Contrasting Protest and Mundanity:

Centralization & Cohesion in Opposition Communication Networks


Benjamin Lind , Ph.D. ( lind.benjamin@gmail.com) 


Social Networks & Social Movements

Traditional Topics
  • Recruitment
  • Diffusion
  • Relationships between organizations

Protests
  • Crucial to social movement research
  • No network studies at protest sites




How do we characterize networks within a protest?

Two Characteristics at Odds


Leadership and formal organization


Solidarity among rank and file

Leadership and Formal Organization

  • Increases movement success and recruitment
    • (Ganz 2009; Lind & Stepan-Norris 2011)
  • Conflicting viewpoints
    • It's necessary (McCarthy & Zald 1977; Oliver et al. 1985)
      • Coordination, organizational skills, and media work
      • Start-up costs
    • It's avoided (e.g., Piven & Cloward 1977; Polletta 2005; Nepstad & Bob 2006)
      • Ideological reasons
      • Co-optation
      • Repression

Solidarity among rank and file

  • Togetherness
    • Movements display "WUNC" (Tilly 2004)
      • Worthy, Unified, Numerous, Committed
    • Collective identity (Snow 2001)
      • “a shared sense of ‘one-ness’ or ‘we-ness’”
      • Commonalities & informal ties bound movement
        • (Diani & Bison 2004; Diani & Pilati 2011)
  • Yet...
  • Not a homogeneous "crowd" (McPhail 2006, 2008)
  • Patchwork gatherings: "withs" and "singles" (McPhail 2008)
    • Groups defined by the foci of their interaction

Brief Case Background

  • Origins
    • Contested 2011 State Duma Election
    • 4 December 2011
  • Turning point
    • 6-7 May 2012 violence at protest
    • Putin's presidency

Data


Underlying Assumptions


Protests occupy a specific location (Fillieule 2012)


A protest's beginning, middle, and end refer to occupation periods.


Occupation affects the location's demographic and communicative structure. 

Data Collection

  • Case selection: Moscow Opposition Protests
    • Demonstrations lead by opposition activists
    • Sanctioned and/or widely announced in news
    • Determine location, date, start and end times
  • Source and process
    • Twitter: 140 character "micro-blog" updates
      • Available on desktop and mobile devices
      • "Following" = directed communication
    • Collect all updates from location during protest
    • Construct follower network
    • Repeat during same time and place one week later 

Date

Time

ºC

Issue

Type

People

Users

Points

Radius

12/6

12-4

23

General

March

22k

345

4

1km

26/7

7-10

26.7

Prisoners

Rally

800

117

1

1km

15/9

2-10

16.8

General

March

14k

1204

7

.75-1km

20/10

12-6

12.8

Elections

Rally

600

317

1

1km

21/10

3-9

11.7

Elections

Rally

600

166

1

1km

30/10

7-9

-0.8

Prisoners

Rally

500

135

1

1km

15/12

3-5:30

-14.2

Prisoners

Gathering*

400

178

1

.5km

13/1

1-4:30

-11.9

Adoptions*

March

9k

316

4

1km

June 12, 12-4pm (March)


June 19, 12-4pm


Methods



User differences



Network differences

User Differences

Account information
  • Followers (n)
  • Following (n)
  • Lists (n)
  • Status updates, "tweets" (n)
  • Account age (days)
  • Language: Russian or other (proportion)

User Differences

Measurements
  • Hedge's g, standardized difference of means
    • Following, followers, lists, tweets, account age
  
  • Difference of proportions (Fisher's stabilization)
    • Proportion of Russian language users


User Differences

Measurements
    Combine Theta (g and θ expressed as Ti)


      Combine p-values using Stouffer's method


      Network Differences

      • Leadership  by centralization
        • Closeness, out
        • Betweenness
        • Eigenvector
      • Solidarity by cohesion: reach and transitivity
        • Strong components (n)
        • Weak transitivity (proportion)
        • Average path length
      • Formal organization by Krackhardt's (1999) typology
        • Connectedness,  Hierarchy, Efficiency, LUBness
      • Meas(Protest) - Meas(Non-protest) 

      Network Differences

      • Conditional Uniform Graph (CUG) tests
          • Create two random networks for each pairing
            • Base each upon characteristics of the observed networks
            • #Vertices, degree distribution, dyad census, nodal reciprocity norms
            • Randomly rewire dyads: 10 * #dyads
        • Take measurements and subtract difference
        • Repeat 1000 times to generate null distribution
      •  Combine Z-scores using Stouffer's method
        • Raw
        • Jack knife corrected

      Findings


      User differences


      Findings


      Network differences

      June 12, 2012. 12-4pm.

      July 26, 2012. 7-10pm.

      September 15, 2012. 2-10pm.

      October 20, 2012. 12-6pm.

      October 21, 2012. 3-9pm.

      October 30, 2012. 7-9pm.

      December 15, 2012.

      3-5:30pm

      January 13, 2013. 1-4:30pm.

      March 2, 2013. 1-6pm.

      Leadership Difference

      (Six Observations)
      • Closeness (out)
        • Z =  -3.447***
        • Jackknife bias corrected Z = -1.945
      • Betweenness
        • Z = -0.818
        • Jackknife bias corrected Z = -0.462
      • Eigenvector
        • Z = 0.167
        • Jackknife bias corrected Z = 0.094

      Formal Org Difference

      • Connectedness
        • Z = -12.798***
        • Jackknife bias corrected Z = -7.222***
      • Hierarchy
        • Z = 2.553*
        • Jackknife bias corrected Z = 1.441
      • Efficiency
        • Z = 5.088***
        • Jackknife bias corrected Z = 2.872**
      • LUBness
        • Z = 1.916
        • Jackknife bias corrected Z = 1.082

      Solidarity Difference

      • Weak Transitivity
        • Z = 0.109
        • Jackknife bias corrected Z = 0.063
      • Strong components
        • Z = 8.091***
        • Jackknife bias corrected Z = 4.566***
      • Average path length
        • Z = 7.342***
        • Jackknife bias corrected Z = 4.143***

      Summary of Findings

      • Users
        • Protest users are more engaged with Twitter
          • More followers, following, lists, & tweets
        • No difference:  adoption time, and language
      • Network visualizations
        • Many more people during protests
        • Greater density, but less reciprocity during protests
      • Network CUG tests
        • Leadership: Perhaps less closeness, no different otherwise
        • Organization: Less connected, yet more efficient, perhaps more hierarchical
        • Solidarity : More components, longer paths

      Conclusions

      1. After accounting for popularity and reciprocity effects...
        1. No indication of leadership, maybe even an absence
        2. Organization is efficient, yet highly fractured
        3. Absence of solidarity
          1. Communication pathways typically don't exist
          2. When they do exist, they are very long
      2. "The Tyranny of the Structureless" (Freeman 1971)
        1. Politically impotent
        2. Good for "just talking," poor for getting things done
        3. Problems exacerbated with large groups
      3. Indicators of weak or strong democratic principles?

      Conclusion

      • Protest brings more people, interactions to a location
      • Leadership through reach
      • Solidarity
        • Patchwork--many more clusters
        • Emphasis on sending, avoidance of receiving
        • Avoidance of redundant transitive communication
      • Limitations
        • Representation
        • Trade-offs when conditioning by degree
      • Future research
        • Proximity effects
        • Expand study across differing political contexts

      Thank You!


      Questions and comments?






      Benjamin Lind
      lind.benjamin@gmail.com

      Protest and Mundanity: Visualizations

      By Benjamin Lind

      Made with Slides.com

      Protest and Mundanity: Visualizations

      • 2,804

      More from Benjamin Lind