Banking for Boomers – A Field Experiment on Internet Banking Adoption

We conduct a field experiment with elderly customers of a large German savings bank to study the effect of different education approaches for technology adoption and lifelong learning on internet banking (IB) adoption at the extensive and at the intensive margin. We use traditional information-based approaches and approaches emphasizing non-technical adoption skills in a social learning context. Our target group consists of clients aged 50-85 who have not used internet banking in the recent past.
In total, there are four experimental groups: i) silent control group that we do not contact, ii) baseline treatment group (traditional, technical information-based approach), iii) non-technical treatment group I (coaching at home), iv) non-technical treatment group II (coaching in person). Note that participants never log into their own private internet banking as part of the training for data protection reasons. Instead, we use a demo tool.

In terms of the technical information on internet banking that they provide, all treatment groups use the same curriculum consisting of three modules: The first module covers all key features of the bank’s internet banking platform and the bank’s security architecture. The second module guides participants through a demo version of the bank’s internet banking platform. It illustrates the login process using hypothetical login details and a simulated transaction authentication number (TAN) confirmation and the various functional areas of the platform (financial overview displaying all account balances, transaction history and details of specific accounts, messages, account settings, etc.). A video that walks participants through these modules is made available to all participants in the treatment groups (for in-person participants, this video is made available after the in-person session). In the third module, participants are confronted with five standard internet banking tasks in the demo tool that they are asked to complete.
The two treatment conditions that foster non-technical skills offer non-technical support in addition to the modules outlined above. To strengthen non-technical skills (such as curiosity, confidence, or emotional regulation), they use distinct social settings: Non-technical treatment group I involves relatives or friends supporting the elderly clients in the education intervention. Participants in this group are expected to nominate an IB-savy person from their social circle, such as a relative, friend or neighbor (we will refer to this person as a “relative” from now on for expositional reasons) and to complete a two-hour training session with the relative by a given deadline. We provide the relative with material that guides them in relaying non-technical skills to the treated person. The relative receives a voucher as an incentive. Non-technical treatment group II includes on-site training in small groups of peers with networking opportunities and in a pleasant atmosphere. Participants in this group receive a two-hour training session in the bank's educational center in small groups of no more than 15 peers (since there are up to two groups in parallel, coffee break networking may include up to 30 peers). Our teaching strategies in both non-technical treatment groups revolve around making emotions and required coping mechanisms explicit in the intervention and building on prior knowledge and experiences to boost confidence and curiosity.
We track compliance in all groups by measuring whether and how frequently a client has accessed the learning platform containing the instructional video and material. We track compliance for non-technical group I (relative / friend) by requiring clients in this group to submit a selfie with their friend/relative and answers to questions on the demo tool tasks they completed in the study session to the bank in order to be eligible for receiving the voucher. We track compliance for non-technical group II by documenting attendance at the on-site training.
The effectiveness of the educational approaches is measured by IB activation as well as transactional banking data such as the number and share of online transactions and the date of the most recent internet banking login. Furthermore, we use survey measures to identify mechanisms like obstacles in IB use, openness to technology / technophobia, computer / internet banking skills, perceived benefits of internet banking, and financial literacy, as well as usage intensity.
Note that upon first contact, all three contacted groups received information on how to set up their personal internet banking account upon request by our partner bank. This information was identical for all groups. We do not consider it part of the treatment variation (in the intervention, we rely solely on a demo tool) but will discuss it in comparisons with the silent control group.

2024-09-09

2024-12-02

Internet banking adoption (extensive and intensive margins); treatment take-up, attrition, effects of non-random assignment

Extensive margin: Adoption level 1: Activation (YES/NO) of IB (browser and / or app), adoption level 2 (conditional on activation): login following intervention; any online transactions (YES/NO, conditional on some transaction conducted)
Intensive margin: Number and share of online transactions (conditional on some transaction conducted), days since last login
Treatment take-up: Binary indicators capturing interest in and completion of treatment.

IB beliefs and perception, IB skills, technophobia, non-technical adoption skills (i.e., emotional regulation, coping strategies, ...)
Heterogeneity analysis by gender (see large literature on financial dependence), age (exploratively, to shed more light on the understudied group of elderly in the context of lifelong learning), training completion, and baseline skills.

See PAP for details

Respondents to a call for participation sent out by the bank in July 2024 are randomly assigned (subject to some restrictions outlined in the experimental details below) to one of three treatment conditions (baseline, non-technical treatment I and non-technical treatment II) with a limited number of slots available in the non-technical treatments due to budget constraints.
Those in the baseline treatment group receive training material containing all technical information about the key functions of internet banking. Those in the non-technical groups receive additional material that addresses non-technical skills in addition and that is designed to be worked with at home (non-technical group I) or in an in-person course with peers (non-technical group II). All material is made available through a designated online platform, which also provides video summaries of the material.
We, thus, use a between-subject design. Survey capacities permitting, we plan to include some survey measures before and after the treatment to use within-subject variation in the analysis of mechanisms.

We conduct a three-stage assignment procedure. First, the partner bank provides us with data on more than 153,000 clients. The sample represents all of the bank’s clients above the age of 40 who have their own bank account and are not excluded for ethical reasons (i.e., not legally competent). Then, we select a sub-sample of more than 27,000 clients who are eligible for our experiment (based on age, recent internet banking use, place of residence, account activity, privacy and data use settings). We will be able to contact roughly half of the 27,000 clients by letter and half by e-mail. At all stages of the study, phone calls are used as an (additional) communication channel. Second, of the 27,000 clients, we assign approximately 4,000 clients to the silent control group (ca. 700 from the e-mail group and 3,000 from the letter group). We do not contact clients in the silent control group at this point. The decision to assign more clients from the mail group to the silent control group was taken due to financial constraints in contacting clients by letter. Thus, we will contact 13,500 clients by e-mail and 10,000 by letter, informing them of the planned interventions and asking them to express their interest and provide some organizational details through an online survey. (Update October 2024: 25,982 clients contacted; 12,482 by letter, 13,500 by email; silent control group: 1725 clients). Given an estimated response rate of 2%, we anticipate up to 500 responses. Third, of those 500 responding clients, we will randomize assignment to the three treatment conditions subject to a number of constraints. Most importantly, only those participants who indicated that they would be able to participate in an in-person or family-based treatment could be randomized into such a treatment. We use the following randomization procedure: Respondents who state that they are unable to participate in any non-technical treatment (i.e., are unable to name a relative who would support them and are unable to attend an in-person course), are assigned to the traditional (baseline) treatment. Since their allocation is non-random, we will – depending on sample size – carefully conduct balance checks (s. pre-analysis plan), control for non-random allocation status and / or exclude this group from the analysis. Respondents who are able to participate in both non-technical treatments are randomly assigned between the baseline and non-technical treatment groups, obeying the capacity constraints for the non-technical treatment groups. Participants who are only able to participate in one non-technical treatment are randomly assigned between this treatment and the baseline treatment.
In case of a higher response rate than expected, we will 1) increase sample size in the baseline treatment group, 2) create a waitlist control group for the non-technical treatments, 3) introduce a third non-technical treatment group (i.e., a highly scalable video-based treatment). In case of an even lower response rate than anticipated, we will contact further potential participants from the silent control group, reducing its size. We will prioritize the groups as following (subject to cooperation restrictions with our partner): 1) non-technical group II, 2) baseline treatment group, 3) non-technical group I.
The relative in non-technical treatment group I receives 100 Euros in the form of a highly versatile gift voucher to account for his/her opportunity costs if the training has been completed. We choose to pay out vouchers rather than cash to comply with the bank’s internal guidelines. Recipients of the voucher can choose from a wide variety of e-commerce platforms, which ensures that a) the value of the voucher is comparable across individuals and b) the value of the voucher to the recipient approximates its nominal value.
The design of the platform for potential participants allows us to identify contacted people who download training material but do not participate in the first survey (and consequently, the actual intervention). If sufficiently large, we will investigate characteristics of this group and their IB use as a passive treatment group.
Stratification: Availability for in-person course during working hours, availability of a relative.
Update October 2024 (before final intervention data is available): We received 551 initial responses that could be merged to bank records. 333 participants met all participation criteria (complete responses, available for at least one non-technical treatment, sufficient language skills to follow training, sufficient access to IB technology) and were randomized into treatment groups (stratification see above). They constitute our main sample, which we check for balance across treatment groups with respect to gender, age, and bank account inflows. The remaining initial respondents were also assigned to treatment groups and constitute an expanded sample used in additional analyses.

The anticipated timeline is as follows:
• June 14, 2024: Collection of login and transaction logs for all clients aged 40 and older by the bank for the month before, transmitted to authors in anonymized form (using project ID).
• June, 2024: Selection of sub-sample eligible for interventions as well as random allocation to silent control group by the authors, transmission of project IDs to be contacted by letter and e-mail to bank.
• July 12, 2024: Initial contact letters / e-mails delivered
• August 5, 2024: Deadline for completing registration survey
• August, 2024: Evaluation of responses, randomization into treatment groups (see above)
• Early October, 2024: Second contact (including pre-survey link and booking platform for non-technical treatment group II)
• October 21 – 30, 2024: On-site interventions (non-technical group II)
• November 04, 2024: Deadline for claiming a voucher after successful training completion (non-technical group I)
• Early December 2024 and January 2025 (expected?): Collection of login and transaction logs for all clients in sub-sample eligible for intervention by the bank for the respective month before (November and December), transmitted to authors in anonymized form
• March 2024 (expected): Third contact (post-survey); if possible: additional collection of login and transaction logs as above

Stratified randomization.
Randomization done in office by a computer.

Individual

No

We invite 23,500 clients who have not recently used internet banking (10,000 by letter, 13,500 by email) to participate. We expect a response rate of 2 percent.
Update October 2024 (before final intervention data is available): We invited 25,982 customers who had not recently used IB (12,482 by letter, 13,500 by letter) to participate.

Update October 2024 (before final intervention data is available): We invited 25,982 customers who had not recently used IB (12,482 by letter, 13,500 by letter) to participate.

Silent control: 1,725 (725 from the group of clients we could have contacted via email)
Main sample:
Baseline treatment: 99 individuals
Non-technical treatment I: 34 individuals
Non-technical treatment II: 200 individuals
Full sample:
Baseline treatment: 252 individuals
Non-technical treatment I: 49 individuals
Non-technical treatment II: 250 individuals

We will use the 10% threshold for statistical significance due to sample size.