The past few decades the American political climates has been witness to increase polarization. As politicians attempt to brave these ever-changing waters, the current President has been able to navigate this uncertainty with success. With Trump’s brute force approach to politics, there remains rigid division in public opinion towards his actions and rhetoric. Representing a stark contrast from any previous President, the question arises what this rhetoric and what effect is there on the President’s approval rating.
Obviously, there findings and outcome from this research could be used to benefit the Trump and his administration to target audiences and modify the message to seek to maximize their approval rating. However, I believe that this research is for the American people at large. I believe that this examination can provide further insights into how nearly half of Americans resonate with this inexperienced politician while also understanding why the other half are so troubled with his election to office. We are hoping to better understand this division and how this division is changing American politics from discourse to policy.
This project will examine Donald Trump's word use in all his Presidential speeches, addresses, and appearances from his inauguration (January 20th, 2017) to March 20th, 2018. We hope to predict whether or his not his speech language (sentiment and policy word use) has any impact on the change in his approval rating.
For this data set, we pulled the approval ratings from the open source FiveThirtyEight (www.FiveThirtyEight.com) dataset. They collect approval ratings from a variety of sources. They sort and weight their sources by sample size and bias to give an adjusted rating score. We then scraped the President’s speeches from the American Presidency Project (http://www.presidency.ucsb.edu/) which have a collection of nearly all the President’s public appearances. We then removed stop words and punctuation and tokenized the speeches to be single n-grams. From there, we counted the number of words in the speeches that were also in our set of sentiment and policy corpora. For the sentiment corpora, we used the nrc and bing lexicons. For the policy corpora, we created our own set of words for what we believe to be the major policy issues. We pulled words that would only be used in the context of that particular policy issue to attempt to limit overlap.
From our examination, we were able to uncover three major findings. The first is that the President’s speech data is incredibly noisy. This is in part due to the method we employed in which we only used words present in a particular corpus. There is a need to further dive into the text data and pull out more consistent phrases and themes, in order to detect a stronger signal. This may mean using a larger proportion of the text per speech, eliminating superfluous documents, and integrating other mediums the President’s use e.g. twitter data.
The second is that we were able to develop a general profile of the President’s rhetoric. When the President is speaking, depending upon the subject, it can be separated into two separate categories. When he is speaking about something positive, he is speaking with surprise, joy, and anticipation. When he is speaking about something negative, there is the use of anger, sadness, fear, and disgust. There are these interesting relationships present in the policy where we see a positive relationship between Foreign Relations and Military/Defense, but a negative relationship between Military/Defense and the U.S. Economy. We believe that this is because of his message of what the government should and should not be involved in i.e. the government should be involved in defending and protecting the American people and their values while remaining indifferent to the workings of the U.S. Economy. However, this seems to contradict the President’s message of getting American workers back to work. When we compare policy and sentiment, we observe that military and defense issues have an underlying sentiment of negativity, fear, and anger, while the U.S. economy has this underlying sentiment of trust and fear, again evidence to seemingly contradict our hypothesis in regards to his message.
The final finding is that Trump’s speeches are not strong predictors of the public opinion or a shift in the public opinion. This may be due to our set of features, and replicability is necessary for any final conclusion on this matter, yet we have evidence to believe that Trump’s words reach an already decided audience. We see that approval rating only changes by about 10 points over the year and half of Trump’s Presidency. It may seem obvious, but one speech is not going to drastically change the public’s opinion on this President, nor does it seem that there is anything the President can say that would alter his approval ratings unless the President drastically changes his opinion on particular policy. We believe that he would not do this because of the incredibly positive and reinforcing support of his base; his approval rating, no matter the controversy, has not sunk below 36 points. In regards to the upcoming two and a half years, we seem incapable of predicting what he will do or how the public will come to react to it.