Determining Factors of Voters’ Participation in Local Elections of Community Head in Armenia

This study attempts to determine the socio-economic factors affect - ing participation in local community head elections in Armenia. To that end, a binary logistic regression model was estimated us-ing the household survey data gathered by the Caucasus Research Resource Center’s (CRRC) regional office in Armenia, within the framework of the Civic Engagement in Local Governance (CEL-oG) Project. The results of the estimation showed that economic condition, settlement type, age, awareness of local government powers, and trust were statistically significant determinants of par - ticipation in local community head elections.

After becoming independent in 1991, Armenia faced socio-economic and political challenges.Among many challenges, Armenia had to deal with the issue of smoothly transitioning to a democratic society.That implied the establishment of democratic institutions and people's awareness of and participation in political processes.According to Grigoryan (2013), "democracy is a type of political system in which power alternates through regular, competitive elections, and citizens enjoy certain basic rights" (p.3).Elections play an important role in any democratic country as a way of expressing people's will and favor for a particular political party.Elections present an opportunity to residents to raise new issues and problems they face and to choose leaders who may possibly undertake actions to resolve them.
The most low-income population in any developing country resides in rural areas (Anriquez 2007, Bokhyan 2017).These people rely on election candidates and hope that the elected candidate will keep his/her promises and their livelihood will change after the elections.According to my personal opinion, people might have more expectations in the case of local community head elections rather than in presidential or parliamentary elections.This can be explained by the fact that local community heads are more familiar with the existing issues their constituents face than the candidates for presidential or parliamentary elections as community heads share the same environment and living conditions.A lack of research on the factors determining the participation in local elections in Armenia is another reason to emphasize the importance of this paper.
The research question of the present study is: What are the socio-economic characteristics influencing voters' participation in the local community head elections?The empirical findings of this research will be significant to policy-makers, government, and international organizations dealing with assisting in establishing democratic institutions in Armenia.
The paper proceeds by first presenting literature review and hypotheses formulation.In the following section, the empirical specification is discussed and data description is presented in the ensuing section.Then, the estimation results are presented and interpreted.The last section includes summary and a set of policy recommendations.

Literature Review Awareness and participation in elections
Awareness, TV-watching, reading newspaper A number of empirical studies (Biswas, Ingle & Roy, 2014;Kleiner, 2015) choose policy awareness through media consumption as one of the determinants of citizens' willingness to vote in elections.However, media consumption has 2 dimensions: active and passive.Newspaper reading may be a form of active media consumption because it involves a person's will to choose, buy and read the newspaper.On the other hand, watching TV may be classified as a form of passive media consumption, because people usually accept the information received from TV and they do not make extra efforts to get that information.Whereas some researchers (Wilkins, 2000;McLeod et al., 1996) found that television use and newspaper use are positively associated with political participation in general, other studies found a negative association or no influence of television use on political participation (Norris, 1996;Viswanath et al., 1990;Kleiner, 2015;Muntean, 2015).
The explanation of local electoral participation also includes awareness variable as a measure of local political interest.According to Mouritzen, Rose and Denters (2014), local political interest and knowledge have positive impact on the likelihood of voting in local elections.
Hypothesis: we expect a positive relationship between a respondent's awareness of local government powers, TV and newspaper usage, on the one hand, and participation in local community head elections, on the other.

Political engagement and participation in elections
Membership in a political party, discussion of personal and community problems with community head, trust Numerous studies suggest that membership in political parties and an attitude toward the candidates may also have significant impact on the person's willingness to vote (Prysby & Scavo, 2002;Kleiner, 2015;Mouritzen, Rose & Denters, 2014).In particular, people who are attached to a political party (members of political party) are more probable to participate in elections when compared to those who do not have party affiliation or psychological attachment.For the same reasons it may be expected that people's cooperation with their local community may also increase the likelihood of voting in a local election (Mouritzen, Rose & Denters, 2014).
Trust towards candidates is another important determinant of political participation (Muntean, 2015).In addition, people who distrust government are not inclined to vote.Researchers (Quintelier, 2007;Muntean, 2015) indicate that especially young people do not trust politicians and they do not believe in the ability of their vote to influence the election results.This leads to the lower participation rate in the local elections.
The positive relationship between the discussion of personal and community problems with community heads and participation in local elections may be explained by the fact that if the community head is eager to discuss and solve any kind of problem of community resident, so he/she might have higher chances to win votes in the elections.
Hypothesis: we expect that the high levels of public trust towards candidates, attachment to a political party, and discussion of problems with community head are in positive relationship with participation in local community head elections.

Socio-economic factors and participation in elections
Age, gender, settlement type, education, and economic condition.
When analyzing the voters' behavior in elections, one set of factors that have an impact on voting in elections include demographic factors such as race, religion, region, social class, gender, and age.Some researchers (Harder & Krosnick 2008;Prysby & Scavo 1993) mentioned that younger people are more likely to vote in elections in general, given the poorer physical health and lower energy levels of older people.In contrast, according to other empirical analyses, older people are more inclined to participate in national and local elections than younger people (Kleiner, 2015;Mouritzen, Rose & Denters, 2014).
Until the 1980s, females had been politically less active and less informed than males.However, recently the picture has changed, women's participation rates in both national and local elections occasionally even exceeded those of men (Kleiner, 2015;Harder & Krosnick, 2008;Mouritzen, Rose & Denters, 2014).A number of empirical studies (Berinsky, 2010;Tenn, 2007;Bredy, Verba & Schlozman, 1995;Mouritzen, Rose & Denters, 2014) show that people with more years of education are more likely to vote in elections compared to ones who have fewer years of education.In addition, people with higher income and better economic conditions are more likely to participate in elections (Filler, Kenny & Morton, 1991) given that people with lower income have less free time to learn about elections, and wealthier people have more interest in participating in elections in terms of getting more psychological and social rewards from voting (Rosenstone, 1980).Some authors (Harder & Krosnick, 2008;Wolfinger & Rosenstone, 1980) also argue that people from rural areas have higher participation rates than people from urban areas, because the farmers think that after elections they may get some farm subsidies from the government.
Hypothesis: we expect that being in a better economic condition, education level, and residing in rural areas will have a positive impact on citizens' participation in local community head elections.Also, we expect that older people are more prone to participate in local community head elections than young generation.

Empirical Specification
To identify the socio-economic variables influencing the voters' participation in local community head elections in Armenia, a binary logistic regression model was estimated, where the dependent variable, participation in local community head elections, was modeled as a function of a set of socio-economic characteristics (variables).The empirical specification of the binary logistic regression model estimated in this study is as follows: Pr(elect= 1) where Pr is the probability of the respondent participating in local community head elections; F is the logistic cumulative density function; elect is a dummy dependent variable taking on 1 if the respondent participated in local community head elections and 0 otherwise; agei is the respondent's age; malei is a dummy variable taking on 1 if the respondent is male and 0 otherwise; good_econ_condi is a dummy variable reflecting economic condition of respondent taking on 1 for having enough money for food and clothes and also for expensive durables like a refrigerator and washing machine and 0 otherwise; urbani is a dummy independent variable taking on 1 if the respondent is from urban areas other than Yerevan and 0 otherwise; Yerevani is a dummy variable taking on 1 if the respondent lives in Yerevan and 0 otherwise; prim_sec_voc_educi is a dummy variable taking on 1 if the respondent has completed secondary technical education and 0 otherwise; high_educi is a dummy variable taking on 1 if the respondent has completed higher education and 0 otherwise; awarei is a dummy variable taking on 1 if the respondent is aware of the local government powers and 0 otherwise; politicalpartyi is a dummy variable taking on 1 for membership in a political party and 0 otherwise; newspaperi is a dummy variable taking on 1 if the respondent reads newspapers and 0 otherwise; TVi is a dummy variable taking on 1 if the respondent watches TV and 0 otherwise; discussioni is a dummy variable taking on 1 if the respondent discusses some personal or community problems with community heads and 0 otherwise; low_trusti is a dummy variable taking on 1 if the respondent has a low level of trust towards community heads and 0 otherwise; high_trusti is a dummy variable taking on 1 if the respondent has a high level of trust towards community heads and 0 otherwise and βs are the parameters to be estimated.
The model was estimated using the STATA 10 software package.First, by observing the statistical significance of the parameter estimates associated with independent socio-economic variables key characteristics were determined.Then, by using the magnitudes of these parameter estimates, the percent change in odds ratios was calculated.Odds ratios were computed through the exponentiation of the logit coefficients (i.e., expβi), and the percent change in the odds ratios were calculated as (expβi -1)*100.

The Issue of Multicollinearity
One of the diagnostic issues that needs to be addressed is related to possible multicollinearity present in the data.To address this issue, the data were checked for the presence of multicollinearity using a set of criteria.The measures used for checking for multicollinearity are presented in the appendix (Table 1).

Data Description
To conduct the analysis, household survey data gathered by the Caucasus Research Resource Center's (CRRC) regional office in Armenia, within the framework of the Civic Engagement in Local Governance (CELoG) Project were used.The sampling method of the survey was multilevel cluster sampling.Stratification was done by region and area of residence, combined with purposed sampling of target pilot communities.These data are available on the CRRC-Armenia's website and they contain all the necessary information to successfully complete the research.The sample used in this study contains information on Armenian respondents who were at least 18 years old at the time when the survey was done.A total of 1,463 observations for Armenia were used in the analysis.To analyze the factors affecting voters' participation in local community head elections in Armenia the following set of household socio-economic characteristics (variables) were analyzed: household member's age, gender, household economic condition, respondent's education level, settlement type, awareness about the local government powers, membership in a political party, media consumption, discussion of issues with community head, and trust towards community heads.

Dependent variable ■ Participation in a local community head elections
The dependent variable for the logit model is the participation in local community head elections.The dependent variable was created based on the answer to the following question: "Did you participate in the last elections to vote for the head of the community?"The possible answers were no (coded 0) and yes (coded 1).

Independent variables ■ Awareness of local government powers
This variable was constructed based on the answers to the following questions: "1.Are you familiar with the decisions passed at your LSG bodies?2. Have you ever heard of an announcement by your LSG bodies inviting the public to monitor the regulations accepted by them? 3. Have you ever inspected the regulations passed by your LSG? 4. Do you know any assessment tool that the government uses to rate the performance of LSGs? 5. Do you know how local taxes, property rates, fees, fines and licenses are determined by the LSG?" Possible answers no (coded 0) and yes (coded 1).

■ Newspaper
The newspaper variable was developed using the answers to the following question: "How often do you use newspapers?"The answers were coded as 0 for "never" and 1 for "1-2 times a month", "1-2 times a week", and "every day".

■ TV
The TV variable was constructed using the answers to the following question: "How often do you use TV?"The answers were coded as 0 for "never" and 1 for "1-2 times a month", "1-2 times a week", and "every day".

■ Discussion of issues with community head
The discussion variable was developed based on the answers to the following question "In the past 6 months, did you contact the head of your community for a personal or a community problem?"The answers to this question were coded as 0 for no, and 1 for yes.

■ Membership in a political party
The variable accounting for the respondents' membership to a political party was developed using the respondents' answers to the following question: "Could you please indicate whether you are a member or not of a political party or its local branch, and if YES, under what terms?"The answers were coded as 0 for a non-member, and as 1 for otherwise.

■ Trust towards community head
The three trust variables reflecting the level of trust towards the community head were formed using the answers to the following question: "How much do you trust the head of community?"The first trust variable represented no trust ("do not trust at all"), the second one represented a low trust level ("very little", "little", and "neither mistrust nor trust"), and the third one represented the high level of trust ("a lot" and "fully trust").

■ Age
The age variable represented the actual age of the respondent at the time of the survey.

■ Gender
The gender of the respondent was accounted for through the gender variable.

■ Settlement type
The respondents' settlement type was included in the model with three dummy variables representing rural area, urban area (excluding the capital, Yerevan), and the capital Yerevan.

■ Economic condition
To account for the economic status of the respondents, variables were developed based on the answers to the following question: "Which of the following best describes your family's economic situation?"The possible outcomes were the following: 1. Family income is not enough for food 2. Family income is enough for food, but not for clothes.3. Family income is enough for food and clothes, but is insufficient for buying expensive household items, such as refrigerator or washing machine.4. We can afford to buy expensive items, such as refrigerator or washing machine. 5. We can afford to buy anything we want.The answers were recoded into two categories: poor_econ_cond (Family income is not enough for food, family income is enough for food, but not for clothes) and good_econ_cond (Family income is enough for food and clothes, but is insufficient for buying expensive household items, such as refrigerator or washing machine.We can afford to buy expensive items, such as refrigerator or washing machine.We can afford to buy anything we want).

■ Education
Respondents' education level was incorporated into the analysis based on the information given in an answer to the following question: "What is the highest level of education you have accomplished?"The possible answers were 0-Have not attended primary school, 1-Primary (complete or incomplete) 2-Secondary (incomplete), 3-Secondary (complete), 4-Vocational, 5-Higher education (incomplete), 6-Higher education (complete), 7-PhD.These have been recoded into three categories: incomplete (have not attended primary school); primary or secondary or vocational (primary (complete or incomplete), secondary (incomplete), secondary (complete), vocational); and higher (higher education (incomplete), higher education (complete), PhD).
Percentages of respondents by socio-economics characteristics in Armenia are shown in Table 1.

Estimation Results
The estimated coefficients, the associated p-values and percent change in odds ratios from the binary logit model are presented in Table 2.The statistical significance of the coefficients was evaluated at the 5% significance level.The interpretation of the estimation results was done in terms of statistically significant percent change in odds ratios.Based on the p-value of the likelihood ratio chi squared statistic, which is equal to zero, it can be concluded that all the parameter estimates were jointly statistically significant at the 5% significance level.Also, the estimation results showed that the odds of participating in local community head elections were lower for respondents who were in good economic condition by 37.2%, compared to those who reported worse economic condition, everything else held constant.This finding is opposite to our hypothesis, and can be possibly explained by the fact that relatively wealthier people are busy spending most of their time working and they do not have time to participate in elections.Consistent with our hypothesis, the odds of participating in local community head elections for people living in Yerevan were by 69.2% lower, compared to the ones who lived in rural areas, ceteris paribus.Each additional year of age increased the odds of participating in local community head elections by 1.8%, other things being equal, which is in accordance with our hypothesis.
Another statistically significant determinant of participation in local community head elections was awareness of local government powers.Consistent with our hypothesis, the odds of participating in local community head elections increased by 64.2% if the respondent was aware of local government powers, compared to those who were not aware, ceteris paribus.The odds of participating in local community head elections were 103.4% higher for the voters who had a low level of trust towards the local com-munity head, compared to those who had no trust towards community head, everything else held constant.Similarly, consistent with our hypothesis, for those respondents who reported a high level of trust towards community head, the odds of participating in local community head elections were 232.4% higher compared to those who did not trust community head, other things being equal.
The impact of such characteristics as gender, education, use of newspapers and television, respondents discussing issues with community head, and membership to a political party was not statistically significant.

Summary and Policy Recommendations
The objective of the present study was to shed light on the characteristics that affect participation in local community head elections in Armenia.To that end, a binary logit model was estimated using the CELoG dataset collected by the CRRC-Armenia in 2015.
The estimation results from the logistic regression indicated that economic condition, settlement type, age, awareness of local government powers, and trust were statistically significant determinants of participation in local community head elections.
Particularly, compared to respondents who reported bad economic condition, being in a good economic condition reduced the odds of participating in local community head elections.Living in Yerevan also reduced the odds of participating in local community head elections, compared to those living in rural areas.Every additional year of age and being aware of the local government powers increased the odds of participating in local community head elections.At the same time, gender, education, use of newspapers and television, respondents discussing issues with community head, and membership to a political party did not influence participation in local community head elections.
Based on the estimation results, the following policy recommendations are suggested geared towards an increase in participation in local community head elections: ■ Increase respondents' awareness of local government powers.This can be accomplished by promoting local government powers through various outlets.

Table 1 .
Percentage of Respondents by Socio-Economic Variables in Armenia

Table 2 :
Binary Logit Coefficients, Associated p-values and Percentage Change in Odds Ratios 2Asterisk indicates statistical significance at the 5% level.