What Makes Politics Interesting?

:
How Political Context Shapes Political Interest

Seonghui Lee
Rice University

EITM 2014
How interested would you say you are in politics?
 % Interested in Politics
30 40 50 60 70

france

luxembourg

belgium

ireland

finland

norway

uk

Country
austria

germany

sweden
ESS (country mean from w1-w5)

45% in France vs. 68% in Denmark

switzerland

iceland

netherlands

denmark
 ESS (country mean from w1-w5)
czech rep spain croatia portugal latvia greece
80
60
40
20

lithuania italy romania slovakia cyprus hungary

80
60
40
20
% Interested in Politics

estonia poland turkey slovenia france luxembourg

80
60
40
20

russia belgium ireland finland bulgaria ukraine

80
60
40
20

norway uk austria israel germany sweden

80
60
40
20
2002 04 06 08 2010 2002 04 06 08 2010

switzerland iceland netherlands denmark

80
60
40
20
2002 04 06 08 2010 2002 04 06 08 2010 2002 04 06 08 2010 2002 04 06 08 2010

Year
Y-axis: % Quite Interested + Very Interested
WVS (wave 4 or 5)
thailand
norway
viet nam
tanzania
israel
mali
ethiopia
switzerland
saudi arabia
china
japan
iraq
burkina faso
united states
germany
sweden
australia
new zealand
zambia
rwanda
georgia
nigeria
uganda
netherlands
philippines
canada
ukraine
ghana
brazil
kyrgyzstan
macedonia
iran
south africa
slovenia
great britain
india
bulgaria
cyprus
puerto rico
poland
serbia and montenegro
andorra
jordan
bangladesh
albania
bosnia and herzegovina
turkey
south korea
russian federation
finland
italy
singapore
moldova
uruguay
france
indonesia
trinidad and tobago
egypt
malaysia
mexico
morocco
serbia
spain
romania
zimbabwe
pakistan
taiwan
venezuela
argentina
algeria
chile
guatemala
colombia
peru
hong kong

10 20 30 40 50 60 70 80 90
% Interested in Politics
Research Question

I How can we explain the cross-national difference?

I Why politics is interesting to some people but not to others?
Motivation

I Probably the most important predictor of political sophistication

and participation
I Key aspect to understand political attitudes and political behavior
necessary for democratic citizenship, thus the quality of democracy
What do we know about political interest?

“... We do not understand where political interest comes from and could
thus not recommend how to increase it. We do not even know if political
interest has the stability of a personal trait or the volatility of a regularly
updated reflection of the contemporary political situation.” Prior 2010 JOP

“... We know relatively little about individuals’ interest in politics

beyond the simple correlation between interest and turnout or other
forms of participation. ...” Horner 2007
What do we know about political interest?

I A kind of political engagement: motivational factor of political

attention and learning e.g., Luskin 1990
I A kind of personal characteristics: stable over time Prior 2005,
developed in the formative years, influenced by parental
socialization e.g., Luskin 1990, Jennings et al 2004, Neundorf et al 2012
What do we know about political interest?

Luskin 1990
What do we know about political interest?

I A kind of political engagement: motivational factor of political

attention and learning e.g., Luskin 1990
I A kind of personal characteristics: stable over time Prior 2005,
developed in the formative years, influenced by parental
socialization e.g., Luskin 1990, Jennings et al 2004, Neundorf et al 2012

I Influenced by contexts: political and media environment (e.g.,

media systems, campaign effects, information costs) influence the
level of political interest Political communication literature, Media system
studies (e.g., van Aelst et al. 2012)

I Varies across countries: dramatic difference across countries (but

within country variance is much smaller)
What do we know about political interest?

ESS

Variable Mean Std. Dev. Min Max Obs.

Political Intest overall 2.359 .235 1.877 2.898 N = 130
between .224 1.989 2.802 n = 34
within .075 2.128 2.676 T-bar = 3.824
Prop. Interested overall .452 .119 .177 .718 N = 130
between .113 .219 .675 n = 34
within .037 .322 .613 T-bar = 3.824
Research Question

How can we explain the cross-national difference?

Possible Explanations for the Cross-national variation

I Measurement incomparability
I Socioeconomic differences (Cost-Benefit) e.g., van Deth & Elff 2004
I Political contexts
Research Question

How can we explain the cross-national difference?

Possible Explanations for the Cross-national variation

I Measurement incomparability
Tested the comparability of the survey instrument (DIF), using
anchoring vignettes technique in Western European countries
I Socioeconomic differences (Cost-Benefit) e.g., van Deth & Elff 2004
SES well explains variations at indi. level, but not cross-national
I Political contexts
A typical individual will be more interested in politics in some
contexts, but less in other. What are the contexts and why?
Outline

1 Theory of interest (micro- & macro level)

2 Testing the mechanism of micro-level theory
3 A comparative research
What is Interest?

I The feeling of wanting to know or learn about something or

someone (Oxford dictionary)
I A feeling of having your attention held by something, or of wanting
to be involved with and learn more about something (Cambridge)
I A feeling of intentness, concern, or curiosity about something
(Collins)
I A feeling of wanting to investigate, become involved, or extend or
expand the self by incorporating new information and having new
experiences with the person or object that has simulated interest
(Izard 1977)
Interest as an Emotion

A discrete experience of interest as a kind of positive emotions

I Nicknamed “knowledge emotion” or “curiosity emotion”

I Involves physiological expressions
Langsdorf et al. 1983, Libby et al. 1974, Banse & Scherer 1996
I still and tilt the head
I eyes and muscles in the forehead move (typical attention and
concentration gesture)
I a faster rate of speech, greater range of vocal frequency,
dilated pupils
Interest as an Emotion

A discrete experience of interest as a kind of positive emotions

I Nicknamed “knowledge emotion” or “curiosity emotion”

I Involves physiological expressions
Langsdorf et al. 1983, Libby et al. 1974, Banse & Scherer 1996

I Unique appraisal structure

e.g., Arnold 1960, Berlyne 1960, Roseman & Smith 2001, Sherer et al. 2001, Silvia
2006
Interest as an Emotion

Appraisal theory

I Explains why we experience a certain emotion

I We experience an emotion only after we (consciously or
subconsciously) appraise the object as matching a particular
abstract structure.
I Appraisal: an “intuitive” assessment of the “here and now” aspects
of situations Arnold 1960; a perceptual and cognitive process, often
proceeds effortlessly and generates emotions automatically
I Appraisal theorists attempt to identify the evaluative dimensions
or criteria that predict the emotion
e.g., Anger occurs when: 1) the event is relevant to my goal, 2) the event is
incongruent with the goal, 3) the event is a threat to my social or
self-esteem, and 4) there is someone I blame for the threat.
Appraisal Structure of Interest

Two-dimensional Structure (Paul Silvia 2006)

I Collative dimension: Novelty*, Complexity*, Uncertainty ...
I Coping potential (or Comprehensibility) dimension

Collative Dimension

Stimuli Feeling of
(Paintings,Text,
Music, etc.)
Interest
Comprehensibility/
Coping Potential
Appraisal Structure of Interest

CollativeNew?
Dimension

Stimuli Feeling of
(Paintings,Text,
Music, etc.)
Interest
Comprehensibility/
Can I cope with it?
Coping Potential
Appraisal Structure of Interest

New?

Ultracold Reactions Probe the Frontiers of

Quantum Chemistry

Can I cope with it?

Appraisal Structure of Interest

Interest = F(Θ, Π | stimulus)

Θ: a set of appraisals of collative variables

Π: appraisal of comprehensibility
Appraisal Theory of “Political” Interest

Key question: What makes politics “comprehensible”? (Π)

I Conventional thoughts (implied in the previous studies):

Knowledge and previous experiences (e.g., education and age)
Appraisal Theory of “Political” Interest

Appraisals of
Collative Variables
Features of Interest in
Political Political
Messages Messages
Appraisals of
Comprehensibility/
Coping Potential Feedback
Loop

Knowledge from
the Past

Interest = F(Θ, Π | stimulus) =⇒ F(Θ, K | stimulus)

Π: a function of previous knowledge
Appraisal Theory of “Political” Interest

Key question: What makes politics “comprehensible”? (Π)

I Conventional thoughts (implied in the previous studies):

Knowledge and previous experiences (e.g., education and age)

I An alternative path:
When political heuristics are available and useful, politics (events
and messages) can be more likely to be appraised as
comprehensible – something I can “cope with”.
Using heuristics, politics is “understandable” even without detailed
knowledge (vs. “understood”)
e.g., Chaiken 1980, Gigerenzer 2008, 2011a, 2011b, Tversky and Kahneman 1974
Heuristics and Comprehensibility

Heuristics
A simple rule that guides people to map an abstract feature and solve
(simple or complex) problems

Political Heuristics

1 Party label: Mapping candidate’s position

2 Ideological label: Mapping political parties’ position; Policy
outcome
3 Party size and Ideological proximity: Predicting coalition partner(s)
Heuristics and Comprehensibility

Appraisals of
Collative Variables
Features of Interest in
Political Political
Messages Messages
Appraisals of
Comprehensibility/
Coping Potential Feedback
Loop

Availability of Knowledge from

Heuristics the Past

Interest = F(Θ, Π | stimulus) =⇒ F(Θ, K, H | stimulus)

Π: a function of knowledge (K) & the avail. of heuristics (H)
Heuristics and Comprehensibility

Further Extension for Comparative Research

Appraisals of
Collative Variables
Features of Features of Extent of
Political System (Typical) Interest in
and Context Political Political
Messages Appraisals of Messages
Comprehensibility/
Coping Potential Feedback
Loop

Extended Availability of Knowledge from

Model Heuristics the Past
Heuristics, Comprehensibility, and Interest

Hypothesis
When heuristics are available, an object (e.g., political event or
message) will be more likely to be appraised as comprehensible, thus
more interesting.

Interest = F(Θ, Π | stimulus) =⇒ F(Θ, K, H | stimulus)

Empirical Test
Designed an experiment to test the mechanism
Experimental Design: Guessing Task

Considerations for Design:

a No prior knowledge or experience should be useful (ruling out K).
b Manipulate the usefulness/availability of heuristic (H).
c Minimize the confounding effects of collative variables (Θ).
d Ideal if it allows us to detect feedback chain (repeated trials).

Basic Setting:
I Task: “guessing” (canceling out K).

I Repeated trials (perhaps make Θ identical across groups).

Question 1: Two of the circles will colored in RED. Guess which TWO?
Answer 1
Question 2: Two of the circles will colored in RED. Guess which TWO?
Answer 2
Question 3: Two of the circles will colored in RED. Guess which TWO?
Answer 3
Experimental
game_g2 Design: Guessing Task

Guess  which  two  circles  will  be  colored  in  RED.  Choose  TWO  circles  from  the  below.
 

Guess which two circles will be colored in RED. Choose TWO from the below.

E
 <Screen'1>

Experimental Design: Guessing Task

Guess%which%two%circles%will%be%colored%in%RED.%Choose%TWO%circles%from%the%below.

Check if your guess was correct!

<Screen'2>

Check%if%your%guess%was%correct!

Circle A B C D E
Your Choice X X
Correct Answer X X

Did%you%correctly%guess%BOTH%of%the%red%circles?
o%%Yes,%I%got%both%correct.
o%%No,%I%didn’t%get%both%correct.
Experimental Design: Treatment

Do this 70 times.
Experimental Design: Treatment

Manipulation: The availability of heuristics

Manipulate the proportion a logical rule applied to the outcome
*Rule (Embedded): the largest circle and the closest to the largest one

High availability-of-heuristics group 87%

54%
24%
Low availability-of-heuristics group 10% (random prob.)

Thus, in the Low Availability group, there is no cue that helps to get the task easy and
comprehensible, whereas in the High Availability group, there is a regularity that makes
the task more comprehensible once a subject recognizes it.
Experimental Design

Activity: Guessing task, 70 trials

Subjects: 120 subjects (30 for each group), from Amazon MTurk

Compensation: Flat rate fee (no performance-based bonus)

Measurements
Interestingness (outcome variable): Ratings on Boring–Interesting
dimension, 11pt scale, at 13 time points over 70 trials
Comprehensibility (self-reporting): Ratings on Easy–Difficult
dimension, 11pt scale, at 13 time points
Comprehensibility (behavioral): Performance – an objective
measure from the number of correct guesses
Experimental Design

Measurements

Shown after trials 9, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, and 70.
Expectations

Interest = F(Θ, K, H | stimulus)

Yitg : individual i’s level of interestingness at time t, i in group g
Nt (Θ component): appraisal of Novelty (collative variable) at time t
Hg : availability/ usefulness of Heuristics in group g (0.1 ≤ Hg ≤ 1)
* Citg : coping potential (appraisal of comprehensibility)

I Individual level: Yitg = βHg + ρt + εi , (β > 0, ρ < 0)

PT PT
Group level: t=1 Yi,g=Hi > t=1 Yi,g=Lo (T=13)
I
Results: Individual Level

Yitg = βHg + ρt + εi , (β > 0, ρ < 0)

Variable Coefficient S.E.

Availability of Heuristics 2.422 (0.288)
Time (trial) -0.044 (0.005)
Intercept 3.924 (0.235)
N = 1560

Relaxing the linearity assumption for β (group effect) and ρ (time effect),

Yitg = (β1 H1 + ... + β4 H4 ) + (ρ1 t1 + ρ2 t2 + ... + ρ13 t13 ) + εi

Results: Individual Level

Group Effects (base=High)

group 2
group 3
group 4 (Low)
Time Effects (base=p1)
period 2
period 3
period 4
period 5
period 6
period 7
period 8
period 9
period 10
period 11
period 12
period 13
Constant
-4 -2 0 2 4 6
Estimated Effects on Interestingness
Results: Group Level

When heuristic is available, they perform better.

Proportion Correct Guesses (per period)

High Availability
.5

Low Availability
.4
.3
.2
.1
0

1 2 3 4 5 6 7 8 9 10 11 12 13
Time
Results: Group Level

The High Availability group feels more interested in the task than the
Low group.

Mean Ratings on Interestingness

6
4
2

High Availability
Low Availability
0

1 2 3 4 5 6 7 8 9 10 11 12 13
Time
Manipulation Checks

More than half of the subjects in high availability group recognized the
presence of the pattern and explicitly described what the pattern was,
but none in other group(s) did.
Some written responses from the High Availability Group...

I Biggest one was always red and the one closest to it was as well. There
was probably a 95% success rate with this and 5% random order if it
wasn’t.
I It seemed like the big one was always selected, and then the one that is
closest to that one.
I For the most part, it seemed to be that the ones in red were the largest
circle and the next closest circle.
I The big circle was almost always colored red. The other circle seemed
almost random.
...
One Simple Finding

The availability of heuristics helps to enhance (or maintain) levels

of interest, by increasing the appraised coping potential of
individuals.
Cross-National Research on Political Interest
Theory

Remember that the core mechanism of the micro-level theory suggests:

When political heuristics are available and useful (a)

→ Politics in general is more comprehensible (b)
→ People are more likely to be interested in politics (c)

The link between (b) and (c) can be explored by looking at the
relationship between comprehensibility and level of interest, using
survey questions “politics complicated” and “political interest.”
Relationship between Comprehensibility and Interest
ESS (2002-2010)

2.8 Western Europe denmark

Eastern Europe
Southern & Other germany
netherlands
switzerland
2.6

sweden
austria
Political Interest

ukraine norway
israel
uk
2.4

finland
russia france
estonia bulgaria
luxembourg
slovenia
poland belgiumslovakia ireland
hungary cyprus
2.2

italy
turkey
croatia
greece
2

czech rep
spain
portugal
3.4

3.2

2.8
Politics Complicated

Comprehensibility and Interest are positively associated.

Relationship between Comprehensibility and Interest
Latinobarometro (2005-2010)

2.4
uruguay
2.2

mexico venezuela
Political Interest

dominican republic
brazil argentina
2

colombia
paraguay
panama
honduras costa rica
el salvador peru
bolivia chile
guatemala
1.8

ecuador
nicaragua
1.6
.65

.6

.55

.5

.45
Politics Complicated

Comprehensibility and Interest are positively associated.

More on Theory

Remaining questions
I What are the (availability of) heuristics that systematically vary
across countries
I How to measure the availability of heuristics

Political heuristics are contextually generated. That is, a country’s

political system and its political context make (or do not make) a
specific kind of political heuristics available and/or useful.
More on Theory

Theoretical Framework for Comparative Research on Political Interest

Appraisals of
Collative Variables
Features of Features of Extent of
Political System (Typical) Interest in
and Context Political Political
Messages Appraisals of Messages
Comprehensibility/
Coping Potential Feedback
Loop

Extended Availability of Knowledge from

Model Heuristics the Past
More on Theory

To answer the question of “how political heuristics make politics more

comprehensible”, we need start to figure out “what heuristics can
substitute the kind of detailed knowledge when we deal with political
events (which predominantly delivered through media messages).”

1 What are the stimuli (types of political news)?

2 What should citizens know to understand politics?
3 How much effort they need to get the necessary information?
4 Are there heuristics available to substitute the kind of information
they need?
More on Theory

Political Messages
I Horse race: parties, leaders, individual candidates...
I Government formation: the connection between electoral
outcomes and government formation (who will take the office?)
I Policy: issues, process, conflicts, prediction of outcome
I (Gossip and scandal...)
How Heuristics Makes Politics More Comprehensible?

Type of Available Heuristics Measurements

Information
Individual Party Label (when parties are Party Unity Score (e.g. Rice In-
Politicians’ more coherent; parties are dex); Ages of Parties, the Num-
Standpoint more institutionalized; parties ber of New Parties); Propor-
play a center role in producing tion of Independents; % Party
electoral candidates) Switchers
Political Left-Right Heuristics (when Importance of LR dimension;
Parties’ party competition centered on Importance of LR issues; Effec-
Standpoint LR dimension; LR dimension tive number of LR dimension
composed of fewer issues)
Prediction Coalition Formation Heuristics Electoral system; Regularity
of Policy (when it is easy to predict and predictability of coalition
Outcomes winners from election results; formation
there are regularities (patterns)
in coalition formation)
Summary and Hypothesis

My theory suggests that the availability of heuristics will better explain

the cross-national difference (because it us systematically driven by
political systems and contexts), whereas the knowledge (and the kind of
SES variables) will be a good predictor of levels of interest at the
individual level but not at the country level. Controlling for the latter, I
expect:

The level of political interest will be higher where contextually

generated political heuristics are available and useful.
Data and Measurements

Dependent variable: Political interest (measured on 1 to 4 pt scale),

data from various cross-national surveys

Explanatory variables: the political contexts where political heuristics

are more likely to be available and useful (listed in the Table), data from
various sources
Data and Measurements
A quick glance: relationship between Interest and Contexts

Effective Number of Issue Dimensions (−)

2.8

denmark
austria
sweden norway
Politics Complicated
3

germany
ireland france
luxembourg netherlands

belgium
3.2

uk
italy

spain
portugal
3.4

finland
greece
2.4

2.6

2.8

3.2
Average Effective Number of Issue Dimensions Since 1945 (ENID, Nyblade 2004)
Data and Measurements
A quick glance: relationship between Interest and Contexts

Number of New Parties (−)

2.8

denmark

netherlands germany
switzerland
2.6

sweden
austria
Political Interest

norway
uk
2.4

finland

luxembourgfrance
belgium
ireland
2.2

italy

greece
2

portugal spain
0

6
Average Number of New Parties from Tavits (2006)
Data and Measurements
A quick glance: relationship between Interest and Contexts

Importance of LR Issues (+)

2.8

israel

denmark
austria
norway
sweden
Politics Complicated
3

bulgaria
germany switzerland
ireland
france
netherlands
luxembourg
estonia

slovenia slovakia
belgium
hungary
3.2

croatia
ukraine
russia
uk
italy

spain poland
portugal
turkey
3.4

czech rep finland

greece
40

50

60

70
Average Importance of LR Issues Since 1991 (CMP)
Data and Measurements

Dependent variable: Political interest (measured on 1 to 4 pt scale),

data from various cross-national surveys

Explanatory variables: the political contexts where political heuristics

are more likely to be available and useful (listed in the Table), data from
various sources

Control variables
I Level 1: Demographic and SES variables for individuals
I Level 2: Socio-econ vars (e.g., GDP, % tertiary education, etc.),
Electoral and Political systems (from the literature, e.g., PR vs.
SMD, ballot type, ENOP, District Magnitude, etc.)
e.g., Gordon & Segura 1997, van Deth & Elff 2004
Modeling Strategy

Approach 1: a general form

Yik = β0k + β1k Ck + β2k Xik

Yik : level of interest of an individual i in country k

Ck : contexts
Xik : characteristics of individuals
Modeling Strategy

Approach 2: two stage estimation strategy

First stage: Hierarchical random intercept model
For the observed response yi ,

if yi∗ ≤ κ1

 1
if κ1 < yi∗ ≤ κ2

2

yi = , where yi∗ is latent response.
 3 if κ2 < yi∗ ≤ κ3
if κ3 < yi∗

4


Random intercept model for subject i nested in survey (year) j nested in country
k, including random intercepts µj and µjk ,

∗
yijk = β 0 Xij + µj + µjk + ijk ,
where µj ∼ N(0, τ 2 ), and µj , µjk ⊥ ijk .
Modeling Strategy

From the random intercept model, the residual (not explained by a set of
individual level covariates X) are obtained with uncertainty measure. The
country (and survey) level random intercept is the outcome variable in the
second stage estimation.

Second stage model

CEk = α0 + α1 Ck + α2 Zk

CEk : context-specific random effects (random intercept) obtained from the first
stage model
Ck : contexts relevant to heuristics use
Zk : other contexts
Next Step

- More on comparative theory/ hypotheses

- Estimation strategy, variable selection

 What Makes Politics Interesting?:
How Political Context Shapes Political Interest

Seonghui Lee
Rice University

EITM 2014