Social Network and Voting Behavior in reference to a differential effect of a multi-party versus a two-party system: With a Comparison between Japan and US Election data

1
Social Network and Voting Behavior in reference
to a differential effect of a multi-party versus
a two-party systemWith a Comparison between
Japan and US Election data

• ELECTORAL SYSTEMS AND ELECTORAL POLITICS
  Bangalore CSES Workshop Planning Committee
  Meeting
• 2- 4 November 2006
• Kenichi Ikeda
• The University of Tokyo

2
Abstract

• In a comparative analysis of American and
  Japanese voting behavior, the effects of social
  networks on voting behavior and the expression of
  political opinion were investigated.
• Using an Interpersonal Political Environment
  (IPE) approach, which assumes that social
  networks form influential environments with
  respect to such behavior, a consistent pattern of
  influence via social networks was detected in
  both countries.
• A similar pattern is also observed with respect
  to voluntary expressions of opinion about
  politicians.
• Further analyses on Japanese data assures that
  this IPE effect reflects voluntary judgment of
  voters.
• Then, the effect of social networks seems to be
  caused by the acceptance of viewpoint of others
  within IPE as a political reality (which helps
  voluntary decision to vote).
• Effect was larger in the US than in Japan
• The difference is partly attributable to
  institutional difference between the US and
  Japan, specifically the political party
  configuration

3
Interpersonal influence in U.S.

• Findings of interpersonal influence on voting
  patterns have been well established in the US
  (Lazarsfeld et al., 1948)
• The influence is more systematically investigated
  in recent studies using social network
  batteries to measure daily interpersonal
  political communication (Huckfeldt Sprague,
  1995).

4
Japanese Voting Peculiarity

• Japan is often described as having collectivistic
  culture, which means interpersonal influence is
  substantial.
• In this sense, the same pattern of interpersonal
  influence as in US could be observed. However,
  the meaning of influence might have been
  different.
• Previous studies of Japanese elections
• Voters are often described as passively
  mobilized to vote in traditional contexts
• Richardson (1991) points out that in Japanese
  elections in the 1970s, peoples voting support
  is sought via mobilization of social networks and
  group affiliations by influence communications.
  
• ?These communications are interpersonal and
  organizational communications, designed to
  directly mobilize and manipulate voting support
  through activation of personal obligations,
  feelings of deference, or other kinds of
  sentiment pertaining to specific ongoing social
  relationships that extend well beyond any given
  election campaign (p. 339).
• ?Often attributed to factors that are peculiar
  to Japanese collectivistic culture. Is it true?

5
Generalization and more

• Exploration of the mechanism of any similarities
  and differences is needed.
• Because,
• Decline of collectivism in Japan
• Japanese are no more collectivistic than
  Americans (Oyserman et al., 2002 Matsumoto,
  1999)
• Huge structural changes in Japan
• From multi-member single nontransferable vote
  system to mixed but basically SMD system
• Beyond peculiarity contention, the
  micro-interpersonal structural effect on both
  Japanese and American voting behavior could be
  similar.
• In this context, I will analyze the interpersonal
  effects in Japan and US.

6
The IPE effect as a hypothesis of network effect

• An individual is embedded in a network, which is
  composed of others who are connected to the given
  individual through ties.
• We refer to this network as
• the IPE (Interpersonal Political Environment).

7
The IPE effect as a hypothesis of network effect

• Others in the network are individuals who not
  only surround A, but also form an environment
  that contributes to As social and political
  reality. Although others in the network form a
  collection of individuals, their productthat of
  the IPE they representis a collective entity.
• In an IPE in which As political discussants are
  all LDP supporters, there will be an information
  configuration that filters out most unfavorable
  information about the LDP. In this situation, the
  IPE promotes an LDP-favorable social reality for
  A. This effect is not caused by any one
  individual, for instance opinion leader, but is a
  collective result of As IPE.

8
The IPE effect as a hypothesis of network effect

• This approach emphasizes the fact that others in
  the network are agents of a variety of
  information resources to A, and play an important
  part in constructing As social reality.
• IPE is more than an agent of information
  resources. Our social reality is a joint product
  of individual belief, collective support by
  others, and mass media information (Ikeda 1993,
  1997), i.e. IPE is an essential part of our
  social reality.
• In this sense, people maintain their sense of
  reality collectively in their personal
  environment, including their voting behavior.
• Hypothesis 1 IPE effect
• In both the US and Japan, the higher the
  ratio of IPE of a specific Party X, the higher
  the probability that the respondent will vote for
  Party X.

9
Effect of social network Spontaneity (1)

• However, supposing collective influence of IPE
  does not mean that people are influenced by
  conformity pressure.
• The phenomenon which has not been well
  investigated but important focus in the context
  of interpersonal effect, i.e. spontaneity
  dimension passiveness vs. activeness.
• Passiveness The effect of network derives from
  a passive and conforming power mechanism.
• Activeness Spontaneous influence acceptance by
  an informational or referent power mechanism.
• ? The latter was only emphasized by Putnam.
• Passive influence such as coercive power which
  makes people conform should be more focused on.
• ? In classical studies (Asch, 1951 Deutsch
  Gerard, 1955),
• group influence has often been thought of as
  a result of passive and conforming pressure (even
  in the US context).
• ? Is it still true in the contemporary world?

10
Effect of social network Spontaneity (2)

• Passive vs. spontaneous distinction is especially
  consequential for actual democratic practices
• The conformity pressure could be a factor for the
  formation of the dark side of social capital
• It may restricts members of a given group to
  commit a specific behavior or subscribe to a
  specific political orientation, thus disturbing
  open discussion and making the members intolerant
  to non-conforming deviants.
• If the mechanism is spontaneous (informational or
  referent), the same concurrent seeking is
  interpreted quite differently
• Trust on others is essential here and it goes
  together with positive social capital
• ? Open discussion, and tolerance to
  heterogeneous ideas due to the lack of coercive
  power.

11
Effect of social network Spontaneity (3)

• As was shown, historically in Japanese society,
  social networks played negative roles in
  modernizing Japanese political attitude and
  behavior (Abe, Shindo Kawato, 1990 Richardson,
  1991).
• This claim presupposes that the group influence
  mechanism under Japanese society is by conformity
  enforcement, i.e. non-spontaneous power is the
  major engines, especially to mobilize people to
  vote some specific candidate.
• There seems to be a good reason, however, that
  this is not the case anymore due to the huge
  structural changes of the society in 1990 and
  after (as mentioned before).
• Lets empirically check this point.

12
Hypotheses and strategies on analyses (1)

• Hypo 2)
• Spontaneous effect hypothesis
• 2A) Spontaneous expression on PM Koizumi or
  Bush/Gore
• Positive and negative open-ended responses on
  politicians these are based on spontaneous
  cognitive process.
• If these spontaneous expressions are a positive
  function of the IPE's political color, IPEs
  effect is not from conformity.
• ?Conformity pressure functions are only
  limited to the very situation under specific and
  concrete group pressure (voting situation).

13
Hypotheses and strategies on analyses (2)

• Hypo 2) Spontaneous effect hypothesis
• 2B) Spontaneity of political participation (JP
  only)
• When conformity pressure is working facilitates
  Egos campaign participation in accordance with
  the IPEs political color, while oppressing
  participation in discordance with that political
  color
• ? Conformity pressure is a force that coerces
  human behavior in a specific direction which at
  the same time oppresses behavior going to the
  opposite direction.
• When spontaneous power is working IPE
  facilitates political participation consistent
  with, but without suppressing opposing directions
  (i.e. tolerant).
• ?IPE is only providing information or a
  behavioral referent, which is adopted/chosen by
  ego spontaneously (This is not testable in voting
  behavior because of its trade-off nature).
• 2C) Controlling mobilization variable (JP only)
• Mobilization attempts often utilize traditional
  power of social network, i.e. conformity (as was
  old Japanese voting theories).
• If mobilization is predictive of vote and at the
  same time reduce the effect of IPE, spontaneity
  hypothesis will be in danger.

14
Before testing Hypos How we measure IPE?

• social network battery with using name
  generators
• Components of network Japan-US

JP has less network members if it is significant
others/ political discussants
15
Relationship with net-other
JPs predominance of spouse and small number of
neighbors
16
Talk and guess on political matters
talk (less) and less guess on politics in JP
17
The test of hypotheses

• National sampling surveys with network batteries
  in both Japan and the US, particularly Japanese
  JES3 and US NES2000 data.
• Data source
• JES3 2001-2005 panel survey with 9 waves
• mainly focus on 2001 election with additional
  analyses on 2003 and after elections
• NES2000 pre-post survey on the Presidential
  Election 2000

18
JESIII 2001 House of Councilors Election

• Pre FtF Survey
• Target period 19 July to 28 July, 2001 (29
  July was the Election day
• Sample 3,000 Japanese voters (over 20 years
  old)
• based on 2 stage stratified sampling
• Response rate 2,061(68.7)
• Post Telephone survey
• Target period 1 August to 5 August, 2001
• Sample panel from the previous wave.
• Response rate 1,253 (41.8 against the original
  sample 60.8 against the 1st wave respondents
  791. against those who provided their phone
  numbers)

19
JESIII 2003 House of Representatives Election

• Pre FtF survey
• Period 29 October to 8 November (9 Nov was the
  election day)
• Respondents
• Total Sample3759, Response 2162, Resp
  rate 57.5
• Panel 2334 1340 57.4
• New Recruit 1425 822 57.7
• (addition of sample by random sampling)
• Post Ftf survey
• Period 13 November to 24 November
• Respondents
• Total Sample3573 Response 2268, Resp
  rate 63.5
• Panel 2356 1828 77.6
• New Recruit 1217 440 36.2
• Those who were accessible in both of the pre and
  post survey 1769

20
Testing Hypothesis 1 (IPE effect on voting
behavior)

• In the Japanese 2001 House of Councilors
  election, voters had two tickets ? Ordered logit
  analysis
• In the US Presidential Elections Gore 0 and
  Bush 1
• ? Logit analysis

Japan
U.S.
IPE effects are estimated by controlling
demographic vars, and party ID (source of
bias). By controlling party ID, we can reduce
projection effect statistically.
Independent vars.
Independent vars.
21
The result of the test on Hypothesis 1

• Even by controlling for party ID variables, the
  results show a significant net effect of
  perceived votes.

22
The result of the test on Hypothesis 1

• The IPE effect hypothesis was supported in Japan
• The IPE effect in the US showed the same pattern
  as in Japan.

23
Further data on Japan

• The IPE effect hypothesis (Hypothesis 1) was
  supported
• People vote consistently with the net-others
  perceived votes.

24
Testing Hypothesis 2A (the spontaneity effect of
discussants)

• Dependent Variables
• Japan Open-ended evaluations on the Koizumi
  cabinet.
• We counted the number of positive and
  negative responses.
• ? 88 of the voters gave one or more positive
  answers
• 36 did so on negative comment
• ? More than 30 gave two or more positive
  answers,
• while 6 gave two or more negative
  answers.
• US Four open-ended answers
• Positive and negative statements about the
  candidates Gore and Bush.
• ?More numerous than in the Japanese survey,
• ?Around 30 gave two or more answers.

25
The result testing for Hypothesis 2A
26
The result testing for Hypothesis 2A

• Japan Without discussants support, two or more
  positive comments were made by only 25 of the
  voters, whereas with full discussants support,
  the number rose by 10. The reverse was true for
  negative comments 42 and 33 respectively,
  resulting in a difference of nine percentage
  points. In total, there was a difference of 19
  percentage points between the effect in the
  positive and negative responses.

27
The result testing for Hypothesis 2A

• If discussants were 100 Democrat, the
  probability of zero-positive comment for Gore was
  only 17, but if discussants were 0 Democrat,
  the probability was 71.
• The corresponding probability for Bush was 25
  versus 76. Both of the cases had more than 50
  percentage point differences.

28
Further data on Koizumi

• Even after controlling verbal fluency (knowledge)
  and political bias (party support), IPE effect
  survived consistently.
• ? Support Hypothesis 2A(2001-2005).

29
Campaign participation as dependent variable
spontaneity effect (H2B)

• Campaign-related participation activities as
  dependent variables.
• The stronger LDP-IPE, the more voters are
  inclined to join LDP campaign related activities.
  The reverse effects do not appear, i.e. DPJ-IPE
  does not decelerate participation for LDP.
• The same is true for DPJ-IPE.
• H 2b was supported strongly in this test.

30
Campaign participation as dependent variable (H2B)

• Do these spontaneous effects survive even after
  controlling mobilization variables?
• Mobilization attempts are effective throughout
  the 4 cases, but still spontaneous effect remains
  valid (based on the post-hoc simulation of the
  equations with mobilization vars to Table 4
  results (1 page before).

31
Effect of mobilization on vote (H2C)

• Mobilization in terms of solicitation to vote for
  some specific party has been thought of in
  Japanese context as the main factor.
• ? We added mobilization variables on to the
  models in Hypo 1
• The results show that Japanese voting behavior
  for conservative party is no more the results of
  conformity pressure.

32
Conclusion on spontaneity

• Spontaneity effect hypothesis (H2)
• Clearly well supported.
• IPE push up voluntary comments on the PM or
  Presidential candidates as well as political
  participation in line with the IPE. Also the
  effect of IPE on political participation was not
  suppressive on incongruent participation with the
  IPE color.
• These findings were strengthened when we
  controlled the mobilization variable.
• The power process working in IPE is not
  conformity pressure, but more voluntary in
  nature.
• All in all, people behave under the high
  influence of interpersonal environment, but it is
  not through the conformity mechanism, which has
  been supposed to be the main interpersonal force
  in Japanese political behavior for long time.
  Their behavior and attitude are constructed
  through a more voluntary and spontaneous basis,
  which is a good news for Japanese democracy and
  social capital theory.

33
Comparison of the patterns between Japan and the
US

• Go back to Hypo 1 and see the difference between
  the countries.
• ?The overall direction of the IPE effect was
  consistent in both countries. However, the effect
  size was much larger in the US.

34
RQ Institutional difference (two-party system vs.
multi-party systems) 1

• Institutionalized visibility problem The
  difference between the Japanese and American data
  is attributable to the difference between the
  two-party system and the multi-party system with
  one dominant and other smaller parties?
• Limit the districts where the number of
  candidates receiving most of the votes was equal
  to the number of seats plus one.
• Close competition occurs when the number of
  popular candidates is slightly higher than the
  number of seats in the given district.
• ? This makes the competitive parties highly
  visible, i.e. US-like institutional context
  arises.

35
RQ Institutional difference (two-party system vs.
multi-party systems) 2

• We selected competitive 22 districts out of 47
• Mt 80 of votes went to seat1 (with DPJ
  candidate)
• (test 1) In these competitive districts, as the
  substantial competition increased the visibility
  of the candidates, the IPE effect would be larger
  than when the same analysis was carried out for
  all respondents?
• (test 2) To limit the districts where only one
  seat was being contested, and two candidates
  received 80 of the vote
• (More comparable to the American situation).
• In both of the two tests, we changed the
  dependent
• variable slightly. Because we analyzed the
  effect
• of seat competitiveness on a district
  level, we only focused on the district vote
  (removed PR vote).

36
RQ Institutional difference (two-party system vs.
multi-party systems)

• The results were clear. As Table 4 shows, in
  these two types of tests, the IPE effect on the
  LDP vote or DPJ vote was larger in the
  competitive districts than in the less
  competitive ones

gt
gt
gt
gt
37
RQ Institutional difference2003 result

• (The 2003 election was for H. of Representative
  Election, i.e. SMD is smaller).
• The results were clear again. As Table shows, in
  this test, the IPE effect on the LDP vote or DPJ
  vote was larger in the competitive districts than
  in the less competitive ones.
• competitivemore than 88 vote for the top 2
  candidates in the single seat district (not PR)

gt
gt
38
Discussion and conclusion for RQ

• Further analyses revealed that whether politics
  is organized around a two-party or multi-party
  system with small parties may well effect the
  impact of local personal political discourse.
• It was particularly revealing that Japanese
  electoral districts that in some respects mimic a
  US-type two-party system showed a stronger IPE
  effect on LDP voters as well as DPJ voters.
• Other analyses (not shown) prompted by the
  research questions suggested that political
  cultural interpretations are not viable.

39
Discussion and conclusion

• In conclusion, the analyses showed differences in
  the effect of IPE, the direction of which
  contradicts the collectivism interpretation of
  Japanese political behavior.
• These differences may be largely attributable to
  institutional difference between the US and
  Japan, specifically the political party
  configuration (the two-party system versus the
  multi-party system with small parties).
• Institutional context could be very
  consequential.
• ?Context change makes party competition
  different, which in turn changes the power of IPE
  on vote.
• ?Japanese Electoral system change (from
  Multi-member single nontransferable vote system
  to SMD dominant system) may have increased the
  IPE influence.