratum, individuals are assigned with equal probability to one of the four treatment cells (Anonymous–Local, Anonymous–Global, Public–Local, Public–Global). No cluster-level randomization is used.

individuals

No

3367 individuals

3367 individuals

Approximately equal numbers of individuals are assigned to each of the four treatment arms, with about 840 to 844 individuals per arm, for a total sample size of 3,367 individuals.

The study is individually randomized with no clustering. The planned sample size is 3,367 individuals allocated approximately equally across four arms in a two by two design. The primary contrast pools across the framing dimension and compares assignment to the public recognition condition versus the anonymous condition, yielding approximately half the sample per main arm. A formal minimum detectable effect size is not pre specified because it depends on the baseline level and variance of the main outcomes in this fundraising campaign, in particular the donation rate, the distribution of donation amounts, and the variance of link click behavior.