- Abstract
- Motivations
- Key Questions and Hypotheses
- Data Collection
- Data Set
- Code Explanation
- CSV Files
- Findings
- Conclusion
- Contributions
- Maia
- Amara
- Sarmad
- Bibliograpy
This project explores words used in tweets put out by members of the United States Congress. We aimed to explore whether there were any disparities between Republicans and Democrats in terms of words frequently used.
We used Twitter's API, with elevated access, to gain access to almost 3 million tweets posted in the past 30 days from all 535 legislators in the US Congress. We then extracted key words through spacy, excluding certain words (such as prepositions), then grouped them by Twitter handle and then counted the 50 most frequent words used by a given legislator. We found significant evidence for our hypotheses, with there being significant variations in the language used by members of different parties. Some of this difference was predicted in our hypotheses but we also find additional dimensions that we had not predicted.
The first part of this page highlights our group's motivations to choose this topic. Next, we formally define key questions and our hypotheses that legislators from the same party will use similar words, with Democrats using more rights-based language and Republicans using more traditional-based language. Then, we explain our data collection process, followed by our data analysis process. Lastly, we highlight our findings and conclusion.
Our group chose this topic area our final project for several reasons. All of our group members were increasingly aware of significant polarization in not only the US Congress (Jeong & Lowry, 2019) but also the wider US society (Iyengar et al., 2019). Having conducted their studies during the COVID-19 pandemic, an era known for what felt like unprecedented political polarisation on multiple policy positions in the US, our group developed a natural interest onto what American legislators have on their minds, whether or not that aligns with their party, and whether or not that contradicts members of other parties.
Some of us also had specific reasons to be interested in this particular topic. For example, one of our members, Maia, is American and wanted to gain further insight into the polarisation in her country, while Sarmad studies the philosophy of language and wanted to learn more about the relationship between words, thoughts and political ideologies. Amara also studies politics for half of her degree. We believe that language is so much more powerful than most people think and that we might not realise how much of an effect it has on us and how it shapes our political opinions. However, people with high amounts of influence, like US politicians, may be aware of its power and therefore chose their words in a way which will influence people towards their personal agendas on big platforms like Twitter. By analysing which words different parties use the most over certain periods, we may be able to see more clearly what they were trying to achieve at that time.
In doing research for this project, we were surprised to learn that such an analysis had, to our knowledge, not been conducted yet. In contributing this project, we hope to add to knowledge about political polarisation in the US and grow awareness about the potentially salient role played by words, language and Twitter in the political landscape.
H1: We hypothesise that Democrats will be more likely to use words such as "green" and "rights".
This hypothesis was informed by the fact that the Democratic party is relatively more focused than Republicans on climate change (Kennedy & Johnson, 2020) and increasing rights for groups such as racial and sexual minorities (Jones, 2020), respectively.
H2: We hypothesise that Republicans will be more likely to use words such as "family", "border" and "crime".
We predict this because Republicans are generally known for their emphasis on traditional family values (Gronbeck-Tedesco, 2022), border security (Oliphant & Cerda, 2022), and crime (Gambino & Greve, 2022).
Data Set: We collected tweets from all of the current Congresspeople's Twitter accounts. Our time-frames were between 7th November to 7th December and 10th December to 10th January. This amounted to almost 3 million tweets. We got the list of Twitter handles from a Excel spreadsheet titled Congressional Twitter Accounts created by the University of California San Diego (UCSD) (Smith, 2022). Our data set is comprised of 223 Democrats (including 4 Delegates) and 215 Republicans (including 1 Delegate and the Resident Commissioner of Puerto Rico), and 3 vacant seats.
This map illustrates the distribution of Congressional representatives throughout all 50 states.
Code Explanation: The code we used to gather our data can be divided into four key sections. First, implementing the Twitter API to make queries. Second, converting the Twitter JSON response to a dataframe. Third, extracting key words from each tweet. Lastly, grouping and counting keywords per user.
First, we implement the Twitter API to retrieve Twitter IDs and pages of tweets:
One of the initial obstacles we had to overcome for this project were limits posed by the Twitter API. The API has three types of access levels. The most basic level allows users to retrieve up to 500,000 tweets per month and have 25 requests per 15 minutes. These limits would hinder our ability to gather the amount of data needed so we decided to apply for elevated access. At this level, we were able to retrieve up to 2 million tweets per month and have 50 requests per 15 minutes. However, since we had to retrieve more than 2 million tweets for our analysis, we had to wait a full month to finish gathering all of them. Additionally the maximum number of tweets per request is 100 and it takes 15 minutes to retrieve 5,000 tweets. To put it into context, the average number of tweets per member of Congress in our data set is 2,842 and most politicians tweeted more than 3,000 for the 30 day time period we used. This means it would take around 10 minutes per legislator. To maximise time-efficiency and avoid reaching the request limit, we decided to use Comma Separated Values (CSV) files to store our data. This would circumvent the need to ask the Twitter API for data we previously requested, as well as re-running the code more than necessary.
Second, we convert JSON to dataframe:
We had to extract the necessary data from the Twitter JSON response by creating a name value pair dictionary.
def get_tweet_dict(tweet, handle, name):
metrics = tweet["public_metrics"]
return {"handle": handle,
"name": name,
"tweet_id": tweet["id"],
"author_id": tweet["author_id"],
"lang": tweet["lang"],
"replied_to": ",".join(tweet['edit_history_tweet_ids']),
'created_at': tweet['created_at'],
'tweet_text': tweet['text'],
'possibly_sensitive': tweet['possibly_sensitive'],
'conversation_id': tweet['conversation_id'],
"retweet_count": metrics["retweet_count"],
"reply_count": metrics["reply_count"],
"like_count": metrics["reply_count"],
"quote_count": metrics["quote_count"]}The JSON response is a tree structure and we needed to create columns per tweet. Therefore, this function created a name value pair dictionary that could be used to create an array of consistent dictionaries to be used creating our panda data-frame.
Third, use spacy to extract key words from tweets
nlp = spacy.load("en_core_web_sm")
nlp.disable_pipe("parser")
nlp.add_pipe("sentencizer")To make the code run faster, we used the sentencizer rather than the default parser since we were only using a limited number of functions from Spacy.
include_types = ["ADJ", "NOUN", "PROPN", "VERB", "ADV"]
def get_tokens(doc):
return [token.lemma_.lower() for token in doc if token.is_alpha and token.pos_ in include_types and token.lemma_.lower() not in exclude_words]A second barrier we faced was that the most frequently tweeted words were primarily prepositions, interjections, and conjunctions, such as "the", "at", and "in". However, these words do not really give us context as to what the Members of Congress are tweeting and thinking about, and do not offer evidence supporting or opposing our hypothesis. To overcome this, we used Spacy's natural language process to extract only adjectives, nouns, proper-nouns, verbs and adverbs. Furthermore, to group past tense, plurals, and similar variables of the same word we used the lemma to extract only the base word. For example, "history", "historical", and "histories" would all be grouped into "history". This would ensure we capture the concepts focused and thoughts expressed by the Members, rather than the particular word used.
exclude_words = ["rt", "amp"]It is important to note that we decided to exclude "rt" because , while it may provide interesting information on the frequency of retweeting in a given month, our project's scope is limited only to the individual words in tweets.
Fourth, we group and count keywords per user and list all of their tweets
def add_word_count(row):
word_freq = Counter(row["key_word_list"])
common_words = word_freq.most_common(50)
df = pandas.DataFrame(common_words, columns = ['Word', 'Count'])
df["handle"] = row["handle"]
return df[["handle","Word","Count"]]The last major step was to group all the keywords by Twitter handle and to gather all the keywords from each tweet into one array to count. Finally, we used a Counter to count the keywords and then find the 50 most frequently used word per legislator, which we used to create a new data frame and csv file.
CSV Files:
Because the CSV files were too large to upload to GitHub, we linked the grouped.csv and tweets.csv files here.
grouped.csv: https://drive.google.com/file/d/1dQA9-0dUVCP86vxsk16WZewj3J7u6yGM/view?usp=drive_web
tweets.csv: https://drive.google.com/file/d/1PgatNy2y8jExcTvWxDlkcaUwN9YTWHFV/view?usp=drive_web
We visualise our results through bar graphs and word clouds.
Below: Figure 1a: The 50 most frequently used words by members of the Democrat Party
Below: Figure 2a: The 50 most frequently used words by members of the Republican Party
Below: Figure 1b: The 50 most frequently used words by Democrats that were not used in a similar frequency by Republicans
Below: Figure 2b: The 50 most frequently used words by Republicans that were not used in a similar frequency by Democrats
Below: Figure 1c: Word cloud of the Democrats' most frequently used words
Below: Figure 2c: Word cloud of the Republicans' most frequently used words
The figures above demonstrate considerable evidence for our hypotheses. In line with H1, we find that Democrats are more likely to use words that align with their party's political ideologies. For instance, as we predicted, we find that the Democrats use rights-based language such as "care" and "protect". Similarly, we find evidence for H2, with Republicans more likely to use words such as "border". However, we also find additional language patterns that we did not protect. For example, Republicans frequently use the word "business", which may be linked to their policies of protecting businesses and the free market (Furhmann, 2022). Similarly, we find the word "woman" as a top word for Democrats, which makes sense given the Democrats' ideological goal of empowering woman by taking policy positions such as reducing the gender pay gap and increasing feamle representation (Horowitz, 2020).
Moreover, another thing we had not considered was the potential usage of language as part of a party's election tactics. For example, we see that 'Biden' is a very frequently used word by the Republicans, and 'Trump' is frequently used by the Democrats, rather than vice versa. This gives us an insight into the parties' approaches to gaining support. It is safe to assume that each side is not using the name of the other's leader in a positive light, so we can only gather than each party adopts the tactic of tearing down the other side in order to gain support from the public. This is not a totally even split, 'Biden' was the number one most frequently used word exclusively by the Republicans, whereas 'Trump' was down at number eight for the Democrats. Noth parties seem to have engaged in this strategy. 'Biden' being the most frequent word for the Republicans may suggest that Republicans may have spent more time (or more words) trying to discuss individuals in the Democratic Party, rather than talking about any one issue which is central to their policies and ideologies.
This usage of language may also tell us something about the users of Twitter. Perpetuated hate speech and echo chambers have been an issue on Twitter for quite some time (Frenkel & Conger, 2022), and some expect this to get worse after Elon Musk's recent purchase of Twitter. Is it the nature of Twitter and it's algorithms which cause this and mean that this tactic of attacking the opposition has been deemed most successful by the two main political parties in the US? Or is it perhaps that these politicians are the part of the cause of Twitter's problem? This may be an interesting topic for further research, however, for now, we have found that US politicians generally see this approach as an efficient enough way to rally support for their party that it is one of the most frequent occurences in their tweets.
Despite our overall considerable evidence, our hypotheses are not fully substantiated. We did not find evidence of the Democrats being more likely to use, for example, climate-focused language, nor did we find the Republicans more focused on crime. This suggests that there may be certain policy areas that are more important or perhaps more polarised that legislators feel a greater need to discuss on Twitter. Alternatively, it may simply be that temporal factors play a role here - for example, closer to a climate-focused protest or event, legislators may discuss climate more on Twitter.
Overall, we saw considerable disparities between the language used by the Democrats as opposed to the language used by the Republicans. Many of our observations aligned with our initial hypotheses concerning the ideologies that each party would be expressing through their language and we also gained insight into the strategic side of politicians' tweets. If we were to continue our investigation, we would be interested to investigate the relationship between the nature of Twitter and the nature of its users (e.g. US politicians, as in this case), in terms of which influences the other more stongly (like the chicken or the egg question) or whether it is a mutual shaping of behaviour.
Maia: Maia created the code to collect the Twitter dataset showing the 50 most common used words per Twitter handles. She also wrote the Index, Motivation, Data Collection, and Bibliography sections of the README.md. Lastly, she created the repository, webpage and added a theme.
Amara: Amara analysed the data, transforming it into multiple more easily useable dataframes. She then created some more easily interpretable bar graphs and wordclouds. She interpreted the data and drew conclusions based on our observations.
Sarmad: Sarmad wrote the Abstract, Motivations, Key Questions & Hypotheses, and Bibliography sections of the README.md, while also copy-editing the rest of it to ensure cohesiveness and consistency. He also did empirical research for all the substantive claims made about the Democrats and Republicans throughout the README.md. Lastly, he converted the README.md to an Index file for webpage creation.
Brush, M. (2010) White House not concerned about new census numbers, Michigan Radio. Michigan Radio. Available at: https://www.michiganradio.org/politics-government/2010-12-21/white-house-not-concerned-about-new-census-numbers (Accessed: January 21, 2023).
Frenkel, S. and Conger, K. (2022) Hate Speech’s Rise on Twitter Is Unprecedented, Researchers Find. Available at: https://www.nytimes.com/2022/12/02/technology/twitter-hate-speech.html (Accessed: January 26, 2023).
Furhmann, R. (2022) Republican and Democratic approaches to regulating the economy, Investopedia. Investopedia. Available at: https://www.investopedia.com/ask/answers/regulating-economy.asp (Accessed: January 27, 2023).
Gambino, L. and Greve, J.E. (2022) Democrats try to flip narrative amid barrage of 'soft on Crime' attack ads, The Guardian. Guardian News and Media. Available at: https://www.theguardian.com/us-news/2022/oct/27/republicans-crime-midterm-election (Accessed: January 21, 2023).
Gronbeck-Tedesco, J.A. (2022) The GOP has revived its 1970s "Traditional family values" playbook, Slate Magazine. Slate. Available at: https://slate.com/news-and-politics/2022/04/the-gops-new-american-family-values.html (Accessed: January 21, 2023).
Horowitz, J.M. (2020) Wide partisan gaps in U.S. over how far the country has come on gender equality, Pew Research Center's Social & Demographic Trends Project. Pew Research Center. Available at: https://www.pewresearch.org/social-trends/2017/10/18/wide-partisan-gaps-in-u-s-over-how-far-the-country-has-come-on-gender-equality/ (Accessed: January 27, 2023).
Iyengar, S. et al. (2019) “The origins and consequences of affective polarization in the United States,” Annual Review of Political Science, 22(1), pp. 129–146. Available at: https://doi.org/10.1146/annurev-polisci-051117-073034.
Jeong, G.-H. and Lowry, W. (2019) “The polarisation of energy policy in the US congress,” Journal of Public Policy, 41(1), pp. 17–41. Available at: https://doi.org/10.1017/s0143814x19000175.
Jones, B. (2020) Democrats far more likely than Republicans to see discrimination against blacks, not whites, Pew Research Center. Pew Research Center. Available at: https://www.pewresearch.org/fact-tank/2019/11/01/democrats-far-more-likely-than-republicans-to-see-discrimination-against-blacks-not-whites/ (Accessed: January 21, 2023).
Kennedy, B. and Johnson, C. (2020) More Americans see climate change as a priority, but Democrats are much more concerned than Republicans, Pew Research Center. Pew Research Center. Available at: https://www.pewresearch.org/fact-tank/2020/02/28/more-americans-see-climate-change-as-a-priority-but-democrats-are-much-more-concerned-than-republicans/ (Accessed: January 21, 2023).
Milroy, L. and Margrain, S. (1980) “Vernacular language loyalty and Social Network,” Language in Society, 9(1), pp. 43–70. Available at: https://doi.org/10.1017/s0047404500007788.
Oliphant, J.B. and Cerda, A. (2022) Republicans and Democrats have different top priorities for U.S. immigration policy, Pew Research Center. Pew Research Center. Available at: https://www.pewresearch.org/fact-tank/2022/09/08/republicans-and-democrats-have-different-top-priorities-for-u-s-immigration-policy/ (Accessed: January 21, 2023).
Smith, K.L. (2022) Libguides: Congressional twitter accounts: Home, Home - Congressional Twitter Accounts - LibGuides at University of California San Diego. University of California San Diego. Available at: https://ucsd.libguides.com/congress_twitter (Accessed: January 21, 2023).
Van Swol, L.M. and Kane, A.A. (2018) “Language and group processes: An integrative, interdisciplinary review,” Small Group Research, 50(1), pp. 3–38. Available at: https://doi.org/10.1177/1046496418785019.