Using Regression Discontinuity Design to measure the effect of political ideology on conspiracy belief
This is a research proposal to measure whether narrowly electing a Republican over a Democrat to the House of Representatives has any effect on the prevalence of belief in conspiracy theories.
I’m not really doing this research, this is an assignment for the ‘Causal Analysis in Data Science’ module. But I think it’s incredibly timely and quite interesting.
The proposed study would seek to find out if right wing representatives increase the prevalence of conspiracy belief in their districts. Lots of research has shown that conspiracy belief is rife at the extremes of both ends of the political spectrum; some research has shown it’s higher on the right and some have found no difference between left and right. It’s hoped this study could add some causal evidence to the debate.
The causal analysis method to be used is RDD. RDD looks at situations where there are treatment and control conditions — here the treatment is when a district elects a Republican, the control is when they don’t (they elect a Democrat). It looks specifically at the times when the assignment of treatment and control was very close; i.e. when a candidate wins by a very narrow margin. At the margin, also known as the cutoff, it assumes that the winner was chosen by chance, a coin flip. The result was random. The assumption of randomness means that we also assume that the candidates running against each other either won or lost are very similar in all variables of interest, this means our results will not suffer from selection bias (bias when groups have pre-existing differences that make them incomparable). Once we have made all these assumptions, we are able to use regression to calculate an outcome score (my outcome of interest is the prevalence of conspiracy belief) for those in the treatment and control groups. The difference between them will tell you whether the treatment has an impact on the outcome — do Republicans make a difference to conspiracy belief? I would plan on measuring conspiracy belief using some very cool analysis of Twitter data.
I think this kind of study is particularly of interest in the current political climate. We’ve recently seen leaders, further towards the right, in some cases supporting debunked theories about 5G, COVID-19 and Q. This study could uncover if there are any lasting ideological effects of certain leaderships. I also chose to design this study because I find conspiracy theories really fascinating, they’re wild, often ridiculous and baseless but they are able to grab people and change their behaviour. A study last year showed that a belief in conspiracies about COVID-19 and vaccines predicted reduced use of preventative measures and vaccine refusal (Romer and Jamieson, 2020). They don’t just make an entertaining read they have the power to change societies and affect us all, hence why research in this field is so critical.
I do consider myself a real scientist (sort of) so I should mention there are a number of limitations in relation to the assumptions of the RDD design and the use of Twitter but overall this would be a largely reliable, doable and interesting piece of research.
