Immigration Reform Power Tweeters: Part 2

By Randy Lynn and James Witte

Beginning in February 2013, researchers at the Institute for Immigration Research (IIR) at George Mason University began tracking the national immigration reform conversation on Twitter. In Part 1, we revealed that our analysis had uncovered five clusters of power tweeters (see Figure 1): (1) Mainstream, (2) Reformers, (3) Progressive Media, (4) Institutional Conservatives, and (5) the Tea Party. In this article, we’ll delve further into our data to look at the similarities and differences between these groups, and present several additional findings.

Figure 1: Clusters of Immigration Reform Power Tweeters


To begin, we examine two broad measures of influence to see if any of these clusters is more or less influential than the others. We then look more deeply at the characteristics of the tweeters in each cluster — beginning with the length of time on Twitter, the number of tweets, and the number of followers and others followed — to better understand the nature of their influence. Finally, we examine the relationships between the power tweeters in each cluster, to look at how relationships among themselves contribute to each group’s overall role in the immigration reform conversation.

The basis for this analysis is found in the statistics reported in Figure 2. For each statistic, asterisks are used to indicate when the value for a cluster is statistically different from the value for the Tea Party cluster.

As an overall measure of influence, we can compare the average Klout scores for the power tweeters making up each cluster. A Twitter user’s Klout score is a number ranging from 1 to 100 and is widely accepted as an indicator of an individual’s online influence.

As Figure 2 indicates, the average Klout score for Tea Party members is lower than those for each of the other clusters. In all but one case (the Reformers), this difference is statistically significant.

Klout scores are representative of a tweeter’s overall influence, but the measures found in the next two rows in Figure 2 — betweenness centrality and impact — are based specifically on our analysis of tweets containing the word “immigration.”

  • Betweenness centrality captures the load and importance of an individual tweeter. It indicates how well-positioned the tweeter is in the overall immigration conversation. While there are relatively large differences between the clusters (e.g., 731 for the Mainstream compared to 516 for the Tea Party) these differences are not statistically significant, because there are also large differences within each cluster.
  • Impact is based on betweenness centrality, and refers to the average ranking (first through tenth) of members of each cluster when a member is among the top ten tweeters in term of betweenness centrality in a given Twitter snapshot. As described in Part 1, we used the top ten tweeters by betweenness centrality over several hundred snapshots in a four month period to generate our group of “power tweeters.”


With our measure of impact, a low value indicates greater influence. As you can see in Figure 2, the impact score for members of the Mainstream group was on average about 25 percent more influential than power tweeters in other groups, including the Tea Party.

Figure 2: Measures of Influence for Five Clusters of Immigration Tweeters


To better examine the relative influence of these clusters, Figure 3 provides additional indicators of the Twitter presence of the members in each cluster. First, with the number of days on Twitter, we see that on average influential members of the Tea Party cluster are newer to Twitter than members of the other four clusters. Then, in the following row, we note that the number of tweets from the Tea Party cluster members is roughly similar to those from the other clusters-with the exception of the Reformers, who tweet significantly less often than all other clusters.

We found important differences between the Tea Party cluster and other clusters, however, in the next two rows of Figure 3, which provide the average numbers of Twitter followers and users followed by the members of each cluster. Here, too, we see an important difference in how these groups seek to become influential.

The Mainstream, Reformer, Progressive Media, and Institutional Conservative clusters all amassed huge numbers of followers, but didn’t do much following of others by comparison. The average number of followers among these four groups ranged from approximately 165,000 to 867,000, while the average number of follows ranged from approximately 5,000 to 13,000.

By contrast, the Tea Party power tweeters had an average following of 47,000, far fewer than any of the other groups–but they averaged nearly 35,000 follows themselves, a figure three to seven times greater than the other groups.

One reason for this difference is a movement among conservative tweeters referred to as #tgdn, or the Twitter Gulag Defense Network, in which, according to the founder Todd Kincannon, “conservatives find and follow each other in order to defend against spam blocking liberals who get conservative Twitter users suspended.” This strategy builds on the premise that Twitter is less likely to suspend a user with several thousand followers.

Figure 3: Measures of Influence for Five Clusters of Immigration Tweeters


This suggests that the influence of the Tea Party group is largely attributable to activity within its own cluster. Support for this interpretation may be seen in Figure 4. The clustering coefficient indicates that the members of the Tea Party cluster are significantly more closely tied to one another. In other words, this cluster is “denser” than all other clusters of influential immigration tweeters.

In addition, the reciprocity ratio found in the final row reveals that members of the Tea Party cluster are significantly more likely to both follow and be followed by one another than members of all other clusters.

Figure 4: Measures of Influence for Five Clusters of Immigration Tweeters



Based on the findings presented here, when it comes to the immigration reform discussion, one may be tempted to treat the cluster of influential Tea Party tweeters as a tempest in a teapot. Though tweeting at a pace comparable to members of other clusters, they do not have much influence on the overall Twitter conversation on immigration. But we believe that dismissing this apparent internal soliloquy as wholly unimportant would be premature.

First, the tweets we have analyzed here only cover one slice of time, between February and June of 2013. These tweets were made as the Senate was successfully passing a comprehensive immigration reform bill, prior to Speaker of the House John Boehner’s July 2013 statement that he would refuse to put the Senate bill up for a floor vote by House members unless a majority of Republicans would support it. In short, it was a time when the Reformers and Progressive Media were on the “offense” with regards to immigration reform, while a significant number of conservatives were on “defense.”

Second, this analysis just looks at power tweeters, not the whole of the discussion about immigration reform on Twitter. It is possible that the cumulative impact of many individuals expressing the same views as the Tea Party power tweeters could actually have more impact than that of the subset of most influential tweeters in the Tea Party cluster. This would require a different type of analysis to assess.

Finally, and perhaps most importantly, unlike Vegas, what happens on Twitter does not necessarily stay on Twitter. Tweeters with a Tea Party orientation are likely to have an impact on the immigration reform discussion quite apart from the conversation on Twitter. Being active on Twitter does not preclude offline forms of activism and influence. In fact, having a significant social media presence is increasingly a means to claim credibility in offline arenas. Our analysis of influential Twitter networks does not capture this sort of influence.

It remains to be seen how the interplay of online and offline dynamics might affect the ultimate outcome of immigration reform. We hope to investigate this in more detail as the movement — or lack of movement, as seems to be the situation for now–toward immigration reform continues to develop. Social scientists have long been interested in public discussion around politics and policy. The difference today, when much of this discussion takes place in public digital spaces, is that the conversation leaves plainly observable digital footprints. These are the tracks we will trace as we hope to better understand the voices of the individual actors and networks that promote or hinder immigration reform.

This post originally appeared on the Huffington Post, December 9, 2013.

To see part one of this post click here.

Immigration Reform Power Tweeters: Part 1

By Randy Lynn and James Witte

Beginning in February 2013, researchers at the Institute for Immigration Research (IIR) at George Mason University began tracking the national immigration reform conversation on Twitter. Our view: Twitter would be one of several social media venues where the evolving immigration reform debate could be tracked and analyzed. Since then, using DiscoverText and NodeXL, two different data mining applications, we’ve collected nearly 3 million tweets containing the word “immigration.”

With a lot of data at our fingertips, we sought to identify who were the most influential tweeters about immigration reform, and how they interacted with one another. To do this, we identified the most influential daily tweeters about immigration between March and May 2013, as assessed by betweenness centrality — one of several important statistical measurements of influence within a social network. We then looked at their connections with one another as of June 2013, using a Clauset-Newman-Moore algorithm — a complex formula that sorts similar actors in a network into groups.

Below we reveal and describe five clusters of immigration reform “power tweeters.” In a related post, “Immigration Reform Power Tweeters: Part 2,” we delve further into the data to note some of the major similarities and differences between these clusters.

Cluster 1: The Mainstream

The first cluster, pictured in the image below, is dominated by mainstream news organizations. Also included in this group are significant numbers of celebrities and influential online commentators. John McCain, a well-known, pro-reform Republican, is the only politician included.

The primary shared attribute for members of this group is their influence in a broad range of topics. Not all of them tweet very frequently about immigration reform — but when they do, their tweets are particularly influential, because these tweeters have many Twitter followers, and therefore have high visibility. Because of this, they are influential no matter what they tweet about.


To better understand the composition of this group, we created a word cloud based on the words these tweeters use in their Twitter profiles. Most describe themselves as “news” sources on a wide variety of topics. Note that the word “immigration” does not stand out in these tweeters’ profiles.


Cluster 2: The Reformers

The second group is dominated by pro-reform individuals and organizations. These include organizations that embrace a variety of other progressive causes, such as the Human Rights Watch, as well as several that specialize in immigration only. Although they tend to have fewer followers overall (as indicated by the smaller images in the figure below), they are influential tweeters about immigration because of their greater focus on that issue. Extremely popular figures, such as President Obama, are the exception in this group.


The word cloud based on these tweeters’ profiles makes clear that many of them are especially concerned with immigration issues.


Cluster 3: Progressive Media

The third group was the smallest of the five clusters. It consists of left-leaning news organizations, like NPR and Mother Jones; left-leaning bloggers and online journalists, like Markos Moulitsas, Josh Marshall, and Matty Iglesias; and left-leaning mainstream media commentators, like George Stephanopoulos, Joe Trippi, and Donna Brazile.


These tweeters tend not to explicitly highlight their political orientation in their Twitter profiles. Most identify themselves as political commentators, but most do not associate their commentary with a particular political agenda.


Cluster 4: The Institutional Conservatives

This group is a “who’s who” of influential conservatives in the media and in Congress: Donald Trump, Marco Rubio, Michelle Malkin, Mike Huckabee, and Tom Coburn are just some of the individuals central to this network. Influential organizations, including Fox News, the Drudge Report, and the Heritage Foundation are also prominent members of this group.


Unlike the left-leaning media, the members of this group tend not to be shy about making their conservative political views explicit in their Twitter profiles.


Cluster 5: The Tea Party

The final cluster we found in our analysis of immigration related tweets consists of conservatives who also identify or sympathize with the Tea Party. Unlike the other clusters, this one consists almost entirely of individuals. Furthermore, the individuals aren’t affiliated with any prominent organizations: they are “ordinary” Americans. Cluster 2, the pro-reform group, has a significant minority of tweeters who fit this description–but Cluster 5 is dominated by them.


Moreover, the word cloud for this group reveals its members are even more adamant about touting their conservative viewpoint in their Twitter profiles.


Conclusions: Power Tweeters and the Immigration Reform Conversation

What can we learn from these clusters? Not surprisingly, we see that political affiliation is a major feature that distinguishes clusters of power tweeters from one another. There are two progressive groups, two conservative groups, and a mainstream group that, depending upon whom you ask, represents a mix of viewpoints or is biased one way or the other.

The second major feature appears to be the presence or absence of institutional affiliations. The reformer group is the most diverse in this regard, with a mix of Democratic politicians, progressive political organizations, and unaffiliated individuals passionate about immigration reform. Three of the other four groups, however, are dominated by institutions, or individuals connected to these institutions (e.g., journalists and politicians). By contrast, the Tea Party power tweeters are distinctive in that they are almost all individuals unaffiliated with any particular institution.

Does this mean that Twitter is primarily a social media platform dominated by elites, or that the Tea Party has harnessed Twitter in a way that other groups of individuals have not? Not necessarily. This particular data just looks at differences among the most influential immigration tweeters. It doesn’t address the many thousands of other immigration tweeters, how these tweeting masses interact with the most influential tweeters, or how they interact with each other.

But other questions remain. Have each of these clusters attained their influence on Twitter in the same way? Do one or more of these clusters have more influence on the conversation than the others? And what about the connections between these clusters–are they all following and interacting with one another, or do they largely exist in their own little bubbles? We explore the answers to these questions in Part 2.

This post originally appeared on the Huffington Post November 5, 2013

Immigration Reform, the Boston Marathon Bombings, and Twitter

By James Witte and Randy Lynn

Beginning in February 2013, researchers at the Institute for Immigration Research (IIR) at George Mason University began tracking the national immigration reform conversation on Twitter. Our view was that Twitter would be one of several social media venues where the evolving immigration reform debate could be tracked and analyzed. Since then, using DiscoverText and NodeXL, two different data mining applications, we’ve collected over 750,000 tweets containing the word “immigration.”

The Boston Marathon bombings on April 15, 2013 provided an unexpected opportunity to examine how this event has affected the immigration reform debate. We compared conversations about immigration on Twitter before, during, and after April 15.

Below we provide a series of NodeXL network graphs that illustrates the shifting Twitter conversation. Each visualization depicts the relationships between tweets containing the word “immigration” over the course of a four-hour period. NodeXL not only collects the content of the tweet and the identity of the tweeter, but also who follows whom, who is retweeting whom, and who is having a conversation with whom.

To represent these relationships, NodeXL uses an algorithm to group the tweeters into clusters. Tweeters belonging to the same cluster are highly interconnected; they tend to follow each other, retweet each other’s tweets, have conversations with each other, and use similar words in their tweets. Green lines represent connections between tweeters in different clusters, while the text overlaying each cluster represents the most common content among the tweets in that cluster.

Two Weeks Before the Bombing

The first visualization below is typical of what we saw in early April, before the bombing. In the lower right corner, there are small clusters with very few connections to other tweeters. In the lower middle area, there are tweeters who aren’t connected to anyone else tweeting about immigration.

Most of the tweeters, however, belong to one of three main clusters. The cluster in the lower left corner is a liberal cluster, as evidenced by some of its most common content (e.g., “timeisnow”). The cluster in the upper right corner is a very conservative cluster dominated by the Tea Party. These two clusters have very few connections between them, indicating that there are very few conversations between their members.

The third cluster, in the upper left corner, includes such diverse tweeters as the RNC, the AFL-CIO, and Bloomberg. It’s best described as a mainstream cluster that includes both liberal as well as conservative viewpoints. These tweeters tend not to be as extreme, and they have many connections with the two more ideological clusters, which suggests that this the mainstream cluster acts in part as an informational bridge between them.


The Day of the Bombing

This visualization, based on tweets collected between 4 and 8 p.m. on April 15, 2013, tells a very different story. Tweeters in all clusters were tweeting about the bombing in relation to immigration — even though there was not yet any evidence of such a connection. As the left half of the visualization shows, there was a large influx of tweeters who weren’t connected to any other immigration tweeters. Words like “deportation,” “amnesty,” “terrorism,” and “borders” point to possible risks associated with immigration seen by these newcomers, as well as moderates and conservatives (lower middle), but not among liberals (upper right).

The Tea Party cluster (middle right) seems to have diminished in size. Tea Partiers on Twitter were still tweeting about immigration, but it appears that more moderate, mainstream tweeters and more conservative tweeters had converged into a single cluster. However, it was unclear at this point whether this change would persist, or whether it was a temporary shift in the conversation.


Ten Days After the Bombing

Our evidence suggests that after a temporary deviation in the days immediately following the bombing, the immigration conversation on Twitter has largely returned to its pre-bombing appearance. The visualization below is from April 25, and it shows a decrease in the number of unconnected tweeters (lower left), as well as the reestablishment of a liberal cluster (lower middle), a conservative/Tea Party cluster (upper right), and a bridging, mainstream cluster (upper left). Note, however, that “bostonbomber” is still one of the most frequently used words in conjunction with immigration in the most conservative cluster. On the other hand, the bombing does not appear as prominent in immigration tweets among unconnected, mainstream, and liberal immigration tweeters.


Twitter and Immigration Reform in the Future

In the mainstream media, the evidence that Chechen immigrants Tamerlan and Dzhokhar Tsarnaev are responsible for the Boston Marathon bombing has led to speculation that comprehensive immigration reform might be delayed, derailed, or even abandoned in Congress. But in the short run, we found no indication that this connection has left a permanent mark on the immigration reform conversation on Twitter. In fact, we found that the immigration conversation quickly returned to patterns similar to those we saw before the bombing.

To the extent that Twitter activity reflects sentiment among influential politicians, journalists, and activists, proponents of comprehensive immigration reform can feel encouraged. On the other hand, conversations on Twitter are not necessarily representative of public opinion. In fact, the Pew Research Center has recently released a study that highlights notable disconnects between public opinion polls and views expressed on Twitter.

As the national conversation on immigration reform continues, so too will our efforts to monitor this debate as it takes place in various venues, including Twitter and other social media forums. Clearly there are many passionate, influential individuals and organizations that regularly tweet about immigration and immigration reform from many varying points of view. Continued analysis of these conversations is important for tracking the progress of immigration reform, but also for better understanding the relationship between social media and public opinion.

This post originally appeared on the Huffington Post May 1, 2013