Some things live forever in social media. In my circles, one article that comes up all the time is the Marten and Gilens study of legislative influence that is often interpreted this way: “US No Longer an Actual Democracy” or “Princeton Concludes What Kind of Government America Really Has, and It’s Not a Democracy.”
Part of the problem is that I think there are serious political inequality issues in the US. But the study fails to prove it, and it’s been blown out of proportion in lots of disempowering ways. The study is basically a victory for the null hypothesis: on reflection, we don’t have much evidence that policies are determined by a single group.
Here’s how it works: the study compares the policy preferences of four groups: median voters, voters in the top 10 percent of income, business-oriented interest groups, and “mass-based” interest groups. Using data from survey responses, they note that policy outcomes more often correlate with the rich and business-oriented interest groups than with the median voter and mass-based interest groups. Case closed, right?
No so fast: It’s still the case that the best predictor of the success or failure of any policy initiative is the median voter. But the median voter and the 90% voter usually agree. When they don’t, the 90% voter usually either a) prefers the status quo or b) prefers the socially liberal position. And so over the last 20 years, the status quo has tended to win out (in aggregate) except when it is broadly favored by median and 90% voters or is socially liberal and favored by the 90% voter.
A lot of this paper depends on obscuring some important issues behind difficult math. Their R2 is .074; that’s actually, pretty close to random. It means that median voters, economic elites, and interest groups, together, determine 7.4% of the variability in policy outcomes! Almost all of that is elites and interest groups, it’s true! But when you’re dealing with such small numbers, it’s easy to get *statistically* significant results that aren’t actually all that significant in human terms. In 92.6% of the policy outcomes, none of these independent variables was predictive.
That means that no group predominates most of the time, but “US System of Government is 95% Democratic” doesn’t get people to link to your study on social media.
What’s worse is that most of my progressive readers will disagree with the policy issues that the median voter has lost on.
A Metafilter friend who looked at the SPSS file wrote: “Of the 1779 polls, 105 ended up with policy being what rich people wanted and what median-income people didn’t, out of 189 polls where rich people and regular folks disagreed. The only times there was more than a 20-point gap (ie 60% of rich wanted one side but only 40% of regular people did) were the 10 or so questions about NAFTA. The rest of the time it was a slight majority of rich people favoring something and a smidge under 50% of regular folks favoring it.
Rich people got their way mostly on social issues — RU486 legality, IDX legality, gays in the military, various abortion/birth control restrictions. But also stuff like outright banning immigration for five years, the legality of public-sector strikes, outright bans on military or domestic aid to any foreign country, and so on.”
The median voter has sometimes been prone to serious mistakes or moral failings, as you can see. So the real title should be “US System of Government is 95% Democratic and a Lot of the Rest of The Time the Demos are Assholes and Deserve to Lose.”
Then, too, nothing in their data can disaggregate the top 1% or the top 0.1%, as the authors freely admit. It also doesn’t ask about the bottom 10% or the bottom 1%, which is a perspective I much prefer.
But that means that their conclusions regarding elite domination are even less well-supported. As a progressive I’m primed to suspect that very wealthy interests usually dominate: what’s interesting is how hard it is to prove what “everybody knows.” If you believe in their conclusions, then you should actually downgrade your priors on the basis of this study. Mostly, this is a study that demonstrates how hard it is to do serious empirical work on this question. It ignores selection bias issues like which policies pollsters poll on. (Hint: they tend to poll on popular policies that are not yet in place.) If there aren’t any polls on the vast majority of things affluent and median disagree on, we’d be in the dark about affluent influence. As longtime readers know, my favorite policy initiative is the basic income guarantee, but it’s very rarely polled and when it is polled, it’s fairly unpopular.
So I do think “median voters and affluent voters generally agree in polls” is the most significant finding here. Democracy may often lead to this confluence of elite and mass interests, either through propaganda or elite-deference to voters. Perhaps voters judge outcomes rather than policy inputs, so politicians aim to guarantee good economic and social outcomes even when these contradict the policy preferences that would have unintended consequences. Perhaps it’s even true that the petite-bourgeoisie still finds its class interests allied with capital most of the time. The real victims can’t vote, right?