A fuzzy dumbbell and I’m pretty sure I saw it in a popular science piece
If you enter all of this into DAGitty, you find that you can actually identify the model with 10 control variables.
Next step is to convert it into a DAG. Because election takes place at specific moment in time, we can represent causation in discrete stages (bolded variables indicate there’s a link between the lag and current version of the variable).
Those loops give important insights. Mainstream party success is self-limiting because there’s a cost of governing and that radical right success is self-limiting because it encourages mainstream parties to reduce immigration which in turns reduces the salience of immigration and RR success.
We used that to construct a causal graph of our problem. You’ll notice that this is very much NOT a DAG. There are a lot of cycles in there!
That feeds into a network visualization of those claims to explore (still very much in testing phase right now and there’s going to be a lot more functionality)
We entered any existing causal claims in the social science literature about the causes and consequences of our dependent and independent variables.
This paper is also acting as a template for how to do better observational causal analysis and was also a chance to try out a new causal analysis platform I’ve been building.
Voter data shows the dilemma is real in voter attitudes. Leave voters are more favorable towards military spending than other voters, BUT are less favorable towards spending on peacekeeping and only as supportive of spending on NATO cooperation as Remain voters.
Short answer: there’s no consistent effect of populist right electoral success on any military spending measure (i.e. major parties don’t rush out to boost or cut the military in the same way as they’ve been shown to do on immigration or euroscepticism).