Stop trying to prove conservatives are intellectually inferior to liberals (part 8)

Yet another study claims to prove that conservatives are inherently inferior to liberals. This one goes further than any other I’ve seen by claiming that moving all state policies to the most liberal possible orientation could have saved 171,030 lives.

The study measures mortality by looking at deaths from all causes between 1999 and 2019 as well as deaths from Cardiovascular disease (CVD), alcohol-induced causes, suicide and drug poisoning among adults aged 25-64. These causes are assessed specifically as they have been identified as key contributors to the disparity in life expectancy between the U.S. and other countries. This data was merged with state-level data on eight policy areas e.g. labour, taxes etc. Each domain is scored from 0-1 where 0 is the most conservative and 1 is the most liberal. These scores are used to build regression models that use policy scores to predict the number of deaths.

Let’s consider the data first. The authors only look at policy data. They justify this by citing a study by Couillard et al. which argues that policies are one of the biggest influences on mortality. But Couillard et al cite other key influences, for example population behaviour such as levels of smoking. Couillard et al. highlight the risk of considering policy without population behaviour because of ‘the likely feedback between these two “inputs” into the health-capital framework’.

Now let’s look at the labelling. The authors provide the following information on how they labelled their data: ‘Policies were considered liberal if they expand state power for economic regulation and redistribution, protect the rights of marginalized groups, or restrict state power to punish deviant behavior’ – There are several questionable terms here. Conservatism tends to be associated with limiting the reach of the state. So why would restricting state power to punish deviant behaviour be liberal rather than conservative? And what is ‘deviant behaviour’ exactly? Who says protecting the rights of marginalized groups is inherently liberal?

Now let’s look at the modelling. Recall that the authors claim that 171,030 lives could have been saved if all states moved their policies to a liberal orientation. They arrive at this number by hardcoding policy values to represent hypothetical scenarios. One scenario involves setting all the policy values to 1 to represent moving all states policies to a liberal orientation. These hardcoded values are passed to the model to get the predicted number of deaths in a certain scenario. This is called extrapolation, where a model is used to predict on a data point that is outside the range of its training data. Extrapolation is generally discouraged because we have no guarantee that a model can predict properly on a kind of data it’s never seen. Consider a model for predicting a persons’ salary based on their age that is trained on data from 20-45 year olds. We wouldn’t apply this model on a 60 year old.

Finally, let’s consider the language used in the paper. The authors do state that their results shouldn’t be considered causal but the rest of the paper strongly implies causality. References to saving lives are made multiple times in the paper. Multiple media outlets have read this study as being causal, like this article in The Guardian titled ‘Americans die younger in states run by conservatives, study finds’ or this article in Truthout titled ‘People Who Live in Red States Die Younger. Conservative Policies Are to Blame’.

I’ve said it many times before: statistics can’t prove that conservatives are inherently inferior to liberals