Individual Decision Making 01

There are three sessions to the experiment. Session 2 is the central experiment and is described first.

Session 2: The objective of this experiment is to test the empathy–altruism hypothesis for charitable giving. The giving is to a children’s hospital’s social and emotional support program. The control group reads a simple three-sentence description of the program. One treatment arm reads a fundraising e-mail. The second treatment arm sees a video showing the program in action.

Session 1: The objective of this experiment is to investigate whether three non-substantive aspects of the experimental instructions affect the measures of altruism and warm-glow preferences. If not, great.
If so, we will use the Session 1 results to decide which of the three versions of the experimental instructions we will use.

Session 3: The objective of this experiment is to obtain additional sample size for the measurement of emotional responses to the same treatments we use in Session 2.

Complete details are described in the Measurement/Analysis plan: “eac-Pre-reg-002-v02a-Measurement-AnalysisPlan-IndvDecMake01.docx”. That document also describes additional objectives that will be achieved using the data generated by the protocol.
All participants see the simple three-sentence description of a hospital-based social and emotional support program for the children. For the control group, that’s all they see. The second group (the e-mail group) sees a fundraising e-mail produced by the hospital that is more emotionally evocative than the three-sentence description. The third group (the video group) sees a one minute 43 second video produced by the hospital describing children who are ill and in hospital.
A pilot study demonstrated that the video raises empathic reaction, and that the e-mail raises both empathic reaction and distress/sadness.
After seeing the messages (control, video, e-mail) the participants make six decisions about splitting an endowment between themselves and the children’s hospital. The decisions vary the endowment ($40, $46) and the amount an organization will give to the hospital ($4, $10, $28, $34). The six-decision part of the protocol is the same as Ottoni-Wilhelm, Vesterlund, and Xie (2017). 2023-08-16: Endowments increased to $52/$63; organization amounts to $4, $15, $28, $39.

The specific objectives for Session 2 are:
Test the empathy–altruism hypothesis: empathic reactions evoke altruistic motivation.

Test the distress/sadness–egoistic hypothesis: distress/sadness reactions evoke egoistic (warm-glow) motivation.

Session 1 participants will all see the fundraising email but, before they make their six decisions, will be randomized into three conditions in which the experimental instructions have them check:

Nothing
The amount they will be paid (endowment – how much they give + $5 show-up fee).
The amount they will be paid and how much will be going to the hospital (how much they give + how much the organization is giving in that decision)

The specific objectives for Session 1 are:
Compare the altruism and warm-glow parameters (α and β) of participants assigned to T0, T1, and T2.
Describe the outcomes—altruism and warm-glow parameters (α and β), altruistic and warm-glow motivation holding generosity constant (τ and γ), generosity (g ̅ ; α + β)—of participants in response to a typical direct mail fundraising message (Email). Compare the outcomes to the strong altruism (weak warm-glow) in the experiment by Ottoni-Wilhelm et al. (2017).
Compare the outcomes of Indiana-44 participants to Indiana-48 perimeter country/IL, KY, MI, OH participants.

Sessions 1 and 2 use participants in the SSRS Opinion Panel and the Ipsos panel---both are probability-based panels. 2023-08-16: \sout{Session 2 also uses participants from nonprobability panels maintained by Prodege and QuestMindShare.} The overall recruitment for Sessions 1 and 2 is managed by SSRS.
Session 3 uses participants from a nonprobability panel recruited by Qualtrics Research Services (QRS), which can be recruited at lower cost (than SSRS participants), to improve first-session precision in IV analyses via TwoSample-IV. This can be done only if the quality of the QRS participants is adequate. The specific objectives for Session 3 are:
(1) Test whether the quality of the QRS participants is similar to that of the SSRS participants from Sessions 1 and 2.
(2) Test whether the first-sessions (Empathic State; Distress-Negative State) of the QRS participants are similar to that of the SSRS participants from Sessions 1 and 2.

2021-09-29

2021-12-31

The primary outcomes are the amounts given to the children’s hospital in the six decisions. From these measures of generosity, altruistic motivation, and egoistic motivation can be constructed. See the Measurement/Analysis.

Longer-term effects: At the end of the experiment each participant will be given the option to begin a longer-term relationship with the hospital. The outcomes will be: (1) subscribe to the hospital’s email newsletter, (2) make a $5 one-time donation that is in addition to the amount the participant may have donated from their experimental endowment, or (3) make a monthly $5 donation.

Complete details are described in the Measurement/Analysis plan: “eac-Pre-reg-002-v02a-Measurement-AnalysisPlan-IndvDecMake01.docx”.

From the amounts given a structural model (a Cobb-Douglas impure altruism model) can be estimated yielding measures of generosity, altruistic motivation, and egoistic motivation. A second measure of generosity is simply the average amount each participant gives across her/his six decisions.

The longer-term outcomes are intended to determine whether the experimental conditions have effects that last beyond the conclusion of the experimental session.

Most of the secondary outcomes were described in AEARCTR-0005478 (“First Session 2: Evaluating Messages”): Empathic State, Distress State, Negative State, Guilt State. Principle of Care State, Inclusion of Other in the Self, Interpersonal Reactivity Index, the Principle of Care disposition, and Attention-Check and Manipulation-Check Questions.

In addition, there will be measures of demographics (age, sex, race/ethnicity, Indiana county or state), questions about the context of the experiment, prior knowledge of the hospital, religiosity, political preference (conservative-liberal), and attention/manipulation-check questions. See the Measurement/Analysis plan for details.

Complete details are described in the Measurement/Analysis plan: “eac-Pre-reg-002-v02a-Measurement-AnalysisPlan-IndvDecMake01.docx”.

Session 2: The experimental design is between-subjects (3 x 1) and in each condition: within-subject (the six decisions).

The participants will be recruited by SSRS (and their partners) and invited to complete a study on “Individual Decision Making”. The inclusion criteria are: (1) resident in the \sout{44} 92 Indiana counties \sout{close to Indianapolis, and not close to another city that has a children’s hospital}, (2) 18 years or older, and (3) an equal number of women and men (to the best of SSRS’s ability to balance gender from the probability panels and the nonprobability panels).

2023-08-16: All participants will complete five real-effort, bibliographic data entry tasks (10 to 15 minutes) to earn their payment (either $52 or $63).

Approximately one-third of the participants will be randomly assigned to each of the three conditions: control, e-mail, and video.

After watching the conditions, all participants fill out the Emotional Response Scale, the Principle of Care state measurement, and the Inclusion of Other in the Self. They then make their six decisions.
Then they answer the Interpersonal Reactivity Index, the Attention-Check question, the Principle of Care dispositional scale, gender, race/ethnicity, the Manipulation-Check questions, the experiment context questions, the questions about familiarity with the children’s hospital, the religion questions, and the political preference question.
At the very end, they decide whether they want a longer-term relationship with the hospital.

Session 1: The experimental design is also between-subjects (3 x 1) and in each condition: within-subject (the six decisions). The participants will be recruited by SSRS and invited to complete a study on “Individual Decision Making”. The inclusion criteria are: (1) a resident of the Indiana-48 perimeter counties, IL, KY, MI, or OH, (2) 18 years or older, and (3) an equal number of women and men (again, to the best of SSRS’s ability to balance gender from the probability panel).

Approximately one-third of the participants will be randomly assigned to each of the three conditions: check nothing, check own payment, and check own payment as well as how much is going to the hospital.

Session 3: The experimental design is also between-subjects (3 x 1), but there is no within-subject variation (participants do not make six donation decisions). The three conditions are the same as in Session 2: control, e-mail, and video.

The participants will be QRS panel members invited to complete a study on “Individual Decision Making”. The inclusion criteria are: (1) resident of IL, KY, MI, or OH, (2) 18 years or older, (3) an equal number of women and men, and (4) an age quota, so that the Qualtrics respondents match (approximately) the US age-sex distribution.

Complete details are described in the Measurement/Analysis plan: “eac-Pre-reg-002-v02a-Measurement-AnalysisPlan-IndvDecMake01.docx”.

Here are the three Session 2 conditions in more detail:

1. Control condition:
RILEY HOSPITAL FOR CHILDREN

Riley Hospital for Children located in Indianapolis, IN is a national leader in providing child- and family-centered care to children.

Child- and family-centered care creates a supportive environment that leads to better health outcomes.

To provide a supportive environment for healing, Riley has a wish list containing books, toys, and craft supplies that a child would love to have.


2. E-mail condition.

Same message as above plus: We would like you to read an email produced by Riley.

The next screen displays the e-mail.


3. Video condition.

Same message as above plus: We would like you to watch a video produced by Riley.

The next screen displays the video.



Here are the three Session 1 conditions in more detail:

0. Check nothing.

1. What is the amount you will receive as payment?
(This is the amount you assign to yourself, plus the $5 we promised you as minimum earnings in this study).
2. Same as T1, plus also:

What is the amount to be spent on the child’s books, toys, craft supplies (emotional support)?

(This is the amount you assign to the child, plus the amount the organization will donate to the child.)

Randomization by computer.

Individual.

No

No clusters.

Session 1: N = 270 SSRS participants (in Indiana-48 perimeter counties and IL, KY, MI, OH). Probability-based sample. We concluded Session 1 with N = 87 after the results first indicated that approximately one-half of the participants made decisions at the top-corner, and our Modification 1 did not affect that result. N = 68 were from a probability-based sample, and N = 19 were from a non-probability-based sample (the N = 19 non-prob sample was a Modification 1 idea to determine whether the top-corner problem was specific to the probability-based sample). Session 2: N = 575 to 600 SSRS, Ipsos, \sout{Prodege, QuestMindShare} participants (in Indiana-44+48=92 counties). N = 600 Probability-based sample (SSRS and Ipsos). \sout{N = 240 nonprobability (Prodege and QuestMindShare).} Session 3: N = 897 Qualtrics participants (in IL, KY, MI, OH).

Session 1: N = \sout{90} Approximately 30 per arm.
Session 2: N = 200
Session 3: N = 300

Emotional response power analysis: “eac-Pre-reg-003-v02a-PowerAnalysis-IndvDecMake01.docx” Our analysis indicated that with N = 180, we are powered to detect differences between conditions of 30 percent of a control group standard deviation in both ESi and DNSi (Ottoni-Wilhelm, 2021). At N = 90 we are powered to detect differences = 42 percent of a control group standard deviation. In our pilot study, Video (N = 74) moved ESi 57 percent of sES (relative to control sES, N = 94), and Email (N = 110) moved DNSi 76 percent of control-group sDNS. α and β power analysis: “- - - Power analysis-IndivDecisionMaking-v01a-2018-10-22.docx” In Session 2 we have budget for N = 600, or N = 200 in each treatment Video, Email, and Paragraph (control). Our power analysis indicates that, with N ≥ 180 in each treatment arm, we will be able to detect differences in α in the range of .06 to .14, and β in the range of .04 to .08 (Ottoni-Wilhelm, 2018). SSRS has told us that N = 600 may not be possible, and that we may end up with less. Our power analysis indicates that, with N = 90 in each treatment arm, we will be able to detect differences in α in the range of .12 to .28, and β in the range of .08 to .16 (Ottoni-Wilhelm, 2018). Session 3 power analysis: “eac-Pre-reg-002-v01a-Measurement-AnalysisPlan-IndvDecMake01.docx” Project 12. Our power analysis indicates that, with N ≈ 300 in each treatment arm, we will be able to detect (testing at .05, with power .80): 1. Substantive-differences—in the first-stage regressions—of about one-half the SSRS effect size (see Project 12, Table 12.2). 2. Quality-differences between QRS and SSRS of about three to 6.5 percentage points (see Project 12, Table 12.3).