Unveiling the Data Behind 2022 Election Advertising: Insights from Meta and Google
In an era where digital advertising shapes political landscapes, understanding how campaigns targeted voters during the 2022 general election is crucial. This article dives into the data construction process behind the election advertising datasets from both Meta and Google, revealing the intricate steps that researchers undertook to analyze this data.
The Data Construction Process: Three Key Steps
The construction of our datasets involved three essential stages: data collection, data processing, and data classification. Each step plays a pivotal role in ensuring the effectiveness of our analysis.
Data Collection: A Deep Dive
The first step revolves around gathering the raw media files and metadata from both Meta and Google’s public ad libraries.
From Google Ads: The Google Transparency Report is an invaluable resource, providing tables in Google BigQuery with relevant advertising data shortly after election cycles. However, our review revealed that ads could be added post-election. This nuance is crucial as variations in data collection dates can lead to discrepancies.
- Navigating Meta’s Library: The procedure here is less straightforward, as it includes both election and non-election ads. Using specific search terms related to federal offices, we identified election-related content and meticulously compiled the IDs of relevant ads.
Data Processing: Transforming Raw Data into Usable Insights
Once collected, the raw data underwent extensive processing. We utilized advanced technologies such as:
- Speech-to-Text: The Google Cloud Speech-to-Text API enabled us to transcribe video ads, ensuring we captured core campaign messages accurately.
- Optical Character Recognition: The Amazon Rekognition API allowed us to extract text from images and detect faces, giving us a complete picture of each ad’s content.
This combination of technology provides a robust framework for analyzing the nuances of campaign messaging, enabling us to categorize everything from ad tone to electoral focus.
Data Classification: Generating Scholarly Insights
In the final classification stage, we produced critical variables tailored for electoral studies. These include:
- Ad Tone: Our classification identifies ads as either "promotional," "attack," or "contrast," allowing scholars to discern the emotional tenor of campaigns over the election cycle.
- Ad Goals: We analyzed whether ads aimed for voter persuasion, fundraising, or mobilizing supporters, which sheds light on campaign strategies.
Engaging with the Data: Insights and Implications
By restricting the datasets to the 2022 general election period—specifically from September 2022 through Election Day—we captured 377,721 ads from Meta and 80,247 ads from Google. This targeted approach not only refines our analysis but also enables easier validation of results.
The Impact of Political Advertising
Understanding the race of focus for each ad gives insight into how different electoral offices are targeted. For instance, ads for U.S. Senate candidates are classified according to the races they are contesting, enriching our comprehension of campaign strategies and voter engagement techniques.
Conclusion: A Treasure Trove of Data
As digital advertising continues to evolve, so does the necessity for transparency and analysis in political campaigns. The methodologies used to create datasets from Meta and Google not only push the boundaries of research but also offer a comprehensive view into how advertising shapes voter perceptions and electoral outcomes.
For further reading on political advertising and its impact, check out FEC regulations and insights on issue ownership in elections.
Engaging with this data not only informs our understanding of political dynamics but also equips scholars, policymakers, and citizens with the knowledge they need to navigate the complex landscape of electoral advertising in the digital age.
