Bryan Caplan  

Exploring Elitist Democracy: The Latest from Gilens and Page

PRINT
The paradoxes of applying nati... Thoughts on Krikorian...
In light of the attention my Gilens posts are getting, now's a perfect time to examine his latest research on who actually runs America.  Co-authored with Benjamin Page, "Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens," (Perspectives on Politics, forthcoming) weighs the relative importance of four models of ultimate power:
While this body of research is rich and variegated, it can loosely be divided into four families of theories: Majoritarian Electoral Democracy, Economic Elite Domination, and two types of interest group pluralism - Majoritarian Pluralism, in which the interests of all citizens are more or less equally represented, and Biased Pluralism, in which corporations, business associations, and professional groups predominate) Each of these perspectives makes different predictions about the independent influence upon U.S. policy making of four sets of actors: the Average Citizen or "median voter," Economic Elites, and Mass-based or Business-oriented Interest Groups or industries.
The dependent variables are straight out of Gilens' earlier work:
Gilens and a small army of research assistants gathered data on a large, diverse set of policy cases: 1,779 instances between 1981 and 2002 in which a national survey of the general public asked a favor/oppose question about a proposed policy change. A total of 1,923 cases met four criteria: dichotomous pro/con responses, specificity about policy, relevance to federal government decisions, and categorical rather than conditional phrasing. Of those 1,923 original cases, 1,779 cases also met the criteria of providing income breakdowns for respondents, not involving a Constitutional amendment or a Supreme Court ruling (which might entail a quite different policy making process), and involving a clear, as opposed to partial or ambiguous, actual presence or absence of policy change.
As far as I can tell, the interest group preferences are also straight out of Gilen's earlier work:
For each of the 1,779 instances of proposed policy change, Gilens and his assistants drew upon multiple sources to code all engaged interest groups as "strongly favorable," "somewhat favorable," "somewhat unfavorable," or "strongly unfavorable" to the change. He then combined the numbers of groups on each side of a given issue, weighting "somewhat" favorable or somewhat unfavorable positions at half the magnitude of "strongly favorable or strongly unfavorable positions.
This coding is the project's main leap of faith.  When Gilens measured the relative influence of voters of different income levels, he had public opinion data for each income level.  For interest groups, in contrast, Gilens and assistants impute opinions.  Still, I don't have a better idea.  If we run with their approach, what do we find?

First, middle-class Americans' preferences are highly correlated with rich Americans' preferences - but not interest groups' preferences:
It turns out, in fact, that the preferences of average citizens are positively and fairly highly correlated, across issues, with the preferences of economic elites (see Table 2.) Rather often, average citizens and affluent citizens (our proxy for economic elites) want the same things from government. This bivariate correlation affects how we should interpret our later multivariate findings in terms of "winners" and "losers." It also suggests a reason why serious scholars might keep adhering to both the Majoritarian Electoral Democracy and the Economic Elite Domination theoretical traditions, even if one of them may be dead wrong in terms of causal impact. Ordinary citizens, for example, might often be observed to "win" (that is, to get their preferred policy outcomes) even if they had no independent effect whatsoever on policy making, if elites (with whom they often agree) actually prevail.

But net interest group stands are not substantially correlated with the preferences of average citizens. Taking all interest groups together, the index of net interest group alignment correlates only a non-significant .04 with average citizens' preferences! (See Table 2.) This casts grave doubt on David Truman's and others' argument that organized interest groups tend to do a good job of representing the population as a whole. Indeed, as Table 2 indicates, even the net alignments of the groups we have categorized as "mass-based" correlate with average citizens' preferences only at the very modest (though statistically significant) level of .12.
Second, each theory looks good in isolation:
When taken separately, each independent variable - the preferences of average citizens, the preferences of economic elites, and the net alignments of organized interest groups - is strongly, positively, and quite significantly related to policy change. Little wonder that each theoretical tradition has its strong adherents.
Third, average citizens lose almost all apparent influence if you do multiple regression:
These results suggest that reality is best captured by mixed theories in which both individual economic elites and organized interest groups (including corporations, largely owned and controlled by wealthy elites) play a substantial part in affecting public policy, but the general public has little or no independent influence.
Check out the multiple regression results:
gilenspage3.jpg

Fourth, if you break interest groups into "business" and "mass" categories, the former have more power than the latter:
The influence coefficients for both mass-based and business-oriented interest groups are positive and highly significant statistically, but the coefficient for business groups is nearly twice as large as that for the mass groups. Moreover, when we restricted this same analysis to the smaller set of issues upon which both types of groups took positions - that is, when we considered only cases in which business-based and mass-based interest groups were directly engaged with each other - the contrast between the estimated impact of the two types of groups was even greater.
Punchline regressions:

gilenspage2.jpg

Conclusion:
By directly pitting the predictions of ideal-type theories against each other within a single statistical model (using a unique data set that includes imperfect but useful measures of the key independent variables for nearly two thousand policy issues), we have been able to produce some striking findings. One is the nearly total failure of "median voter" and other Majoritarian Electoral Democracy theories. When the preferences of economic elites and the stands of organized interest groups are controlled for, the preferences of the average American appear to have only a minuscule, near-zero, statistically non-significant impact upon public policy.

The failure of theories of Majoritarian Electoral Democracy is all the more striking because it goes against the likely effects of the limitations of our data. The preferences of ordinary citizens were measured more directly than our other independent variables, yet they are estimated to have the least effect.

Nor do organized interest groups substitute for direct citizen influence, by embodying citizens' will and ensuring that their wishes prevail in the fashion postulated by theories of Majoritarian Pluralism. Interest groups do have substantial independent impacts on policy, and a few groups (particularly labor unions) represent average citizens' views reasonably well. But the interest group system as a whole does not.
Gilens and Page's latest paper is hardly the final word.  Their results for interest groups hinge on their own coding - an inherently opaque process.  Still, they offer novel evidence on big questions.  They've definitely got my attention.  They deserve yours, too.



COMMENTS (14 to date)
vikingvista writes:

The bimodal correlation in the USA, at least, is largely due to the fact that most of the US rich spent most of their lives as average. I.e., they are essentially the same people.

But the point still holds that democracy is essentially the nonauthoritarian authoritarianism. Thus democracy is a religious conviction for the antiauthoritarian authoritarians.

ed writes:

I noticed that in the popular coverage of this study, there was basically no description of the actual specific issues on which high income and average income Americans appear to disagree. Strange, since reporters usually want to lead with an anecdote to give the story some color.

There is a reason for this: if you download the data (which is freely available), you'll see that the biggest differences tend to be either on religious and cultural issues (elites tend to be more in favor of abortion and against school prayer for example), and economic issues (where the average American says she favors policies like federal price controls for natural gas, or requiring a years notice before closing a workplace.)

It's understandable that Gillens and Page want to talk in general terms about the oligarchy, and don't really want to spend a lot of time defending the actual specific policies that they predict would arise if our democracy was "healthier."

vikingvista writes:

The bimodal correlation in the USA, at least, is largely due to the fact that most of the US rich spent most of their lives as average. I.e., they are essentially the same people.

But the point still holds that democracy is essentially the nonauthoritarian authoritarianism. Thus democracy is a religious conviction for the antiauthoritarian authoritarians.

Pajser writes:

So, wealthy people are the ruling class, Marx was right and John Galt was wrong. It appears we have some progress here.

Brandon Berg writes:

A hugely underreported fact about this paper is that "economic elite" means people at the 90th percentile of the income distribution. The media are trying to sell this as being proof that the rich are buying their preferred policy outcomes, but that's not plausible when we're taking about the merely well off rather than the truly rich.

Garth Zietsman writes:

The R-square for the whole model is a mere 7%. This suggests that NO group has much influence over the political process. Politicians are doing their own thing regardless of what the elite want.

gwern writes:
The R-square for the whole model is a mere 7%. This suggests that NO group has much influence over the political process.

Measurement error and discretizing variables will bias the estimates downwards, no? (I don't think this even measures other avenues of influence like donations; at least, Caplan doesn't mention money.) And 7% is small... compared to what, exactly? Are there a bunch of other factors which predict much more of variance? What are they?

Norman Pfyster writes:

With no mention of the Illuminati, it's hard to take such a study seriously.

KPres writes:

What about the correlation between income and IQ/education? My guess is higher income people get their way more because they're smarter and the policies they prefer are better. This is reinforced by the fact that what they're calling "wealthy" is the top 10%, not the top 1% or .01%. That means lawyers, physicians, etc. All of these people have interests that align more with Joe Six Pack's then they do with Bill Gates and Warren Buffet.

Furthermore, the entire purpose of representative democracy, as opposed to a direct democracy, is to put a set of elites in control. The "will of the people" is supposed to be ignored when it's stupid.

Jason Harner writes:

Be careful interpreting the R-Squared, remember that this a logit regression, and it's very difficult to generate models that can predict binary events with near certainty.

I also think it's interesting they classified the top 10% income bracket as "economic elites". It doesn't come as very surprising that this group, who likely have far more time and opportunity to study political issues and form strong opinions, may strongly influence the public narrative and thus influence the median voter even under Majoritarian Electoral Democracy.

I've always operated on the assumption that a large portion of society simply doesn't care about public policy that doesn't substantially impact them. I think that's a problem of surveying the general population on all issues, specifically ones for which they don't have skin in the game, so to speak.

Chris Koresko writes:

If the 'elite' is the top 10% of the income distribution, I'd guess they're fairly distinctive in the age distribution as well. So it may be that they are influential because they've been around long enough to develop connections that can influence politics... that might be more important than mere money.

Just speculating, of course.

Mark V Anderson writes:
I've always operated on the assumption that a large portion of society simply doesn't care about public policy that doesn't substantially impact them. I think that's a problem of surveying the general population on all issues, specifically ones for which they don't have skin in the game, so to speak.

Yes, this is the most important point in this whole discussion. Of course the opinion leaders in the media, business, and politics are all in this top 10% of income. The average folks follow the lead of these opinion leaders because they have no interest in developing their own opinions. The top 10% are not opinion leaders because they are wealthy, they are wealthy because they are opinion leaders.

Jordan M writes:

R-squared aside, the main model in this paper is inherently broken -- it's not robust to specification, and the multicollinearity among LHS variables all but invalidates the conclusion about the significance of average citizen preferences the multiple regression.

The statistical work here is so bad that one can't help but wonder what type of referees read this. Certainly not ones that are even moderately well-versed in advanced quantitative methods.

Jordan M writes:

To elaborate on my previous comment: average citizen and economic elite prefs are correlated with a factor of 0.78. The authors are either not aware of what happens when you put this type of variables as predictors in the same regression, or choose to simply ignore it to push a conclusion.

Comments for this entry have been closed
Return to top