Exploring Australian Political Party Affiliations On Twitter

Background

In the following exploration, we will inspect the relationship between the major Australian political parties and their adherents. Also, we will explore the influence of politics in a social system such as that of Australia using data obtained from Twitter, a social media platform.

Social systems are defined by relationships among people. With the increasingly socialized world of today, the media plays a critical role for evaluation of different social trends. Some of the areas in which social media can be used for monitoring include:

1. Media trends
2. Political trends

• Business trends

1. Music trends, etcetera

 Over the years, digital and technological innovations have been on what can be considered “inflation”, with almost each day recording an innovation. This has simultaneously enabled new methods of collecting useful data, exploring the data and coming up with important insights that have greatly contributed to the amount of knowledge on different social issues.

Li et al. (2010) defines social media as “a conversational, distributed mode of content generation, dissemination, and communication among communities…” According to their study, social media has defined a whole new meaning of social confines and interactions. This broadly interprets to freedom from traditional geographical limitations that hindered interactions and therefore monitoring of social media interactions among different groups can provide insightful information of:

1. Peoples’ consumer preferences
2. Peoples political interests and affiliations

• The society’s view on current and emerging trends on political

In this study, we will explore different political affiliations for different groups of people given three political leaders of different political parties. We will explore data from imported from twitter to establish whether different people keep relations with people from different political parties or whether they are enemies owing to their difference in political allegiance. Through this, we will establish the truth of the common fallacy that: “people with different political affiliations are actually enemies even in real life”

Data for this project is extracted from 10 twitter friends from each of:

• Malcolm Turnbull of the Liberal Party of Australia
• Bill Shorten of the Australian Labor Party
• Michael McCormack of the National Party of Australia

In order to explore relationship between different friends of different political leaders, we will further mine data on 1000 friends of the first ten friends of a each politician. 

1. Twitter apps
2. R statistical software

• R dependent packages

In this project we will test the following hypotheses:

Null hypothesis

H₀: Political parties are disjoint

Alternative hypotheses

H₁: Political parties are non-disjoint

H₂: People are still friends despite different political affiliations

In the cause of our research project, we assume that:

1. Every friend of a politician is interested in politics and has a political opinion
2. Any friend of a politician shares the politician’s political ideology

• Every friend is unique, I.e holds a definite political view.

library(httr)

library(twitteR)

library(twitteR)

library(“openssl”)

library(“httpuv”)

library(“base64enc”)

consumerKey <- “qqNvwX4KY5uuP3dwUovKxbaTB”

consumerSecret <- “7PZVZB4dX1eiL0713YWG5rdaDhcoGZPzw71l1iWxYqH3ptu2Nd”

 accessToken <- “818345262221590528-nn33pinbmGIQ7TsiA3v7QDCrXr87fPw”

 accessTokenSecret <- “fVsUi0djqxc2vqENm6aUoB1eCFt2SCAeTl57dgoUU7wXM”

 setup_twitter_oauth(consumerKey, consumerSecret, accessToken, accessTokenSecret)

#importing twitter id’s

#Malcolm Turnbull

tanbull = getUser(“@TurnbullMalcolm”)

tanbullfrends = tanbull$getFriends(10)

Social Media’s Role in Monitoring Social Trends

tanbullfrends

$`36275767`

[1] “CommGamesAUS”

$`103999368`

[1] “sydneyroosters”

$`2300779218`

[1] “MarisePayne”

$`111449574`

[1] “alannahmadeline”

$`208528305`

[1] “tweetinjules” 

$`36872712`

[1] “SenatorRyan”

$`224846346`

[1] “SouthernStars”

$`49574044`

[1] “nthqldcowboys”

$`263378242`

[1] “TheMatildas”

$`110855776`

• “Socceroos” 

#Bill Shorten

billshort = getUser(“@billshortenmp”)

billshortfrends = billshort$getFriends(10)

billshortfrends

$`1368753678`

[1] “DomLorrimer”

$`135057136`

[1] “WSFM1017”

$`3168147002`

[1] “GOFoundationAU”

$`129402565`

[1] “joelcreasey”

$`889348386779836416`

[1] “LoveIslandAU”

$`20356090`

[1] “adamrichard”

$`21654257`

[1] “Briggs”

$`216876629`

[1] “lisa_fernandez”

$`37621265`

[1] “steenrasko”

$`110138384`

[1] “MahaliaBarnes”

mccormac= getUser(“@M_McCormackMP”)

mccormacfrends = mccormac$getFriends(11)

Mccormacfrends  

$`851828803219410944`

[1] “4roadsafety” 

$`39937499`

• “FarmOnline” 

$`36275767`

• “BeddingtTrol” 

$`987679153`

[1] “ALGAcomms” 

$`17939037`

[1] “ProfBrianCox” 

$`626111072`

[1] “astroduff” 

$`48927175`

[1] “TomMcIlroy”

$`2393844469`

[1] “lindareynoldswa” 

$`2602764942`

[1] “maculardisease” 

$`359750088`

[1] “ItaButtrose”

#comparing if there are any similar friends shared by the politicians

#between Tanbull and billshort

names(tanbullfrends)%in%names(billshortfrends)

 [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE 

#between Tanbull and Mccormac

names(tanbullfrends)%in%names(mccormacfrends)

• FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

#between Billshort and Mccormac

names(billshortfrends)%in%names(mccormacfrends) 

 [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE 

From the data we collected, we find that there were 60 entries for the three candidates combined, however there were no shared friends among the three politicians given the first 10 friends. This implies that, taking the assumption of the most staunch supporters being the first friends of each politician. Then, there are little chances of a supporter having allegiance to two politicians. Therefore, we can conclude that, in politics support and allegiance are key indicators of the ideologies one believes in. 

library(igraph)

library(dplyr)

##Collecting data for comparing relationships between friends

compare1<-tanbullfrends [[1]]$getFriends(1000)

 compare2<-tanbullfrends [[2]]$getFriends(1000)

compare3<-tanbullfrends [[3]]$getFriends(1000)

compare_1<-mccormacfrends [[1]]$getFriends(1000)

compare_2<-mccormacfrends [[2]]$getFriends(1000)

compare_3<-mccormacfrends [[3]]$getFriends(1000) 

compare_1<-mccormacfrends [[1]]$getFriends(1000)

compare_2<-mccormacfrends [[2]]$getFriends(1000)

compare_3<-mccormacfrends [[3]]$getFriends(1000) 

#friendslist

frendlist<-names(compare1) 

summary(frendlist)  

Length     Class      Mode

499    character   character  

combndfrends<-names(Combinedfriendlist) 

#comparing similar friends among the friends

library(compare) 

 comprsn <- compare(combndfrends,frendlist,allowAll=TRUE)

comprsn$tM

 [1] “103999368”          “110138384”          “110855776”         

 [4] “111449574”          “129402565”          “135057136”         

 [7] “1368753678”         “17939037”           “20356090”          

[10] “208528305”          “21654257”           “216876629”         

[13] “224846346”          “2300779218”         “2393844469”        

[16] “2602764942”         “263378242”          “3168147002”        

[19] “359750088”          “36275767”           “36872712”          

[22] “37621265”           “39937499”           “48927175”          

[25] “49574044”           “626111072”          “851828803219410944”

[28] “889348386779836416” “987679153”

#plotting relationships

deta<-as.matrix(get.adjacency(graph.data.frame(relashionshpdta))

grp<-graph_from_adjacency_matrix(deta, mode =”undirected”,weighted=NULL,diag=FALSE)

 plot.igraph(grp,layout=layout.circle,edge.color=”skyblue”) 

Elsewhere, Given 1000 friends from each of the 10 friends, we found out that there were 499 common friends at least among two politicians and 44 shared friends shared among the three politicians. Implying that despite difference in political affiliations among different people, they still do maintain friendships among themselves and therefore politics does not cause volatility among citizens. As such, it indicates a relative amount of independence in adoption and translation. However, the low percentage of interrelations may indicate that politics has a role in volatizing the masses.

Hypotheses

#calculating edge

edge_density(grp, loops=F) 

0.02429379

 reciprocity(grp)

• 1

transitivity(grp) 

[1] 0

#calculating size neighbourhood of each edge

neihbourhood<-as.matrix(neighborhood.size(grp,1000))

Neihbourhood

  [,1]

 [1,]    3

 [2,]    2

 [3,]    2

 [4,]    2

 [5,]    2

 [6,]    2

 [7,]    2

 [8,]    2

 [9,]    2

[10,]    2

[11,]    2

[12,]    2

[13,]    3

[14,]    3

[15,]    2

[16,]    3

[17,]    5

[18,]    5

[19,]    2

[20,]    2

[21,]    3

[22,]    5

[23,]    5

[24,]    2

[25,]    2

[26,]    2

[27,]    2

[28,]    2

[29,]    2

[30,]    2

[31,]    2

[32,]    2

[33,]    3

[34,]    2

[35,]    2

[36,]    2

[37,]    2

[38,]    2

[39,]    2

[40,]    2

[41,]    2

[42,]    2

[43,]    2

[44,]    2

[45,]    3

[46,]    2

[47,]    3

[48,]    5

[49,]    2

[50,]    2

[51,]    2

[52,]    2

[53,]    2

[54,]    2

[55,]    2

[56,]    2

[57,]    2

[58,]    3

[59,]    2

[60,]    2

 #modularity

modularity(medntlgenceceb)

• 9356409 

highsocialclp <- cluster_label_prop(grp)

highsocialclp

IGRAPH clustering label propagation, groups: 27, mod: 0.94

+ groups:

  $`1`

  [1] “A”         “103999368” “36275767”   

  $`2`

  [1] “B”         “111449574”  

  $`3`

  [1] “C”          “1368753678”  

  $`4`

Additionally, according to Moreno et al. (2017), “High modularity for a partitioning reflects dense connections within communities and sparse connections across communities…” in our case we obtain a modularity of 0.9356409 which is relatively high while our density is 0.02429379. The neighborhoods fall within 2 and 5. Additionally, there were 27 nodes having high social capital.

#defining function to be used

ehomphil<-function(graph,vertex.attr){

    #ego-hemophili

nodehomo<-as.vector(rep(NA,vcount(graph)))

   #repeat for every node

   for(t in 1:vcount(graph)){

     if(degree(graph,V(graph)[t])>0){

       #get the neighbors

       node.alters<-ego(graph,1,nodes=V(graph)[t],mode=”out”)

       Node_ec<-get.vertex.attribute(graph,vertex.attr,V(graph)[t])

       #extract neighbors with matching attribute value       nodehomo[t]<-sum(get.vertex.attribute(graph,vertex.attr,node.alters[[1]])==ec)/length(node.alters[[1]])

     }

   }

   return(nodehomo)

 } 

 data 

 [1] “1368753678”         “135057136”          “3168147002”        

 [4] “129402565”          “889348386779836416” “20356090”          

 [7] “21654257”           “216876629”          “37621265”          

[10] “110138384”          “36275767”           “103999368”         

[13] “2300779218”         “111449574”          “208528305”         

[16] “36872712”           “224846346”          “49574044”          

[19] “263378242”          “110855776”          “851828803219410944”

[22] “39937499”           “987679153”          “17939037”          

[25] “626111072”          “48927175”           “2393844469”        

[28] “2602764942”         “359750088”  

labour_party 

 [1] “1368753678”         “135057136”          “3168147002”        

 [4] “129402565”          “889348386779836416” “20356090”          

 [7] “21654257”           “216876629”          “37621265”          

[10] “110138384”  

combined_party

 [1] “36275767”           “103999368”          “2300779218”        

 [4] “111449574”          “208528305”          “36872712”          

 [7] “224846346”          “49574044”           “263378242”         

[10] “110855776”          “851828803219410944” “39937499”          

[13] “987679153”          “17939037”           “626111072”         

[16] “48927175”           “2393844469”         “2602764942”        

[19] “359750088”          

#calculating hemophily

V(nodegg)$Politics<-sample(c(“Labour party”,”Other Party”),100,replace=T)

V(nodegg)$Social<-sample(c(“None”,”Does not relate”,”Relates”),100,replace=T)

V(nodegg)$ehomphil<-ego_homophily(graph = nodegg, vertex.attr = “Social”)

hemophili<-V(nodegg)$ehomphil

#dealing with missing values

hemophili[is.na(hemophili)]<-0

hemophili 

 [1] 0.6666667 0.3750000 0.2857143 0.1428571 0.1111111 0.5000000 1.0000000

  [8] 0.4000000 0.5000000 0.6000000 1.0000000 0.6363636 0.7500000 0.0000000

 [15] 0.5000000 0.2500000 0.5000000 0.2500000 0.5000000 0.3333333 0.5000000

 [22] 0.7500000 0.5000000 1.0000000 0.2500000 0.2857143 0.5000000 0.1666667

 [29] 0.4285714 0.1250000 0.8000000 0.3333333 0.5000000 0.6666667 0.4000000

Data Collection and Analysis

 [36] 0.3333333 0.2857143 0.3750000 0.5000000 0.6000000 0.5000000 0.2500000

 [43] 0.4444444 0.7142857 0.6666667 0.2000000 0.3333333 0.4000000 0.6666667

 [50] 0.5000000 0.2000000 0.6000000 0.4000000 0.2500000 0.7500000 0.8571429

 [57] 0.5000000 0.5000000 0.6000000 0.4285714 0.2500000 0.6000000 0.4166667

 [64] 0.3333333 0.6000000 0.2500000 0.3333333 0.4285714 0.6000000 1.0000000

 [71] 0.5000000 0.4000000 0.1428571 0.4444444 0.0000000 0.6666667 1.0000000

 [78] 0.4285714 0.6666667 0.3333333 0.5714286 1.0000000 0.5000000 0.5000000

 [85] 0.5000000 0.4000000 0.1666667 0.6666667 0.1818182 0.2857143 0.2727273

 [92] 0.3750000 0.2000000 0.2000000 0.2222222 0.3333333 0.5000000 0.5714286

 [99] 0.4000000 0.1250000

mean(hemophili)

[1] 0.4572861

sd(hemophili)

• 2259754 

#plotting graph for homophilly

hist(hemophili,main=”Histogram for Network Homophily”,

  xlab=”Network Homophily”,freq=T,xlim=c(0,1))

lines(density(hemophili),lty=”dotted”,lwd=2,col=”skyblue”)

abline(v=mean(hemophili),lwd=2,col=”red”)

From our study, the average Homophily between labor party supporters and the supporters of both the Liberal and National parties combined given an alpha-level of 0.05 is 0.4572861

medntlgenceceb <- cluster_edge_betweenness(grp)

dendPlot(medntlgenceceb, mode=”hclust”)

plot(medntlgenceceb, grp)

From the graph we realize the network is a weak network due to the hierarchical distribution of different nodes that represent connection among members of the test society. Despite it being a weak network, there are no outliers in the data. I.e. no abnormal relationships. Generally because of the disjoint among supporters of different parties, this is true from the fact that the politicians do not share common supporters.

Yes, from our analysis we establish that different adherents of different parties are loyal and hence Australian politics can be assumed to be disjoint. In that no single friend of a politician is partisan among the political parties.

True, as noted from our analysis, different members are friends with other members despite the other members political affiliations. Therefore, it would be a fallacy to assume disjointedness among members of political parties. In that respect, we can argue that, despite difference in political ideologies members can still be friend and the few “outliers” are only probably acting for themselves.  

Conclusion

We can confide that social media intelligence is one of the better modern ways that can be used to gather Intel that will aid in gaining insight on contemporary social issues. Nevertheless, despite the tremendous benefits embed in social media intelligence, it is also wrought with some negative shades. This include:

1. The potential of gathering an authorized third-party information, which may lead to legal action
2. Willful manipulation of target audience given background insights on their preferences, etcetera

However, despite this side of this new digital development, we can rake benefits if properly utilized. For instance, from our research we realize that, it is not always that people having different political ideologies fail to hold relationships, however limited the frequency. This can help counter the rising claims of Australia becoming more polarized and help quieten naysayers while actually promoting social integration despite political differences. It may therefore be considered illogical and assumptive to conclude that members of different parties hate each other.

Hoover, JN (2007). US spy agencies go Web 2.0 in effort to better share information. InformationWeek, 23 August. Available at: https://www.informationweek.com/news/internet/showArticle.jhtml?articleID=20

1801990 Google Scholar

Mandel, M (2010).British jokester freed after learning police not amused. Toronto Sun, 1 July: 8. Google Scholar

Van Dijck, J, Nieborg, D (2009).Wikinomics and its discontents: a critical analysis of Web 2.0 business manifestos. New Media & Society 11(5): 855–874. Google

Scholar, SAGE Journals, ISI

Joseph, K & Benjamin, B (2018) The Political Significance of Social Media Activity and Social Networks, Political Communication, DOI: 10.1080/10584609.2018.1426662 

Best, S. J., & Krueger, B. S. (2005). Analyzing the representativeness of Internet political participation. Political Behavior, 27(2), 183–216. [Crossref], [Web of

Science ®] ,[Google Scholar]

Bond, R. M., Fariss, C. J., Jones, J. J., Kramer, A. D. I., Marlow, C., Settle, J. E., & Fowler, J. H. (2012). A 61-million-person experiment in social influence and political mobilization. Nature, 489(7415), 295–298.[Crossref], [PubMed], [Web of Science ®],[Google Scholar]

Gibson, R., & Cantijoch, M. (2013). Conceptualizing and measuring participation in the age of the Internet: Is online political engagement really different to offline? Journal of Politics, 75(3), 701–716.[Crossref], [Web of Science ®], [Google

Scholar]

Ognyanova,K.(2016).Network Analysis and Visualization with R and igraph.Available at: https://www.kateto.net/wp-content/uploads/2016/01/NetSciX_2016_Wor

kshop.pdf  

Al-Qaheri, H., Banerjee,S. and Ghosh G.(2103) “Evaluating the power of homophily and graph properties in Social Network: Measuring the flow of inspiring influence using evolutionary dynamics,” 2013 Science and Information Conference, London, 2013, pp. 294-303.