Princeton Study Concludes US is an Oligarchy not a Democracy

[eyeball]

...a self-deprecating sardonic critique?

[-1=e^ipi]

@ cybercoma - multicollinearity understating the effect of the income group is only true under a number of questionable assumptions.

But in any case, multicollinearity and low R-squared are only two of the 8 points I made.

My primary concern is that they way they test for Majoritarian Electoral Democracy (see point 1).

[cybercoma]

It's a good point, but not something that this study was looking to explore. You've identified a problem with the American de facto 2-party system more generally: that people's opinions don't necessarily fit nicely into two boxes. But that's the point of the paper isn't it? When the majority of people lean a certain way, it doesn't matter because the two parties that are making policy tend to gear their policy towards the economically advantaged groups. Your point here I think underscores the conclusions that are made by the paper and speaks more to the mechanisms at play, rather than simply testing the relationships as they did. Either way, your point doesn't undermine the findings of the paper. It's another topic altogether that's well worth exploring.

[cybercoma]

Also, I believe their response to your point is found in footnote 33.

To make an approximate assessment of the fit between the preferences of the median preference

respondent and those of the median-income respondent we calculated the median

preference within each of five income groups (at the 10th, 30th, 50, 70th and 90th percentiles).

For example, if, on a particular non-monotonic item, the five income groups had imputed median

preferences of .50, .60, .70 .65 and .55 (for the 10th to 90th income percentiles, respectively),

then the estimated over-all median preference would be .60 (in this case equaling the preference

at the 30th income percentile). In most cases the over-all median preference is the same as the

median income (50th percentile) preference. When it is not, the preference differences across

income levels tend to be small. Using this technique, the median over-all preferences and the

median-income preferences track each other very closely: r=.997.

[-1=e^ipi]

It's a good point, but not something that this study was looking to explore.

... They are testing 4 different political theories, that is the main point of the paper. If the way they test for Majoritarian Electoral Democracy is flawed, then that has huge implications on their paper, because they main conclusion that they make is a rejection of Majoritarian Electoral Democracy.

You've identified a problem with the American de facto 2-party system more generally: that people's opinions don't necessarily fit nicely into two boxes. But that's the point of the paper isn't it?

... The paper assumes political unidimensionality in order to use median voter theorem to test for Majoritarian Electoral Democracy...

When the majority of people lean a certain way, it doesn't matter because the two parties that are making policy tend to gear their policy towards the economically advantaged groups.

Even if I pretend that their test for Majoritarian Electoral Democracy made sense, there are other explanations for their result (that do not reject Majoritarian Electoral Democracy).

Maybe the policy makers do mostly consider the public when making decisions and Majoritarian Electoral Democracy is the closer of the 4 models to reality, but the reason we get a non-significant result is because the public is most uninformed or indifferent about most policy issues.

When I ask a number of friends or family members about how they feel about a certain policy (ex. what do you think about the oil sands?), most of them don't have an opinion or don't care. Something like 40% of the voting population doesn't vote. So when you gauge public opinion the result will be heavily influenced by those that do not vote, care or have an opinion (where as interest groups will obviously be informed and care about an issue). Add to this the fact that large segments of the population believe in nonsense like the lump-of-labour fallacy and things can get very problematic since policy makes will tend to listen more to constituents that care or are informed about a particular issue that those that do not care or are uninformed.

The US has representative democracy, which means that people elect people to make decisions for them. The elected representatives are supposed to be informed and listen to all constituents. Maybe elected representatives are more informed than the general public, and in addition the economic elites and interest groups are also more informed than the general public. Then you could get a regression result that seems to contradict Majoritarian Electoral Democracy even though Majoritarian Electoral Democracy maybe be true, since the correlation between economic elites & interest groups with the decisions of policy makes might be due to each of these factors being correlated with being informed (i.e. informed people are more likely to come up with the same conclusion).

Their study doesn't account for how much people care or inform themselves about a particular issue. For all we know, the opinion of median voter that cares and is informed about a particular issue might be closer to the opinion of the 90th percentile than it is to the 50th percentile. Of course what I state here is testable, so if they a future group can account for how informed or how much individuals care about a particular issue (by using survey data for example) then they might able to exclude the care/informed explanation.

Either way, your point doesn't undermine the findings of the paper.

Even though their main finding is the rejection of Majoritarian Electoral Democracy?

Also, I believe their response to your point is found in footnote 33.

With respect to the footnote, why didn't they just directly use the median preference respondent rather than the 50th percentile to represent the median voter in order to test for median voter theorem? Using the 50th percentile doesn't make as much sense.

While on the topic, why didn't the authors also try to regress the 10th percentile or other percentiles to test to see if the 'poor' people have a significant impact on policy decisions even when other factors are included? It would have been useful for the sake of consistency.

Furthermore, have they provided anyone with the data set for replication? This whole thing seems a bit suspicious. I wonder if the group chose to do many different regressions and picked the regression that gave the result that fitted with their conclusions the most. Such actions are not uncommon in academia, especially when there is a high emphasis on publishing results to get funding. Heck, even Milikan won the nobel prize for his oil drop experiments after he threw out all the data that didn't fit with his model (and didn't tell anyone).

[jbg]

Link to the study here.

Authors Martin Gilens and Benjamin I. Page from Princeton University conducted a multivariate analysis on the influence of various groups on US government policies. They found that average citizens and mass-based interest groups had little to no effect on policy outcomes, whilst economic elites and organized groups representing business interests had a large and significant effect on policy. The article clarifies what they mean by economic elite (90th percentile income earners) and what kinds of interest groups fall under their different categories (in the appendices). Essentially, the article concludes that the US policy is driven by the economic elite and business interests, while the average person has little to no influence whatsoever. Looks like the government truly is for sale to the highest bidders.

Is an alternative explanation possible, that companies that are active in certain businesses may know more, or have a larger stake, and thus be more vocal and organized?