When we think about political parties, we think about the fanfare we see in the media, politicians attached to them, or topics they parade. When we begin to look under the hood of the political parties, we begin to see trends through text analysis that provide in-depth findings on how sentiment analysis, words, and topics might impact the outcome.
Party Platforms are published every Presidential Election year. These party platforms serve the core beliefs of the party for the next four years. These platforms are the leg that forms and supposedly influences the politicians who represent the parties. I analyzed the platforms for both parties from 2000 to 2016. Before my findings, things to note are that the length of the document varies from year to year, with variation between the political parties. I utilized sentiment analysis to look at how the political platforms might shape election results.
Democratic and Republican Platform
At first, I want to look at the common words of the party platforms. Some words that stand out between both parties are America, American, and nation. A common misnomer between both parties is that one is more patriotic and more nationalistic. Yet, the word clouds expose the falsehood of these sentiments. One can conclude that both parties are equally patriotic as their main commonality falls alongside these three words.
Furthermore, this builds upon the topics that we see between both parties. The differing beliefs of parties are supposedly broad but reveal the commonality between their words to discuss the same topics. After a series of running correlations, I found that both parties might be able to agree as they both describe and have strong correlations around words that seemingly resemble the topics of immigration and national security. With the constant discussion and the slow-paced government that we have, the data reveals that these conversations might be more counterproductive as they don’t reflect what the text is saying to us. This analysis will be discussed further when looking at the topics of discussion by recent presidential candidates.
Public outlook is how a political party wins an election and entices its voter base. Sentiment analysis shows us how positive or negative a text is perceived through machine learning. Looking at the Figure 1 and Figure 2, we see a drastic difference in the perceptions of the political platforms between both parties.
The overall sentiment means are 76 and 45 respectively for the Democratic and Republican parties. We see that the Democratic platform is overall positive when matched up against its counterpart. When looking at these platforms in their respective election year, despite the overall depressed mean, the Republican party won 3 of 5 elections within this timespan. This indicates a significant finding sentiment text analysis might not be a perfect way to display results of an election or even reflect the public point of view. The tumultuous 2016 election presents a very skewed data point in the sentiment analysis. Both parties have an incredibly depressing sentiment analysis on their platforms. At times we wish the data can explain the factors that come to this conclusion, but I think it might be essential to think about the external factors that might have influenced the language. The 2016 election is a turning point in politics. We had polar opposite candidates; neither found favor in the general public and shifted ideals of whether a candidate’s experience was more important than a new perspective. Although the overall sentiment for the 2016 Democratic party platform was positive, this didn’t ensure the party’s win, but the negative sentiment of the opposing party took the cup. Yet, I must point out that the popular vote for 2000 and 2016 exceeded the winning party’s votes, i.e., the electoral college delivered the win for the election. When we look at the sentiment analysis, we can conclude that it is an indicator for the winner of the popular vote showing the overall sentiment rather than looking at the sole winner of the election.
Politicians are representatives of political parties, and they should reflect the same sentiments and variety o topics they discuss in their speeches. I scraped the transcripts for every speech given by Joe Biden and Donald Trump from this past election cycle for this analysis. It is crucial not only to have a scope of the party but also to speak for the parties. We need to look at if self-interest or the depictions of specific topics over others present a deviation from the party’s core beliefs.
As the political candidate of the democratic party, we would like to see a similar sentiment analysis in the democratic party platform. I first began analyzing the sentiment of the speeches given during the campaign trail. Below we will see in the graph the upward trajectory with the speeches given by Biden. Yet, when we devel into the analysis, we can see there lies a finding.
The mean sentiment score for speeches given by Biden is 28, a large gap between the average sentiment score recorded for the general Democratic political platform; Biden score has a deficit of 48 points. This might be a huge shock, but we need to realize the compact nature of speeches and the shorthand information/beliefs they are trying to disseminate. This also provides and delineates a more exciting finding. In essence, we expect politicians to touch on every point that their Democratic party platform, but we know that buzzwords or topics carry a politician and their campaign, especially in the media. Specifically for the Democratic platform, I identified trained the data to ten issues: healthcare, economy, employment, public safety, security, education, protection, unity, middle class, and constitutionality. Across the various speeches given, Joe spoke about the economy about 65% of the time. The finding is indicative of the concerns surrounding the economy, especially post-COVID, alongside raising wages and general growth outlook for the economy. The emphasis on this topic in both the platform and the politics reflects the criticisms given to Democrats because of their willingness to provide more fiscal stimulus to Americans versus their counterparts. Although there is an overall lower mean for the sentiment of Biden’s speeches, we can learn the key issues that are reflective of the party’s platform.
Alongside the same vein of my analysis of Biden’s speeches, I conducted a similar study with Donald Trump’s speech transcripts as well. My first finding was that the mean sentiment score was higher than Biden’s. The data shows a contrasting view between the media and the public outlook of Trump’s addresses.
We can point to different factors as to why this might be, but we will devel a bit later — Trump’s trendline as shown as a higher trajectory than that of Biden’s. With Trump having a higher mean sentiment score, he still falls short of reaching the average sentiment of the overall party. Though we might conclude that it might be the topics that are factoring in here, we might want to look at the politics surrounding the data results. Trump may represent the Republican party, but by the institutional Republican party, he is not a representative of what a “Republican” might be. This nonacceptance breeds two agendas, but I want to clarify this doesn’t necessarily mean that they’re wildly different. Despite this conclusion, I wanted to look at which issues correlated most between the two different agendas. I trained the Republican platform on ten various topics: taxes, economy, state affairs, national security, congressional affairs, laws, middle class, public system, terrorism, and constitutionality. Over 40% of Trump’s addresses talked about national security; this shouldn’t come as a shock as both the candidate and the party align heavily on the vision of how national security should look, especially on the immigration front. I think it is fascinating that despite the sentiment analysis that separates the agenda of the party and the agenda of the politician, there is still commonality. Even though we see a contrasting display at times in the media, the issues and the sentiments are more alike than we think.
The legislature is the final key of delivery for political parties. I decided to crack the bills’ text and how their sentiment might identify some insights for us. I will note this text analysis was done on bill summaries and not their text. Looking at the sentiment score over time for the bills, we can see a bell-shaped graph. I deduced that this is a normalization between the different parties agendas. We can see the split between their platforms but can also see it in the distribution of the sentiment scores of the bills.
The normalized graph for the sentiment of bills renders the balance between the two parties. The normalization of the graph let’s us see the comprise that must come about in order to have these bills. If the graph were skewed we might want to devel into which party is writing more bills. Furthermore, because of the constant shifting of powers in congress this normalization will most likely continue as compromises will also serve as a factor in the normalization of the sentiment scores across different bills.
Throughout this data analysis, I believe these are key insights
- Sentiment Analyses can deviate from human perception but can also serve as a predictor.
- Correlation between different texts reveal commonality where we might think there are wide differences
- Political Parties are just a vehicle for ideas because a bill or person will never fully display the information that you need