In their analysis of survey responses regarding proposed federal policies, Martin Gilens and Benjamin Page compare the policy preferences of “average citizens” versus “economic elites. They use the policy preferences of median-income survey respondents as a proxy for the preferences of the average citizen and the preferences of survey respondents in the top 10% income bracket as a proxy for the very wealthy. They found that if only elites favored a policy proposal, it was adopted by the US government a bit less than a third of the time. But if only average citizens favored a policy proposal, it was almost never adopted. Gilens and Page conclude that “[w]hen a majority of citizens disagrees with economic elites or with organized interests, they generally lose” (576)  – a state of affairs under which  “America’s claims to being a democratic society are seriously threatened.” (577).

So, somehow, the average citizen has become a majority of citizens, which is odd since “average” doesn’t mean “most”. Gilens and Page also seem to assume that if politicians don’t adopt policies favored by most respondents in opinion surveys, they are not truly representing the majority of the electorate. But are survey responses a sufficient gauge of public support for government policy? Doesn’t the public’s behavior in the voting booth matter? After all, politicians need to secure a majority of votes to get elected. If the majority of voters don’t like the policies they get from the politicians they voted for, aren’t these politicians likely to lose their jobs?

Put simply, Gilens and Page are making confident claims about the implications of their research that go way beyond their findings. In doing do, they are violating several basic principles of good research:

“Good research is cautious about drawing conclusions, careful to identify uncertainties and avoids exaggerated claims. It demands multiple types of evidence to reach a conclusion. It does not assume that association (things occur together) proves causation (one thing causes another). Bad research often contains jumps in logic, spurious arguments, and non-sequiturs (‘it does not follow’).” Todd Litman

References:

Gilens, M. and Page, B.I. “Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens” Perspectives on Politics September 2014 | Vol. 12/No. 3, 564-58.

Litman, Todd “Evaluating Research Quality:  Guidelines for Scholarship” 22 February 2012; Victoria Transport Policy Institute.

In their paper Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens (2012), Martin Gilens and Benjamin Page analyze survey data on public support for proposed federal policies. They found that median-income “average citizens” and affluent citizens (“economic elites”) agreed on policy preferences 78% of the time and when they disagreed, policies supported by the economic elites were later adopted a bit less than a third of the time.  When average citizens supported a policy but the economic elites did not, the policy was almost never adopted. How big this latter data set was, I can’t say:  Gilens and Page do not provide us the figures.

Gilens and Page conclude that average citizens lack influence on government policy. If average citizens get policies they support, it’s only because economic elites also favor the policies. Although they decline to speculate how elites actually exercise influence, Gilens and Page approvingly quote the Communist Manifesto on how the modern capitalist state exists to serve the material interests of the dominant classes.  They further argue that their findings support theoretical models of “Economic-Elite Domination” and “Biased Pluralism”, in which the economic self-interest is central to elite policy support. Bottom line: policy support reflects self-interest and average folks lack influence because the state serves the rich.

Gilens and Page do not discuss potential confounders in their analysis of the data. For example, they identify respondent income as the variable of interest, but income may be a proxy for other factors, like age or education. After all, affluent Americans are more likely to be of prime working age than people at lower income levels and are more likely to be college graduates, as the following chart shows:

Given that federal policy makers are usually college educated and in their prime working years, is it possible that the greater political alignment of affluent citizens and policy makers reflects similarities other than income?* Gilens and Page refer obliquely to this possibility but only to ridicule it:

“A possible objection to populistic democracy is that average citizens are inattentive to politics and ignorant about public policy; why should we worry if their poorly informed preferences do not influence policy making? Perhaps economic elites and interest group leaders enjoy greater policy expertise than the average citizen does. Perhaps they know better which policies will benefit everyone, and perhaps they seek the common good, rather than selfish ends, when deciding which policies to support.

But we tend to doubt it. We believe instead that— collectively—ordinary citizens generally know their own values and interests pretty well, and that their expressed policy preferences are worthy of respect.” p.156

Here Gilens and Page are essentially shrugging off serious consideration of confounding variables by using a strawman argument. Instead of addressing whether economic class or factors associated with class impact how much individuals pay attention to and understand policy issues, which in turn may affect policy preferences, they present a cartoonish version of the implications of such consideration: that ordinary citizens are ignorant, their views unworthy of respect, and it’s a good thing elites are calling the shots. Who could possibly agree with that?

Nor do Gilens and Page examine the assumption of a rigid class structure in the US, where average citizens and elite citizens are distinct groups with stable membership. This is simply untrue.  Income varies enormously across the life span, typically peaking in middle age (45-54).  According to Auten, Gee and Turner in “Income Inequality, Mobility, and Turnover at the Top in the US, 1987-2010”, most Americans will spend time in both the bottom and top income quintile. At least 39 % of Americans will reach the top 5 % of earners, 56 % will get to the top 10 % and a whopping 73 % will spend at least a year in the top 20 % of earners. And more people fall out of the top income brackets than stay stuck at the bottom.

So when Gilens and Page say…

 “When a majority of citizens disagrees with economic elites or with organized interests, they generally lose.” (576)

…the majority of which they speak are not the same people from year to year. Who knows? It’s possible that a good number of the surveyed average citizens and economic elites had switched places within the four-year post-survey follow-up period that Gilens and Page allowed for policy adoption.

Next: how did the median-income average citizen morph into a “majority” of citizens and can we really conclude that the majority of US citizens have no independent political power?

*As it turns out, income is very much associated with education, civic engagement, and political knowledge.     In “Fault Lines in Our Democracy: Civic Knowledge, Voting Behavior, and Civic Engagement in the United States”, Richard Coley and Andrew Sum found that “the oldest, most highly educated, and highest income” individuals were by far the most engaged and knowledgeable. For instance,  just 7% of college graduates expressed no interest in political affairs. compared to 30% of high school graduates.   Furthermore, as documented by  William Galston in “Civic Education and Political Participation”,  civic knowledge can alter opinions on specific issues, such as immigration policy.

 References:

Auten, G., Gee G. and Turner, N.  “Income Inequality, Mobility, and Turnover at the Top in the US, 1987-2010.” American Economic Review, 2013; 103(3): 168-72.

Coley, R.J. and Sum, A. (2012) Fault Lines in Our Democracy Civic Knowledge, Voting Behavior, and Civic Engagement in the United States. Educational Testing Service.

Galston, W.A. “Civic Education and Political Participation,” Political Science Online, April 2004.

Gilens, M. and Page, B.I. Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens Perspectives on Politics September 2014 | Vol. 12/No. 3, 564-58.

“Good research is cautious about drawing conclusions, careful to identify uncertainties and avoids exaggerated claims. It demands multiple types of evidence to reach a conclusion. It does not assume that association (things occur together) proves causation (one thing causes another). Bad research often contains jumps in logic, spurious arguments, and non-sequiturs (‘it does not follow’).” Todd Litman

Good research papers are imbued with the spirit of humility. Their authors seriously consider alternative explanations for their research findings and do not overstate the significance of these findings. Based on these two criteria alone, let’s look at Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens by Martin Gilens and Benjamin I. Page, which created quite a stir in 2014 when it was published.

In this paper, the authors looked at pro/con survey responses regarding support for various federal policy proposals and then determined if the proposals were later adopted as US government policy. Survey respondents were classified as either median-income “average citizens” or “economic elites” (top 10% income bracket, serving as a proxy for the very wealthy). The authors’ data analysis revealed that economic elites and average citizens had the same policy opinions about 78% of the time. When both groups supported a policy change, the change was later adopted a bit less than half the time.  When they disagreed, policy changes supported by elites were occasionally adopted but policies supported by average citizens alone were almost never adopted. Thus, according to Gilens and Page, when “a majority of citizens disagrees with economic elites…they generally lose”* (p. 576). They conclude that “majorities of the American public actually have little influence over the policies our government adopts” and “a small number of affluent Americans” have too much influence. (pp. 576-577)

Next up: did the authors seriously address alternative interpretations of their data and are their conclusions warranted by their findings?

* The Gilens and Page paper also looks at the influence of business and mass public interest groups on government policies, which I will not be addressing in these posts.

References:

Martin Gilens and Benjamin I. Page Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens  Perspectives on Politics September 2014 | Vol. 12/No. 3, 564-581 doi:10.1017/S1537592714001595

Todd Litman Evaluating Research Quality:  Guidelines for Scholarship  22 February 2012 Victoria Transport Policy Institute