• Nenhum resultado encontrado

Order in chaos: Ballot order effects in a post-conflict election?

N/A
N/A
Protected

Academic year: 2017

Share "Order in chaos: Ballot order effects in a post-conflict election?"

Copied!
4
0
0

Texto

(1)

Research and Politics October–December 2014: 1 –4 © The Author(s) 2014 DOI: 10.1177/2053168014559425 rap.sagepub.com

Creative Commons Non Commercial CC-BY-NC: This article is distributed under the terms of the Creative Commons

Attribution-NonCommercial 3.0 License (http://www.creativecommons.org/licenses/by-nc/3.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (http://www.uk.sagepub.com/aboutus/openaccess.htm).

Introduction

Ballot order effects in democratic elections are well docu-mented across a number of different countries and election settings (Kim et al., 2014; Koppell and Steen, 2004; Krosnick et al., 2004; Miller and Krosnick, 1998). Ballot order effects entail that candidates placed higher on the bal-lot receive more votes solely due to their balbal-lot position. Here we test for the presence of ballot order effects in the 2010 parliamentary election in Afghanistan. The election provides a first look at ballot order effects in a high stakes, post-conflict setting.

Voters could expect to incur substantial risk by turning out for this election since the period surrounding the presi-dential election in the previous year had seen the highest number of civilian casualties since the fall of the Taliban in 2002 (AIHRC-UNAMA, 2009). On Election Day in 2010, news agencies reported that at least 14 people had been killed (BBC Online, 2010). Partly as a consequence of the increased threat level, only 3.9 million of the 10 million eligible voters turned out.

What might be the underlying mechanism of ballot order effects in the case of Afghanistan? Kim et al. (2014) find that ballot order effects are stronger for voters who 1) have little information about the candidates, 2) feel ambivalence about candidates, 3) have limited cognitive skills, and 4) allocate little effort to candidate evaluation. We argue that

some of these factors are more likely to explain potential ballot order effects in the Afghan setting. On the one hand we can expect the high illiteracy of the Afghan electorate and the lack of experience with democratic elections to leave the voters with limited candidate information and cognitive skills to deploy for candidate selection. For instance, 82.4 percent of all Afghan women are estimated to be illiterate (World Bank, 2011). This should translate into substantial ballot order effects.

On the other hand we can expect the post-conflict setting of Afghanistan to provide strong incentives for picking the right candidate. This argument is in line with Ho and Imai (2008) who find that ballot order effects are much weaker in high-stakes US elections. We may view Afghanistan as an extreme case of a high-stakes election. Perhaps more importantly, the potential life-threatening act of voting should work to filter out the least dedicated voters by rais-ing the expected costs of turnrais-ing out. On balance, we there-fore have mechanisms at work which should both increase and reduce ballot order effects in the case of Afghanistan.

Order in chaos: Ballot order effects

in a post-conflict election?

Bertel Teilfeldt Hansen

1

and Asmus Leth Olsen

1

Abstract

Ballot order effects are well documented in established democracies, but less so in newly democratizing countries. In this research note we analyze ballot order effects in the 2010 parliamentary election in Afghanistan. The election provides a first look at ballot order effects in a high stakes, post-conflict election. In this setting, we argue that limited cognitive skills and information are more likely explanations of potential ballot order effects than mechanisms of lacking effort or ambivalence of choice. However, we find no clear evidence of a positive effect on the vote share of a higher ballot position. This raises the broader question of how applicable anomalies found in political behavior are to post-conflict democracies.

Keywords

Ballot-order effects, post-conflict elections, Afghanistan

1Department of Political Science, University of Copenhagen, Denmark

Corresponding author:

Bertel Teilfeldt Hansen, Øster Farimagsgade 5, Copenhagen, 1353, Denmark.

Email: [email protected] 559425RAP0010.1177/2053168014559425Research & PoliticsHansen and Olsen

research-article2014

(2)

2 Research and Politics

If we find ballot order effects in Afghanistan, limited infor-mation and cognitive skills are more likely explanations than lacking effort and ambivalence.

Data and design

On September 18th, 2010, Afghan parliamentary election were held. The elections determined who were to be mem-bers of the Wolesi Jirga (the House of the People). Our data consist of ballot information and election results from each of the 34 provinces for this election. We have obtained the total number of votes received for each candidate in every province along with their ballot position (IEC, 2013). In the smallest district the ballot contained 12 candidates while the largest district of Kabul had 664 (Median = 47). As we exclude the country-wide Kuchi constituency, our full sam-ple consists of 2427 candidates.

Afghanistan is a particular well-suited case to study bal-lot order effects due to its single non-transferable voting system (SNTV). Each voter casts one vote for a candidate on a list with a potentially unlimited number of candidates. The candidates receiving the most votes then fill the prede-fined number of seats allocated to each electoral district. SNTV impedes the development of political parties as votes cannot be transferred among candidates (Goodhand and Sedra, 2007; Tadjbakhsh and Schoiswohl, 2008). Therefore, elections to the Wolesi Jirga involve large num-bers of independent candidates. This leads to very long bal-lots and to voters being tempted to employ heuristic shortcuts such as the order of the candidates in their choice.

Each candidate on the Afghan ballot is distinguished by name, “party affiliation” (all of the elected candidates were independents), profile picture, and a unique black and white symbol (such as a motorcycle, three candles, or two books). The latter identifies them to illiterate voters.

Ballot order in Afghanistan is determined by a random lot-tery which allows the data generating process to be treated as a randomized experiment. Specifically, 35 separate lotteries where held which correspond to a lottery for each of Afghanistan’s 34 provinces and one for the Kuchi constitu-ency (IEC, 2010). However, if strong ballot order effects exist and if these are known there will be equally strong incentives to manipulate the ballot order by political or administrative elites (Meredith and Salant, 2013). This poses a particularly serious problem in a highly corrupt state such as Afghanistan.1 Any manipulation of the order would invalidate the natural experimental setting by inducing unobservable factors affect-ing both order an election outcomes. One way of testaffect-ing for manipulation is to investigate whether candidate order is uncorrelated with candidate background characteristics.

We conduct two such tests. First, we look for irregulari-ties in the distribution of women on each ballot. In Afghanistan’s highly patriarchal society we would expect that manipulation of the ballot order should result in lower average ballot position for women. However, the data show

that female candidates are on average placed less than one position lower than men, and this difference is not signifi-cant (d = 0.7, p = 0.913).2 Second, we observed whether or not the elected candidates were incumbents. If manipula-tion of ballot order had occurred we would expect elected incumbents to have a higher ballot position than elected non-incumbents as the former would have been able to lev-erage their position. Again the data do not suggest that this is the case as incumbents are in fact placed somewhat lower than non-incumbents (d = 21.5, p = 0.288, n = 225). In sum-mary, we find it plausible to assume that candidate order was not manipulated and was therefore exogenous to any candidate attributes which could affect their vote share. We proceed to estimate the following model:

Vote Shareij =β0+β1Ballot Position pct( .)ijij (1)

Where Vote Share is the percentage of votes received by candidate i in province j, Ballot Position (pct.) is the rela-tive ballot position of candidate i on the ballot of province

j, and εrepresents unobserved determinants of Vote Share. The results presented come from estimations using OLS, but the same conclusions obtain with Tobit.

Empirical findings

The main findings are presented in Figure 1. Here the didates’ relative ballot position is plotted against each can-didate’s vote share. For the full sample we see no significant effect of ballot position on a candidate’s vote share (b = 0.001, p = 0.409). This null result is robust to the exclusion of the Kabul district, which has a much longer ballot than the others (b = 0.002, p = 0.409, n = 1778). Furthermore, the results hold when using logged ballot position and con-trolling for logged ballot length (b = 0.015, p = 0.740). If we restrict the model to districts with ballot length below the median of 46, we estimate a small and barely significant positive coefficient (b = 0.01, p = 0.086, n = 489). This cor-responds to a 0.01 percentage point increase in vote share for each percentage point moved down the list.

In a model with logged vote share and province fixed effects we find a marginally significant, but very small, negative effect (b = -0.002, p = 0.095). For candidates with the median vote share of 0.5 percent, for instance, the effect translates into an estimated decrease in vote share from being moved from first to last on the list of only 0.11 percentage points. Estimating an interaction between ballot order and logged ballot length results in both the main and the interaction term being insignificant (b

= 0.015, p = 0.156, b = -0.003, p = 0.151). Finally, we include dummies for each decile of relative ballot position. None of these dummies turn out to be significant.

(3)

Hansen and Olsen 3

everybody else (Meredith and Salant, 2013). Here we con-tinue to control for logged ballot length to account for the fact that smaller lists come with larger expected vote shares for each candidate and thus increase the probability of a given candidate being at the top of the list.

On average the first candidate on the ballot got a some-what larger vote share than everybody else but the differ-ence is highly insignificant (d = 0.141, p = 0.776). The difference from the rest is marginally significant for the top ten (d = 0.461, p = 0.045), but not for the top five (d = 0.148, p = 0.525). However, the top ten-effect is driven entirely by two observations. If these are removed the effect disappears (d = 0.342, p = 0.131, n = 2425). On balance we do not find any clear support for the idea of ballot order effects in Afghanistan’s 2010 election.

Conclusion

The findings presented here show no or very marginal sup-port for ballot order effects in Afghanistan’s 2010 election. The absence of ballot order effects in the chaotic post-con-flict setting of Afghanistan points to the limits of the phe-nomenon, which has previously been identified for elections of varying importance in stable democracies.

We have argued that the high stakes, clear conflict lines, and especially the selection effect induced by extreme levels of electoral violence in Afghanistan should lead to a high level

of effort and a low level of ambivalence for the average voter selecting into turnout. At the same time the high Afghan illit-eracy rate and lack of experience with democratic elections should imply that the electorate has somewhat limited levels of information and cognitive skills to apply. The analysis indi-cates that these were not sufficient conditions for the electorate to employ candidate order as a shortcut.

An important limitation of the study is, however, that we are unable to determine if ballot order effects exist for sub-sets of the ballot. For instance, voters may conduct a first search based on ethnic or clan-based cues and then after-wards select according to ballot position. Nevertheless, the findings raise the broader question of how generalizable anomalies found in political behavior in the developed world are to post-conflict democracies.

Acknowledgements

The authors thank Jacob Gerner Hariri, Michael Cohen, Anders Wivel, and two anonymous reviewers for their helpful comments. All errors are our own.

Declaration of conflicting interest

The author declares that there is no conflict of interest.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

0

5

10

15

20

25

Pct. of Votes

0 20 40 60 80 100

Ballot Position (pct.) Lowess Smoother

0

5

10

15

20

25

Pct. of Votes

0 20 40 60 80 100

Ballot Position (pct.)

Lowess Smoother (ballot length < median)

0

5

10

15

20

25

Pct. of Votes

0 20 40 60 80 100

Ballot Position (pct.) Lowess Smoother (without Kabul)

−5

0

5

10

15

20

Residualized Pct. of Votes −6 −4 −2 0 2

Residualized Ballot Position (log) Controlled for Logged Ballot Length

(4)

4 Research and Politics

Supplementary Material

The replication files are available at: http://thedata.harvard.edu/ dvn/dv/researchandpolitics

Notes

1. According to the most recent report from Transparency International Afghanistan is the third most corrupt country in the world.

2. Here and in the following, “d” refer to the difference in means between two groups, “p” gives the p-value, and “b” refers to the unstandardized coefficients from a regression. n

gives the sample size and when it is not noted it implies that the full sample is being used.

References

Afghan Government (2010) Ballot lottery for 2010 Wolesi Jirga elections. Available at: www.iec.org.af/pdf/wsfactsheets/ fs_ballot_lottery.pdf (accessed 22 July 2014).

AIHRC-UNAMA: Afghanistan Independent Human Rights Commission-United Nations Assistance Mission in Afghanistan (2009) Joint monitoring of political rights, presidential and pro-vincial council elections, Third Report, UNAMA.

BBC Online (2010) ‘Brave’ Afghan voters hailed amid Taliban threats. Available at:www.bbc.co.uk/news/world-south-asia-11360050 (accessed 10 July 2014).

Goodhand J and Sedra M (2007) Bribes or bargains? Peace condi-tionalities and ‘post-conflict’ reconstruction in Afghanistan.

International Peacekeeping 14(1): 41–61.

Ho DE and Imai K (2008) Estimating causal effects of ballot order from a randomized natural experiment the California

alphabet lottery, 1978–2002. Public Opinion Quarterly

72(2): 216–240.

Independent Election Commission of Afghanistan. (2013) Results by ballot order. Wolesi Jirga election final certified results. Available at: www.iec.org.af/results_10/eng/index.html (accessed 21 July 2014)

Kim N, Krosnick J and Casasanto D (2014) Moderators of candi-date name order effects in elections: An experiment. Political Psychology, FirstView: 1–18.

King A and Leigh A (2009) Are ballot order effects heterogene-ous? Social Science Quarterly 90(1): 71–87.

Koppell JG and Steen JA (2004) The effects of ballot position on election outcomes. Journal of Politics 66(1): 267–281. Krosnick JA, Miller JM, and Tichy MP (2004). An

unrecog-nized need for ballot reform. In A. Crigler, M. Just, & E. McCaffery (Eds.), Rethinking the Vote, Oxford University Press, New York: 51–74.

Mantonakis A, Rodero P, Lesschaeve I and Hastie R (2009) Order in choice: Effects of serial position on preferences.

Psychological Science 20: 1309–1312.

Meredith M and Salant Y (2013) On the causes and consequences of ballot order effects. Political Behavior 35(1): 175–197. Miller JM and Krosnick JA (1998) The impact of candidate name

order on election outcomes. Public Opinion Quarterly 32: 291–330.

Tadjbakhsh S and Schoiswohl M (2008) Playing with fire? The international community’s democratization experi-ment in Afghanistan. International Peacekeeping 15(2): 252–267.

Imagem

Figure 1.  Ballot position and vote shares.

Referências

Documentos relacionados

WKHQ GHILQHG DV WKH PD[LPXP GHYLDWLRQ RI WKH UHVSRQVH FXUYH IURP WKH %HVW)LW 6WUDLJKW /LQH %)6/  PHDVXUHG RQO\ RQ LQFUHDVLQJ ORDG 1RQOLQHDULW\ LV XVXDOO\ H[SUHVVHG DV D

Tabela (5) F11 Deficiência na área de projeto do produto Falta de tempo para atendimento das exigências Conscientização do pessoal para a qualidade

O rendimento de grãos das cultivares (Tabela 5) foram submetidos ao teste de Scott e Knott (5%). Os resultados revelaram que os genótipos não se distinguiram

Segundo Campbell et al (2001) a classificação dos diversos modelos em função do tipo de decisão a tomar pode ser feita da seguinte forma: Modelos que abarcam a definição de

No que se refere aos resultados obtidos, utilizando escala de CoS, foram tratados e analisados os comportamentos de conformidade de segurança físico (CoSCF) e

Através da nova designação, «sociedade gestoras de participações sociais», abreviadamente SGPS, pretende-se retratar mais fielmente o objeto das sociedades em causa. Reduz-se

Em cada uma das três categorias em que as postagens foram organizadas – assuntos relacionados à prática do esporte e à atividade profissional em clubes e nas seleções

De qualquer forma, espera-se que os artigos sirvam como ponto de partida para novas abordagens e que contribuam para que se continuem a registar avanços quanto às pesquisas sobre