The Effect of Polarization on the US Elections
Authors: Priyal Sharma (project lead), Vidusha Adira, Ryan So, Sean Gee, Megha Velakacharla
Introduction
The “United States” is a name that now rings with irony as our country is the most polarized, divided and angry more than ever. Both sides of the political spectrum move increasingly towards the ends of their respective political spectrum, and the middle ground of overlap grows fainter into nonexistence. A country brewing in political opposition and political antipathy can often lead to a society that breeds negativity, hatred and an overall loss of empathy of other human beings on the other side of the political spectrum. Furthermore, from a bureaucratic standpoint, an unstable state of this sort leads to inefficient policy making due to intense congressional gridlock and divided government.
Today, we aim to contextualize our country’s polarized state. How exactly is it polarized? How did it increase over time? Why did this happen — and how does social media play a part?
And finally, we want to contextualize these ideas in a very relevant manner: What president would our country vote for, in its divided state?
How can we measure the polarization of our country?
To understand and measure polarization, it’s essential to define what polarization is in the context of the United States. Polarization in the U.S. refers to the growing ideological divide between political parties and how it has influenced the nation’s policies and historical trajectory. Today, we aim to measure this trend of polarization throughout U.S. history to predict the future 2024 voting results.
To predict voting results, we first need to discuss how general polarization has affected the primary institutions of the U.S. government, the House of Representatives, and the Senate. Throughout U.S. history, policies that shaped America can be traced back to how willing many lawmakers and officials were willing to spend and how it will affect the overall country. Consequently, this has sparked debates and separation within both the Senate and House over what policies were allowed to pass and what didn’t. This ideological shift can be measured using NOMINATE scores, which quantify political ideology in Congress by analyzing roll-call voting data to position legislators on a liberal-conservative spectrum.
As shown in the graph, we illustrated the average NOMINATE scores of each year, with red (Republican) and blue (Democrat) indicating the leaning factor for each year. This will not only reveal the prevailing political ideology of that current year but also which party held more influence and how their ideologies have affected the nation as a whole. For instance, both the House and Senate leaned Democratic between the 1920s and 1960s, a period that included both World War II and the Cold War, indicating a Democratic consensus on combating fascism and communism. The green trend line reveals that, over the years, polarization in the U.S. has increasingly leaned Democratic in the House and more Republican in the Senate, reflecting the current political climate where Republicans prioritize controlling the Senate.
As we approach the 2024 election, recognizing these trends allows us to anticipate potential shifts in voting behavior and policy directions. With this foundation, we can now explore specific voting patterns in the U.S. This next section will delve into how historical and current polarization influences voter preferences and election outcomes, providing a comprehensive understanding of the forces driving the American electoral landscape.
We can then look at this next visualization to gain a better understanding of how polarization has changed across the country over time. The visual uses data of elected presidents, senators, and state representatives post 1960. The data was then separated by Republican and Democrat to analyze polarization by party, by examining the likelihood of being nominated for each official. The range for the color scale was determined by using the first and third quartiles of each subset to create the most accurate visual representation. To illustrate this concept, we can look at Florida in 2000, which is colored light green for the Democratic Party yet dark green for the Republican Party despite having similar numbers.
For both visuals it is evident that as time progresses, the average likelihood of nomination increases. This directly reflects voter unity on whom they vote for as political candidates lean more into conservative and liberal ideology. After understanding how polarization is measured in the previous sections, we can now look to find underlying factors that play a role in this shift.
What part did social media play into this?
To understand the role social media has played into this, we need to analyze how politicians are using it. Politician use of social media has only grown over time, in particular with twitter, as seen below. There is an increased reliance on short form social media sites such as twitter to get their message across. While polarization has only been increasing since the 1960s, the advent of social media in the 1990s and its entrance into mainstream popular culture in the early 2000s has seen an extreme increase in polarization, measured by the difference in average NOMINATE scores over time. The slope has increased from 0.003 from the 1960 to 1990, to 0.007 in 1990 onward, indicating an increasingly rapid divergence between the two parties’ ideals currently representing a majority of the US. There can be a connection drawn between the increase of polarization and social media as social media entered the mainstream in the early 2000s with the creation of MySpace, the first social media platform to pass 1 million users.
Part of this gap can be attributed to the increased distrust in conventional news sites by conservatives, with a negative correlation in trust from conservatives the more mainstream a news site is, while liberals tend towards a positive correlation. This means that conservatives are more likely to seek their information from unconventional sources which they could dictate is “true” due to the closeness of the source. In other words, social media posts from trusted leaders. The increase in tweets and other social media posts over the past years also means that there is an almost information overload from lawmakers, giving the impression of being well informed from these social media sites. This inherent distrust of mainstream news also encourages disagreements between liberals and conservatives, as the information that each side uses in discussions is dismissed by the other side as “incorrect”, furthering the divide.
Taking a look at tweets written in 2015 by lawmakers who were attacking other representatives, there is a distinct use of partisan language, with the words “GOP”, “Republican”, and “Obamacare”, a notably partisan and highly divisive program that has since faced multiple attempts by the Republican powers since its creation to dismantle it. In comparison to the most common words used in neutral worded tweets, which were “American”, “will”, “great”, and others similar. With such limited word count when attacking lawmakers, these representatives cannot delve into the specifics of their arguments and thus devolve into using generalizing terms and vague language that oversimplifies their arguments. The identification of the opposing side or the representative’s own side plays into party politics, as the constituents reading it are more likely to identify with whoever is more politically aligned with you. Thus, they approach the dangerous combination of a lack of nuance and overgeneralization, leading to black and white mentality, and rather than one side being a committee or a specific group of representatives, it becomes one of the two parties, driving the partisan divide.
By analyzing the terms used by members of Congress on social media that are associated with the highest spikes in engagement, we can gain further insight into the role social media plays in exacerbating political polarization. The most influential terms displayed below indicate the topics that received the highest increases in engagement from users on both Facebook and Twitter from 2015 to 2020. Here, engagement is measured by reactions/shares on Facebook and favorites/retweets on Twitter. While there is some overlap in the topics that engaged both Democrats and Republicans, there are also clear differences in the issues that each party tends to prioritize and the language they use to discuss them. The popularity of posts containing terms like “impeachment” and “kavanaugh,” across both parties suggests these topics were highly polarizing and garnered significant attention from both sides of the political spectrum. Interestingly, users are more likely to interact with posts in which a congressperson mentions a political opponent or a prominent political figure from the opposing party. Users tend to react to Democratic members’ posts mentioning then-president Donald Trump, former Republican Secretary of Education Betsy DeVos, and Supreme Court Justice Brett Kavanaugh. Republican members’ posts that gain more traction mention former Democratic Speaker of the House Nancy Pelosi, and Democratic Californian Representative Adam Schiff, who was the lead prosecutor in Trump’s first impeachment trial.
Moreover, the distinct terms such as “president elect” and “nobillnobreak” for Democrats, and “china” and “fbi” for Republicans indicate differing priorities and concerns within each party. ‘No bill, no break!’ refers to a sit-in staged by Democrats in the House of Representatives in 2016 to protest the lack of progress on gun control. “China” was most mentioned in 2020 following the outbreak of COVID-19. According to a Pew Research Center survey, unfavorable views of China in the U.S. reached new highs in 2020, with Republicans and Republican-leaning respondents being far more critical of China (83% held unfavorable views in 2020) compared to Democrats and Democrat-leaning participants (68% held unfavorable views). The spike in mentions of the FBI is likely linked to Republicans’ disapproval of their handling of Hillary Clinton’s email server and its inquiry into the alleged coordination between the Trump campaign and Russia.
Facebook and Twitter users are far more likely to interact with posts from members of Congress that criticize events or individuals. This trend emphasizes how online political discourse often revolves around contentious issues and people, reflecting the broader political divide in society. Users are possibly seeking out content that aligns with their political beliefs, especially critiques that resonate with their views. Social media platforms may also be amplifying condemnatory and polarizing content because it tends to receive more engagement.
The belief-affirming content that has been made readily available through social media has facilitated a rise in distrust in traditional news sources, more so among Republicans. In 2021, a Pew Research Center survey found that Republicans who say their main news source is ‘mainstream’ have less trust in it, while the opposite is true for Democrats. This phenomena indicates that information is being spread on social media in a way that promotes possible misinformation and reinforces existing biases, escalating polarization.
Who would win our 2024 Election?
To contextualize our results today, our team built a machine learning model based on MIT datasets on past elections and election winners to predict the 2024 election. The code imports historical poll and election results data, cleans and merges them. A linear regression model is trained on the data to predict election outcomes. The model’s accuracy is evaluated, and it’s applied to predict the 2020 and 2024 elections. Finally, the electoral votes are calculated to determine the predicted winner of each election.
Before we present the results, there are a few key ideas to highlight. Our visualizations and data demonstrate that America is becoming increasingly polarized. When comparing the two candidates for the 2024 election, Trump appears more extreme in his beliefs, while Biden is seen as a more moderate figure on the left. This suggests that a polarized America might favor the more outspoken and radical candidate on their respective side of the spectrum.
According to our machine learning model, Trump will be winning our 2024 election.
Final Thoughts
The United States’ polarization, characterized by a growing ideological divide, has led to societal discord and legislative gridlock. Social media exacerbates this divide by amplifying partisan content and creating echo chambers, with politicians’ use of platforms like Twitter deepening the rift. Language and engagement analysis show that contentious topics draw more interaction, perpetuating polarization. Our machine learning model predicts a Trump victory in the 2024 election, reflecting current trends. Addressing this polarization requires promoting media literacy, encouraging bipartisan dialogue, and fostering empathy to heal the divisions undermining our democracy.
