Brute force effects of mass media presence and social media activity on electoral outcome

  • View
    82

  • Download
    0

  • Category

    Science

Preview:

Citation preview

Brute force e�ects of mass media presenceand social media activity on electoral

outcome

Marko Kovic1; Adrian Rauch�eisch2; Julia Metag3; ChristianCaspar4; Julian Szenogrady5

May 11, 2016

ZIPARZurich Institute of Public Affairs Research

1ZIPAR / marko.kovic@zipar.org2ZIPAR; Institute of Mass Communication and Media Research, University of Zurich3Institute of Mass Communication and Media Research, University of Zurich4Research Institute for the Public Sphere and Society, University of Zurich5Institute of Mass Communication and Media Research, University of Zurich

ZIPAROutlineWhat I will be talking about

I Introduction: What are «brute force e�ects»?I The Swiss electoral systemI Data and methodsI ResultsI Discussion

ZIPAROutlineWhat I will be talking about

I Introduction: What are «brute force e�ects»?

I The Swiss electoral systemI Data and methodsI ResultsI Discussion

ZIPAROutlineWhat I will be talking about

I Introduction: What are «brute force e�ects»?I The Swiss electoral system

I Data and methodsI ResultsI Discussion

ZIPAROutlineWhat I will be talking about

I Introduction: What are «brute force e�ects»?I The Swiss electoral systemI Data and methods

I ResultsI Discussion

ZIPAROutlineWhat I will be talking about

I Introduction: What are «brute force e�ects»?I The Swiss electoral systemI Data and methodsI Results

I Discussion

ZIPAROutlineWhat I will be talking about

I Introduction: What are «brute force e�ects»?I The Swiss electoral systemI Data and methodsI ResultsI Discussion

ZIPARIntroductionWhat are brute force e�ects?

ZIPARIntroductionWhat are brute force e�ects?

ZIPARIntroductionWhat are brute force e�ects?

We are witnessing a historic event!

I Many a conventional wisdom has been proven wrong.I One of them: Spending on advertising.

ZIPARIntroductionWhat are brute force e�ects?

We are witnessing a historic event!

I Many a conventional wisdom has been proven wrong.

I One of them: Spending on advertising.

ZIPARIntroductionWhat are brute force e�ects?

We are witnessing a historic event!

I Many a conventional wisdom has been proven wrong.I One of them: Spending on advertising.

ZIPARIntroductionWhat are brute force e�ects?

ZIPARIntroductionWhat are brute force e�ects?

Donald Trump has done something interesting!

I Maximize presence in mass media news reporting.

I Disregard whether they love or loathe you.

I Be very active on social media.I ... win!

ZIPARIntroductionWhat are brute force e�ects?

Donald Trump has done something interesting!

I Maximize presence in mass media news reporting.

I Disregard whether they love or loathe you.

I Be very active on social media.I ... win!

ZIPARIntroductionWhat are brute force e�ects?

Donald Trump has done something interesting!

I Maximize presence in mass media news reporting.I Disregard whether they love or loathe you.

I Be very active on social media.I ... win!

ZIPARIntroductionWhat are brute force e�ects?

Donald Trump has done something interesting!

I Maximize presence in mass media news reporting.I Disregard whether they love or loathe you.

I Be very active on social media.

I ... win!

ZIPARIntroductionWhat are brute force e�ects?

Donald Trump has done something interesting!

I Maximize presence in mass media news reporting.I Disregard whether they love or loathe you.

I Be very active on social media.I ... win!

ZIPARIntroductionWhat are brute force e�ects?

Could it be that the sheer volume of mass media presence and socialmedia activity matters?

I We humans «su�er» from many cognitive heuristics.I Most of our thinking is «fast»: Automated and prone to errors.I One family of such biases is related to exposure to information.

I Most famously: Mere exposure e�ect.

ZIPARIntroductionWhat are brute force e�ects?

Could it be that the sheer volume of mass media presence and socialmedia activity matters?

I We humans «su�er» from many cognitive heuristics.

I Most of our thinking is «fast»: Automated and prone to errors.I One family of such biases is related to exposure to information.

I Most famously: Mere exposure e�ect.

ZIPARIntroductionWhat are brute force e�ects?

Could it be that the sheer volume of mass media presence and socialmedia activity matters?

I We humans «su�er» from many cognitive heuristics.I Most of our thinking is «fast»: Automated and prone to errors.

I One family of such biases is related to exposure to information.

I Most famously: Mere exposure e�ect.

ZIPARIntroductionWhat are brute force e�ects?

Could it be that the sheer volume of mass media presence and socialmedia activity matters?

I We humans «su�er» from many cognitive heuristics.I Most of our thinking is «fast»: Automated and prone to errors.I One family of such biases is related to exposure to information.

I Most famously: Mere exposure e�ect.

ZIPARIntroductionWhat are brute force e�ects?

Could it be that the sheer volume of mass media presence and socialmedia activity matters?

I We humans «su�er» from many cognitive heuristics.I Most of our thinking is «fast»: Automated and prone to errors.I One family of such biases is related to exposure to information.

I Most famously: Mere exposure e�ect.

ZIPARIntroductionWhat are brute force e�ects?

Brute force e�ects: Positive electoral impact of the mere volume ofmass media presence and social media activity.

I Brute force e�ects are not rational!

I Purely rational voting: Utility maximization.I Purely non-rational voting: Preference generation through

heuristics.I Overlap: Mobilization through heuristics.

I Brute force e�ects are either purely non-rational, or an overlap ofrational and non-rational.

ZIPARIntroductionWhat are brute force e�ects?

Brute force e�ects: Positive electoral impact of the mere volume ofmass media presence and social media activity.

I Brute force e�ects are not rational!

I Purely rational voting: Utility maximization.I Purely non-rational voting: Preference generation through

heuristics.I Overlap: Mobilization through heuristics.

I Brute force e�ects are either purely non-rational, or an overlap ofrational and non-rational.

ZIPARIntroductionWhat are brute force e�ects?

Brute force e�ects: Positive electoral impact of the mere volume ofmass media presence and social media activity.

I Brute force e�ects are not rational!I Purely rational voting: Utility maximization.

I Purely non-rational voting: Preference generation throughheuristics.

I Overlap: Mobilization through heuristics.

I Brute force e�ects are either purely non-rational, or an overlap ofrational and non-rational.

ZIPARIntroductionWhat are brute force e�ects?

Brute force e�ects: Positive electoral impact of the mere volume ofmass media presence and social media activity.

I Brute force e�ects are not rational!I Purely rational voting: Utility maximization.I Purely non-rational voting: Preference generation through

heuristics.

I Overlap: Mobilization through heuristics.

I Brute force e�ects are either purely non-rational, or an overlap ofrational and non-rational.

ZIPARIntroductionWhat are brute force e�ects?

Brute force e�ects: Positive electoral impact of the mere volume ofmass media presence and social media activity.

I Brute force e�ects are not rational!I Purely rational voting: Utility maximization.I Purely non-rational voting: Preference generation through

heuristics.I Overlap: Mobilization through heuristics.

I Brute force e�ects are either purely non-rational, or an overlap ofrational and non-rational.

ZIPARIntroductionWhat are brute force e�ects?

Brute force e�ects: Positive electoral impact of the mere volume ofmass media presence and social media activity.

I Brute force e�ects are not rational!I Purely rational voting: Utility maximization.I Purely non-rational voting: Preference generation through

heuristics.I Overlap: Mobilization through heuristics.

I Brute force e�ects are either purely non-rational, or an overlap ofrational and non-rational.

ZIPARIntroductionWhat are brute force e�ects?

R NR

BF

Figure: R stands for the set of rational voting, NR for the set of non-rationalvoting, and BF for the subset of brute force e�ects.

ZIPARIntroductionWhat are brute force e�ects?

R NR

BF

Figure: R stands for the set of rational voting, NR for the set of non-rationalvoting, and BF for the subset of brute force e�ects.

ZIPARElections in SwitzerlandA seismograph of voter preferences

I Latest national elections: October 18, 2015.I Lower and upper house.

I Lower house: 200 seats; proportional vote; cantons are votingdistricts.

I Upper house: 46 seats; majority vote; cantons are votingdistricts.

I The lower house elections are of interest in this study.

ZIPARElections in SwitzerlandA seismograph of voter preferences

I Latest national elections: October 18, 2015.

I Lower and upper house.

I Lower house: 200 seats; proportional vote; cantons are votingdistricts.

I Upper house: 46 seats; majority vote; cantons are votingdistricts.

I The lower house elections are of interest in this study.

ZIPARElections in SwitzerlandA seismograph of voter preferences

I Latest national elections: October 18, 2015.I Lower and upper house.

I Lower house: 200 seats; proportional vote; cantons are votingdistricts.

I Upper house: 46 seats; majority vote; cantons are votingdistricts.

I The lower house elections are of interest in this study.

ZIPARElections in SwitzerlandA seismograph of voter preferences

I Latest national elections: October 18, 2015.I Lower and upper house.

I Lower house: 200 seats; proportional vote; cantons are votingdistricts.

I Upper house: 46 seats; majority vote; cantons are votingdistricts.

I The lower house elections are of interest in this study.

ZIPARElections in SwitzerlandA seismograph of voter preferences

I Latest national elections: October 18, 2015.I Lower and upper house.

I Lower house: 200 seats; proportional vote; cantons are votingdistricts.

I Upper house: 46 seats; majority vote; cantons are votingdistricts.

I The lower house elections are of interest in this study.

ZIPARElections in SwitzerlandA seismograph of voter preferences

I Latest national elections: October 18, 2015.I Lower and upper house.

I Lower house: 200 seats; proportional vote; cantons are votingdistricts.

I Upper house: 46 seats; majority vote; cantons are votingdistricts.

I The lower house elections are of interest in this study.

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.I Voters can do a number of things with lists:

I Vote for a list as is.I Remove candidates from a list.I Add candidates from other lists→ panachageI Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.

I Voters can do a number of things with lists:

I Vote for a list as is.I Remove candidates from a list.I Add candidates from other lists→ panachageI Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.I Voters can do a number of things with lists:

I Vote for a list as is.I Remove candidates from a list.I Add candidates from other lists→ panachageI Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.I Voters can do a number of things with lists:

I Vote for a list as is.

I Remove candidates from a list.I Add candidates from other lists→ panachageI Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.I Voters can do a number of things with lists:

I Vote for a list as is.I Remove candidates from a list.

I Add candidates from other lists→ panachageI Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.I Voters can do a number of things with lists:

I Vote for a list as is.I Remove candidates from a list.I Add candidates from other lists→ panachage

I Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

How lower house elections work:

I There are voting lists for each canton.I Voters can do a number of things with lists:

I Vote for a list as is.I Remove candidates from a list.I Add candidates from other lists→ panachageI Put a candidate’s name twice on a list→ cumulation

ZIPARElections in SwitzerlandA seismograph of voter preferences

This results in very interesting data!

For each candidate, we know:

I Whether they got elected or not.I How many votes they received.I How many panachage votes they received.I How many cumulation votes they received.

ZIPARElections in SwitzerlandA seismograph of voter preferences

This results in very interesting data!

For each candidate, we know:

I Whether they got elected or not.I How many votes they received.I How many panachage votes they received.I How many cumulation votes they received.

ZIPARElections in SwitzerlandA seismograph of voter preferences

This results in very interesting data!

For each candidate, we know:

I Whether they got elected or not.

I How many votes they received.I How many panachage votes they received.I How many cumulation votes they received.

ZIPARElections in SwitzerlandA seismograph of voter preferences

This results in very interesting data!

For each candidate, we know:

I Whether they got elected or not.I How many votes they received.

I How many panachage votes they received.I How many cumulation votes they received.

ZIPARElections in SwitzerlandA seismograph of voter preferences

This results in very interesting data!

For each candidate, we know:

I Whether they got elected or not.I How many votes they received.I How many panachage votes they received.

I How many cumulation votes they received.

ZIPARElections in SwitzerlandA seismograph of voter preferences

This results in very interesting data!

For each candidate, we know:

I Whether they got elected or not.I How many votes they received.I How many panachage votes they received.I How many cumulation votes they received.

ZIPARElections in SwitzerlandA seismograph of voter preferences

Accordingly, we test a number of hypotheses:

I H1: BFE have no substantial impact on the probability of beingelected.

I H2: BFE have no substantial impact on the total amount ofvotes.

I H3: BFE have a substantial impact on the amount of panachagevotes.

I H4: BFE have a substantial impact on the amount of cumulationvotes.

I H5: Mass media presence exerts a stronger BFE than socialmedia activity.

ZIPARElections in SwitzerlandA seismograph of voter preferences

Accordingly, we test a number of hypotheses:

I H1: BFE have no substantial impact on the probability of beingelected.

I H2: BFE have no substantial impact on the total amount ofvotes.

I H3: BFE have a substantial impact on the amount of panachagevotes.

I H4: BFE have a substantial impact on the amount of cumulationvotes.

I H5: Mass media presence exerts a stronger BFE than socialmedia activity.

ZIPARElections in SwitzerlandA seismograph of voter preferences

Accordingly, we test a number of hypotheses:

I H1: BFE have no substantial impact on the probability of beingelected.

I H2: BFE have no substantial impact on the total amount ofvotes.

I H3: BFE have a substantial impact on the amount of panachagevotes.

I H4: BFE have a substantial impact on the amount of cumulationvotes.

I H5: Mass media presence exerts a stronger BFE than socialmedia activity.

ZIPARElections in SwitzerlandA seismograph of voter preferences

Accordingly, we test a number of hypotheses:

I H1: BFE have no substantial impact on the probability of beingelected.

I H2: BFE have no substantial impact on the total amount ofvotes.

I H3: BFE have a substantial impact on the amount of panachagevotes.

I H4: BFE have a substantial impact on the amount of cumulationvotes.

I H5: Mass media presence exerts a stronger BFE than socialmedia activity.

ZIPARElections in SwitzerlandA seismograph of voter preferences

Accordingly, we test a number of hypotheses:

I H1: BFE have no substantial impact on the probability of beingelected.

I H2: BFE have no substantial impact on the total amount ofvotes.

I H3: BFE have a substantial impact on the amount of panachagevotes.

I H4: BFE have a substantial impact on the amount of cumulationvotes.

I H5: Mass media presence exerts a stronger BFE than socialmedia activity.

ZIPARElections in SwitzerlandA seismograph of voter preferences

Accordingly, we test a number of hypotheses:

I H1: BFE have no substantial impact on the probability of beingelected.

I H2: BFE have no substantial impact on the total amount ofvotes.

I H3: BFE have a substantial impact on the amount of panachagevotes.

I H4: BFE have a substantial impact on the amount of cumulationvotes.

I H5: Mass media presence exerts a stronger BFE than socialmedia activity.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.I 873 candidates.I 35 seats in the lower house.I Final four campaign weeks.I Media sample: 12 newspapers; 6 TV news shows; 3 radio news

shows.I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.

I 873 candidates.I 35 seats in the lower house.I Final four campaign weeks.I Media sample: 12 newspapers; 6 TV news shows; 3 radio news

shows.I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.I 873 candidates.

I 35 seats in the lower house.I Final four campaign weeks.I Media sample: 12 newspapers; 6 TV news shows; 3 radio news

shows.I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.I 873 candidates.I 35 seats in the lower house.

I Final four campaign weeks.I Media sample: 12 newspapers; 6 TV news shows; 3 radio news

shows.I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.I 873 candidates.I 35 seats in the lower house.I Final four campaign weeks.

I Media sample: 12 newspapers; 6 TV news shows; 3 radio newsshows.

I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.I 873 candidates.I 35 seats in the lower house.I Final four campaign weeks.I Media sample: 12 newspapers; 6 TV news shows; 3 radio news

shows.

I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Data collection:

I Canton of Zurich: Largest canton by population.I 873 candidates.I 35 seats in the lower house.I Final four campaign weeks.I Media sample: 12 newspapers; 6 TV news shows; 3 radio news

shows.I Social media sample: Facebook, Twitter.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.

I Total: 1’685 individual occurrences.

I Social media activity:

I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:

I Total: 49’869 Facebook resonance activities; 15’887 Twitterresonance activities.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.

I Total: 1’685 individual occurrences.

I Social media activity:

I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:

I Total: 49’869 Facebook resonance activities; 15’887 Twitterresonance activities.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.I Total: 1’685 individual occurrences.

I Social media activity:

I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:

I Total: 49’869 Facebook resonance activities; 15’887 Twitterresonance activities.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.I Total: 1’685 individual occurrences.

I Social media activity:

I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:

I Total: 49’869 Facebook resonance activities; 15’887 Twitterresonance activities.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.I Total: 1’685 individual occurrences.

I Social media activity:I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:

I Total: 49’869 Facebook resonance activities; 15’887 Twitterresonance activities.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.I Total: 1’685 individual occurrences.

I Social media activity:I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:

I Total: 49’869 Facebook resonance activities; 15’887 Twitterresonance activities.

ZIPARData and methodsWhat and how

Variables for data analysis:

I Mass media presence of individual candidates.I Total: 1’685 individual occurrences.

I Social media activity:I Total: 1’595 Facebook posts; 7’036 tweets.

I Social media resonance:I Total: 49’869 Facebook resonance activities; 15’887 Twitter

resonance activities.

ZIPARData and methodsWhat and how

Additional variables for data analysis:

I IncumbencyI GenderI AgeI Voting list rank

ZIPARData and methodsWhat and how

Additional variables for data analysis:

I Incumbency

I GenderI AgeI Voting list rank

ZIPARData and methodsWhat and how

Additional variables for data analysis:

I IncumbencyI Gender

I AgeI Voting list rank

ZIPARData and methodsWhat and how

Additional variables for data analysis:

I IncumbencyI GenderI Age

I Voting list rank

ZIPARData and methodsWhat and how

Additional variables for data analysis:

I IncumbencyI GenderI AgeI Voting list rank

ZIPARData and methodsWhat and how

Data analysis:

I Multilevel-models with varying intercepts.

I Voting list = group level variable.

I Bayesian models with vague priors.I 1 x binary response; 3 x metric.

ZIPARData and methodsWhat and how

Data analysis:

I Multilevel-models with varying intercepts.

I Voting list = group level variable.

I Bayesian models with vague priors.I 1 x binary response; 3 x metric.

ZIPARData and methodsWhat and how

Data analysis:

I Multilevel-models with varying intercepts.I Voting list = group level variable.

I Bayesian models with vague priors.I 1 x binary response; 3 x metric.

ZIPARData and methodsWhat and how

Data analysis:

I Multilevel-models with varying intercepts.I Voting list = group level variable.

I Bayesian models with vague priors.

I 1 x binary response; 3 x metric.

ZIPARData and methodsWhat and how

Data analysis:

I Multilevel-models with varying intercepts.I Voting list = group level variable.

I Bayesian models with vague priors.I 1 x binary response; 3 x metric.

ZIPARResultsBeing elected

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−0.1 0.0 0.1 −0.2 −0.1 0.0 0.1

0.0000 0.0025 0.0050 0.0 2.5 5.0 7.5 10.0

−1 0 1 0.0 0.2 0.4 0.6

0 2 4 6 −0.10 −0.05 0.00

0.00 0.01 0.02 0.03 −0.4 −0.2 0.0

ZIPARResultsBeing elected

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−0.1 0.0 0.1 −0.2 −0.1 0.0 0.1

0.0000 0.0025 0.0050 0.0 2.5 5.0 7.5 10.0

−1 0 1 0.0 0.2 0.4 0.6

0 2 4 6 −0.10 −0.05 0.00

0.00 0.01 0.02 0.03 −0.4 −0.2 0.0

ZIPARResultsBeing elected

Parameter Mean 2.5% quantile 97.5% quantile R̂

Media - broadcasting −0.15 −0.83 0.63 1.00Media - newspapers 0.24 0.11 0.39 1.00Facebook - activity −0.03 −0.12 0.05 1.00Facebook - resonance 0.00 0.00 0.00 1.00Twitter - activity −0.02 −0.06 0.01 1.00Twitter - resonance 0.01 0.00 0.02 1.00Age 0.01 −0.05 0.07 1.00Incumbency 4.36 2.23 6.67 1.00Voting list rank −0.11 −0.27 0.00 1.00Sex (female) 2.24 0.54 4.24 1.00

ZIPARResultsNumber of votes

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−4 0 4 −20 0 20

1 2 3 4 14000 16000 18000 20000

0 500 1000 100 150 200 250 300

−100 0 100 200−10 −5 0 5 10

−2.5 0.0 2.5 5.0−35 −30 −25 −20 −15

ZIPARResultsNumber of votes

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−4 0 4 −20 0 20

1 2 3 4 14000 16000 18000 20000

0 500 1000 100 150 200 250 300

−100 0 100 200−10 −5 0 5 10

−2.5 0.0 2.5 5.0−35 −30 −25 −20 −15

ZIPARResultsNumber of votes

Parameter Mean 2.5% quantile 97.5% quantile R̂

Media - broadcasting 362.94 −70.09 1009.30 1.00Media - newspapers 165.94 114.21 232.92 1.00Facebook - activity −7.33 −18.58 5.46 1.00Facebook - resonance 2.14 1.88 2.56 1.01Twitter - activity 1.23 −3.72 7.05 1.00Twitter - resonance −0.16 −2.19 2.24 1.00Age −0.52 −3.42 2.33 1.00Incumbency 16784.52 15623.83 18023.62 1.00Voting list rank −23.24 −27.25 −19.32 1.00Sex (female) 34.39 −40.94 110.68 1.00

ZIPARResultsPanachage votes

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−3 −2 −1 0 1 −10 −5 0 5 10

−0.3 0.0 0.3 0.6 2000 4000 6000 8000 10000

−100 0 100 200 25 50 75 100 125

0 50 100 −2 0 2 4

−1.0 −0.5 0.0 0.5 1.0 −10 −8 −6

ZIPARResultsPanachage votes

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−3 −2 −1 0 1 −10 −5 0 5 10

−0.3 0.0 0.3 0.6 2000 4000 6000 8000 10000

−100 0 100 200 25 50 75 100 125

0 50 100 −2 0 2 4

−1.0 −0.5 0.0 0.5 1.0 −10 −8 −6

ZIPARResultsPanachage votes

Parameter Mean 2.5% quantile 97.5% quantile R̂

Media - broadcasting 76.13 −15.88 155.54 1.00Media - newspapers 82.67 57.47 102.72 1.00Facebook - activity −1.36 −5.30 2.75 1.00Facebook - resonance 0.32 0.03 0.54 1.01Twitter - activity 0.21 −0.55 1.69 1.00Twitter - resonance −0.01 −0.35 0.28 1.00Age −0.49 −1.63 0.63 1.00Incumbency 6615.75 4749.08 8259.01 1.05Voting list rank −8.12 −9.79 −6.48 1.00Sex (female) 25.07 −6.26 57.20 1.00

ZIPARResultsCumulation votes

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−3 −2 −1 0 1 2 −10 −5 0 5 10

−0.4 0.0 0.4 0.8 4000 5000 6000 7000 8000

−200 0 200 400 0 20 40 60 80

0 50 100 −2 0 2 4

−2 −1 0 1 −10 −8 −6 −4

ZIPARResultsCumulation votes

Age Facebook − activity

Facebook − resonance Incumbency

Media − broadcasting Media − newspapers

Sex (female) Twitter − activity

Twitter − resonance Voting list rank

−3 −2 −1 0 1 2 −10 −5 0 5 10

−0.4 0.0 0.4 0.8 4000 5000 6000 7000 8000

−200 0 200 400 0 20 40 60 80

0 50 100 −2 0 2 4

−2 −1 0 1 −10 −8 −6 −4

ZIPARResultsCumulation votes

Parameter Mean 2.5% quantile 97.5% quantile R̂

Media - broadcasting 90.18 −30.80 231.92 1.10Media - newspapers 41.72 22.00 64.35 1.00Facebook - activity −1.20 −5.62 3.69 1.01Facebook - resonance 0.28 0.00 0.49 1.01Twitter - activity 0.43 −1.25 2.66 1.01Twitter - resonance −0.16 −0.98 0.55 1.01Age −0.41 −1.59 0.78 1.00Incumbency 5724.66 4060.00 7748.60 1.01Voting list rank −7.32 −8.97 −5.73 1.00Sex (female) 24.80 −6.36 56.11 1.01

ZIPARDiscussionNot so brute and not so forceful?

What have we learned? Take home messages:

I In terms of being elected or not, BFE are unlikely to matter.I However, for the total amount of votes, the amount of

panachage and the amount of cumulation votes, BFE mightmatter!

I Traditional media is still king.

I Newspapers show the most substantial and most persistente�ect.

I Social media resonance seems to be more important than directactivity.

ZIPARDiscussionNot so brute and not so forceful?

What have we learned? Take home messages:

I In terms of being elected or not, BFE are unlikely to matter.

I However, for the total amount of votes, the amount ofpanachage and the amount of cumulation votes, BFE mightmatter!

I Traditional media is still king.

I Newspapers show the most substantial and most persistente�ect.

I Social media resonance seems to be more important than directactivity.

ZIPARDiscussionNot so brute and not so forceful?

What have we learned? Take home messages:

I In terms of being elected or not, BFE are unlikely to matter.I However, for the total amount of votes, the amount of

panachage and the amount of cumulation votes, BFE mightmatter!

I Traditional media is still king.

I Newspapers show the most substantial and most persistente�ect.

I Social media resonance seems to be more important than directactivity.

ZIPARDiscussionNot so brute and not so forceful?

What have we learned? Take home messages:

I In terms of being elected or not, BFE are unlikely to matter.I However, for the total amount of votes, the amount of

panachage and the amount of cumulation votes, BFE mightmatter!

I Traditional media is still king.

I Newspapers show the most substantial and most persistente�ect.

I Social media resonance seems to be more important than directactivity.

ZIPARDiscussionNot so brute and not so forceful?

What have we learned? Take home messages:

I In terms of being elected or not, BFE are unlikely to matter.I However, for the total amount of votes, the amount of

panachage and the amount of cumulation votes, BFE mightmatter!

I Traditional media is still king.I Newspapers show the most substantial and most persistent

e�ect.

I Social media resonance seems to be more important than directactivity.

ZIPARDiscussionNot so brute and not so forceful?

What have we learned? Take home messages:

I In terms of being elected or not, BFE are unlikely to matter.I However, for the total amount of votes, the amount of

panachage and the amount of cumulation votes, BFE mightmatter!

I Traditional media is still king.I Newspapers show the most substantial and most persistent

e�ect.I Social media resonance seems to be more important than direct

activity.

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...

I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.

I Human cognition and, with it, heuristics and biases areuniversal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPAROutlookWhat’s next in light of the results?

What should happen next?

I More research is needed...I But this time, that’s true!

I Brute force e�ects are universal phenomena.I Human cognition and, with it, heuristics and biases are

universal as well!

I A lot of what is going on in voting behavior can be described interms of behavioral economics.

I So let’s do so more often!

ZIPARThanks!

Thank you for your attention!

ZIPARThe modelsBinary outcome

yi ∼ Bernoulli(µ)µ = logistic(β0i|j + β1x1i + ...+ βkxki)

logistic(x) = 1(1+exp(−x))

ZIPARThe modelsMetric outcome

yi ∼ t(µ, σ2, ν)µ = (β0i|j + β1x1i + ...+ βkxki)

σ2 = 1τ

ZIPARThe modelsPriors

β0i|j ∼ N (0, 0.001)βk ∼ N (0, 0.001)τ ∼ γ(0.01, 0.01)

ν ∼ exp(1/29, shift = 1)