Decoding the Inauguration Effect: An NLP Analysis of Changes in Immigration Rhetoric in Facebook Political Advertising Open Access

Abstract

This thesis investigates the relationship between issue-area agenda setting and political advertising behavior, using the case of President Trump's inauguration and the subsequent changes in immigration-related Facebook political advertising to identify and measure this effect. To analyze the proportion of ads related to immigration and emotional sentiment, I assessed this relationship via three separate models of Natural Language Processing: two data dictionaries, a Small Language Model (ollama 3.2 3B), and a Large Language Model (ollama 3.3 70B). I systematically validated each method against hand-coded samples to assess its efficacy for political advertising text. In doing so, I found that a data dictionary was most effective at identifying immigration-related advertisements, while the Large Language Model performed best at identifying sentiment and emotion (91.6% accuracy on immigration-specific tone and 83.4% accuracy on emotions), though at significantly greater time cost. 

Using the best-performing methods, I then conducted a difference-in-difference analysis of the two relationships, treating President Trump's inauguration as an exogenous political shock. I compared changes in immigration ad volume and sentiment across five actor categories—nonprofits, politicians/political parties, PACs, media, and corporations—and across states with high versus low immigrant population shares. Nonprofits significantly increased their immigration advertising following the inauguration, while corporations and media showed no change. Politicians and PACs also shifted their behavior around immigration in their advertising, although those findings were causally inconclusive. At the geographic level, I found that political advertisers meaningfully shifted immigration content toward high-immigration states post-inauguration, a statistically significant result suggesting that the inauguration caused actors to adjust their behavior by intensifying geographic targeting in the areas most directly affected by the new administration's immigration policies.

Table of Contents

I. Introduction 1

II. Literature Review 2

Agenda setting 2

Presidential transition as a media shock 4

How agenda setting impacts the advertising media space 5

Why immigration? 6

III. Argument 9

Foundation 9

Defining the Actors 11

Framework 13

Reasoning 14

Hypotheses 17

IV. Methodology 20

V. Results 35

Parallel Trends Analysis 35

Actor-by-Actor Analysis 40

High vs. Low Salience Immigration State Comparison 47

VI. Conclusion 51

Works Cited 56

Appendix 60

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Data and Decision Sciences
Degree	B.S.
Submission	Honors Thesis
Language	English
Research Field	Political Science, General Statistics
Keyword	Natural Language Processing Facebook Advertising Agenda Setting Large Language Model
Committee Chair / Thesis Advisor	Cuttner, Allison, Emory University
Committee Members	Semetko, Holli, Emory University McAlister, Kevin, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Decoding the Inauguration Effect: An NLP Analysis of Changes in Immigration Rhetoric in Facebook Political Advertising ()	2026-04-07 11:15:58 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions