ABSTRACT
A wealth of research examines the relationship between digital media consumption and political participation. Research typically defines participation broadly and focuses on Western contexts. We seek to add to the understanding of this relationship by focusing more directly on the relationship between digital media consumption and the propensity to vote among young people in a less democratic context. To do so, we examine a set of Central Asian countries (Kazakhstan, Kyrgyzstan, Tajikistan and Uzbekistan) that have varying degrees of democratization. We test whether digital media consumption stimulates voting among respondents aged 18–30, and if this is contingent on how free and fair are the elections. Our results suggest that in the most democratic country, Kyrgyzstan, the relationship between digital media use and the propensity to vote is relatively flat while digital media use in less democratic countries, overall, is associated with a decrease in the propensity to vote.
Acknowledgements
We thank Tolganay Umberaliyeva and Mereilim Kalenova for sharing the data with us.
Disclosure statement
No potential conflict of interest was reported by the authors.
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
Notes
1 In October 2020, following election results the citizens of Kyrgyzstan viewed as fraudulent, protest erupted across the country, eventually resulting in the President of Kyrgyzstan being forced to step down from his position and casting a shadow on the future of the country with both China and Russia having urged for a return to order (Kim, Lee, and King Citation2020).
2 We use the Varieties of Democracy (V-Dem) index of electoral democracy to classify each country.
3 According to the United Nations Children’s Fund (UNICEF) (https://www.unicef.org/), young people under 30 years of age comprise 60% in Uzbekistan, 54% in Kazakhstan, 68% in Tajikistan and 49% in Kyrgyzstan.
4 Sairambay (Citation2022) defines hypeization as ‘when new media assist to generate political participation because of hype and/or fashion to boost one’s popularity and/or to earn followers in social media’.
5 The data were collected by Public Opinion (a research institute in Kazakhstan) and Shark (a research institute in Tajikistan).
6 For a detailed description of the sampling procedure in Kazakhstan, see https://library.fes.de/pdf-files/bueros/kasachstan/13343.pdf. Unfortunately, the Friedrich Ebert Foundation – an initiator and a sponsor of this comparative youth study – did not prepare reports on the other countries in English. Nonetheless, they have reports for each country in Russian (see https://opinions.kz/ru/issledovaniya/molodezh-tsentralnoj-azii). There were very few missing values as is typical with face-to-face surveys. In fact, our propensity to vote measure (see below) was the only variable with any missing values in our models. There were only 4% of cases missing or the full sample (9% for Kazakhstan, 1% for Kyrgyzstan, 4% for Tajikistan and 1% for Uzbekistan). Even though the missing values are low, to prevent bias in our estimates assuming the data were missing at random (MAR), we decided to replace the missing values using multiple imputation. Through this process, we replicate 30 datasets where missing data are substituted with draws from the posterior distribution of the missing value conditional on observed values assuming a normal distribution (for a full description of the multiple imputation methodology, see Little and Rubin Citation2014). These observed values are based on all variables in our analyses. Instead of pooling our model results and using Rubin’s Rules to correct for any deflation in standard errors, we decided to circumvent this necessity by taking the average across each imputed value to replace the missing case. Because the number of replicate datasets is high, 30 replicate sets, across imputation error should approach a normal distribution, making the mean imputation value for any missing case an unbiased estimate.
7 Kazakhstan (mean = 2.05, SD = 0.96), Kyrgyzstan (mean = 2.12, SD = 1.04), Tajikistan (mean = 2.27, SD = 1.04) and Uzbekistan (mean = 2.44, SD = 0.91).
8 See https://www.v-dem.net/ for a complete description.
9 See https://www.v-dem.net/ for information about regime cut-points.
10 See https://databank.worldbank.org/ for individuals using the internet (% of population).
11 The distribution of each of the ordinal variables by country is graphically presented in the Appendix in the supplemental data online. There is some country-level variation.
12 Before concluding the ordered logistic was the best fit, we fit the model using ordinary least squares (OLS) and tested whether the assumptions were violated. We plotted the residuals, and it was clear that the relationship is not linear and the error is heteroskedastic (see the Appendix in the supplemental data online for the plot available at https://jasongainous.academia.edu/research#dataappendicesetc). This is confirmed with the Breusch–Pagan test where we reject the null hypothesis of homoskedasticity with a p-value close to 0.