VT How is the long list of causal factors of turnout best understood Flashcards
Overall plan
- literature talks past each other
- give examples
- literature talks past - meta-analyses
- variables
- measurements
- areas included
- individual-level core turnout model
- aggregate-level core turnout model
intro
The long list of causal factors is best understood as a long list of literature that talks past each other by looking at different geographical regions; using different controls; and measuring variables in different ways. I first discuss some of these issues with the literature. I then utilise one individual-level meta-analysis and four aggregate-level meta-analyses in order to assess the extent to which some of these factors may be considered a core model. I argue that even these meta-analyses also talk past each other to some extent and the lack of consensus amongst them illustrates the complete lack of a core model. I argue that one variable, compulsory voting, has found consistent support in the literature and could form part of a core model but it is very unclear which other variables should be included.
lit talks past each other (non-meta)
The literature on the causal factors which affect voter turnout is extremely extensive. Some of the key models include the resource model, the mobilisation model, the psychological model, the political institutional model and the rational choice model (Smets and van Ham, 2012). For each causal factor, consensus on whether it positively affects turnout varies. For example, there is largely consensus that compulsory voting increases turnout but the magnitude varies, with Blais (2006) finding estimates between 10 and 15 percentage points. In contrast, the direction of the effect of unicameralism on turnout is much more contested with some reporting positive effects such as Jackman (1987), Jackman & Miller (1995), and Fornos et al. (2004). However, Blais & Carty (1990), Black (1991), Radcliff & Davis (2000), and Perez-Linan (2001) indicate no effect. Whilst these indicative findings demonstrate some variation between studies, the extent to which the literature talks past each other is best illustrated by meta-analyses as they compare large numbers of studies.
lit talks past each other (meta)
The meta-analyses on turnout suggest that there are three main ways that the literature talks past each other which has generated the long list of causal factors. Firstly, the variables used differ substantially. In Frank and i Coma’s (2021) aggregate meta-analysis of turnout, they find that of the 127 distinct variables used in studies they included, less than half (44%) appear more than once in the studies. Similarly, in Smets and van Ham’s individual-level meta-analysis of 90 studies of voter turnout, they find over 170 independent variables used to explain voter turnout in these studies alone. Only 8 of these were included in more than 25% of the studies they reviewed and even the two most common variables were included in 72% and 74% of studies. This illustrates that the literature largely talks past each other as they analyse different variables and very few are including all of the same control variables which may explain the high number of causal variables that different researchers find have effects on turnout. Furthermore, not only is there wide variation in the variables included, there are also discrepancies in how they measure these variables. Frank and i Coma (2021) find that in their meta-analysis of turnout, among the 55 distinct variables which appear more than once, over half (57%) are measured in more than one way. Hence, even when the same variables are used, they are often measured differently which will affect the results. There is also variation in the inclusion of certain geographical regions, levels of development and years of democracy which will affect the results. For example, Smet & van Ham (2012) exclude studies in new democracies. Therefore, there is a long list of causal factors which are claimed to affect turnout due to a large body of literature using different variables, measurements of variables and the inclusion of different geographical regions.
individual-level core turnout model?
Whilst the literature talks past each other in important ways, I will use several meta-analyses to discuss the extent to which the long list of causal factors may be viewed as a core model. The core model may vary based on whether it is an individual-level or aggregate-level model, so I discuss these types of models. Smets and van Ham (2012) provide an individual-level meta-analysis of turnout and find several factors to have a consistent effect on voter turnout: media exposure, region, residential mobility, education, age, political knowledge, political interest, party identification, voting in previous elections and mobilisation (partisan and non-partisan). Whilst it may be suggested that these factors may be used to create a core individual-level model of turnout, the analysis excludes new democracies, so the extent to which this model may be applied must be carefully considered. Furthermore, some of the common findings in developed nations and older democracies do not hold in other regions. For example, Kuenzi and Lambright (2010) use Afrobarometer survey data to assess the determinants of voting for over 17,000 voting-age adults in 10 African countries and they find contrary to other socioeconomic models that poorer individuals are more likely to vote than wealthier. But they do find support for age, education and party idenififcation. Hence, the core variables given by Smets and van Ham need further research in order to assess whether they could be used for a core individual-level model for other regions.
aggregate-level core turnout model?
In order to assess the extent to which the long list of causal factors could be used to create an aggregate-model of voter turnout, I compare four aggregate-level meta-analyses by Geys (2006); Frank & i Coma (2021); Stockemer (2017); and Cancela & Geys (2016). Meta-analyses allow us to assess the extent to which there is consensus in the literature which could be utilised to create a core model of turnout. There are some variables which are supported by multiple of these meta-analyses. Frank and i Coma (2021) and Stockemer (2017) both find that compulsory voting is positively associated with turnout. Also, Geys (2006) and Stockemer (2017) both find that small countries are positively associated with higher turnout. For other variables there are contrary findings between the meta-analyses. Stockemer (2017) finds that the effect of electoral closeness is inconclusive at best whereas the other three find this positively affects turnout. Furthermore, Stockemer (2017) finds that PR is only important for a minority of cases which is contrary to Cancela & Geys (2016). There are other factors which are only reported as having an effect by one of the meta-analyses such as Frank and i Coma (2021) finding economic globalisation important. These comparisons between meta-analyses suggest that compulsory voting and population size may form part of an aggregate-level core model of turnout but the other variables are disputed.
conclusion
In conclusion, the long list of causal factors which affects voter turnout can be best understood as a large body of literature which talks past each other by using different variables, different measurement and using different inclusion criteria in relation to geographic regions, economic development and age of the democracy. I used meta-analyses to assess the extent to which the long list of causal factors could be seen as a core model but argue that whilst some factors such as compulsory voting and population size could become elements of a core model, others such as election closeness are disputed even amongst the meta-analyses. Whilst meta-analyses are good in assessing consensus in the literature, they do have limitations. For example, the vote counting procedure used by Smets and van Ham (2012) does not allow them to estimate effect size. Therefore, other forms of analysis may be needed to supplement in order to create a causal theory. Overall, the long list of causal factors can be explained as a clear lack of consensus over the factors that affect turnout but also how to analyse turnout in terms of variables and measurements.
