The Impact of Ask String on Charitable Giving: An Online Experiment in Japan

We will conduct an experiment using oTree to measure the impact of ask string on charitable giving. In the experiment, we will intervene by using a screen where participants decide how much to donate to an NGO engaged in environmental conservation activities in Japan. We will randomly assign participants to one of three groups.

We will conduct an experiment using oTree to measure the impact of ask string on charitable giving. In the experiment, we will intervene by using a screen where participants decide how much to donate to an NGO engaged in environmental conservation activities in Japan. We will randomly assign participants to one of three groups.

- Control Group: Participants assigned to this group will not see preset donation options; only an open entry field will be displayed.
- Lower Default Amounts Group: Participants assigned to this group will see three options (10 points, 20 points, 30 points) displayed above an open entry field.
- Higher Default Amounts Group: Participants assigned to this group will see three options (20 points, 30 points, 40 points) displayed above an open entry field.

Participants begin with an initial endowment of 50 points. In all groups, they can donate any amount from 0 to 50 points. In the intervention groups, participants can also select one of the displayed options to make their donation. If a participant enters a value outside the permissible range, a warning message will appear, preventing them from proceeding to the next screen. If they wish not to donate, they must enter 0 points; leaving the entry field blank will also prevent progression to the next screen.

2025-03-12

2025-03-13

1. Whether to donate
2. Unconditional Average Donation Amount

1. Whether or not to donate at least one point.
2. The average donation amount, including those who did not donate

1. Conditional Average Donation Amount
2. Total Donation Amount
3. Expected Average Donation Amount by Other Participants
4. Expected Donation Amount Desired by the Charity

1. The average donation amount among those who donated at least one point
2. The total sum of donations collected in each group
3. Estimated average points donated by Other participants
4. The amount guessed as what the charity hoped participants would donate

1. Experimental Setup
This study aims to examine the impact of a list of default amounts on charitable giving in Japan, using a donation experiment structured as a dictator game. The experiment is conducted on oTree. Participants are residents of Japan aged 18 or older and are recruited through Yahoo! Crowdsourcing, a Japanese crowdsourcing service. Participants receive two types of compensation: a fixed reward and a performance-based reward. A fixed reward is guaranteed for all participants upon participation in the experiment and completion of the survey, while the performance-based reward is determined by their performance in the donation experiment. The maximum amount for the performance-based reward is 50 JPY. The compensation is provided in the form of points usable through PayPay, a Japanese payment service, instead of cash.

2. Procedure
The experiment will proceed as follows:
1. Obtain Informed Consent
2. Pre-experiment Survey
3. Experiment Instructions & Comprehension Test
4. Donation Experiment
A donation experiment structured as a dictator game will be conducted. Participants are initially given 50 points and instructed to enter how much to donate to the target charity.
5. Post-experiment Survey
6. Compensation Payment
Additionally, to identify participants who may answer the questionnaire without closely reading the instructions, the Directed Questions Scale (DQS) will be included twice in the pre-experiment survey. A warning will be displayed only if the first DQS answer is incorrect. If the second DQS answer is incorrect, the participant will not receive a warning, and they will be excluded from the analysis.

3. Intervention
We will conduct an experiment using oTree to measure the impact of ask string on charitable giving. In the experiment, we will intervene by using a screen where participants decide how much to donate to an NGO engaged in environmental conservation activities in Japan. We will randomly assign participants to one of three groups.

4. Hypotheses
This section will be disclosed after the experiment.
5. Analysis
This section will be disclosed after the experiment.

1. Experimental Setup
This study aims to examine the impact of a list of default amounts on charitable giving in Japan, using a donation experiment structured as a dictator game. The experiment is conducted on oTree. Participants are residents of Japan aged 18 or older and are recruited through Yahoo! Crowdsourcing, a Japanese crowdsourcing service. Participants receive two types of compensation: a fixed reward and a performance-based reward. A fixed reward is guaranteed for all participants upon participation in the experiment and completion of the survey, while the performance-based reward is determined by their performance in the donation experiment. The maximum amount for the performance-based reward is 50 JPY. The compensation is provided in the form of points usable through PayPay, a Japanese payment service, instead of cash.

2. Procedure
The experiment will proceed as follows:
1. Obtain Informed Consent
2. Pre-experiment Survey
・ Socioeconomic demographics: Age, Gender, etc.
・ Environmental Awareness
・ Impression for Charitable Organizations
・ Past Donation Experience
・ Awareness and Donation Experience with the Target Charity
・ General Trust
・ Cognitive Reflection Test (CRT)
・ Conformity Scale
3. Experiment Instructions & Comprehension Test
Participants will receive an explanation about the donation experiment and the target organizations they can donate to. After the instructions, we will give a five-question comprehension test about the experiment. If a participant answers even one question incorrectly, they will be redirected back to the instruction screen and they must retake the comprehension test. If they fail the comprehension test three times consecutively, their participation will end at that point, and only a fixed compensation will be provided.
4. Donation Experiment
A donation experiment structured as a dictator game will be conducted. Participants are initially given 50 points and instructed to enter how much to donate to the target charity.
5. Post-experiment Survey
・ Simple questions regarding understanding of the experiment
・ Norms regarding donation amounts
・ Evaluation of the default
6. Compensation Payment
Additionally, to identify participants who may answer the questionnaire without closely reading the instructions, the DQS will be included twice in the pre-experiment survey. A warning will be displayed only if the first DQS answer is incorrect. If the second DQS answers is incorrect, the participant will not receive a warning, and they will be excluded from the analysis.

3. Intervention
We will conduct an experiment using oTree to measure the impact of ask string on charitable giving. In the experiment, we will intervene by using a screen where participants decide how much to donate to an NGO engaged in environmental conservation activities in Japan. We will randomly assign participants to one of three groups.

- Control Group: Participants assigned to this group will not see preset donation options; only an open entry field will be displayed.
- Lower Default Amounts Group: Participants assigned to this group will see three options (10 points, 20 points, 30 points) displayed above an open entry field.
- Higher Default Amounts Group: Participants assigned to this group will see three options (20 points, 30 points, 40 points) displayed above an open entry field.

Participants begin with an initial endowment of 50 points. In all groups, they can donate any amount from 0 to 50 points. In the intervention groups, participants can also select one of the displayed options to make their donation. If a participant enters a value outside the permissible range, a warning message will appear, preventing them from proceeding to the next screen. If they wish not to donate, they must enter 0 points; leaving the entry field blank will also prevent progression to the next screen.

4. Hypotheses
This study aims to analyze the impact of a list of default amounts on charitable giving in Japan by testing the following hypotheses:

H1: Compared to other experimental groups, the donation rate will increase in the Lower Default Amounts group.
H2: Compared to other experimental groups, the unconditional average donation will increase in the Higher Default Amounts group.

In relation to these hypotheses, an exploratory analysis will investigate whether the amounts within the ask string were more frequently selected. In addition to testing these hypotheses, we plan to conduct exploratory analyses using data obtained from pre- and post-experiment surveys. Specifically, these analyses will aim to elucidate the mechanisms through which default settings influence donation behavior and examine heterogeneity in these effects. Furthermore, we will employ machine learning methods to estimate the Conditional Average Treatment Effect (CATE) and test optimal policies to maximize the impact of the ask string.
The detailed hypotheses and their theoretical background are outlined in the attached file, "Theoretical Background”.　

5. Analysis
In this study, we will conduct analyses using statistical hypothesis testing and generalized linear models, including Ordinary Least Squares (OLS). We will employ statistical hypothesis testing to validate the proposed hypotheses and provide the result of regression analysis as supportive evidence.

1) Statistical Hypothesis Testing
We will conduct tests to identify differences in population proportions regarding donation rates (whether to donate). For an unconditional average donation amount, a conditional average donation amount, and a total donation amount, we will use tests for differences in population means. Based on these results, we will assess the validity of the hypotheses. We will apply Holm's method to adjust for multiple comparisons in these tests.

2) Regression Analysis
To account for potential imbalances in covariates between groups, we will conduct regression analysis to support the test results. OLS will serve as the baseline method. For binary outcomes (e.g., whether to donate), we will use logit or probit model. For discrete outcomes (e.g., an unconditional average donation amount), we will employ regression techniques suitable for the outcome distribution, such as the Tobit model. Additionally, regression analysis will be used to explore mechanisms and analyze heterogeneity of the effect.

Furthermore, as part of our exploratory analysis, we plan to estimate the CATE using machine learning methods to validate optimal policies. Specifically, we plan to use the Generalized Random Forest.
The details are documented in the attached file "Pre-Analysis Plan,"

Randomization is done by oTree

Individual

No

N/A

2,000 individuals

Approximately 666 participants will be included in each group

Based on the results of Study 4 by Moon and VanEpps (2023), we calculated the effect sizes for donation rates and average donation amounts concerning the two effects of ask string settings and differences in set amounts. For donation rates, the effect size of setting a list of default amounts was Cohen’s h = 0.25, while no effect was observed for differences in option amounts. Regarding average donation amounts, the effect size of setting ask string was Cohen’s d = 0.387, and the effect size for differences in option amounts was Cohen’s d = 0.26. Using these calculations, we designed the sample size for this study by assuming an effect size of 0.25 to simplify calculations. We set the significance level α to 0.05 and the power to 0.80. The primary outcomes of this study are the donation rate and unconditional average donation amount. To test the hypotheses, we need to make comparisons across all three groups for both outcomes, resulting in a total of six tests. While Holm's correction will be used for multiple testing in the experiment, the Bonferroni correction, the most stringent criterion, was employed in sample size design. Under these conditions, the calculated sample size required for each group is approximately 390. To allow for exclusions due to inappropriate responses and to conduct heterogeneity analyses, we set a target sample size of 2,000 for the entire experiment. Reference： Moon, A., & VanEpps, E. M. (2023). Giving suggestions: Using quantity requests to increase donations. Journal of Consumer Research, 50(1), 190–210. https://doi.org/10.1093/jcr/ucac047