Effects of Fraud on Digital Services Use

How to cite

Kubilay, Elif; Spantig, Lisa; Raiber, Eva; Barasa, Laura Nelima. 2026. "Effects of Fraud on Digital Services Use." AEA RCT Registry. AEARCTR-0019148. https://doi.org/10.1257/rct.19148-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Elif Kubilay* , Lisa Spantig , Eva Raiber , Laura Nelima Barasa

* Primary investigator

Registry ID
AEARCTR-0019148
DOI
https://doi.org/10.1257/rct.19148-1.0
First registered
Last updated
Status
In development
Trial dates
2026-07-15 – 2027-02-28
Countries
Private
JEL codes
D14, D18, G53, O12
Registry record
http://www.socialscienceregistry.org/trials/19148

Abstract

The expansion of digital financial services in developing countries has brought substantial economic benefits but has also led to an increase in consumer protection problems. While previous research has focused on the direct costs of scams, less is known about how fraud perceptions influence broader digital engagement. We study two channels through which these perceptions affect behavior: beliefs about the prevalence of scams (the environment channel) and beliefs about one's own ability to detect scams (the self channel). In a large-scale randomized online experiment with Kenyan respondents, we experimentally vary these beliefs using objective information on scam prevalence and personalized feedback on scam identification ability. We then measure effects on beliefs, digital service use, responses to genuine and fraudulent SMS messages, demand for scam prevention training, consumer protection engagement, and trust in digital institutions. We expect the findings to provide new evidence on how fraud perceptions shape digital participation and inform the design of consumer protection policies.

Intervention

To study how fraud perceptions influence digital behavior, we implement a large-scale online experiment with Kenyan respondents. Participants are randomly assigned to one of five experimental arms that vary perceptions of scam prevalence and susceptibility to fraud. The interventions are designed to study two distinct channels. The first is the prevalence (environment) channel, whereby beliefs about the frequency of scam attempts influence trust in and adoption of digital services. The second is the susceptibility (self) channel, whereby beliefs about one's own ability to identify scams shape perceived vulnerability and subsequent digital behavior. Further details on the intervention are provided in the pre-analysis plan.

Experimental design

The study consists of a large-scale online randomized experiment administered to a panel of Kenyan adults. Approximately 3,500 respondents are randomly assigned at the individual level to one of five experimental arms. The five arms include a pure control condition and four intervention conditions that vary two dimensions of fraud perceptions. One set of interventions provides respondents with feedback designed to update beliefs about their own ability to identify fraudulent messages (high susceptibility vs. low susceptibility). The other set provides information designed to update beliefs about the prevalence of scam attempts in Kenya (high prevalence vs. low prevalence). The treatments are designed to isolate the effects of these two channels on subsequent beliefs and digital behavior. Following the treatments, all respondents complete a common set of outcome measures capturing beliefs, digital service use, responses to genuine and fraudulent messages, consumer protection behavior, and trust in digital services. Full details on the treatment implementation, randomization, and information provided to respondents are included in our pre-analysis plan.

Design details

Randomization method
Randomization is done within Qualtrics
Randomization unit
Individual
Clusters
Treatment is not clustered in this study.
Observations
3500
Arms
(1) Pure Control: app. 700 respondents (2) Low susceptibility arm: app. 700 respondents (3) High susceptibility arm: app. 700 respondents (4) Low prevalence information: app. 700 respondents (5) High prevalence information: app. 700 respondents
Primary outcomes
(1) Expected Usage of Digital Services, captured by (1.1) Digital Service Use Index and (1.2) Digital Platform Engagement Index, (2) Adoption of Digital Services and (3) Take-up of Digital Opportunities
Secondary outcomes
We examine two groups of secondary outcomes. The first consists of policy-oriented outcomes, including reporting behavior, demand for fraud prevention, support for consumer protection measures, responsible consumer behavior, and responses to fraud attempts. The second consists of mechanism outcomes, including posterior beliefs about scam identification ability, scam prevalence, and the likelihood of experiencing fraud-related losses; perceived consequences of victimization; perceived trustworthiness of digital services; responses to suspected scam messages; and digital platform engagement by perceived platform risk.

Cited by

No citation data available for this registration yet.

Keywords