Intro
Melvin Carter and Jacob Frey were inaugurated as Mayor of St. Paul, MN and Minneapolis, MN respecitvely in January 2018. Both mayors stressed increasing affordable housing and improving police-community relations as major elements of their campaign platforms.
Nearly one year into their mayoral terms, we are examining a snapshot of tweets and news content to try to find out:
What do Twin Cities residents care about?
How do people think the two new mayors are doing?
To get a sense of how people have been reacting to Jacob Frey and Melvin Carter, we collected all tweets from November 24th, 2018 to December 13th, 2018 that included the Twitter handle of either of the mayors. This meant collecting all tweets that included @Jacob_Frey or @MayorCarter_
In exploring the Twitter data, one recent incident stood out in particular. In late November, Minneapolis police officers in the city’s fourth precinct put up a christmas tree decorated with garbage and found items that are stereotypical of People of Color. Additionally, the officers wrapped the tree with crime scene tape as opposed to tinsel. This racist act quickly became a nationwide story and residents from the Twin Cities and beyond offered their opinions on a variety of media platforms including Twitter and the comments section of the Star Tribune (a major Twin Cities newspaper).
Using the Twitter API, we were only able to collect tweets from the few days immediately preceding each API call. To get a better understanding of what issues are talked about during a regular news cycle, we looked at the headlines of articles in the Star Tribune whose main text mention either mayor. These articles were published between mid-June 2018 and mid-December 2018. We also examined comments on articles published during the same period as our Twitter data to compare sentiments across different media platforms.
Star Tribune
The word clouds below give us an idea of what sorts of issues have been written about in the Star Tribune over the past six months. These word clouds highlight the most popular words in headlines of articles that mention either Mayor Jacob Frey or Mayor Melvin Carter. The combined word cloud in the colors of the Star Tribune show that common topics between both mayors include police, city council, the mayors themselves, and issues such as homelessness and housing.
Looking at Jacob Frey in particular, we see that “police” is far and away the most popular word in article titles, followed by (city) council, and homelessness. Words related to the racist tree incident do show up, but they are less prevalent since this cloud presents a broader temporal range.
The word cloud of headlines whose body text includes Melvin Carter is dominated by the words “mayor,” “city,” and “council.” Looking beyond these words that are common to both cities, we see that the issue of $15 minimum wage is a major topic in article titles over the past six months. The closeness of Minneapolis-St.Paul is clearly illustrated in this word cloud of St. Paul, as the word “Minneapolis” is one of the more popular words.
The bar plots below provide an alternative visualization to the word clouds above. These plots highlight the 12 most prominent words in the titles of articles whose main text mention either Melvin Carter or Jacob Frey.
We decided to gather Star Tribune comments and compare them with our data from Twitter. We wanted to compare the sentiments expressed on these different media platforms. The Star Tribune’s website format for comments was not compatible with Selector Gadget, so instead, we copy-pasted the comments into different documents and saved them in a plain text format.
We analyzed comments on articles posted on the Star Tribune from November 24th to December 13th because these were the days we scraped tweets. We chose the four articles that mention Jacob Frey with the most comments and the three articles that mention Melvin Carter with the most comments. It’s important to note that articles that mentioned Jacob Frey overall had more comments than those that mention Melvin Carter. If we add the number of comments of the top 3 most-commented articles that mention Jacob Frey, there is a total of 813 comments. If we add the number of comments of the top 3 most-commented on articles that mention Melvin Carter, there is a total of 283 comments.
This word cloud shows the most frequently used words in comments on an article titled “North Minneapolis police commander demoted after Christmas tree controversy.” We made this word cloud to explore how people were interacting with the article – what they were talking about. The larger the word, the most often it was used in the comments. So, from this visualization, we can gather that many comments mentioned the words “police,” “officers,” and “fired” as well as “mayor.” It seems that people were expressing their opinions on the decision to fire the police officers involved in decorating the tree and mayor Jacob Frey’s support in that decision.
We created this bar graph to explore the sentiments most frequently expressed in the comments. This visualization shows the number of occurances of each sentiment. It’s important to note that many words have multiple sentiments attached to them; for example, some words have both “fear” and “negative” in the nrc lexicon. From this bar graph, we can gather that “trust” was a commonly expressed sentiment as well as “anger” and “fear.” Surprise is the least commonly expresed sentiment. Negative and positive sentiments are almost equal.
This word cloud illustrates the words mentioned in comments on the article titled: “Vido of offensive tree at Minneapolis police Fourth Precinct headquarters is being reviewed.” One limitation of this visualization is that we don’t see phrases – the words are taken out of context. When we see the word “racist,” we don’t know whether someone is saying they think the tree is racist or they don’t think the tree is racist. What we can draw from this visualization is that the larger words were frequently used in conversation. It’s interesting that the words “offended” and “privilege” are fairly large. It seems like people are having conversations about the privilege on this media platform. It’s also interesting that the word “shooting” is so large since this article was not about a shooting. It seems that people are talking about racism in general, not just the Christmas tree incident.
This visualization shows the frequency of the words used in the comments on the Star Tribune article titled: “Homeless camp residents, American Indian leaders discuss tensions, plans to move this week.” It’s interesting that “Frey” is so frequently mentioned – people seem to be invoking him often when talking about this issue.
This visualization shows the frequency of certain words in the comments on a Star Tribune article titled: “Minneapolis City Council right to approve a municipal ID program.” This word could is interesting because it looks much different than the word clouds from our twitter data. It seems that people have not tweeted as much about this issue as they have commented on this article about this issue.
This visualization shows the most frequently used words in comments on a Star Tribune article titled: “St. Paul council set to approve amnesty for overdue book fines.” We think this wordcloud really illustrates the limitations of this visualization because it’s hard to tell how people feel about this new policy. The only words that give us a sense of public opinions are “accountable;” perhaps commentor’s feel people should be held accountable for their fines.
This visualization shows the frequency of words used in comments on the Star Tribune article titled: “Property taxes climbing, but needs are many, activists say.” This word cloud shows that the comments are generally very on topic – the most used words “property,” “taxes,” “tax,” “money,” etc. clearly fit that article. Perhaps words like “income” and “affordable” describe the many needs the article’s title mentions.
This visualization shows the frequently used words in comments on a Star Tribune article titled: “St. Paul City Council reaches deal on Mayor Melvin Carter’s budget, with double-digit tax levy increase.” It looks very similar to the other article on tax increases, but in this word cloud, “schools” and “school” are larger. This makes sense since this article is on other tax increases besides property taxes.
In comparing the two platforms – Twitter and the Star Tribune – we wanted to see if there are certain issues people seem to tweet more or comment more about. We compared the number of comments on the 3 most commented on articles with the number of tweets front those days. To do this, we made an two excel sheets with the number of comments for the top 3 most commented articles for each mayor and the number of tweets at each mayor on the days the articles were published.
These visualization shows the number of comments on 3 specific articles compared to the number of tweets at the respective mayors on the days the articles were published. It’s significant that the only time there were more tweets at one of the mayors than comments on a specific top-commenting article is on December 4th for Jacob Frey. The article that came out on December 4th was about the racist Christmas tree incident; it seems that people tweeted about this issue more than other issues that were also hot topics on the Star Tribune.
The tf-idf plot below shows the most prevalent and unique words in both the Star Tribune over the past six months and in tweets over several weeks in November and December of this year. Looking at this plot, we see that the words “racism” and “white” are the two most prevalent unique words on Twitter whereas the most unique and prevalent word in the Star Tribune is “15” (for the $15 minimum wage). While this plot certainly provides a nice comparison between two types of media, it is important to keep in mind the different time periods of the data. In future, a more comprehensive plot would include Twitter data that matches the time period we scraped Star Tribune data from.
In this graph, we can see that overall, Jacob Frey received many more mentions on Twitter than Melvin Carter did during the time period observed. This is likely due to multiple factors. Minneapolis, with roughly 420,000 residents, is more populous than St. Paul, which is home to 300,000 people. Assuming that people are most concerned with the actions of the mayor of their own city, we could expect more tweets at Jacob Frey in general. However, the explosion of twitter mentions of Jacob Frey seen during the range of dates we observed corresponds to a particular incident that got many twitter users talking across the Twin Cities and beyond.
We can see that mentions spike around December 2, when news coverage of the racist Christmas tree event in Minneapolis’ fourth precinct was also escalating. The large number of tweets mentioning Jacob Frey around this time, and their strong focus on the Christmas tree incident, largely defined the results of our sentiment analyses.
Sentiment of Tweets
The word cloud below displays the most common words in tweets that mention Jacob Frey from November 24th, 2018 to December 13, 2018. This cloud shows the 300 most frequent words, all of which have a minimum frequency of five occurrences. The recent scandal in Minneapolis’s 4th police precinct features heavily in this word cloud – words such as “racist” and “tree” are the most popular. Additionally, issues such as housing affordability appear to be prevalent in tweets that mention Jacob Frey.
The word cloud below displays the most common words in tweets that mention Melvin Carter from November 24th, 2018 to December 13, 2018. Compared to the Jacob Frey word cloud, this cloud appears quite positive – some of the most popular words include “happy” and “community.” This visualization uses the same criteria as the previous word cloud.
The wordcloud below displays the most common words in tweets that mention either Melvin Carter or Jacob Frey. Comparing this visualization to the previous two, it is evident that the number of tweets mentioning Jacob Frey far outnumbers those that Mention Melvin Carter. This cloud has some strong similarities to the Jacob Frey word cloud with words such as “racist” and “tree” featuring prominently.
This graph shows the tf-idf scores of words used in tweets that mention Melvin Carter or Jacob Frey. Tf-idf scores convey words that are both prevalent in tweets and unique to tweets at each individual mayor. For example, the word “racist” is very prevalent in tweets at Jacob Frey and not prevalent in tweets at Melvin Carter. Many of the Jacob Frey’s words, such as “racist”, “tree”, “Christmas”, “racism” and “white”, are probably in reference to the racist Christmas tree decorated by police officers in Minneapolis’s fourth precinct.
We wanted to understand the different emotions and overall attitude towards Melvin and Jacob, so we decided to do a broad sentiment analysis of all tweets. After cleaning up all the tweets we removed common stop words and joined the words to the NRC sentiment lexicon. This chord diagram shows a normalized sentiment percentage by sum of words per mayor so Frey’s high comparative number of tweets wouldn’t crowd out trends in tweets at Melvin. It shows that tweets at Jacob usually contain words that are more commonly associated with emotions like anger and disgust while tweets at Melvin usually contain more words associated with joy, anticipation, and trust. While this gives us a better understanding of how tough it might be a communication staffer for Jacob, we should take all sentiment analyses with a grain of salt. In the NRC lexicon the word “white” would be categorized as positive while “black” was categorized as negative. Especially with Melvin being the first Black mayor in St. Paul history, this racist categorization is a major limitation of our sentiment analysis.
This graph shows the sentiment of the most favorited tweets that mentioning each mayor. To make this graph, the top 25 most favorited tweets mentioning Jacob Frey and the 25 most favorited tweets mentioning Melvin Carter were gathered. The text of these tweets was then analyzed, characterizing each word of the tweets according to its sentiment. We can see that overall, the most favorited tweets mentioning Jacob Frey overwhelmingly reflect more negative sentiments, such as disgust and sadness, than those mentioning Melvin Carter, which reflect sentiments such as anticipation and joy to a greater degree.
Who’s Tweeting
These are maps of where people in the US are tweeting about Frey or Carter. Because location settings are frequently hidden, a joke (“i.e. “the swamp”), or impossible to geocode (“downtown”) we were limited to plotting 61% of tweet locations. We hand cleaned location name then used the free online geocoder “Geocodio” after grouping and adding the number of tweets by location. While the majority of tweets for both mayors are unsurprisingly tweeted out from the Twin Cities, Frey got a lot more attention from out of state tweeters. More than a third of geocoded tweets at Frey were from outside of Minnesota, most of them retweets or quoted tweets about the racist christmas tree incident. The 32 tweets at Melvin Carter from out of state tweeters were mostly positive and excited about his appointment. These maps show how the racist Christmas tree incident went viral across the country.
This density plot shows the account creation date for users who tweet at Melvin Carter and Jacob Frey. Most accounts that tweet at both mayors were created in 2009. 2012 and 2018 also saw increases in the creation of accounts that mention Jacob Frey. Creation of accounts that mention Melvin Carter has generally decreased over time. Twitter was created in March 2006.
This histogram shows the number of tweets that unique users tweet. It appears that for tweets at both mayors, most users tweet once or twice. However, this graph is a bit hard to read because one user, @TheRealGOP, tweeted 116 times. The second highest number of tweets per single user is 50, which is less than half that of @TheRealGOP.
We decided to look more into the @TheRealGOP account to see if they were a real person or a twitter bot. Twitter bots’ tweets usually sound like gibberish and don’t make much sense, so we predicted that @TheRealGOP is most likely a real person. Here is a sample of tweets from @TheRealGOP:
## [1] "@VoteHealth @vankapro @myserenity69 @Slate @Jacob_Frey @MayorCarter_ Incredible that the government prevents people from using their own property as they see fit. The rule should be the state shouldn't step between commercial acts between consenting adults."
## [2] "@VoteHealth @vankapro @myserenity69 @Slate @Jacob_Frey @MayorCarter_ Incredible that the government prevents people from using their own property as they see fit. The rule should be the state shouldn't step between commercial acts between consenting adults."
## [3] "@northxnorthside @jeremiah4north @CunninghamMPLS @Jacob_Frey Funyons are racist. Fight the power! #Funyongate (Cheetos, we're coming for you!)"
## [4] "@Jacob_Frey @deanbphillips Please take a moment away from Hanukkah to declare various Christmas trees racist. We will all be thankful."
## [5] "@Jacob_Frey showed he is racist by assuming those products were just for black people. Apparently us crackers don’t eat Popeyes, drink beer, smoke menthols. Assuming only black people do is way more racist than the act itself. Gotta love having a racist mayor."
## [6] "@HaileyHenjum @Jacob_Frey Because Frey doesn't want to hear anyone's opinion. He's a typical liberal politician. Minneapolis is just a stepping stone for him. He's vacuous like them all."
## [7] "@webster @Jacob_Frey @MinneapolisPD Did MPD ever catch the Somali man who randomly stabbed the woman in Uptown 14 times? Or the pizza delivery driver shot in the face? Glad they are prioritizing Funyans on the Yule tree over those trivial things."
## [8] "@emmdee22 @Jacob_Frey Not just tents. Also free needles son they can shoot heroin. Frey probably also thinks getting high on the taxpayer dime is a right."
## [9] "@webster @Jacob_Frey @MinneapolisPD Minneapolis declares that Funyans are racist. Good job guys. Keeping us safe."
## [10] "@kurta59 @northxnorthside @jeremiah4north @CunninghamMPLS @Jacob_Frey Leftists don't care about logic. It's all about power for them."We wanted to see what our unique users data would look like without tweets from @TheRealGOP and @Q_SpecialForces, whose high tweet counts were skewing our data. Again, this graph shows that for tweets at both mayors users are most likely to tweet once or twice.
Conclusions
This two week snapshot analyzing tweets at Melvin Carter and Jacob Frey showed us how a single event can sway public discourse. While most online users seemed to reflect positive sentiments towards Carter, the racist Christmas tree incident made Frey’s twitter a battleground. Tweets directed at the Minneapolis Mayor went viral across the country and even attracted the attention of a very active troll. From twitter to the Star Tribune, relatively positive news involving the mayors like affordable housing and municipal IDs initiatives were out commented by negative comments about the racist Christmas tree. Even if the tweets weren’t explicitly directed at Frey, it reflects poorly on his administration and dominates multiple online mediums. They say all news is good news, but this was definitely not the case for Frey.
Data Limitations & Future Research
One major limitation of our research is the limited range of dates that we were able to gather Twitter data for. We encountered difficulty trying to gather data beyond the few days immediately preceding each API call using Twitter’s API. It would be interesting to trace how Twitter trends may have changed over the course of the mayors’ campaigns, first year in office, and going forward into the future.
For this research project, we were able to examine Star Tribune article headlines over a several month period, and were also able to analyze the comment sections of several relevant articles. In future research, it would be interesting to anayze the content of articles themselves as well.
Another limiation of our research methods has to do with the sentiment analysis methods available to us. Given that text is analyzed word by word, differentiation between a negative word such as “racist” and a positive phrase such as“not racist” would be obscured. Bias within the sentiment analysis packages we used also reflected bias and lead to misleading categorizations of certain terms, such as “Black” and “White.”