Skip to main content
Search form
  • 00:00


  • 00:09

    RACHEL BLUM: I'm Rachel Blum.I'm a professor of political scienceat Miami University of Ohio, where I'm in my third year.I teach political parties, methods, and other fun stuff.My biggest project so far is on the Tea Party--that's what my book manuscript has been on.

  • 00:32

    RACHEL BLUM [continued]: And a lot of the projects I'm doing right nowspin off of that.So it's usually something to do with party factions, partyideology.And I stumbled into data science through that project.So it came out of my dissertation,which is a multi-year process.And when I first started thinkingabout this as a dissertation, I was in a different field

  • 00:54

    RACHEL BLUM [continued]: within political science.I was a political theorist.So we study texts and the history of ideas.And I did that because I was bad at math--like, never even really learned math bad at math.And so I just wanted to work with ideas.And as I was doing that, I realized that you can't justtalk about ideas if you're dealing

  • 01:14

    RACHEL BLUM [continued]: with a contemporary movement.So I try to do interviews, but you can only do so many.And then I realized that the way to trace the things I was trulyinterested in-- which were relationships among Tea Partygroups and between Tea Party groups and other elites,and patterns in what they were talking about in their text--

  • 01:35

    RACHEL BLUM [continued]: that these kinds of things actuallyrequired more computational techniques-- which, of course,turns out to be data science.So I ended up going from somebodywho did zero math, zero stats to someone whoalmost entirely uses what we'd call computational methods.So most of my methodology is using network analysis.

  • 01:57

    RACHEL BLUM [continued]: So tracing relationships among actors or text analysis.Trying to find ways to use a computer to help summarizewhat's going on in texts.Or to-- if I know what I'm looking for in a set of texts--replicate that in another set.

  • 02:22

    RACHEL BLUM [continued]: The original purpose of the researchwas to answer a substantive question, whichis still how I always start.Substantively, I wanted to know whatthe relationship was between the Tea Party and the RepublicanParty.So in the US, if you are a conservative-- there's alreadya party that ostensibly does all the things you want.So why would you mobilize as this faction that

  • 02:45

    RACHEL BLUM [continued]: then primaries your party and tries to ousttheir congressional leaders?What does that mean?And what does that look like?And that was the original purpose,which, of course, spirals in all these other directions.I used several different research methods

  • 03:07

    RACHEL BLUM [continued]: in this project.The beginning one was actually doing interviews.I would contact people, and follow through on connections,and show up places with consent forms and a voice recorder.And try to get people to tell me what was going on.And then based on that, I started branching offin other directions.

  • 03:28

    RACHEL BLUM [continued]: So the next method I used was I conducted a surveyand I did some basic analysis of that survey.I didn't think that was enough so thenI created a list of all the Tea Party groups that had websites.Then I scraped their web pages using webscraping techniques to get a list of everyone they linked toon their website.So every outgoing hyperlink.

  • 03:49

    RACHEL BLUM [continued]: Then I constructed a network of thatand that was kind of the first computational thingI ever tried to do.I wanted to see of all the Tea Party groups, whatnews outlets, what political figures arethey constantly linking to?Are they the same?Or are they different than our typical, like, Fox News,Wall Street Journal is some kind of conservatism?

  • 04:12

    RACHEL BLUM [continued]: Is it a different set?And based on that, I ended up branchinginto a set of other methods.I did everything from geospatial analysis--plotting the locations of groups--to scraping their blog posts and doing a text analysisto uncover what the topics were they were actually discussing.

  • 04:33

    RACHEL BLUM [continued]: And the language they were using to discuss those topics.Many unexpected challenges in this.Given that I was scared of all things statsfor the beginning part of this--I went from someone who didn't know any stats to someone

  • 04:54

    RACHEL BLUM [continued]: who had to learn the statistical program R. And not only that,I wasn't trying to just run regressions or things thatare more typical.I was trying to write scripts thatwould download entire webpages, that would extract strings,that would convert formats, that could iterate through 48,000

  • 05:15

    RACHEL BLUM [continued]: blog posts to try to find a theme.So the challenges ended up being that every time Ilearned a new thing, I realized that therewere about five extra things I could do with that.And I had to learn those, too.And these methods are now more acceptedor at least, something like this, peopleare using them probably better than I ever could.

  • 05:38

    RACHEL BLUM [continued]: But at the time, there weren't a lotof resources for those methods.So I was teaching myself how to program in ways that I'd neverimagined one could program.So if a student were trying to do a similar piece of research,

  • 06:05

    RACHEL BLUM [continued]: I would have a lot of different recommendations.The first might be if you think that you want to describe,like, a major phenomenon in a party systemrealize that's going to be a whole ton of work.So maybe I would just caution that student firstabout the scope of their project.But the biggest thing that I think I learned from this

  • 06:27

    RACHEL BLUM [continued]: and that I use in my future work isthat you shouldn't let yourself beinhibited by existing structured data sets.And what I mean by that is that data comes in a lot of format,obviously.You have the nice curated spreadsheetsyou might get from survey results.Everything's already in a column,you can run regressions-- it's great.

  • 06:50

    RACHEL BLUM [continued]: But a lot of data are not formatted like that.So you may be interested in who's retweeting whomor you may be interested in prioritiesin congressional press releases.And you can't get these in pre-formatted data sets,but you can probably conceive a place wherethey exist in the hypothetical.

  • 07:10

    RACHEL BLUM [continued]: Like, well, I guess I would need to collect all the tweetsor the press releases.These kinds of methods--the computational methods-- make it possiblefor you to get data they really aren'tin any kind of analyzable format and make them analyzable.So my biggest piece of advice for someonewould be not to stop and only use

  • 07:33

    RACHEL BLUM [continued]: the data that's easy to use.If you have a big question you want to answer,find the data that actually would answer that questionbecause there probably is a creative way to analyze it.The US party system is really unique

  • 07:55

    RACHEL BLUM [continued]: and we all kind of realize this in a certain way.We only have two parties--that annoys a lot of people, but we can'tseem to do anything about it.Well, that's structural.We have districts where there's only one member--a single member district-- and theyhave to just get the plurality of the vote.This works against multiple parties.And what that means is that if there's ever a dissident group,

  • 08:18

    RACHEL BLUM [continued]: if there's ever someone who's like the big umbrellaof conservatism doesn't capture what I believe--they have no real political outlet.They could be a third party, but theyknow that they're not going to achieve anything that way.So what I was trying to argue is that there is this biggerhistorical trend of a faction mobilizingwithin a major party with the goal

  • 08:40

    RACHEL BLUM [continued]: not of just getting their own candidate-- so say, like,the Sanders Democrats, they just wanted a candidate.Not even with the goal of just shifting their party--we can think maybe the Christian rightwith the Republican Party--but with the goal of completely reshaping their party.And so I called that a party within a party.It's a faction that wants to take over its existing party

  • 09:02

    RACHEL BLUM [continued]: because it thinks that's the only way it can have influence,win office, and make policy.So I explained the Tea Party as the most recent exampleof this, but I think it also could illuminatesome weird historical episodes, like,when the Southern Democrats triedto take over the Democratic Party right after the New Deal.So we've seen a lot of this in the US.

  • 09:23

    RACHEL BLUM [continued]: And it can help us understand how ideological diversity canexist within a party, even without havingthis separate official outlet.So one of the spin-offs from that projectis a paper that I'm working on moving towards publication.

  • 09:46

    RACHEL BLUM [continued]: And it involves looking at whether there actuallyis a distinct faction within the Republican Partythat starts with the Tea Party, thatmay be leading into Trump and a different styleof conservatism.A lot of the Tea Party groups werelinking to local resources on their website.The county commissioner-- that kind of stuff,

  • 10:06

    RACHEL BLUM [continued]: just for their members.But there were some national thingsthat they really converged on.So most of the groups linked to a setof five to 20 elite figures and theseinclude some the national Tea Party groups, like,Freedom Works.They also included Breitbart, Glenn Beck's TheBlaze,

  • 10:27

    RACHEL BLUM [continued]: Michelle Malkin--these kind of voices that were a little bit moreon the outside of traditional conservatism.We're not seeing the National Review, The Wall StreetJournal in here.We're not seeing Republican elected officials.We are seeing what became the alt right--people who Trump later staffs his administration with.And then if you expand that a little bit, think well

  • 10:51

    RACHEL BLUM [continued]: if they're really looking at these different figuresin their network, this alternative conservativevoice--do we see evidence of that somewhere else?Are they listening and taking that in?So that's where I looked at the blog posts I scraped.I had about 48,000 blog posts from the local Tea Partygroups.And so I used just a basic unsupervised text analysis

  • 11:14

    RACHEL BLUM [continued]: method.I used a model called LDA, which,essentially, goes over a corpus or a body of text.And it asks, well, what's the probabilitythat this set of words would occur togetherbased on all the words?We call that a topic--a distinct group of words.And usually they have substantive interpretationsyou can make of them.

  • 11:34

    RACHEL BLUM [continued]: So like one of the topics would be things like tax, payroll,cut, business.Based on those words, I'd say that we're picking upon the fact that somewhere in those blog poststhey're talking a lot about taxes.And I grouped together the different policy topicsthat emerged and of course, they have some of that focus--the traditional Republican focus.

  • 11:56

    RACHEL BLUM [continued]: But a lot of what they ended up talking aboutseemed very much like what we later hear Trump talking about.There is a topic that had words like make, America, great--I call it American greatness.There was a topic about immigration-- immigrants beingthreats.There's a topic about guns.There are a lot of topics about elitesand them being distrustworthy.

  • 12:18

    RACHEL BLUM [continued]: So we get this picture of this alternative threat-centric kindof conservatism emerging, really predatingTrump-- my data's 2009 to 2015.So we can use these methods like social network analysisand text analysis to get at bigger substantive questionsabout the nature of ideologies and party evolution.


Rachel Blum, PhD, Professor of Political Science at Miami University of Ohio, discusses her research of Tea Party groups using network and text analysis, including current research interests, the goal of the current research, methods used to collect and analyze data, challenges faced and overcome, advice to students wanting to do this type of research, key findings of the research, and future research on the topic.

Looks like you do not have access to this content.

Researching Tea Party Groups Using Network Analysis and Text Analysis

Rachel Blum, PhD, Professor of Political Science at Miami University of Ohio, discusses her research of Tea Party groups using network and text analysis, including current research interests, the goal of the current research, methods used to collect and analyze data, challenges faced and overcome, advice to students wanting to do this type of research, key findings of the research, and future research on the topic.

Copy and paste the following HTML into your website