[PPT] - Ideology Estimation, Media Slant, and Opinion Segregation: Facebook PowerPoint Presentation

SLIDE 1

Ideology Estimation, Media Slant, and Opinion Segregation: Facebook as a Social Barometer

Keng-Chi Chang

National Taiwan University

2017-09-15

1 / 57

SLIDE 2

Motivation: Bond and Messing (2015)

2

2 0.5 1 1.5 2

Politicians Individuals

Facebook ideology score Density

2 / 57

SLIDE 3

Highlights

Specify potential ideological universe
Select possible US users
Place different political actors on the same ideological spectrum

(politicians, gures, news outlets, parties, and interest groups)

Replicate mass ideology distribution at national and state level
Allow time and text dimensions to explore
All using publicly available open data

3 / 57

SLIDE 4

Introduction

Best place to get interactions between differnt actors
Extract information from the major action: “like”
e.g. What can we infer if two pages share many users?
Joint measure of ideologies of political elites, news outlets,

interest groups, and ordinary citizens

cf. surveys: Revealed preference, low cost, real time
Past papers only look at following of pages, we look at like of

posts (adds time and post content dimension)

4 / 57

SLIDE 5

Literature: Ideology Measures

Need “bridges” to connect different actors

▸ Politicians: Poole and Rosenthal (1985), Clinton et al. (2004) ▸ Media-Politicians: Groseclose and Milyo (2005) ▸ Media-Citizens: Gentzkow and Shapiro (2011) ▸ Politician-Citizens: Bonica (2014)

Lack of joint ideological measures across all these actors
Social media acts as brigdes for different actors
Both Bond and Messing (2015) (Facebook) and Barberá (2015)

(Twitter) only consider political elites

5 / 57

SLIDE 6

Model and Traditional Estimation

Assume that user i’s ideological position is θi and

politician/media j’s position ϕj

Assume also that the probability i likes j’s post is proportional to

the negative distance between θi and ϕj P (yi j = 1∣αi, βj, γ, θi, ϕj) = logit−1 (αi + βj − γ ∥θi − ϕj∥

2)

Traditionally this is solved by Markov-Chain Monte Carlo

(MCMC) to maximize posterior density (MLE) {ˆ θi, ˆ ϕj} = argmax

θi,ϕ j

∏

i∈user

∏

j∈page

logit−1 (πij)

yi j (1 − logit−1 (πi j)) 1−yi j

This is slow for large numbers of parameters

6 / 57

SLIDE 7

Methodology: Dimension Reduction

Similar to Heckman and Snyder (1997) (Congress), Bond and

Messing (2015) (Facebook), and Barberá (2015) (Twitter), we use dimension reduction to recover the latent ideological space

Barberá et al. (2015) also shows that simulations and dimension

reduction (Correspondence Analysis) generate very similar results (ρ = 0.98)

We show that Correspondence Analysis and Principal

Component Analysis (2 stage) generate similar results (ρ > 0.94)

Drawbacks:

▸ What are the dimensions? ↝ Guess and verify ▸ How many dimensions to consider? ↝ Scree plot 7 / 57

SLIDE 8

Select Meaningful Data

Select fan pages mentioned two major presidential candidates
Calculate likes, comments, shares and select top 1000 pages
Also include past and present national politicians (Sen, Rep, Gov)
Facebook open data do not give any personal information
Select users ever liked national politicians in 2015 and 2016

8 / 57

SLIDE 9

Data Summary (Main Sample)

Time Period 2015-01-01 to 2016-11-30 Total Reactions 19,085,783,534 US User Likes 16,180,488,916 Total Users 366,840,068 US Users 29,412,610 Total Posts 24,788,093 Total Pages 2132 Politician 1225 News Outlets 560 Political Groups 211 Other Public Figures 93 Others 43

9 / 57

SLIDE 10

Estimation: Afliation Matrix

First we build the afliation matrix A, which contains number of

shared users between pages

Trump FoxNews TeaParty Clinton CNN NYTimes Trump 2243216 1078513 128225 32731 120963 25842 FoxNews 1078513 2449174 148016 87084 186850 63401 TeaParty 128225 148016 242089 1528 10738 2162 Clinton 32731 87084 1528 1768980 351210 367021 CNN 120963 186850 10738 351210 1201156 216163 NYTimes 25842 63401 2162 367021 216163 986613

10 / 57

SLIDE 11

Estimation: Agreement Matrix

Agreement matrix G is computed by gij = ai j/aii
Ex. 0.44 is (Trump & Fox) / Fox
Can interpret each column as feature and row as observaton
Interpretation: Col 1 is how each row similar to “Trump” feature

Trump FoxNews TeaParty Clinton CNN NYTimes Trump 1.00 0.48 0.06 0.01 0.05 0.01 FoxNews 0.44 1.00 0.06 0.04 0.08 0.03 TeaParty 0.53 0.61 1.00 0.01 0.04 0.01 Clinton 0.02 0.05 0.00 1.00 0.20 0.21 CNN 0.10 0.16 0.01 0.29 1.00 0.18 NYTimes 0.03 0.06 0.00 0.37 0.22 1.00

11 / 57

SLIDE 12

Estimation: Compute PCA

Compute the principal components of G after standardizing
The rst principal component is the dimension that can explain

the laregest variance, guess that positions on this dimension can represent “ideology” of pages

Calculate the ideological position of users by computing the

means of the ideologies of the pages they like (minimizer under Euclidean norm)

12 / 57

SLIDE 13

5 10 15 20 0.00 0.02 0.04 0.06

Scree Plot for Principal Component Analysis

k-th Principal Component Proportion of Variance Explained

13 / 57

SLIDE 14

Trump Clinton NYTimes Fox News

0.00 0.25 0.50 0.75

2
1

1 2

Estimated Facebook Ideology Score Density

Public Figure Political Groups News Outlets

14 / 57

SLIDE 15

0.0 0.5 1.0 1.5

2
1

1 2

Estimated Facebook Ideology Score Density

Magazine Newspaper Radio TV Website

15 / 57

SLIDE 16

The New York Times Politics and Washington The New York Times Opinion Section USA TODAY The New York Times The Wall Street Journal Washington Post Chicago Tribune Boston Herald The Christian Post The Washington Times 0.0 0.3 0.6 0.9

1.5
1.0
0.5

0.0 0.5 1.0

PC1 (First Principal Component) Density

PC1 Density of Newspaper Pages

16 / 57

SLIDE 17

New Republic Mother Jones The Economist The Atlantic The New Yorker Forbes The Hollywood Reporter The Nation Magazine National Review High Times Charisma News 0.00 0.25 0.50 0.75 1.00 1.25

1

1

PC1 (First Principal Component) Density

PC1 Density of Magazine Pages

17 / 57

SLIDE 18

CNN Breitbart MSNBC PBS Fox News Opinion The Rachel Maddow Fan Page. Fox News NRA News The Federalist Papers ABC News 0.0 0.5 1.0 1.5 2.0

2
1

1 2

PC1 (First Principal Component) Density

type_sub

radio tv website

PC1 Density of TV, Radio, Website Pages

18 / 57

SLIDE 19

Thom Hartmann Bill Moyers Bernie Sanders Ted Cruz Gary Johnson George Takei Ron Paul Elizabeth Warren Rand Paul Newt Gingrich Bill O'Reilly Matt Kibbe 0.0 0.1 0.2 0.3 0.4 0.5

1

1 2

PC1 (First Principal Component) Density

type_sub

journalist politician

PC1 Density of Public Figure Pages by Type

19 / 57

SLIDE 20

The Liberal Fallout Shelter The Party of Scrooge - History of American Politics Occupy Healthcare Criminalize conservatism NARAL Pro-Choice America Occupy Wall St. Reform Immigration FOR America Brady Campaign to Prevent Gun Violence GMO Dangers The Cato Institute National Pro-Life Alliance My Favorite Gun Christians Against Electing Leists Positively Republican! Stop Obamacare America is RIGHT 0.0 0.1 0.2 0.3

2

2

PC1 (First Principal Component) Density

PC1 Density of Party & Interest Group Pages

20 / 57

SLIDE 21

Democratic Party Green Party The Tea Party Libertarian Party Republican National Committee 0.0 0.1 0.2 0.3

2

2

PC1 (First Principal Component) Density

PC1 Density of Party & Interest Group Pages

21 / 57

SLIDE 22

ρ = 0.92 ρR = 0.50 ρD = 0.22

Schumer McConnell McCain Pelosi Sanders Ryan Rubio Warren Cruz Booker

1.0
0.5

0.0 0.5 1.0

2
1

1 2

Estimated Facebook Page Ideology Score, 2015-01 to 2016-11 DW-Nominate Score of 114th Congress

Democratic Party Independent Republican Party

Using politician and top 1000 page matrix

22 / 57

SLIDE 23

ρ = 0.90 ρR = 0.52 ρD = 0.09

Schumer McConnell McCain Pelosi Sanders Ryan Rubio Warren Cruz Booker

1.0
0.5

0.0 0.5 1.0

2
1

1 2

Estimated Facebook Page Ideology Score, 2015-01 to 2016-11 DW-Nominate Score of 115th Congress

Democratic Party Independent Republican Party

Using only politician page matrix (Bond and Messing 2015)

23 / 57

SLIDE 24

New Yorker New Republic NYTimes WashPost MSNBC BuzzFeed Politico CNN WSJ ABC News USA Today The Hill WSJ Opinion AWM Fox News National Review The Blaze Breitbart 0.5 1

2
1

1 2

Estimated Facebook Ideology Score Share of Republican-Affiliated User

Magazine Newspaper Radio TV Website

24 / 57

SLIDE 25

New Yorker New Republic MSNBC NYTimes BuzzFeed WashPost Politico CNN WSJ USA Today ABC News The Hill AWM WSJ Opinion Fox News National Review Breitbart The Blaze 0.5 1

2
1

1 2

Estimated Facebook Ideology Score Mean User Republican-Prone Index

Magazine Newspaper Radio TV Website

25 / 57

SLIDE 26

0.0 0.5 1.0

1
0.5

0.5 1

Estimated Facebook Ideology Score Density

Extremely Liberal Liberal Slightly Liberal Moderate Slightly Conservative Conservative Extremely Conservative

26 / 57

SLIDE 27

Users Like More than 10 Pages and Posts

0.0 0.3 0.6 0.9 1.2

1
0.5

0.5 1

Estimated Facebook Ideology Score Density

Extremely Liberal Liberal Slightly Liberal Moderate Slightly Conservative Conservative Extremely Conservative 27 / 57

SLIDE 28

Users Like More than 20 Pages and Posts

0.00 0.25 0.50 0.75

1
0.5

0.5 1

Estimated Facebook Ideology Score Density

Extremely Liberal Liberal Slightly Liberal Moderate Slightly Conservative Conservative Extremely Conservative 28 / 57

SLIDE 29

0.0 0.5 1.0 1.5 2.0

1
0.5

0.5 1

Massachusetts

0.00 0.25 0.50 0.75 1.00

1
0.5

0.5 1

Washington

0.0 0.2 0.4 0.6

1
0.5

0.5 1

Michigan

0.0 0.2 0.4 0.6

1
0.5

0.5 1

Pennsylvania

0.0 0.3 0.6 0.9 1.2

1
0.5

0.5 1

Texas

0.00 0.25 0.50 0.75 1.00

1
0.5

0.5 1

Wyoming

29 / 57

SLIDE 30

10 20 30

1
0.5

0.5 1

Massachusetts

0.0 0.5 1.0 1.5 2.0 2.5

1
0.5

0.5 1

Washington

1 2

1
0.5

0.5 1

Michigan

0.0 0.5 1.0 1.5

1
0.5

0.5 1

Pennsylvania

2 4 6 8

1
0.5

0.5 1

Texas

2 4 6

1
0.5

0.5 1

Wyoming

30 / 57

SLIDE 31

Applications

Time Dimension: Polarization and Spatial Voting
Text Dimension: Echo Chambers
Election Forecasting
Opinion or Ideological Segregation

31 / 57

SLIDE 32

ThinkProgress.com MSNBC Washington Post New York Times Bloomberg CNN WSJ ABC News Fox News Breitbart Federalist Papers NRA News CNSNews.com

1

1 2 06-01 2015 08-01 2015 10-01 2015 12-01 2015 02-01 2016 04-01 2016 06-01 2016 08-01 2016 10-01 2016 12-01 2016 02-01 2017 04-01 2017

Estimated Facebook Ideology Score

4-week window

32 / 57

SLIDE 33

ThinkProgress.com MSNBC Washington Post New York Times Bloomberg WSJ CNN ABC News Fox News Breitbart Federalist Papers NRA News CNSNews.com

1

1 2 3 06-01 2015 08-01 2015 10-01 2015 12-01 2015 02-01 2016 04-01 2016 06-01 2016 08-01 2016 10-01 2016 12-01 2016 02-01 2017 04-01 2017

Estimated Facebook Ideology Score

1-week window

33 / 57

SLIDE 34

Warren Clinton Sanders Johnson Trump Ryan Rubio Cruz Warren Sanders Clinton Johnson Trump Ryan Rubio Cruz

1.0
0.5

0.0 0.5 1.0 05-01 2015 07-01 2015 09-01 2015 11-01 2015 01-01 2016 03-01 2016 05-01 2016 07-01 2016 09-01 2016 11-01 2016 01-01 2017

Estimated Facebook Ideology Score

4-week window

34 / 57

SLIDE 35

Warren Clinton Sanders Johnson Trump Rubio Cruz Warren Sanders Clinton Johnson Trump Rubio Cruz

1.0
0.5

0.0 0.5 1.0 04-01 2015 06-01 2015 08-01 2015 10-01 2015 12-01 2015 02-01 2016 04-01 2016 06-01 2016 08-01 2016 10-01 2016 12-01 2016 02-01 2017

Estimated Facebook Ideology Score

1-week window

35 / 57

SLIDE 36

1

1

1

1

Estimated Ideology of User Estimated Ideology of User Like Same Post

0.0% 0.25% 0.50% 0.75% >1%

% of posts

Likes of Facebook Posts on Immigration, 2015-07

36 / 57

SLIDE 37

1

1

1

1

Estimated Ideology of User Estimated Ideology of User Like Same Post

0.0% 0.25% 0.50% 0.75% >1%

% of posts

Likes of Facebook Posts on Cubs

37 / 57

SLIDE 38

1

1

1

1

Estimated Ideology of User Estimated Ideology of User Like Same Post

0.0% 0.25% 0.50% 0.75% >1%

% of posts

Likes of Facebook Posts on JoliePitt

38 / 57

SLIDE 39

1

1

1

1

Estimated Ideology of User Estimated Ideology of User Like Same Post

0.0% 0.25% 0.50% 0.75% >1%

% of posts

Likes of Facebook Posts on Cohen

39 / 57

SLIDE 40

Predicting Election Outcomes

AL AK AZ AR CA CO CT DE FL GA HI ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY

ρ = 0.73 95% CI [0.56, 0.84] 30 40 50 60 25 50 75

Share of Facebook User Closer to Clinton (10-01 to 11-07) 2016 Clinton Vote Share

Rep wins 2016 & 2012 Swings from Obama to Trump Dem wins 2016 & 2012

40 / 57

SLIDE 41

Compare with Major Election Forecasts

State E.V.† Winner FB 538 NYT PEC* Wisconsin 10 Trump ○ × × × Iowa 6 Trump ○ ○ ○ ○ Florida 29 Trump ○ × × × Pennsylvania 20 Trump ○ × × × Ohio 18 Trump ○ ○ ○ ○ Michigan 16 Trump × × × × Maine 2 Clinton × ○ ○ ○ Alaska 3 Clinton × ○ ○ ○ Montana 3 Trump × ○ ○ ○ Trump’s Electoral Vote 306 292 235 216 215

† Electoral Votes. * Princeton Election Consortium.

41 / 57

SLIDE 42

Ideological Segregation

Parallel indexes as in Gentzkow and Shapiro (2011)

Sm = ∑

j∈Jm

( consj consm ⋅ consj visitsj ) − ∑

j∈Jm

( libj libm ⋅ consj visitsj )

For each news outlet j of type m (news outlets, politicians; tv,

newspapers; etc), we can calculate the share of conservative daily visitors and weight by the relative importance of that page inside the conservative or liberal campaign

0: All conservatives and liberals visits the same page
1: Conservatives only visits all conservative pages

42 / 57

SLIDE 43

Ideological Segregation

Jan 05 2015 Apr 06 2015 Jul 06 2015 Oct 05 2015 Jan 04 2016 Apr 04 2016 Jul 04 2016 Oct 03 2016 Jan 02 2017 Mar 27 2017 0.1 0.2 0.3 0.4 0.5 0.6

Weekly Isolation Index by Page Type (Max Like)

newspaper tv magazine website radio

43 / 57

SLIDE 44