Interethnic Tolerance, Demographics, and the Electoral Fate of Non-nationalistic Parties in Post-war Bosnian Municipalities

Transcription

1 Interethnic Tolerance, Demographics, and the Electoral Fate of Non-nationalistic Parties in Post-war Bosnian Municipalities (Work in progress) Rodrigo Nunez-Donoso University of Houston EITM Summer School 06/2014 1

2 Summary Take-home message Motivation Post-conflict developments Research questions A brief introduction to the Bosnian political system Competing theories Hypotheses EITM Approach Data, models Results Conclusion 2

3 Take Home Message The level of inter-ethnic tolerance is a strong predictor of vote choice for non-nationalistic parties, in local level elections at least in one post-conflict society. Its effect is consistent across different model specifications and subsamples. Conversely, ethnic fractionalization (polarization) has an erratic behavior as predictor. Need for better data/model. 3

4 Motivation Post-conflict societies are characterized by intense political competition Externally imposed institutions intended to manufacture electoral democracy Ethnonationalist political elite prevail Most of the theoretical and empirical research is based on national or regional level data, while the municipal level is often overlooked 4

5 Post-conflict Developments Nearly half of all civil wars are due to post-conflict relapses (Collier, Hoeffler, and Soderbom 2008) From civil war to electoral violence: Angola (1992) Burundi (2010) Kosovo (2014) From civil war to stability: Bosnia Herzegovina (1996) Macedonia (2002) Large amount of international aid for peacebuilding US has provided over $2 billion in aid (USAID) 5

6 Research Questions What are the determinants of vote choice for non-nationalistic parties in post conflict societies at the local level? Does the level of interethnic tolerance affect vote choice? Does the demographic geography influence vote choice? If so, is social heterogeneity detrimental or beneficial for non-nationalist parties? 6

7 Politics in Bosnia One country Source: Maphill 7

8 Politics in Bosnia Pre-war ethnic territorial distribution Source: Wikipedia 8

9 Politics in Bosnia 2 Entities, 10 cantons 9

10 Politics in Bosnia 142 Municipalities Source: Wikimedia Commons 10

11 Bosnian Political System Due to the post-war institutional arrangement, in practice we see two semiindependent part system (each entity) Main parties: Bosniak Serb Croat 11

12 Main Non-nationalistic Parties* * Abbreviated as (NNP) and interchangeably referred as non-ethnic parties also. 12

13 Local Elections Outcome (2012)

14 Local Elections outcome (2012)

15 Local Elections outcome (2012)

16 Local Elections outcome (2012) Non-nationalist parties Bosniak ethnic parties Croat ethnic parties Serb ethnic parties

17 Competing Theories Supply side: NNP flourish when political competition is de-ethnified (e.g. Homogenous districts) (Husley, 2011) Demand side: religiosity and right-wing political ideology decrease the probability to vote for NNP (Pickering, 2009) Ethnic distance, resource competition, negative assessments of the political system, and social capital would have no effect 17

18 Competing Theories (cont d) Unresolved issues: Supply side theories based on de-ethnification of political competition cannot explain the emergence of pockets of ethnic authoritarianism Demand side theories have been tested in homogenous samples (single ethnic group) Disconnection between theory and empirical tests: aggregate level measures, inadequate units of analysis, effect of electoral systems 18

19 The Unit of Analysis Problem Political unit \ Degree of heterogeneity Municipality Mono ethnic district (Homogeneous) Ethnic Authoritarianism Mixed district (Heterogeneous) More favorable for nonnationalist parties Canton Entity Federal More favorable for nonnationalist parties (Husley 2011) Ethnified political competition (Husley found stronger effect here, though)* H U S L E Y * Mixed Croat districts split more than mono ethnic Bosniak or Croat (!!)19

20 Argument Ethnic heterogeneity at the local level creates inter-ethnic tolerance Contact Hypothesis Heterogeneity at higher levels is detrimental to inter-ethnic tolerance Threat Hypothesis "Threat is perceptual; it involves what people think is the outgroup proportion and thus can be easily manipulated by political leaders and the mass media. Contact is experiential; it can reduce individual and collective threat as well as prejudice. (Pettigrew et al. 2010) 20

21 Argument (cont d) Individuals living in more heterogeneous municipalities will be more likely to have greater levels of inter-ethnic tolerance, which in turn will increase the likelihood of voting for non-nationalist parties in local elections We must not forget that municipal elections are conducted under a plurality system, which makes the argument even more counterintuitive 21

22 Hypotheses Tolerance hypothesis (H1): more tolerant individuals will have a greater propensity to vote for NNP Contact hypothesis (H2): living in a heterogeneous municipality will increase the probability of voting for a NNP, all else equal Advantages of my approach: Correct unit of analysis Majoritarian electoral system is a tough test Multiple ethnic groups in sample Recent data 22

23 EITM Approach EITM step 1: Intuition: voters would deviate from the expected ethnically motivated voting preference, maximizing their utility regarding inter-ethnic tolerance and context. Behavioral concept: decision making Statistical concept: nominal choice 23

24 EITM Approach EITM step 2: Behavioral analogue: utility maximization Statistical analogue: discrete choice modeling (voting for non-nationalist parties or not) EITM step 3: Unification (see next slide) 24

25 EITM Approach The dependent variable is Y ij = 1 if voting for NNPs 0 if voting for a nationalist (ethnic) party We try to model the vote choice if each individual i in each municipality j Utility model: There is a latent utility consisting of a systematic and random component U ij = V ij + ε ij (1) A person should choose m if its utility exceeds that of the other alternative U m n ij > U ij (2) 25

26 EITM Approach Y ij = β 0j + β 1j X ij + ε ij (3) where β 0j = γ 00 + γ 01 Z j + δ 0j (3.1) and β 1j = γ 10 + γ 11 Z j + δ 1j (3.2) then Y ij = γ 00 + γ 01 Z j + γ 10 X ij + γ 11 Z j X ij + δ 0j + δ 1j X ij + ε ij (4) The behavioral model is: Logit (π ij )= F (β 0 + β 1 Interethnic tolerance + β 2 Serb + β 3 Croat + β 4 Fractionalization + β 5 Eval. Of System + β 6 Eval. of Parties + β 7 Age + β 8 Education + β 9 Income + β 10 Rural + β 11 Population + β 12 GDP ) 26

27 EITM Approach Therefore Tolerance hypothesis (H1) implies that β 1 > 0 Contact hypothesis (H2) implies that β 4 > 0 27

28 Data UNDP Early Warning System survey ( ) emphasis in waves conducted in 2008 Municipal socioeconomic data from UNDP and Analitika s Moje Mjesto website Fractionalization and polarization at municipal level (FBiH only ) were calculated using ethnic distribution data estimated by Bochsler, Schlapfer and Shubiger (2010) Suboptimal data (!) DV Vote for non-nationalistic parties IVs Inter-ethnic tolerance Ethnic heterogeneity (municipal level) 28

29 Descriptive Stats Variable Observations Mean Std. Dev. Min Max Vote choice Inter-ethnic tolerance Bosniak Serb Croat Unkown ethnicity Minority status Returnee status Federation BiH Republika Srspka Brcko district Polarization Fractionalization Rural status Evaluation of the system Evaluation of parties Age Education Income Population (1) Population (Analitika) GDP per capita BAM 5, BAM 2, BAM 1, BAM 29,

30 Models and Results Vote Choice (1) Vote Choice (2) Vote Choice (3) Vote Choice (4) Vote Choice Inter-ethnic tolerance (14.23)*** (10.78)*** (9.15)*** (9.15)*** (5.80)*** Serb (15.99)*** (0.30) (0.22) (0.08) Croat (8.47)*** (7.12)*** (7.13)*** (4.13)*** Polarization (3.51)*** Fractionalization (3.55)*** (2.54)** Pol. Syst. Evaluation (1.74)* Eval. of parties (3.48)*** Age (3.94)*** Education (3.76)*** Income (0.34) Rural status (4.42)*** Population (4.38)*** N AIC BIC 3, , , , (5) 1, * p<0.1; ** p<0.05; *** p<0.01 Note: The estimation method is logistic regression. Odds ratio are reported. Z scores are reported in parentheses 30

31 Results (cont d) Vote choice (6) Vote choice Inter-ethnic tolerance (3.85)*** (5.45)*** Croat (3.64)*** (4.77)*** Fractionalization (1.22) Eval. of parties (2.19)** (2.26)** Pol. Syst. Evaluation (3.94)*** (6.20)*** Age (2.76)*** (2.03)** Education (2.04)** (2.68)*** Income (0.08) (0.09) Rural status (2.55)** (1.95)* Population (3.79)*** Serb (7.57)*** N AIC BIC * p<0.1; ** p<0.05; *** p<0.01 (7) Note: The estimation method is logistic regression. Odds ratio are reported. Z scores are reported in parentheses. Sample includes only waves conducted in

32 Predicted Effect An older citizen living in a rural municipality with the lower level of tolerance, other variables average 95% Conf. Interval Pr(y=1 x): [ , ] Pr(y=0 x): [ , ] 95% Conf. Interval Pr(y=1 x): [ , ] Pr(y=0 x): [ , ] An older citizen living in a rural municipality with the highest level of tolerance, other variables average 32

33 The Effect of Inter-ethnic Tolerance 0.25 Pr(Non-nat party) Inter-ethnic tolerance Bosniak Croat Note: calculated from model 6 (year 2008) 33

34 Results (cont d) Vote choice (8) Vote choice (9) Vote choice (10) Vote choice Inter-ethnic tolerance (6.09)*** (3.57)*** (4.87)*** (2.55)** Serb (8.34)*** (4.95)*** Croat (4.60)*** (3.32)*** (3.62)*** (2.92)*** Eval. of parties (2.39)** (2.58)*** (1.66)* (2.46)** Pol. Syst. Evaluation (6.24)*** (5.29)*** (4.02)*** (4.12)*** Age (4.20)*** (2.39)** (4.03)*** (2.38)** Education (5.06)*** (2.90)*** (4.38)*** (2.74)*** Income (0.29) (0.02) (0.50) (0.05) Rural status (4.66)*** (2.93)*** (4.54)*** (2.96)*** Fractionalization (1.47) (0.57) Population (2.54)** (2.74)*** N LR Test Prob >= chibar2 2, * p<0.1; ** p<0.05; *** p<0.01 1, (11) Note: The estimation method is XTLOGIT. Odds ratio are reported. Z scores are reported in parentheses. Models (9) and (11) correspond to waves conducted in 2008 only. 34

35 Results (Cont d) Vote choice (12) Inter-ethnic tolerance (2.88)*** Croat (4.66)*** Fractionalization (1.50) Eval. of parties (2.24)** Pol. Syst. Evaluation (4.30)*** Age (4.27)*** Education (3.72)*** Income (0.88) Rural status (3.52)*** GDP per capita (0.76) N LR test vs. logistic regression: chibar2(01) Prob>=chibar2 * p<0.1; ** p<0.05; *** p< Note: The estimation method is Mixed Effects logit (MEQRLOGIT). Odds ratio are reported. Z scores are reported in parentheses. 35

36 Conclusions Inter-ethnic tolerance increases the odds of voting for a non-nationalistic party Fractionalization and polarization produce opposite effects when predicting vote choice in logit models More questions than answers: best model? The role of economic conditions? Social capital? 36

37 Questions? Comments? Skepticism? Attacks? Thank you 37

38 Descriptive Stats (2008) Variable Obs. Mean Std. Dev. Min Max Vote choice Inter-Ethnic Tolerance Bosniak Serb Croat DK Minority status Federation BiH Republika Srpska Brcko District Polarization (RQ) Fractionalization Rural status Evaluation of the system Evaluation of parties Age Education Income

39 Inter-ethnic tolerance index 3 sets of 5 questions (one per ethnic group) How acceptable would be: To live in the same state with (ethnic group) Having (ethnic group) as neighbors Your children going to same school with (ethnic group) children To have (ethnic group) as colleagues One of your relatives getting married to (ethnic) Not actual wording. Source: UNDP in Bosnia and Herzegovina 39