3 ICRC donation suggestion and cost-info trial: Project description, summary, timing, background

Authors: Deborah Kistler, David Reinstein, and Jan Schmitz

Randomized trial within the ICRC’s April mailing campaign: suggested donation and cost information treatments

Creators note: This is a very sparse file, being edited in the root directory of the Dropbox folder icrc_grids_output_expt_to_be_public_version, linked here. It is being pulled in, mirrored, or used, in several other places, but should only be edited in the place just mentioned.

3.1 Overview

Randomized trial within the ICRC’s April mailing campaign (i.e., RCT, i.e., field experiment)

3.2 Population

Mailing addressed to ICRC’s existing donor base of ‘active, warm and sleepy donors’, in all parts of Switzerland (German, French and Italian)

More details in pre-registration

3.3 Treatments (outlined)

We randomly assign donors into one of seven different treatment conditions

Treatment conditions vary the ‘request to give’– this is text prominently displayed above the pay-in slip (which is referred to as a ‘BVR’. Donations via BVR are the main source of income for the ICRC from these campaigns, i.e., there are very few online donations via the ICRC website)

The appeal/mailing is always the same (identical content in all conditions) other than the information provided on the pay-in-slip (presumably when many donors make their decisions), which varies between treatments

Treatment conditions independently vary:

  1. “Cost-output-info”: the availability of information about cost for a given unit of suggested output and
  2. “Suggested-donations” the suggested donation amount

Note that where the treatments are crossed, giving both cost information and suggested donation amounts, the ‘quantity provided by the suggested donation’ will be indicated.

3.4 Timing

20 April 2021: This campaign is scheduled to starts around the 20th of April 2021 (the postal mailing is intended to go out on this day). Over April 20-23 all the relevant appeals will be delivered by Swiss Post.

3.4.1 May (unrelated) campaigns: is it a confound?

See discussion and response in pre-registration under “7) How many observations” and “Our analysis will cover (contingent plan)”

3.4.2 Data collection/analysis universe

See preregistration

(pre_registration_and_PAP/preregistration.md)


3.5 Treatment specifics (i.e., ‘experimental conditions’)

  1. control = “DONATE TODAY: your donation can supply food parcels to a Syrian family”

  2. cost = “DONATE TODAY: your donation can supply food parcels (ca. 17CHF/parcel for one month) to a Syrian family”

  3. cost_sug_50 = “DONATE 50CHF TODAY: your donation can supply 3 food parcels (ca. 17CHF/parcel for one month) to a Syrian family”

  4. cost_sug_150 = “DONATE 150CHF TODAY: your donation can supply 9 food parcels (ca. 17CHF/parcel for one month) to a Syrian family”

  5. sug_50 = “DONATE 50CHF TODAY: your donation can supply food parcels to a Syrian family”

  6. sug_150 = “DONATE 150CHF TODAY: your donation can supply food parcels to a Syrian family”

  7. icrc_standard = “DONATE TODAY: With 50CHF you offer 4 Hygiene kits to Syrian families; With 100CHF you offer 14 school kits to Syrian students; With 150CHF you offer 9 food parcels to Syrian families.”

Note that Condition 7 is the standard ICRC condition which is kept as a reference for the ICRC and to compare outcomes (not mechanisms) to conditions 1-6

Note: The text above is an English translation.[Note: This is in translation. For the specific treatment text (and context) in the three languages we are using see ICRC-mailing 0221-corrected -TREATMENTS.pdf.]

3.6 Treatment selection and ssignment (‘randomization’) procedure

The ICRC provided us with their donation data from between 2013 and 2020. Based on this data we proposed our Suggested-donations and Cost-output-info treatments and provided suggested amounts for the latter. These amounts correspond to:

  • 50 CHF = 50th percentile

  • 150 CHF = 85th percentile

… based on the ‘average last positive gift’.

3.6.1 Data and blocked randomization

The ICRC also provided us with a list of donors (169,919) to be sent the mailing. We merged the two data sets (historical giving data and list of donors for the experiment) focusing on the past 4 years of giving (see pre_experiment_code_analysis/experiment_list.do).

Using this dataset we randomized potential donors into experimental conditions within blocks defined by (see randomization.do):

  • ‘average donation between 2017 and 2020’ and
  • ‘donor category’, i.e. ‘warm list’ status (‘active’, ‘warm’ or ’sleepy)

3.6.2 Followup with the ICRC

We sent these assignments to the ICRC as an Excel file (randomized_data_for_ICRC) containing the constituent id (donor id), the treatment number (1-7) and an associated grid for assignment in the ICRC database (01-07).

The ICRC integrated the data and did a final check. Some potential donors were removed from the list (e.g., because…)

  • they died,

  • they removed themselves from the ICRC contact list, or

  • they are large legacy donors that the canvasser picked to contact directly, rather than through the mailing.

Based on this final internal ICRC check they re-sent a document with a final list of donors and their internal treatment allocation.

3.6.3 Final adjustments and checks

We checked this for consistency (see file: pre_experiment_code_analysis/consistencycheck) to check whether our randomization was still balanced. We produced a document with mismatch and those who are dropped (see folder: data/checkup data).

We produced a dataset with the final set of potential donors in the experiment population (see data/final in experiment/final list ICRC.dta). We communicated the experimental treatment mismatches to the ICRC. The ICRC adjusted the mismatched donors.

We performed a final power analysis (see file pre_experiment_code_analysis/consistency_power) to check whether “detectable effect sizes” changed in any substantial way.

3.6.4 Data collection/analysis universe

See preregistration

(pre_registration_and_PAP/preregistration.md)


3.7 Treatment specifics (i.e., ‘experimental conditions’)

  1. control = “DONATE TODAY: your donation can supply food parcels to a Syrian family”

  2. cost = “DONATE TODAY: your donation can supply food parcels (ca. 17CHF/parcel for one month) to a Syrian family”

  3. cost_sug_50 = “DONATE 50CHF TODAY: your donation can supply 3 food parcels (ca. 17CHF/parcel for one month) to a Syrian family”

  4. cost_sug_150 = “DONATE 150CHF TODAY: your donation can supply 9 food parcels (ca. 17CHF/parcel for one month) to a Syrian family”

  5. sug_50 = “DONATE 50CHF TODAY: your donation can supply food parcels to a Syrian family”

  6. sug_150 = “DONATE 150CHF TODAY: your donation can supply food parcels to a Syrian family”

  7. icrc_standard = “DONATE TODAY: With 50CHF you offer 4 Hygiene kits to Syrian families; With 100CHF you offer 14 school kits to Syrian students; With 150CHF you offer 9 food parcels to Syrian families.”

Note that Condition 7 is the standard ICRC condition which is kept as a reference for the ICRC and to compare outcomes (not mechanisms) to conditions 1-6

Note: The text above is an English translation. For the specific treatment text (and context) in the three languages we are using see ICRC-mailing 0221-corrected -TREATMENTS.pdf