Back to History Current Version

Religiosity, supernatural beliefs and economic behavior: Evidence from lab-in-the-field experiments in Bangladesh

Last registered on March 21, 2023

Pre-Trial

Trial Information

General Information

Title
Religiosity, supernatural beliefs and economic behavior: Evidence from a lab-in-the-field experiments in Bangladesh
RCT ID
AEARCTR-0010944
Initial registration date
March 19, 2023

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
March 21, 2023, 4:43 PM EDT

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

Locations

Region

Primary Investigator

Affiliation
Technical University of Munich

Other Primary Investigator(s)

PI Affiliation
Technical University of Munich

Additional Trial Information

Status
On going
Start date
2023-03-10
End date
2023-04-20
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Belief in supernatural beliefs is widespread across countries and pervasive in the daily life of many people. Previous empirical evidence has focused on the relationship of supernatural beliefs and prosocial behavior (e.g., Gershman 2016, 2021; Le Rossignol et al. 2022). Yet, empirical evidence on the relationship between supernatural beliefs, individual and social risk preferences, and honest behavior is missing. These outcomes are fundamental for economic development and many economic decisions such as investment decisions, occupational choices, and health behavior. In this project, we investigate the causal effect of supernatural beliefs on individual risk taking, social risk taking, honest behavior and associated perceived social norms. To do so, we conduct a lab-in-the field experiment in rural Bangladesh. Our experiment consists of a control group and two treatment groups where participants are asked to donate money to agents who are believed to have supernatural powers. In a second project, we investigate reputational concerns when donating money to those agents.
External Link(s)

Registration Citation

Citation
Ahmed, Firoz and Andreas Pondorfer. 2023. "Religiosity, supernatural beliefs and economic behavior: Evidence from a lab-in-the-field experiments in Bangladesh." AEA RCT Registry. March 21. https://doi.org/10.1257/rct.10944-1.0
Experimental Details

Interventions

Intervention(s)

We study the effect of supernatural beliefs on individual risk taking, social risk taking and honest behavior. Participants take part in a lab-in-the field experiment and will be randomly assigned across control and treatment groups. In the treatment groups, participants can voluntary donate an endowment given by the experimenters to local people who are believed to have supernatural powers. We hypothesize that activating beliefs in the existence of supernatural powers through voluntary donations significantly affect behavioural outcomes.
Intervention (Hidden)
Intervention Start Date
2023-03-10
Intervention End Date
2023-04-20

Primary Outcomes

Primary Outcomes (end points)
Risk preferences
Individual risk taking (IR): the amount invested in the risky option (ranges between 0 and 10)
Risk taking in PCR: the amount invested in the risky option (ranges between 0 and 10)
Risk taking in NCR: the amount invested in the risky option (ranges between 0 and 10)

For each of the three variables above, we compute a respective dummy variable which takes on the value of one if people chose the equal split (5), and zero otherwise. We also investigate extreme risk taking (risk aversion) by computing a dummy variable which take on the 1 if people chose 10 (0) in the risky option and zero otherwise.

To compare the effect of supernatural beliefs on social risk taking, we further compute the following variables:
delta_IR_PCR: amount invested in PCR - amount invested in IR
delta_IR_NCR: amount invested in NCR - amount invested in IR
delta_IR_PCR_d: share of equal split in PCR - share of equal split in IR
delta_IR_NCR_d: share of equal split in NCR - share of equal split in IR
The same share is constructed for extreme risk taking (risk aversion)

RAG
The amount allocated to the other person (ranges between 0 and 10).

Religious donation
Donation: Binary variable which takes on the value of one if participants donated to Pir or Gunin, and zero if the money was given back to the university.
Donation to Pir: Binary variable which takes on the value of one if participants donated to Pir, and zero if the money was given back to the university.
Donation to Gunin: Binary variable which takes on the value of one if participants donated to Gunin, and zero if the money was given back to the university.

Social norms
Descriptive norms investment games and RAG: value between 0 and 10
Injunctive norms in investment games: value between 0 and 10
Socially acceptability in RAG: 4-point scale ranging from very socially unacceptable, somewhat socially unacceptable, socially acceptable, very socially acceptable.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Before the experiments start, all participants receive a show-up fee (see Experimental Protocol for details).

Measuring risk preferences and honest behavior
We measure risk preferences by using the elicitation method of Gneezy and Potters (1997). Participants are asked to allocate the amount x between a save option and a risky option (3x). To study social risk taking, we focus on two aspects: positive correlated risk (PCR) and negative correlated risk (NCR). Active participants will be randomly matched with a passive and anonymous player (from another village). In PCR (NCR), the passive player receives the same (the opposite) payment. This leads to the situation where higher risks in the PCR do not influence the equality between the active and passive player (as both receive the same amount), while higher risk taking in NCR is increasing inequality as one player will gain the high amount while the other will gain the low one. We follow a within-subject design: Participants will make three decisions: They always start with the individual decision, followed by PCR and NCR (the order for PCR and NCR will be randomized).

We measure honest behavior by using the Resource Allocation Game (RAG) (Hruschka et al. 2014, Lowes et al. 2017). In this game, participants play in private with 10 tokens, two envelopes and a fair dice with three sides of black color and threes sides of white color. Participants are asked to allocate each token to one of the two envelope. First, they mentally choose one of the envelopes and then roll the die. If one colored side comes up, players are instructed to put the coin into the envelope they mentally chose. If the other color appears, players are instructed to put the token into the opposite envelope from the one they chose. Participants chose between an envelope for themselves and an envelope assigned to a passive and anonymous player (from another village that is adjacent to them).

We follow a within-subject design: Both games are played by all participants. The order of the two games is fixed. People start with the investment game, followed by the RAG. In total, we have 12 different set of protocols (see attachment).

Treatment variation 1: Activating supernatural beliefs
To study the effect of supernatural beliefs on behavioral outcomes, we will implement a control condition (C) and two treatments (T1P and T2G) in a between subject design. In the two treatments, participants can voluntarily donate an endowment (money) given by the experimenters to local agents who are believed to have supernatural powers - the Pir (T1P) or the Gunin (T2G). Before we start with the explanations of the game in the individual sessions, we put the donation in an envelope labelled with the name of local supernatural agents. People will be told that at the end when they have made all their decisions, this donation will be given on behalf of them to the to the local supernatural agent. We also inform them that we ask for their consent about their donation at the end of the games. During the whole time of the session, the envelope will be visible in front of the participant. When the participants have made all their decisions, they will be asked to donate the endowment to the local supernatural agent or give it back to the experimenter (where the money comes from). To collect additional information about donation decisions, participants in the control group (C) will also be asked to donate a given endowment to the local supernatural agent or give it back to the experimenter. However, the donation decision will only be presented at the end of the session.

Treatment variation 2: Public vs private donation decision
To study social image/reputational concerns related to the donations, at the end of their session participants will make the donation decision either in private or in public (i.e., when the research assistant is watching). In each condition of treatment variation 1 (C, T1P, and T2G), participants will be randomly assigned to either the public or private condition. Note that half of the observations in the control group will be asked to donate to Pir, while the other half will be asked to donate money to the Gunin (see also Table 1 below). The public vs private donation decision will be evaluated as a separate treatment variation.

Eliciting social norms
For each behavioral outcome, we elicit social norms. All norm questions will be incentivized (people receive payment for correct guesses). For all decisions we elicit descriptive norms by asking respondents how much they believe other community members invested (risk game) / allocated to the other person (RAG). For injunctive norms, we applied the following elicitation method. For the risk game, we first ask all participants what they think is the amount which should be invested. In a second question, we first inform participants that we’ll know the answers to this first question at the end of day. We subsequently ask them what they think is the amount that most other people report to the first question. Thus, the first question reveals personal normative beliefs, while the second question gives us injunctive norms. For the RAG, we present participants the following scenario: they should assume that out of 10 times, 5 times the color that allocated money to them shows up. Next, we ask them the following five questions: how socially acceptable is it to allocate money 5 (6, 7, 8, 9 10) times to yourself? Social acceptability is measured on a 4-point scale ranging from very socially unacceptable, somewhat socially unacceptable, socially acceptable, very socially acceptable. In a second question, we inform participants that we know the answers to this first question at the end of day. We subsequently ask them what they think what category was chosen by most others in their community. This method is comparable to the elicitation approach of Krupka and Weber (2013).

After the experiments, we collect additional information in a post-experimental questionnaire (see below). We collect information on religious beliefs, religiosity, supernatural beliefs, trust, conformism, confidence, relatedness, envy, zero sum thinking, trust, financial activities and socio-demographic characteristics.

Experimental Design Details
Randomization Method
The randomization of villages will be done by computer. A random shuffle (lottery) will assign participants to treatments.
Randomization Unit
Village level
The randomization of villages will be done by computer. We randomly select 10 Union Parishads (UPs) across three Upazilas (sub-districts) in the area of Khulna and Satkhira. Next, we randomly select two villages per UPs for our study.

Individuals
We will invite people by door-to-door visit. Four enumerators will simultaneously start selecting households from four sides (north, south, west and east). Every third household will be targeted, and an adult member will be asked to participate in the study. We start with recruiting a male member. In the next household a female member will be recruited (and so on). Thus, we aim to have a balanced sample in terms of gender (50 % males, 50 % females). If a male (female) participant denies, enumerators will go to the next nearest household and will invite a male (female) person. As 90 % of the population in Bangladesh are Muslim and Pir and Gunin are more common within Islam, we restrict our sample to Muslims (i.e., Hindus are not selected for the experiments).
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
20 villages
Sample size: planned number of observations
About 720 participants across 20 villages (36 participants per village)
Sample size (or number of clusters) by treatment arms
Control group: 240 participants
T1P: 240 participants
T2G: 240 participants
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
To calculate the minimum detectable effect size, we use data from a pilot (the pilot data will not be used in the analysis). In the pilot, we collected 13 observations for the control group in treatment variation 1 (activating supernatural beliefs). Our power calculation is based on the following variables: mean individual risk (IR: 5.08, SD: 1.04, N=13), mean positive correlated risk (PCR: 5.23, SD: 0.93, N=13), mean negative correlated risk (NCR: 4.62, SD: 2.02, N=13) , mean difference between IR and PCR (delta_IR_PCR: 0.15, SD: 1.46, N=13), mean difference between IR and NCR (delta_IR_NCR: -0.46, SD: 1.81, N=13), and the mean of honesty (mean RAG: 6.38, SD: 1.80, N=13). Below we list the minimum detectable effect size for 240 observations per treatment, alpha 0.05, and a power of 80 (90) percent. We assume that the standard deviation in the control group and treatments (T1P and T2G) are identical. IR: 0.27 (0.31) PCR: 0.23 (0.28) NCR: 0.52 (0.60) delta_IR_PCR: 0.37 (0.43) delta_IR_NCR: 0.46 (0.53) RAG: 0.46 (0.53) Likewise, we compute the minimum detectable effect size for treatment variation 2 (public vs. private donation decision). In the pilot, we collected 22 observations for the private condition (mean fraction of donations: 0.36, SD: 0.49, N=22). Below we list the minimum detectable effect size for 360 observation per treatment, alpha 0.05, and a power of 80 (90) percent. We assume that the standard deviation in the two groups (public vs. private) are identical. Donation: 0.10 (0.12)
Supporting Documents and Materials

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

Request Information
IRB

Institutional Review Boards (IRBs)

IRB Name
Research and Innovation Centre, Khulna University, Khulna-9208, Bangladesh
IRB Approval Date
2022-12-15
IRB Approval Number
KUECC-2022/12/42

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