Poverty and voting behavior: an electoral geographical analysis
Loading...
Date
2016-07-28
Authors
Indhumathy, S.
Journal Title
Journal ISSN
Volume Title
Publisher
University of Peradeniya, Sri Lanka
Abstract
Introduction
Voting patterns in Sri Lanka have tended to follow swings between the United National Party (UNP) and the Sri Lanka Freedom Party (SLFP)/People’s Alliance (PA)/United People’s Freedom Alliance (UPFA). In recent times, the presidential election has seen substantial votes being garnered by an alternative party, the New Democratic Front (NDF). These trends are usually associated with socio-economic- political factors. While these factors have been studied, the role of poverty has not been specifically investigated in explaining election outcomes.
Despite reasonable per capita income growth, the prevalence of pockets of poverty in Sri Lanka have been evidenced and measured in terms of their impact on overall measures of poverty in the country. Surveys by the Department of Census and Statistics have highlighted either the rural or the estate sector as having the highest incidence of consumption poverty at a particular time. It can be speculated that poverty as a distinct socio-economic variable could influence voter behavior.
A person’s socio-economic status has been found to have influenced their vote in some studies. In many countries those who belong to high socio- economic status tent to support conservative parties while people with low socio-economic status tent to be supportive of leftist parties Handelman (1998). In contrast, Inglehart (1990) finds that the impact of social and economic status on elections is very low.
The main objective of this study is to test for a relationship between poverty and voting behavior in Sri Lanka.
Methodology
The approach was to test for any association (correlation) between poverty and (presidential) elections results at district level. Poverty was measured by the official Poverty Headcount Ratio (PHR) compiled by the Department of Census and Statistics through their periodic surveys covering 1990/1991, 1995/1996, 2002, 2006/2007, 2009/2010, and 2012/2013. PHR data for all the electoral districts was available only since 2010, due to the earlier civil war situation. The PHR was matched with data from the Department of Elections for the 1988, 1994, 1999, 2005, 2010 and 2015 elections. This was done by creating PHR and election maps using GIS, and then constructing both variables as categorical (binary) variables. The null hypothesis was specified as no relationship between poverty and voting behavior. Fisher’s Exact Test was used to find test for correlation through cross tabulation, using SPSS.
Results and Discussion
Table 1. District-wise distribution of poverty status and election result
< table >
Results of the test for different years are given in Table 2 below.
As per the probability values at 5% level, the null hypothesis of no correlation between poverty and election results cannot be rejected (no probability value was calculated for 1994 because there are no observed and expected values for one category). It is possible that factors other than absolute (consumption) poverty, such as historical voting patterns, caste, religion, culture, factors specific to each generation, influence of media, education, ethnicity, etc. exert a greater influence on voter behavior. At the same time, the district could be too large a unit to capture the relationship, compared with, e.g. a Grama Niladhari Division. Also, missing data for some years could have distorted the result.
Table 2. Probability Value of correlation
< table >
Conclusion
The study objective was to test for any correlation between poverty at district level and voters’ decision on whether or not to return the ruling party to power at presidential elections. Districts were classified according to whether their Poverty Headcount Ratio fell below or above the national PHR. Fisher’s Exact Test was used to find test for correlation.
According to the test results, there is no correlation between the poverty headcount ratio and the probability of a party winning the election. We are forced to conclude tentatively that voting behavior is not related to poverty. Such a conclusion may hold if factors other than poverty are more important for election results. Alternatively the result may be due to the specific measure of poverty used, or the unit of analysis (district) being too large, compared with, e.g. a GN division. Hence the hypothesis needs further testing by refining the poverty and voting behavior variables, and narrowing down the regional unit of analysis.
Description
Keywords
Poverty , voting behavior
Citation
Proceedings of the International Conference on the Humanities and the Social Sciences (ICHSS) -2016 Faculty of Arts, University of Peradeniya. P.68 - 71