Nudges and the Malleability of Identity: Evidence from Religiosity 2026

After collecting information on participants' demographic characteristics and beliefs, we randomly assigned them to view a brief text message accompanied by an image. The text promotes prosocial behavior and reflection on relationships with others. Different versions of the message use different types of language (e.g., referencing religious or scientific perspectives), while conveying a similar underlying theme. A control group proceeds directly to the survey without seeing any message.

All participants then complete a set of questions about their attitudes and beliefs. The intervention is brief, non-coercive, and involves no deception.

2026-04-22

2026-12-31

The study includes two primary outcomes capturing complementary dimensions of religiosity:

Importance of God in life
Continuous variable measured on a 0–10 scale (0 = Not at all important; 10 = Very important).
Desire to pray often
Binary indicator equal to 1 if the respondent reports praying at least once per day, and 0 otherwise.
Derived from an ordinal question on desired prayer frequency.

We measure religiosity using two preregistered outcomes reflecting both religious self-identification and intended religious practice.

Importance of God in life is measured using the question:

“How important is God in your life?”
Responses are recorded on a 0–10 scale, where higher values indicate greater perceived importance of God. This serves as our main continuous measure of self-reported religiosity.

Desire to pray often is based on the question:

“Ideally, how often would you like to pray?”

Response options are:

Never / almost never
Rarely
Once a year
Only on special religious holidays
Only when attending religious services
Several times a week
Once a day
Several times a day

From this item, we construct a binary indicator equal to 1 if respondents select “Once a day” or “Several times a day,” and 0 otherwise. This measure captures intensive religious practice at the upper end of the distribution.

We collect a set of secondary (non-confirmatory) outcomes capturing religious beliefs, attitudes, behaviors, and broader social and behavioral preferences:

Belief measures (binary indicators)
Belief in: God, afterlife, hell, and heaven (yes/no).
Religious attitudes (ordinal Likert variables)
Agreement with statements on religion vs. science and exclusivity of religion.
Religious behavior (ordinal variable)
Frequency of church attendance.
Moral attitudes (continuous variables)
Perceived justification (0–10 scale) of: abortion, premarital sex, and homosexuality.
Immigration attitudes (ordinal variable)
Self-reported attitude toward immigrants.
Same-sex marriage (ordinal variable)
Position on legalizing same-sex marriage.
Behavioral preferences (continuous variables)
Willingness to take revenge, willingness to forgive unlawful acts, and willingness to punish unfair behavior (toward self and others).
Political ideology (categorical variable)
Self-placement on a left–right political spectrum.
COVID-19 vaccination status (binary variable)
Vaccinated vs. not vaccinated.
Self-comparison tendencies (continuous variable)
Degree of comparison with others in decision-making.
Willingness to sacrifice for the future (continuous variable)

The secondary outcomes provide a broader set of measures to explore potential spillovers of the intervention beyond core religiosity.

Belief measures are elicited using the question:

“Which of the following do you believe in?”
for the items God, afterlife, hell, and heaven. Each is coded as a binary indicator.

Religious attitudes are measured using agreement with the statements:

“Whenever science and religion conflict, religion is always right.”
“The only acceptable religion is my religion.”
Responses follow a four-point Likert scale from Strongly agree to Strongly disagree.

Religious behavior is measured using the question:

“Apart from weddings and funerals, how often do you attend church nowadays?”
with ordinal response categories ranging from Never / almost never to More than once a week.

Moral attitudes are measured using the question:

“For each of the following actions, indicate whether you think it is never justified, always justified, or somewhere in between.”
Responses are recorded on a 0–10 scale for abortion, premarital sex, and homosexuality.

Immigration attitudes are measured using the question:

“What is your attitude toward immigrants in our country?”
with responses on a five-point Likert scale from Very positive to Very negative.

Same-sex marriage attitudes are measured using the question:

“What is your position on legalizing same-sex marriage?”
with responses ranging from Strongly support to Strongly oppose.

Behavioral preferences are measured using 0–10 scales based on the following questions:

“How willing are you to take revenge if you are treated very unfairly?”
“How willing are you to forgive an unlawful act?”
“How willing are you to punish a person for unfair behavior towards you?”
“How willing are you to punish a person for unfair behavior towards someone else?”

Political ideology is measured using the question:

“Which political ideology best represents you?”
with categorical responses from Left to Right, including No ideology / not interested in politics.

COVID-19 vaccination status is measured using the question:

“Were you vaccinated against COVID–19?”
and coded as a binary indicator.

Self-comparison tendencies are measured using the question:

“In your choices, how much do you compare yourself to others?”
on a 0–10 scale.

Willingness to sacrifice for the future is measured using the question:

“Compared with others, how willing are you to sacrifice today to benefit in the future?”
on a 0–10 scale.

Design:
Parallel-arm, between-subjects randomized controlled trial conducted online. Participants are randomly assigned (1:1:1) to one of three arms at the start of the survey:
(A) a brief prosocial message framed in religious language;
(B) a closely matched prosocial message framed in scientific/psychology language; or
(C) no message (control).

Messages are similar in length and readability; only the framing differs. No deception is used.

Eligibility & setting:
Adults (18+) completing a single session on a standard online survey platform. Participation is anonymous and voluntary; respondents may skip questions.

Timing & flow:
After providing consent, participants are randomly assigned to one of the three conditions. The assigned message (or no message) is displayed on its own screen. Immediately afterward, participants complete the outcome measures within the same session.

Primary outcomes (confirmatory):
Self-reported religiosity measured immediately after the intervention using two outcomes:

Importance of God in life (0–10 scale; 0 = not at all important, 10 = very important);
Desire to pray often, constructed as a binary indicator equal to one if respondents report praying at least once per day.

Secondary outcomes (public summary):
A broader set of attitudinal and behavioral measures, including:

religious beliefs (e.g., belief in God, afterlife);
religious attitudes (e.g., views on religion vs. science);
religious behavior (e.g., church attendance);
moral and social attitudes (e.g., abortion, homosexuality, immigration, same-sex marriage);
behavioral preferences (e.g., revenge, punishment, forgiveness);
political ideology and vaccination status;
additional preference measures (e.g., self-comparison, willingness to sacrifice for the future).

Randomization & masking:
Treatment assignment is implemented automatically by the survey platform’s randomizer. Participants are not informed about the existence or content of alternative messages.

Analysis (high-level):
We estimate intent-to-treat effects by comparing average outcomes across groups using two-sided tests at α = 0.05. The primary confirmatory contrast is the scientific framing relative to the control group. Secondary contrasts include religious vs. control and scientific vs. religious framing. Secondary outcomes are analyzed as exploratory with appropriate consideration of multiple testing.

Exposure & quality controls (summary):
A minimum on-screen exposure time is imposed for the message. A simple attention check is included to encourage data quality. Any exclusion criteria are prespecified elsewhere in the preregistration.

Not available

Treatment assignment is implemented automatically by the survey platform’s randomizer.

The unit of randomization is the individual participant. Each participant is independently assigned by the survey platform to one of the three treatment arms (religious framing, scientific framing, or control). There is no clustering or multi-level randomization.

No

1,000 individuals.

1,000 individuals.

Approximately 333 individuals in the control group, 333 individuals in the religious framing group, and 334 individuals in the scientific framing group.

Assuming individual-level randomization, no clustering, equal allocation across the three arms, two-sided tests at the 5% significance level, and 80% power, the minimum detectable effect size for the main confirmatory comparison (Scientific framing vs. Control) is approximately 0.22 standard deviations for continuous outcomes. For a binary outcome, assuming a baseline proportion around 0.50, this corresponds to a minimum detectable difference of approximately 10.8 percentage points. Because the design is not clustered, the standard deviation is defined at the individual level, and the detectable effect is approximately 22% of one standard deviation for continuous main outcomes.