The study explores the hypothesis that President Trump strategically uses Twitter to divert media attention from issues potentially damaging to his presidency. This research is timely and relevant given the unprecedented reliance of contemporary politicians on social media for communication and agenda-setting. Traditional media’s role as the primary driver of political discourse is challenged by the rise of social media platforms like Twitter. This is particularly evident in the case of Donald Trump, whose prolific use of Twitter has altered the communication landscape. The research question focuses on whether Trump leverages Twitter to shift public and media focus away from potentially harmful news stories. The study contrasts the situation with that of authoritarian regimes, where governments systematically employ strategies to control information and distract the public through large-scale social media operations. In contrast, the researchers aim to determine whether, in a democratic context, a politician can effectively manipulate the news agenda using social media. The importance of this study lies in understanding how politicians can utilize social media to manage their image and influence the public narrative, highlighting the complexities of political communication and agenda-setting in the digital age. The introduction notes past research on media influence and agenda-setting, specifically focusing on how media coverage can impact events and public opinion, both positively and negatively. However, it also acknowledges the emerging role of social media, particularly Twitter, in shaping political discourse. The introduction concludes by specifically stating that this article will present the first empirical test of the hypothesis that President Trump's use of Twitter strategically diverts attention from politically damaging news stories.

The paper reviews existing literature on agenda-setting, highlighting the traditional view of media's dominant role in shaping public discourse. This traditional view posits that media, not politicians, primarily determines the agenda of public conversation in liberal democracies. The authors provide various examples from the literature to support this conventional understanding, citing studies demonstrating the causal link between media coverage and events like terrorist attacks, the rise of right-wing populist parties, and election outcomes. However, the authors also acknowledge a shift in this conventional understanding brought about by the rise of social media, particularly Twitter. This section of the literature review discusses instances where social media, including the spread of misinformation, has demonstrably impacted political agendas and public opinion. The discussion specifically touches upon how Twitter enables politicians to directly influence public discourse, citing previous research that indicates the reliance of journalists on Twitter for news generation and information gathering. Finally, the authors acknowledge the existing research on Donald Trump's Twitter usage, noting that while previous studies have primarily focused on the content of his tweets, there's been comparatively less attention to their role in agenda-setting and diversionary tactics. The review concludes by highlighting the anecdotal evidence suggesting that Trump strategically uses tweets to divert attention from potentially harmful issues, providing context for the present study's empirical analysis.

The study employed a quantitative approach using data collected over a 731-day period (from Trump's inauguration to the end of his second year in office). The data sources included all of Donald Trump's tweets from his official account (@realDonaldTrump), all full-text articles from *The New York Times*, and headlines from *ABC World News Tonight*. The researchers used keywords to identify relevant content. For the Mueller investigation (the potentially harmful topic), the keywords "Russia" or "Mueller" were used. For Trump's preferred topics (the diversionary topics), the keywords "China," "jobs," and "immigration" were initially used in a targeted analysis. A broader, expanded analysis considered all pairs of words from Trump's Twitter vocabulary that appeared at least 150 times. The methodology describes two analytical approaches: a targeted analysis focusing on pre-selected keywords and an expanded analysis encompassing Trump's entire Twitter vocabulary. The targeted analysis assessed the relationship between media coverage of the Mueller investigation and Trump's tweets on pre-selected diversionary topics using ordinary least squares (OLS) regression models. The expanded analysis used the same methods but extended the analysis to all word pairs in Trump's vocabulary. Both approaches included numerous control variables such as weekly intercepts, long-term trends, and lagged observations to account for temporal dependencies and other potential confounding factors. A three-stage least squares (3SLS) regression model was also employed to simultaneously estimate the effects of media coverage on diversion and diversion on subsequent media coverage. To enhance the reliability and robustness of the findings, several sensitivity analyses, including randomization, placebo analyses with Brexit-related coverage, and exploration of alternative explanations such as reverse causality, were conducted. The methodological details include descriptions of data cleaning processes, the inclusion of control variables to address confounding factors, and the statistical techniques (OLS and 3SLS) used for analysis. It explains the rationale behind using both independent OLS models and a simultaneous 3SLS approach to capture potential reciprocal causality between media coverage and Trump’s tweets.

The key findings of the study center around the relationship between media coverage of the Mueller investigation and President Trump's tweeting activity, and the subsequent effect on media coverage. The targeted analysis, using pre-selected keywords for diversionary topics ("China," "jobs," "immigration"), revealed a statistically significant positive association between increased media coverage of the Mueller investigation and an increased number of Trump's tweets focusing on the pre-selected diversionary topics. Furthermore, this increase in diversionary tweets was followed by a statistically significant reduction in subsequent media coverage of the Mueller investigation. The magnitude of these effects varied across the media outlets studied (*The New York Times* and *ABC News*), but the overall pattern was consistent. The researchers addressed concerns about potential endogeneity (measurement error, reverse causality, omitted variables) using a range of techniques. They conducted sensitivity analysis and robustness checks to assess the influence of omitted variables. A three-stage least squares (3SLS) model was also employed to analyze the simultaneous effects of media coverage on diversion and diversion on media coverage, showing that the results from this approach were largely consistent with the OLS results. Placebo analyses using Brexit coverage, which presented no political risk to the president, showed no significant relationships between media coverage and Trump's tweeting activity, further supporting the hypothesis. The expanded analysis, using all possible word pairs from Trump's Twitter vocabulary, also demonstrated a similar pattern. A significant portion of word pairs showed that increased media coverage of the Mueller investigation was associated with increased use of those word pairs in Trump's tweets, and that the subsequent media coverage of the Mueller investigation decreased significantly. The researchers compared these observed results to a randomized null distribution, demonstrating that the observed pattern was far more pronounced than would be expected by chance alone. The word cloud visualization highlighted the presence of the pre-selected keywords ("China," "jobs," "immigration") among the most frequent words associated with the observed pattern, providing linguistic evidence supporting the findings.

The findings provide strong empirical support for the hypothesis that President Trump's Twitter activity serves as a strategic tool for diverting media attention away from potentially damaging news coverage. This strategic use of social media to manipulate the news agenda successfully reduces the subsequent attention given to potentially harmful stories. The results support anecdotal evidence of Trump's use of diversionary tactics to counter negative media narratives. The discussion addresses potential alternative explanations for the findings, such as endogeneity (measurement error, reverse causality, omitted variables), emphasizing the robustness of the findings even after accounting for these concerns. The use of placebo analyses further strengthens the conclusions by demonstrating the specificity of the observed effect to topics that pose a political risk to the president. The discussion highlights the implications of these findings for journalistic practice and suggests that the observed interaction between presidential social media communications and mainstream media coverage may represent a defining challenge for journalism in the 21st century. The study links its findings to existing literature on diversionary theories of war and other rhetorical techniques frequently employed by Trump, such as deflection, pre-emptive framing, and inoculation. The overall discussion underscores the potential influence of social media on political communication and the need for further investigation into similar patterns among other political leaders.

The study concludes that President Trump's use of Twitter shows a statistically significant association with media diversion, potentially suppressing negative coverage. While acknowledging limitations in terms of establishing direct causality and generalizability beyond the studied media outlets, the results are robust across multiple analytical approaches and placebo controls. Future research should investigate whether similar patterns exist with other political leaders, expanding the generalizability of the findings. Further work should also explore the cognitive and psychological mechanisms underlying both the president's behavior and the media's response. The authors emphasize the importance of understanding how such patterns impact journalistic practices in the age of social media.

The authors acknowledge limitations in their study. While the findings suggest a strong association, the study's observational nature does not definitively establish causality. The analysis focused on two major media outlets; the findings might not generalize to all media. It also remains unclear whether Trump's behavior is intentional or a result of intuitive responses to media coverage. The specific keywords used in the analysis might influence the results, and the limited time frame of the study (two years) may not capture the full range of President Trump's Twitter behavior. The study also acknowledged the inherent difficulties in establishing absolute certainty regarding the role of omitted variables influencing causal inferences.