Popular Science

So there is this article which has been making ripples on the Internet today. Supposedly it's trying to put some science behind the argument that democracy in the US doesn't work. The punchline is convenient:

But the picture changes markedly when all three independent variables are included in the multivariate Model 4 and tested against each other. The estimated impact of average citizens’ preferences drops precipitously, to a non-significant, near-zero level. Clearly the median citizen or “median voter” at the heart of theories of Majoritarian Electoral Democracy does not do well when put up against economic elites and organized interest groups.

Except the problem is, the conclusion isn't really accurate. A quick look at the article shows the authors conclude that two of their independent variables are strongly correlated (average citizen preferences and economic elite preferences, see Table 3), and then proceed to ignore the ensuing multicollinearity issues in their modelling approach. It's worth reminding that multicollinearity makes individual variable coefficients unreliable. Moreover, all three independent variables are demonstrably positively correlated with the outcome variable (Table 2), suggesting the conclusion should be exactly the opposite of the one pitched . What's "beyond alarming" here are not the findings or the Internet buzz. It's the fact that analysis like this actually passed peer review this way.