Trust in Political Leaders after Electoral Violence

By Michael Holmes

In 2007, post-election violence in Kenya claimed over 1,000 lives and displaced 350,000 people. Electoral fraud, voting irregularities, ethnic tensions, and weak institutions contributed to violence perpetrated by ethnic and political gangs, as well as police with “shoot-to-kill” orders. Many studies focus on causes and responses to electoral violence, but few examine the impact violence has on individuals and their trust in government officials.[1] I ask, did experiencing the 2007 post-election violence lead to lower trust in the president, prime minister, and parliament? By focusing on individuals’ perception of electoral violence, we can examine whether it plays a unique role in determining trust apart from political party or ethnicity. The degree to which electoral violence influences public perception of political leaders may inform how domestic and international groups face these challenges. If political leaders can exercise violence as a means of coercion or repression without losing the trust of the electorate, there are potentially dire consequences for future elections. I predict that Kenyans who experienced the post-election violence will report lower trust in political leaders/institutions.

I use Afrobarometer survey responses to measure trust in the president, prime minister, and parliament (See Table 1).[2] They are scored on a scale from 0 (no trust) to 3 (complete trust).[3] The winning and incumbent president, Mwai Kibaki, was from the Party of National Unity (PNU) and the prime minister, Raila Odinga, and a majority of the parliament were members of the Orange Democratic Movement (ODM). Since people identifying with both parties perpetrated the post-election violence, I chose these three leaders/positions as they are representative of the major parties and the major institutions in government.

     
The main independent variables address direct (personal) and indirect (family member) experiences of electoral violence. Afrobarometer created six questions based on respondents’ personal experience of violence and seven based on their families’ experience. The questions included the categories of personal injury, destruction of businesses, job loss, home loss, eviction, property loss, and death (asked regarding family).[4] I derive one variable from the responses signifying whether an individual experienced electoral violence personally, and another signifying whether violence affected the respondent’s family.[5]

         
In order to isolate the impact of electoral violence on respondents’ levels of trust, I condition the study on variables that potentially influenced their trust of the leaders in government and their chances of experiencing the post-election violence. First, I derive one variable for people who voted for the main opposition party, ODM, and another for people who belong to ethnicities associated with that party.[6]  Since Kenya’s political parties are largely ethnic, support for ODM and membership of the related ethnicities should both relate to higher trust in the prime minister and parliament, but lower trust in the president.[7] Since electoral violence was carried out along ethnic and party lines, these two variables could also have influenced an individual’s experience.

         
I use two variables directly related to the act of political violence. One examines individual beliefs on the appropriateness of violence in politics, and the second records whether an individual wanted to participate or participated in the violence.[8] Politically active individuals may have faced increased risk of violence, so I use a measure for political activity.[9] Lastly, I include variables indicating the respondent’s education, developed environment (urban/rural), sex, age, and economic conditions.[10]

       
I tested three models measuring the relationship between trust and experiencing electoral violence (see Table 2). The only statistically significant relationship was between trust in the president and personal experience of electoral violence (see Figure 1). Respondents who experienced electoral violence exhibited a level of trust in the president that was approximately .214 points lower than those who had not. There were no statistically significant relationships between violence and trust in the prime minister or trust in the parliament. The respondents’ families’ experience did not have a statistically significant relationship with any trust variable.

         
Some control variables possessed significant relationships with the trust variables. Respondents who voted for ODM displayed lower trust in the president, but higher trust in the prime minister. Members of the ODM-related ethnicities also displayed lower trust in the President, but higher trust in the parliament. Respondents who wanted to or actually participated in violence, or viewed violence as “Sometimes Necessary,” showed lower trust in the president. Finally, higher levels of education related to lower levels of trust in the parliament and prime minister.

         
My prediction was partially correct. While I expected the experience of electoral violence to reduce trust in all three figures/institutions, it was only associated with a reduction of trust in the president. This may relate to the power associated with the executive office. Only two months prior to the election, President Kibaki replaced all of the electoral commissioners and appointed three new judges to the High Court, making both the electoral commission and the judicial branch appear less independent and more like extensions of the executive.[11] Since both the commission and the court system played major roles in the voting process, electoral fraud investigation, and judicial review, their failings reflected on the executive more than anyone else.[12] Since claims of electoral fraud sparked much of the violence, many citizens could hold the president responsible for the violence, reducing their trust in him as a leader if they directly experienced it. Since I account for political parties, rival ethnicities, proclivities toward political violence, and actual offenders in the analysis, it appears a personal experience of electoral violence negatively influences an individual’s trust in the president.[13] Still, it is possible that lower trust caused the violence. I avoid this problem by focusing on the effects of violence, not the cause or act itself. Though low trust may contribute to the act of engaging in violence and experiencing an injury, merely distrusting the president did not automatically destroy someone’s home. Therefore, we can rule out the idea that low trust directly caused all experiences of post-election violence. While we must hold the results with caution, we can still conclude that experiencing electoral violence in 2007 negatively influenced trust in the president.  

---

[1] See Susanne D. Mueller, "The Political Economy of Kenya’s Crisis," in Voting in Fear: Electoral Violence in Sub-Saharan Africa, ed. Dorina A. Bekoe (Washington DC: United States Institute of Peace Press, 2012. And Karuti Kanyinga, “Stopping a Conflagration: The Response of Kenyan Civil Society to the Post-2007 Election Violence,” Politikon: South African Journal of Political Studies , 1 (2011).

[2] Afrobarometer Data, Kenya, Round 4, 2008, available at http://www.afrobarometer.org

[3] Afrobarometer asked each respondent how much he or she trusted each political figure/institution. They had five response choices, 0 for “Not at all,” 1 for “Just a Little,” 2 for “Somewhat,” 3 for “A lot,” and 9 for “I don’t know.” I recoded “I don’t know as 1.5. I generally assume that if an individual trusts the figure/institution at all, “Just a Little” provides an adequate answer, making an “I don’t know” response seem similar to a “Not at all” answer. Even so, the answer can simply mean that he or she does “Not know.” For this reason I made it the middle value. There are 80 total “I don’t know” responses for all three variables.

[4] Afrobarometer phrased these as “yes” or “no” questions. ex. “Question: Now I would like to ask you some questions about the events that followed the December 2007 general elections. As you know, there were outbreaks of violence in various parts of the country. Please tell me if YOU PERSONALLY were affected in any of the following ways: Personal injury?” Respondents could refuse to answer this question, but that response was not recorded in the data. For the questions addressing the respondent’s family, “I don’t know” was an option. Since there were very few “I don’t know” responses and an “I don’t know” is very similar to a no in this question, I dropped the “I don’t know” values. It is similar to a “no” because this question ultimately seeks to get the perceptions of individuals whose families experienced electoral violence. If you do not know if your family experienced violence, you are not coming from the same position as someone who knows his or her brother was injured. Since this question seeks to understand the effect of violence, the respondent must know of the experience.

[5] Both of these are binary variables, meaning that I recorded the variable as a “1” if a person experienced one or more of the six categories in electoral violence. Conversely, if respondents did not experience any of the six variables, I recorded the variable as a “0.”

[6] The ethnicities associated with the ODM were the Luo, Luhya, Kalenjin, and Mijikenda.

[7] Sebastian Elischer, Political Parties in Africa: Ethnicity and Party Formation (New York: Cambridge University Press, 2013), 99.

[8] Example Question: “Question: Which of the following statements is closest to your view? Choose Statement 1 or Statement 2.

Statement 1: The use of violence is never justified in Kenyan politics today.

Statement 2: In this country, it is sometimes necessary to use violence in support of a just cause.” This variable is coded from 1-4. 1 indicates strong support for statement 1, 2 general agreement with statement one, 3=general agreement with statement two, 4= strong agreement with statement two. The individuals who answered “I don’t know” were recorded in the middle as 2.5.

[9] This is an index variable based on an individual’s attempts to contact his or her Member of Parliament, contact members of local government, contact official agencies, participate in protests, and vote. Ex. The rating of a two on this index indicates that a respondent partook in two of the above activities, such as participating in a protest and voting. 

[10] The economic conditions variable is not measuring income but rather the respondent’s personal opinion of their living conditions on a scale of 1-5 (Very Bad-Very Good).

[11] Mueller, “The Political Economy of Kenya’s Crisis,” 160.

[12] Ibid., 165

[13] I tested all these variables for multicollinearity using a VIF test. The highest VIF score was on the individual experience of violence variable, with a 1.50.