An Experimental Infrastructure to Investigate the Impact of Online Tracking, Targeting, and Advertising on Consumer Behavior and Consumer Welfare

Participants are assigned to three experimental conditions: control, ad-blocking, and anti-tracking.

2025-05-09

2026-05-08

This is an experimental infrastructure designed to investigate an array of research questions related to the impact of online advertising, tracking, and targeting, as well as protections from advertising, tracking, and targeting, on a variety of consumers' online behaviors and online outcomes. The infrastructure consists in a client-server architecture, with the client component consisting of desktop and smartphone software that participants will install on their devices, and the server component consisting of secure databases that will receive and store de-identified/anonymized information collected from the participants' machines. While the experimental infrastructure itself is described under the "Experimental Design" section of this trial information, in this section ("Primary Outcomes") we discuss the key variables collected by the infrastructure. The infrastructure collects four types of data: desktop browser data, email data, smartphone browser data, and survey data. Before any type of data is sent from the clients to the servers, the data is stripped of any personal identifying information. The infrastructure is built to uphold the highest standards of privacy and confidentiality for its participants.

1. Desktop browser data
Desktop browser data includes multiple data fields. Some of these fields are collected for all websites visited by the participant during the experiment (for instance: visited urls); others data fields are collected only on specific websites (for instance: html content for the shopping, search, and LLM websites visited by the participant). The fields include:

+All Website Visits

Web Navigation Details:
URL (domain, parameters, timestamps), stripped of PII
HTTP Headers (cookies)
Header Bidding
Transition Type: Clicked link, typed, redirect, bookmark

Advertisement Content on Page:
Web requests to known ad domains
Visual Ad Content (HTML, images)
Non-visual Ad Content
Tracking elements (pixels, beacons, cookies, etc.)

Shopping, Website Visits:
Full Page Content (HTML, CSS, images, scripts), stripped of PII
Information Extracted:
Products: Viewed, compared, searched, added to cart, purchased
Timeline: Time spent on each stage of product research and purchasing

Search and LLM Website Visits:
Full Page Content (HTML, CSS, images, scripts), stripped of PII
Information Extracted:
Search: Queries, results, sponsored results, clicks
Timeline: Time spent reviewing results, paging through results

Browser Context and Configuration:
Settings and permissions for both the browser and for specific websites
Participant’s use of multiple windows, tabs, third-party extensions
Interactions with Consent Mechanisms

2 Email data
From the participant’s email, an algorithm identifies and collects data from purchase confirmation emails and promotional emails. Data collected from promotional emails includes subject line, timestamp, and sender. Data collected for shopping emails include:

Email Metadata: Timestamp, sender
Email content: Vendor, product information, price, quantity, stripped of PII

3 Smartphone data
Web Navigation Details:
URL (domain,timestamps)


4.4.3 Survey data
Survey data includes anonymized responses to entry, periodic, and exit surveys. Surveys include questions related to:
Buying experiences and finances
Browsing experiences
Subjective well-being
Ads attitudes
Tracking, targeting, and privacy attitudes
Demographic questions
WTA questions

The raw variables collected by the infrastructure will be used to construct variables to use in the analysis, including (but not limited to):
- Time spent on different websites, including shopping, searching, and LLMs, and so forth.
- Purchase patterns, including amounts spent (total, and average across products), choice of online vendors, rates of product returns, and so forth.
- Patterns of interactions with LLMs, including prompts and answers.
- Interactions with online privacy settings and consent mechanisms across platforms and websites.
- Access to online news, entertainment, and other content.
- Exposure to and interactions with advertising.
- Online advertising ecosystem changes to participants' treatments, including changes in header bidding data.
- Longitudinal patterns in attitudes towards online ads, tracking, targeting, and privacy.
- Longitudinal patterns in satisfaction with online browsing, shopping, and searching.
- Willingness to accept payments to remain exposed or not exposed to online advertising, tracking, and targeting.
- And so forth.

Because this project is designed as a infrastructure (or "platform") for multiple studies, the non-exhaustive list of research questions that we intend to tackle includes:
- Study the extent to which shielding users from advertising, tracking, or targeting influences their online purchase behavior.
- Examine whether exposure to advertising affects users' online spending levels.
- Investigate where and how consumers choose to purchase their products online.
- Analyze whether consumers shop across more or fewer different online merchants under various experimental conditions.
- Assess whether consumers' satisfaction with their online purchases changes across conditions.
- Evaluate how much time consumers spend searching for product information under different experimental conditions.
- Explore participants’ understanding of ad targeting and their perceptions of online advertising.
- Examine how attitudes toward online advertising evolve over a three-month period across different experimental conditions.
- And so forth.

As noted above, this is an experimental infrastructure designed to investigate an array of research questions related to the impact of online advertising, tracking, and targeting, as well as protections from advertising, tracking, and targeting, on a variety of consumers' online behaviors and online outcomes. While the primary goal is to investigate research questions about participants as consumers and buyers of products, the infrastructure allows much broader investigations of participants' other online behaviors. For instance, a non-exhaustive list of additional research questions that we intend to tackle includes:

- Study how participants in the ad-blocking condition respond to websites that restrict access to users with ad blockers.
- Investigate how ads and tracking interact with the usage of large language models (LLMs).
- Explore the relationship between advertising/tracking and the consumption of online entertainment.
- Study how ads and tracking influence the consumption of online media.
- Analyze how advertising and tracking influence the sources and the quality of content consumed by participants.
- Analyze the relationship between advertising, tracking, and political beliefs.
- And so forth.

Participants will be recruited via a variety of recruitment channels and screened through an initial Qualtrics survey. Qualified and consenting participants will complete a baseline entry survey about their online behaviors and preferences/attitudes, and will then install the study software (including the browser extension, email extension, and smartphone apps) on their devices. The software will apply experimental treatments and will collect data passively for three months. Participants will complete monthly surveys and a final exit survey. At the end, they may choose to upload additional data or extend participation for extra compensation. They will then be guided through uninstalling the software.

Not available

Randomization is performed by the software installed by the participants.

Individual participant.

No

1,200 participants planned, but will depend on how many actually get enrolled.

1,200 participants

400 per experimental treatment.