Home » Academia » Commentary on Gilens and Page, “Average citizens have no political influence”

# Commentary on Gilens and Page, “Average citizens have no political influence”

Join 395 other followers

This is an interesting paper, that brings admiral clarity to the competing theoretical models that address the problem, ‘Who governs? Who really rules?’ (Gilens and Page. 2014, p.3). However I’m skeptical as to whether the authors’ dataset provides unequivocal support for the general equation between ‘electoralism’ and oligarchical rule claimed by Yoram Gat in his open letter to Professor Gilens, for the following reasons:

1. Dataset
It’s surprising that a total of 1,932 cases yielded as many as 1,779 instances demonstrating a clear relationship between public preferences and policy change (p.10). Most legislative outcomes involve messy compromises involving trade-offs between the preferences and interests of the various parties involved. What criteria were employed by Gilens’s ‘small army of research assistants’ in order to decide that these 1,779 instances involved a ‘clear, as opposed to partial or ambiguous, actual presence or absence of policy change’ (ibid.)? Are public preferences really as unambiguous as the authors claim? An influential work by Benjamin Page’s frequent collaborator Robert Shapiro used the examples of Bill Clinton’s (failed) healthcare reforms and Newt Gingrich’s ‘Contract with America’ as examples of elite- and partisan-driven policy initiatives (Jacobs and Shapiro, 2000). However in the former case survey evidence was ambiguous: a Gallup Poll conducted in early August 1991 indicated that 91 percent of the public felt there was a ‘crisis in healthcare’ (Gallup, 1991, p. 4) and a large majority (75% of adults polled) wanted the government to provide healthcare (Times, 1992). But it was not clear what the public wanted done about health care, being torn between the desire for comprehensive provision and the deep-seated American aversion to big government: ‘different polls and even successive questions in the same polls turn up seemingly contradictory responses’ (Kosterlitz, 1991, p. 2806). In any event, Clinton’s healthcare reforms were defeated: ‘the policy outcome turned, in the end, on the response of the relatively few centrist legislators to – exactly – the median national opinion as measured by polls’ (Quirk, 2009, p. 6, my emphasis). Similarly the GoP ‘Contract with America’ was entirely driven by the median-voter strategy:

The issues that garnered very favourable ratings with the public were included in the contract and those that did not were left off. There was little discussion about how these policies fit together, rather the concern was maximizing popularity. (Geer, 1996, pp. 34-5, my emphasis).

2. Multivariate Analysis
I’m not a political scientist or a statistician, so I’m not well qualified to comment on this; however the authors acknowledge 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 [.78], average citizens and affluent citizens (our proxy for economic elites) want the same things from government [p. 14]

The problem only arises for the (minority of cases) when the preferences of elites differ from those of the median voter [.22?]. I speculate on these cases in the paragraph below, but I’m not sure why in the (majority of) cases where priorities coincide, the authors still claim that this reinforces elite theory. It may well be the case that average citizens have ‘little or no independent influence on policy’, [p.15, my emphasis] but the implication – that government policy in this majority of cases is determined by elite preferences – is not warranted by the data. If Janet and John both prefer curry to pizza, why attribute the decision to visit the Indian restaurant to John rather than Janet (or vice-versa)? Although politicians are responsive to campaign contributions their ultimate need is to secure votes, so the obvious conclusion, in the two-thirds majority of cases, is that they are following public preferences and elite opinion (with the emphasis on the former). Elections are decided by the number of votes cast.

3. Political Economy
The USA may well be a ‘hyper-responsive’ democracy (King, 1997) but its political economy is robustly capitalist. The ‘American Dream’ is not for the collectivist New Jerusalem but that everyone can aspire to join the ranks of the rich ‘n powerful. Economic success and the resulting general prosperity is believed, rightly or wrongly, to result from the commanding heights of the economy being in the hands of private enterprise, as opposed to ‘big government’. Decision-making in private businesses is by the owners of capital (and/or their appointed executives). Givens and Page acknowledge a loose correlation between the owners of capital and economic elites (the principal focus of their paper), so it is unsurprising if, in a capitalist political economy, those who have demonstrated success in civil society should have a disproportionate influence in public affairs, particularly where economic policy is concerned. One of the first moves of the (UK) Thatcher government was to recruit successful business people into Whitehall, in order to dilute the power of the Oxbridge generalist culture that predominates in the civil service. This has become a bi-partisan policy, as the goal is to harness expertise, the relationship with wealth being seen as entirely contingent. Both economic elites and biased pluralism are natural aspects of capitalist political economy. Givens and Page offer no cross-cultural data, but one might well anticipate that in (European) social-democratic political economies there would be an even more than two-thirds link between median voter preferences and policy outcomes.

4. American Exceptionalism?
Givens and Page rightly entitle their paper ‘Testing theories of American politics’ as it could well be that this is the exception rather than the rule. They only offer US data, thereby providing no empirical foundation for those seeking to make the general connection between electoralism and oligarchy. In fact there are very good reasons for viewing US politics as exceptional:

a) During the period that they consider (1981-2002), US politics has become more partisan, whereas in the same period UK politics has converged on the median voter (‘Mondeo Man’; ‘Worcester Woman’). The latter phenomenon has been attributed to the effect on the Labour Party of the collapse of Eastern block socialism, whereas US politics (which has never included a mainstream socialist party) is still reeling from the fallout from the civil rights legislation of the 1960s and 1970s (Jacobs and Shapiro, 2000, p. 31).

b) The US would appear to be unique in having no effective cap on political funding (for largely constitutional reasons). Party funding in the UK is much more tightly controlled, political advertising is banned on TV and electoral broadcasts are rationed. There is no demonstrable connection between electoral success and party funding, the current prediction being that UKIP, a party with negligible financial resources, will win the May 2014 European Parliament elections.

In conclusion, Gilens and Pages’s analysis is limited to US politics and no general conclusions about electoralism and oligarchy can be drawn from it. Furthermore, unless my reduction of the multivariate analysis to the Janet and John scenario is mistaken, then Figure 1 is highly misleading as, where elite and popular preferences coincide, ‘U.S. federal government policy is consistent with majority preferences roughly two-thirds of the time’ (p.5). The authors’ claim that ‘ordinary citizens get what they want from government only when they happen to agree with elites or interest groups that are really calling the shots’ (p.18, my emphasis), relies more on Mosca and Mills than on a sober analysis of the data.

When it comes to the use and abuse of statistics we would all do well to heed Mark Twain’s advice.

References

Gallup. (1991). Gallup Poll Monthly, August 1991.

Geer, J. G. (1996). From Tea Leaves to Opinion Polls: A theory of democratic leadership. New York: Columbia University Press.

Gilens, M., and Page, B.I. (2014), Testing theories of American politics: elites, interest groups, and average citizens, Perspectives on Politics (forthcoming).

Jacobs, L. R., & Shapiro, R. Y. (2000). Politicians Don’t Pander: Political manipulation and the loss of democratic responsiveness. Chicago: University of Chicago Press.

King, A. (1997b). Running Scared. Atlantic Monthly(279), 41-61.

Kosterlitz, J. (1991). National Journal, November 1991

Quirk, P. J. (2009). Politians do pander: Mass opinion, polarization, and law making. The Forum, 7 (4, Article 10).

1. keithsutherland says:

I note that the status of this paper is unpublished, and would be interested to see if the journal’s editors approve the publication of figure 1: “Average Citizens’ Preferences” — there being no mention that this is the residual dataset once the majority (2/3) correlation with economic elites’ preferences has been factored out. This strikes me as the sort of polemical and misleading graphic that you would expect to see in a tabloid newspaper, not a peer-reviewed journal. By way of analogy this is like saying my wife and I are on the brink of divorce because we disagree over a few things (ignoring the fact that we are in full agreement 2/3 of the time). Hopefully statisticians such as Conal and Yoram will correct me if I have misunderstood this.

Like

2. bernie1815 says:

I just found your commentary. I agree. The headlines derived from the press release of this paper are misleading and indicate to me that few have actually read and thought through the methodology of this paper. First, little of the variance in policy outcomes is explained by the proposed variables. Secondly, how does one fit the platforms of the two parties into this analysis. Thirdly, a large number of high profile interest groups are not listed in Appendix A. Fourth, as you suggest, there is something odd about using a multivariate technique when the variables are so complex and multifaceted. I also look forward to seeing what the more statistically savvy reviewers have to say about the study. Finally, it strikes me that the bottom-line of the study could also be written as “Groups with strong interests in particular policies and pieces of legislation are led by members of the elite and tend to influence the final outcome of legislation more than groups and people who are less interested.” Now that is not very surprising or interesting at all.

Like

3. keithsutherland says:

Thanks Bernie, good to hear I’m not in a minority of one in this matter. Like you I look forward to a response from some of the more statistically-savvy commentators on this list — would be grateful if you keep us in touch with any comments made in the wider public and scholarly domain.

Like

4. keithsutherland says:

Gilens and Page refrain from speculating on the possible mechanisms whereby the preferences of rich voters are more likely to be adopted by legislators (in the minority of cases where they fail to coincide with the preferences of ‘ordinary’ voters). Three possibilities come to mind:

1. Rich citizens ‘buy’ policies via campaign contributions.

2. The preferences of rich voters and legislators are more likely to overlap via common interests and cognitive styles.

3. Legislators follow majority public opinion when they can and only deviate when stymied either by the ‘laws of arithmetic’ (as Dennis Healey put it), conflicting public preferences, party ideology or overriding political norms.

Looking at the threads in the Washington Post etc, most knee-jerk commentators go for option 1, despite the fact that “rich” is defined as 100,000+ (not the income bracket of serious campaign contributors). I would imagine that Yoram and Terry (and followers of Scott Page) would go for option 2. Mainstream political scientists would probably go for option 3, as politicians are viewed as rational actors, motivated primarily by the wish to accumulate as many votes as possible. It would be interesting to hear what mechanisms Gilens and Page suggest. Like 5. bernie1815 says: Keith: You have highlighted an extremely important issue. If there is no mechanism what is this research actually saying? The 3 options you pose presume the meaningfulness, precision and robustness of the proposed categories. Nor do they seem to me to be mutually exclusive across policies or with respect to a particular policy. I think the models distinct independent variables are too loosely defined to carry the explanatory weight that many think they can. Like 6. keithsutherland says: Bernie, Makes sense, probably a combination. Going back to your earlier comment on the absent high-profile interest groups, the glaring omission is the Tea Party (although I don’t think the data covers the relevant period). Given the sheer numbers involved in AARP, Tea Party and even NRA, I don’t see the evidence against majoritarian pluralism as that robust. The whole project appears to be driven by the authors’ own political agenda (I note also that Larry Bartels, the author of the Washington Post piece, is also a member of their team). Who leaked the m/s to the press such a long time ahead of publication? I publish a number of scholarly journals (including History of Political Thought) and I would take a dim view of such a breach of embargo and copyright. What is your own background/speciality? Like 7. bernie1815 says: Keith: I was a management consultant with a background in economics, organizational behavior and statistics. I designed large scale surveys and developed various type of psychometric instruments. Like 8. keithsutherland says: A very relevant skillset. Might I ask you to look over the draft chapter of my PhD dealing with the questions raised by the Gilens and Page paper when it’s ready? I was hoping for a little more feedback from this forum but we seem to be experiencing something of a commentary black-out at the moment. Like 9. bernie1815 says: Sure, but my background in political theory and philosophy is both limited and fading. Like 10. keithsutherland says: No problem, this is the chapter that tackles the empirical literature. Like 11. HH says: I read that paper recently and being a statistician, I find the analysis rather poorly done. The dataset itself seems to be very interesting (minus all the points mentioned above), but in the statistical analysis itself, little data is provided at all. The models are only very roughly described, the most interesting analyses seem not to be included at all. For example, in order to make the claim of the authors – looking into subjects where people stronly disagree (and restricting the analysis to those) would be interesting (i.e. average voter 60% or something along those lines). Restricting to those cases would actually give you a clue – to stick to the Indian restaurant metaphor – if John wants to go but Janet doesn’t and they end up going there, you can much better imply who made the decision. Like 12. Agreed — focusing on the exceptions might give a better idea as to what mechanisms are involved. Given that the ultimate currency of electoral politics is votes (rather than money), one would anticipate that if there was strong disagreement the majority position would win. If you strongly disagree with a politician over an important issue then no amount of advertising will induce you to vote for her. The paper goes against the majority polsci viewpoint and the poor analysis and rush to publicity might make one think that it was written for political, rather than scholarly purposes. I’m surprised that some of the defenders of the Mosca/Mills perspective (the dominant view on this forum) have not rallied to its defence. Like 13. bernie1815 says: HH: I agree. It strikes me more as a unit of analysis issue. The focus should be on individual policy decisions and then look at the pattern across types of issues. Like 14. HH says: I have downloaded the dataset the authors published in 2012 and had a brief look. A couple of things stand out: 1) There are barely and policy initiatives where there is a big disagreement (i.e. one group > 60%, the other 55%, 55%, < 45%), the interesting part is that there are more initiatives where rich subjects agree and average people oppose than vice versa 3) The policy initiatives that rich people favor (and average voters oppose) have are only marginally more likely to be enacted than all initiatives on average 4) Policy initiatives favored by the average voter and opposed by the rich are less likely to be enacted than the average (though I did not check if it is statistically significant; given the low number of these cases very likely not) 5) the area of the initiatives have very different likelihoods of being enacted. E.g. Foreign policy is very likely to be enacted, social welfare isn't. 6) the are that rich people favor are, not surprisingly, very different from the areas the poor favor Overall, given the low number of policy initiatives where people actually disagree, and the large number of covariates to take into account (e.g. all the different policy ares) one cannot draw any firm conclusions from this dataset One interesting ntoe though – it may be the case that policy initiatives favored by the rich are more likely to be included in the dataset in the first place – i.e. rich people have an easier time getting things they are interested in on the agenda. This would also be a mechanism not captured at all by the logistic regression of the paper Like 15. bernie1815 says: HH: I hope others take advantage of the data set and explore it as you are doing. Can you single out some specific policies so that I can concretely understand where the “rich” and “average” differ significantly in their desired outcome? Like 16. Terry Bouricius says: Bernie and HH, Gilens explains his analysis approach more fully in his 2005(or so) paper when he developed the dataset initially.The new material in the current paper is his coding of preferences of various interest groups to add them to the analysis. As I understand it his analysis of the top 10% vs. middle 50% of income earners shows that middle income people generally only get what they want when the richest 10% want that too (which is generally the case). But the key of his analysis is that on those issues where the rich and middle diverge in terms of preferences, the preferences of the middle have virtually no independent influence on the outcome of government action. This does not mean the rich always get their way, but they have a statistically significant independent influence, which poor and middle do not. Like 17. HH says: Terry: My main argument is that Gilens analysis is essentially pointless. When you look into the number of policy initiatives where the rich and the average voter actually disagree (for me that is one group supports with > 55%, the other group supports with 55%), both oppose (i.e. < 45% approval) or at least one of the groups is indifferent (between 45% and 55% approval). Drawing any conclusions from these few cases has no meaning. However, more importantly – analyzing this statistically as if all policy proposals have equal importance is ridiculous. A single large policy initiative (e.g. Obama's healthcare reform) can easily have a larger impact than even hundreds of small policy changes. Like 18. keithsutherland says: HH: 3) The policy initiatives that rich people favor (and average voters oppose) are only marginally more likely to be enacted than all initiatives on average. If that is the case then it’s even more surprising that Gilens (the principal author) draws such strong conclusions from his dataset. Page has indicated in an interview that his own involvement was limited to the opening theoretical section (outlining the four hypotheses). Like 19. HH says: @keithsutherland: Important to note from the Gilens paper: The R-squared is only about 8% – meaning that in his analysis, his predictors explain almost none of the variation in the dataset. Drawing any conclusion based on such a weak model is just not good science. Like 20. HH says: sorry for my garbled comments – this commenting seems to be faulty – it is often not printing parts of my sentences. Like 21. André Sauzeau says: Keith Sutherland comments on the health care problem in the USA, as considered in Clinton times, are very interesting. The ordinary citizen is aware of the problem because it concerns anybody in his personal life, and life of parents, children, or friends. The big majority of ordinary citizens think the current system is bad (and, as far as I know, it seems really bad, given the costs and results, and in such a developed country). But the “public deliberation system” is unable to propose a health care policy able to resist the opposition of the minority who is satisfied by the current health care system. Therefore we must consider how the political system is able to elaborate, propose and organize rational deliberation about policy. Let’s suppose a citizen jury studying the problem during enough time before deciding, and with good deliberation rules. Maybe Keith Sutherland is convinced that such a sovereign body would be unable to make a choice, because of contradiction between practical necessities and “deep-seated ideological aversions”, or because of social and racial antagonistic feelings. But is Keith Sutherland sure of that? Maybe a mini-populus would take a reasonable choice. This choice could be accepted by a public opinion which was unable to elaborate it – not because of the stupidity of ordinary citizens, but because “opinion” has intrinsically low rational efficiency. Dêmokratia is not conformity to the public opinion, but to the will of the dêmos, elaborated through rational deliberation. It gives at least a chance of rational decision. Like 22. keithsutherland says: Andre, you raise two distinct issues: 1. US healthcare policy. I don’t claim any expert knowledge in this area, but from the literature that I’m aware of the problem is not the opposition of a (small rich/powerful) minority. Many/most (?) people who have health insurance are, more or less, satisfied with the status quo and are opposed to the effects that Obamacare would have on their own personal policies. Whilst surveys indicate a desire for universal cover, the problem arises when it affects voters’ existing cover. The minority is those who have no cover. I’m not disputing the power of the industry lobby, merely clarifying the electoral arithmetic. 2. Whilst a deliberative minipublic might make a “reasonable” choice (as might a meeting of Platonic guardians) it needs to have binding authority. If the authorisation criterion is democratic then it would need to demonstrate consistency/repeatability, rationality will not suffice. I don’t see how this is compatible with the warm and fuzzy deliberation proposed by Tom/Ahmed. Like 23. bernie1815 says: Andre: Two quick points. First, the vast majority of the US population was covered by the existing insurance market. Flaws in the system including non-portability, local as opposed to national insurance markets, and litigation induced costs all could have been addressed with incremental changes rather than the wholesale restructuring effort. Second, it is true that interest groups were opposed to changes in the status quo. But “interest groups” were also in favor of and were pushing for changes. Since we are a representative democracy, what exactly is the difference between our current process and a “citizen jury”. I am new to this site, so apologies if this is old ground. Like 24. keithsutherland says: Bernie, >It is true that interest groups were opposed to changes in the status quo. But “interest groups” were also in favor of and were pushing for changes. Exactly. When it comes to their own direct interests, people are not so gullible that they would vote for a system that only represented the interests of the private healthcare industry. Clintoncare was, in the end, killed off by arguments that represented the exact median of public opinion (torn between the desire for universal coverage and the American suspicion of “big government”). >Since we are a representative democracy, what exactly is the difference between our current process and a “citizen jury”. This blog is devoted to the development of large (c.300) deliberative minipublics, whereby legislators are selected randomly as a statistical sample of the whole population. Differences of opinion exist as to what form the deliberation should take — I argue for an analogue of the Athenian legislative jury, in which the citizen jury determines the outcome of a debate between (elected) advocates and experts, whereas most other regular posters on this site argue for a richer Habermasian-style deliberation. It strikes me that this would be in breach of the statistical mandate, that only applies at the aggregate (voting) level, thereby ruling out speech acts by randomly-selected members. I would be very interested in the views of HH and yourself as to what sort of activities a group with a purely statistically-representative mandate (nobody chose them) could undertake in a democracy. Like 25. bernie1815 says: Keith: I see the value of careful consideration of the ways in which collective decisions are made but little value in building systems. Personally, I see many of the issues as being very old and our current parliamentary processes, though flawed in theory and more frequently in practice, are a synthesis of efforts to resolve many of the same limitations that these new discursive solutions and structures are supposedly meant to address. I am somewhat averse to such new systems because much of the weakness of the current process are rooted in the flaws and foibles of individuals that we have elected or appointed to act for us in such deliberative bodies. Alternate systems are simply replacing one set of flawed decision makers with another plus the costs of rediscovering long known problems. My personal inclination is to try to improve the processes incrementally by developing ways to identify “bad actors” and bad information. And to apply stringent reality tests to such theoretical constructs. Nancy Pelosi’s infamous statement about passing health care reform legislation “so that you can find out what is in it, away from the fog of controversy” epitomizes the issues that afflict any deliberative process that stumbles into really big issues. Any proposed modification in our current deliberative process should have to show how it would have handled the issue of health care reform more effectively. Otherwise I see it as largely unproductive. Like 26. André Sauzeau says: Answers to Keith and Bernie *** I have no direct and serious knowledge of the health care problem in the USA. I have only the ideas that its results are not very good for its cost, and that some cases are badly treated –maybe exaggerated ideas? *** I was interested by the political problem appearing from the data from Keith Sutherland: that “a Gallup Poll conducted in early August 1991 indicated that 91 percent of the public felt there was a ‘crisis in healthcare’ (Gallup, 1991, p. 4) and a large majority (75% of adults polled) wanted the government to provide healthcare (Times, 1992)”, and finally no alternative policy was elaborated. It seems somewhat of a paradox in a so-called democracy, and interesting for anybody concerned by the basic “political problem”. *** Bernie says that “the vast majority of the US population was covered by the existing insurance market”. This is not contrary with the idea that a majority is not satisfied by the system, as the information from Keith seemed to demonstrate. First, the cost may be found excessive, or some risks badly covered. Second, some moral interests can be taken into account, and not only material personal interests. I am retired, and I am not satisfied by the high level of joblessness in France, even if I am not materially concerned. I am concerned because of some members of my family, or because generally of my countrymen – I hope it is not to be considered putting me into the lunatic fringe. *** Bernie says that “interest groups were also in favor of and were pushing for changes”. Sure – in a polyarchy you have no debate at all if at least a strong lobby is not pushing for change. The problem is that the choice proposed to the “public opinion” is reduced to a predeterminated set. A mini-populus could choose a policy which is not pushed by a strong lobby, but is deemed reasonable by a majority of ordinary citizens. Like 27. bernie1815 says: Andre: Thank you for replying. I am retired also, which gives me time to frequent these types of discussions. As to dissatisfactions with the previous US system, there were many. However, the proposed reform does not address these dissatisfactions. Health insurance is expensive, but then so is health care. Your notion of a mini-populous may or may not come up with better options. I do not see any a priori reason to assume that it would be an improvement over either the status quo or the current reform. I am also not sure as to the nature and source of your pre-determined set. As to what might be reasonable to a significant majority of ordinary citizens, there certainly were simple changes that would likely dramatically reduce costs but which would be fiercely opposed by powerful interest groups, most notably personal injury lawyers and small insurance companies. I do not think the issue is “moral interests” but simply a utilitarian interest – but the distinction may be mere semantics. A more inefficient system is simply a bad choice. A system that gives greater and unnecessary control to a bureaucracy without clear off-setting benefits is also a bad choice. The assumption that the collective can make better decisions than an individual about individual matters seems to me to be fundamentally anti-democratic. Like 28. keithsutherland says: Bernie, As someone of a similar age and conservative inclination to yourself, I’m also predisposed to incremental fixes — “all life is problem solving” as Karl Popper put it. Unfortunately our democratic arrangements appear to be beyond repair and I think this is a structural problem, rather than just a case of flawed individuals. In the UK parliamentary system there is no longer anything that would pass for serious deliberation, other than in the House of Lords. Backbench members spend most of their time in their offices dealing with constituents’ problems (that are more than often the province of local government) until the division bell is sounded, at which point they simply troupe through the division lobbies as directed by the whips. I don’t even know the name of my own constituency MP. Political leaders compete for office using the skillsets normally reserved for thespians, and spend more time rubbishing their opponents than developing serious policies. But, unlike many other commentators on this blog, I am not proposing sweeping this all away, merely requiring the victors in a general election to present their policies for scrutiny by a minipublic as part of the legislative process. This proposal, while modelling itself of 4th century Athenian practice, takes heart from the recent literature on the wisdom of crowds, which claims that a representative jury of ordinary folk are just as capable of judging the merits of legislative proposals as experts or partisans. It is also using the word “deliberation” in the Latinate sense of “weighing” arguments (and has nothing in common with Habermasian deliberative democracy) so fits well with Rousseau’s model of silent deliberation within. It also presupposes a strict separation of powers, with the executive organised along entirely different lines. Like 29. This reminds me of an article called “Most Published Research Findings are False” http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182327/ Like 30. keithsutherland says: Ha! The problem with the current paper is not that the findings are false, but that the results are portrayed in a manner that is open to partisan manipulation. The outpouring of identical “told you so’s” on Huffington and Washington Post shows how people will eagerly embrace any science that endorses their own prejudices. It will be interesting to see how the authors’ peers treat the results and whether they mirror the scepticism of Bernie and HH. One of the problems in this field is that political theorists still seem to be in awe of Mosca, Michels and Mills (even though their books are rather old now!) and political scientists are desperate to reclaim their domain from their economist colleagues. In my recent PhD upgrade I presented the draft of the chapter that dealt with these issues and was advised to either deal with the political theory literature OR the political science literature as it is entirely contradictory, but this didn’t strike me as a good way to arrive at the truth of the matter. Like 31. bernie1815 says: Keith: Having been subjected to a jury decision in a civil suit I am not so enamored with the wisdom of crowds. Fortunately the judge essentially nullified the jury verdict. The preconditions for the effective execution of such a process are fraught. For example, in my trivial but hopefully apposite case, those with the technical background need to understand the case were excluded from the jury. Like 32. keithsutherland says: I lost an expensive case in the Court of Appeal when the judge ruled against my company on a pure legal technicality (whether a document had been served “correctly”); I doubt whether a jury of my peers would have decided in the same way. Like 33. bernie1815 says: Keith: The problem you describe, I believe, afflicts many branches of the social sciences. Old and new media journalists are equally guilty of not looking at and thinking critically about reports and articles if they remotely align with their preconceptions. Given the general world view of academia, they are not helped by those who put out press releases for Universities. One of my touchstone books is the late great Aaron Wildavsky’s “But Is It True?” – it should be required reading for all who attempt to communicate scientific findings to the general public. Like 34. bernie1815 says: Keith: It looks like our life experiences have encapsulated the dilemmas, paradoxes, ambiguities and challenges we face when trying to establish a “system” to make decisions. Woe to him who thinks he has the true answer. Like 35. Yoram Gat says: > Old and new media journalists are equally guilty of not looking at and thinking critically about reports and articles if they remotely align with their preconceptions. Given the general world view of academia, they are not helped by those who put out press releases for Universities. This is rather comical. Your own (Sutherland, HH and bernie1815) superficial critique of the paper is quite obviously nothing but a reflection of your own prejudices against the findings. (And of course an exhibition of your ignorance of the methodology involved.) If you can refer me to a similar critique that any of you has written of an empirical paper whose findings are to your liking, I’d be rather surprised. Like 36. HH says: @keithsutherland Hmmm, it will certainly be interesting to see how their peers treat it. Of course, that also depends on wether their peers know more statistics than Gilens and Page … Like 37. bernie1815 says: Yoram: So which of my original points do you have trouble with? “First, little of the variance in policy outcomes is explained by the proposed variables. Secondly, how does one fit the platforms of the two parties into this analysis. Thirdly, a large number of high profile interest groups are not listed in Appendix A. Fourth, as you suggest, there is something odd about using a multivariate technique when the variables are so complex and multifaceted. I also look forward to seeing what the more statistically savvy reviewers have to say about the study. Finally, it strikes me that the bottom-line of the study could also be written as “Groups with strong interests in particular policies and pieces of legislation are led by members of the elite and tend to influence the final outcome of legislation more than groups and people who are less interested.” Now that is not very surprising or interesting at all.” Are you saying that you have no questions with the methodology and the way the conclusions have been stated in the mainstream media? Like 38. HH says: @Yoram Gat It is true that I did not go into much depth regarding my critique about the paper. In my defense a) I do have a job and can’t spend days writing critiques of articles, which would likely never get published (my critiques I mean) b) I am a statistician that was educated at a U.S. elite university with a PhD. So I do know a little something on how to make (and report on) statistical models – and from that perspective Gilens and Page did a lousy job. c) It is true – there are few reports that get headlines that I do like when I look into their methodology. But I don’t think that has to be with me being overcritical – merely that there are *very* few surprising results to be found these days – so that most papers that do claim to have such results do have flaws of one sort or the other d) I did actually download and look at the underlying data. As I said in my earlier posts – the main problems are that there are very few policy initiatives were people actually disagree, that the models don’t explain much about the policy outcomes (which Gilens and Page actually admit in their article!; see low R-squared) and therefore the conclusions just are too far reaching given the sparse data. e) Please don’t tell me that I don’t know the methodology if you don’t even know my educational background based on a few paragraphs of posts I wrote – that is rather rude Like 39. keithsutherland says: Yoram: >Your own (Sutherland, HH and bernie1815) superficial critique of the paper is quite obviously nothing but a reflection of your own prejudices against the findings. (And of course an exhibition of your ignorance of the methodology involved.) Whilst I will gladly hold my own hands up when it comes to ignorance of statistical methodology, that is not true of my co-respondents. Also, as you are also the moderator of this blog, it would behove you to be less rude to newcomers who don’t share your political perspective, and to offer a substantive defence of Gilens’s methodology. Like 40. Yoram Gat says: Bernie, Actually none of your comments above makes much sense AFAICS, but just to pick an easy example, what does “there is something odd about using a multivariate technique when the variables are so complex and multifaceted.” even mean? Like 41. Yoram Gat says: HH, > I am a statistician that was educated at a U.S. elite university with a PhD. So I do know a little something on how to make (and report on) statistical models – and from that perspective Gilens and Page did a lousy job. I must say that at this point I am curious to as to which elite university this was. I’d also be interested in a link to your academic work/thesis and to those reports that write. I am sorry to be rude, but your comments have been technically incorrect and (as the comment above still is) rather self-important. About the technical issues. You wrote above: > Important to note from the Gilens paper: The R-squared is only about 8% – meaning that in his analysis, his predictors explain almost none of the variation in the dataset. Drawing any conclusion based on such a weak model is just not good science. Since you have are a highly trained statistician, let me pose a hypothetical example to you: Say that there are elections, in which the vote is split 50/50 between two parties. A researcher takes a sample of N people, half of them women and half of them men. If turns out that the vote split among women is 60/40 and among men is 40/60. Would that be an interesting, valid scientific finding? What would the r^2 of the gender-based voting model be? Like 42. Yoram Gat says: I also note that none of you responded to my query about similar critiques you have written of papers with findings whose political implications are supportive of the existing political system. It seems that your own biases are showing. (This reminds me of the critiques of the Lancet/Johns Hopkins studies of post-invasion mortality in Iraq. When the findings are unpleasant, we suddenly become such critical thinkers.) Like 43. HH says: @Yoram: First – about your last comment: No – have not written similar critiques on polisci papers. Matter of fact – I am a statistician, not a political scientist. That is also why I never criticized Gilens for their polisci conclusions (which I happen to believe). My criticism is that their analysis of their data does not support their conclusion (at least not under their current analysis). Now to your little example. I assume that we know the individual votes of the subjects then, building the model vote ~ gender has an R- squared in your case about 0.04. (depending on the definition of R-squared you use, of which there are many for logisitic regression). For simplicity I used a linear regression, which in this case is close enough. Do I find the gender split interesting – yes! Would I draw conclusions from this data as to whether this gender split even really exists? Probably not. Just to keep playing with your example. Assume furthermore that 60% of males work in a technical job but only 40% of the women. Furthermore assume that every person in the technical field voted no, all people not in a technical field voted yes. Granted – a very extreme example, but I use it to illustrate my point. Would you still think the gender split is interesting? The model that is only based on gender explains little of the underlying variation in your case – just the 4%. This leaves the possibility for other predictions that actually do explain the voting behaviour much better (see job type above). That is why in general one should be wary of models that explain little about the variation in the data, especially if one wants to draw far-reaching conclusions. To get back to the Gilens paper – Have you downloaded the dataset Gilens has linked on his website? I think that was the basis for the article – the number of observations does not match exactly, but the data contained looks reasonably close to what he describes. One other variable in the data is political area the policy initiative was in. That one is actually very interesting – as different political areas seem to have highly different likelihoods of having proposals implemented. Just 2 examples: “campaign finance” based proposals were implemented 12.5% of the time, “foreign policy” proposals 58% of the time. A very cursory analysis seems to suggest that the policy area has actually higher predictive power as to whether the proposal gets implemented than all of Gilens variables combined. Another note – the dataset contains much more data – based on gender, age, geographic location, faith etc. Those could explain some of the variability in the data. Did Gilens use them in the model? From the paper – I think not. It is also not stated why not, which is odd when he had all of them directly at his disposal. In general, if one wants to draw far reaching conclusions – should not all available predictors be used? So getting back to the large different likelhoods of policy proposals being implemented – what if (and I haven’t checked if that is actually true) campaign finance is an area that the rich oppose and the average voter favors and foreign policy is an area where the rich usually favor changes that the poor oppose. In such a circumstance – it would look like the rich more often get their way. But is that now due to the rich having more influence – or them simply disagreeing with the average voter in a political area where change relatively easy? Just another note: As usual in all empirical analyses, one has to be vary not to imply causality when there is correlation. What I mean is – maybe Gilens is right, and policy proposals that the rich favor have a higher likelhood of being implemented *because* the rich favor it. Another possibility however would be that people with a higher income understand the political system better and are better educated and simply don’t have a tendency to agree with positions that they know will have no chance of being implemented in the current circumstances. A further note on the “multivariate model”. Its weakness regarding the modeling of complex data is certainly the inherent assumption of linearity of the predictors (and here specifically of the additivity). Furthermore, by not including the political area the proposal was made in – he does not allow for the possibility that the influcence of the wealthy actually depends on the political area of the initiative (see campaign finance vs. foreign policy above). It is quite possible that they have influence on some but not on others … but from his article it looks like Gilens did not even investigate it? Or did he do so in one of his other articles? I am not familiar with all his work. Anyway – I am happy to discuss statistical implications more, Yoram. But I would appreciate a less aggressive tone. And just to re-iterate. I do not disagree with Gilens opinion of the influence of the rich. I merely think his statistical analysis leaves much to be desired (and you are right – the word “lousy” was a bit strong) Like 44. keithsutherland says: Yoram, >I also note that none of you responded to my query about similar critiques you have written of papers with findings whose political implications are supportive of the existing political system. It seems that your own biases are showing. Yes, that’s a fair point. But the rhetorical dialectic doesn’t work that way — people critique things they don’t agree with and then the onus is on the parties who agree with the original findings to respond. Your strategy would appear to be to ignore commentaries that you disagree with, as you have refused to respond to any of the critiques of the Gilens paper (apart from ad hominem attacks on the commentators and one request for clarification), even though you were the one who made the original post. Like 45. bernie1815 says: Yoram: Like HH, I do not have a problem with Gilens’ conclusions, given his model. I find it totally unsurprising that a business or cultural elite would appear to drive the legislative process and that groups with coherent and cohesive interests would appear to have a greater impact on legislation than categories of individuals with more diffuse interests. As my earlier comments pointed out and as HH also indicated the model strikes me as incomplete. I do not much appreciate your snarky tone either. Like 46. Yoram Gat says: HH, > A very cursory analysis seems to suggest that the policy area has actually higher predictive power as to whether the proposal gets implemented than all of Gilens variables combined. If this is the case then this is a valid critique. And in that case, including the policy area in the model, as you suggest, would be interesting and the right thing to do and does raise the question of why the authors didn’t do this. No point in speculating, however, since the authors have provided their data. (That said, again, your argument as I quoted it above is simply wrong. The value of R^2 is, by itself, not an indicator of the quality of the work.) > As usual in all empirical analyses, one has to be vary not to imply causality when there is correlation. This is either a truism or a falsehood, depending on what you mean. Analysis of observational data by itself does not imply causation. However, as is routinely done (and justly so), observational findings are combined with a causal model to infer causation. If we stick to the letter of your rule, causation can never be learned in any real world setting. The language Gilens and Page use is no different from that of others in their field or in other empirical fields. If, as you do now, you have a compelling competing model to offer, that is one thing. But simply saying “correlation does not imply causation” is not useful. The same is true for the issue of linearity – or more generally, any modelling assumptions. There will always be modelling assumptions made. Again, if you have specific critique of certain assumptions that’s fine, but simply saying “they have made various assumptions that may not be valid” is not useful. > But I would appreciate a less aggressive tone. I’d be happy to a have a substantive discussion and it seems we are moving in this direction. My tone, I feel, was a natural reaction to the facile and presumptuous tone in much of this thread – starting, naturally, from Sutherland’s post. Like 47. Yoram Gat says: bernie1815, > I do not have a problem with Gilens’ conclusions, given his model. That is, I do have a problem is the conclusions of the paper, but I am willing to re-interpret them in a way that I like. > the model strikes me as incomplete. Oh, it does, does it? It is this kind of facile and presumptuous comments that make the snarky tone appropriate. Like 48. keithsutherland says: >[to HH] the facile and presumptuous tone in much of this thread – starting, naturally, from Sutherland’s post. >[to bernie] It is this kind of facile and presumptuous comments that make the snarky tone appropriate. Yoram, you really do need a lesson in internet etiquette, otherwise you will ensure that this blog remains the province of the few of us prepared to put up with your habitual rudeness. Your snarky remarks used to be addressed exclusively to me, but now appear to extend to anyone who doesn’t share your political perspective. An apology would be in order and, while you’re at it, an explanation as to what it was about my post that was “facile and presumptuous”. >However, as is routinely done (and justly so), observational findings are combined with a causal model to infer causation. Page’s four hypotheses were all correlatory, no mechanisms were suggested. Given that the immediate currency of democratic politics is votes, not dollars, the authors fail to hypothesise a mechanism whereby policies preferred by the vast majority of voters were rejected in favour of policies preferred by a tiny minority. In my “facile and presumptuous” commentary I suggested three possible mechanisms, but all the authors provide is evidence of a correlation. Like 49. bernie1815 says: Keith: I have little patience with attitudes like Yoram’s. He must be a real fun guy to have a drink with. I enjoy tough discussions but plain rudeness guarantees that the discussion will be fruitless. My thanks to you and HH for an interesting discussion. George Orwell had it right: Beware the demagogues, regardless of their self-proclaimed benevolence. If you are still interested in feedback on your chapter let me know. Like 50. keithsutherland says: Working on the chapter right now; sorry if Yoram has driven you from this blog. I’ll be in touch in due course. Like 51. HH says: @Yoram: I think we have just very fundamentally different opinions as to what the job of an author and what the job of a reviewer/critique of an article is. Anyway – I thank bernie, keith and everyone else except Yoram for a fruitful discussion. Also, before coming to this blog, I did not know about Kleroterians and while I would not call myself a convert yet – I do find the concept intriguing :-) Like 52. keithsutherland says: HH Many thanks for your timely and valuable contributions, and apologies for the hostile reception. It looks as if the Kleroterians have reverted to type — circling the wagons and talking amongst themselves. This is standard policy for left-wing cabals intent on building the New Jerusalem — a self-selected vanguard seeking to lead the oppressed masses to rise up against their rich ‘n powerful masters. My interest in sortition is more to get the trains running on time, but this puts me in a minority of one and leads to regular beatings from the commissars of the People’s Republic of Aleatoria. I can understand if you don’t share my taste for martyrdom. Like 53. Yoram Gat says: > I think we have just very fundamentally different opinions as to what the job of an author and what the job of a reviewer/critique of an article is. I disagree. This is not just a matter of “opinion”. Selectively setting the bar for plausibility at either impossibly high levels or at ridiculously low levels is a standard maneuver of established power for dismissing dissent. It is not critiquing but simply propagandizing. Again, I think that in your last comment you did move away from simple dogmatic rejection toward a considered argument, and I would be interested to see the analysis that your claim was based on. On the other hand, Bernie and, of course, Sutherland have stayed within the confines of pure dogma. Like 54. André Sauzeau says: *** Keith Sutherland says: “I’m predisposed to incremental fixes — “all life is problem solving” as Karl Popper put it. Unfortunately our democratic arrangements appear to be beyond repair”. *** I agree but I would add that some important progress could not be got by incremental fixes. Karl Popper used the metric system – not the result of an incremental progress. And he used the “Arabic” numerals, the positional numerical system, which was not the previous Roman system bettered by incremental fixes. Sometimes we must jump. Clearly, it is better to think before jumping. Like 55. keithsutherland says: Yoram, It’s strange how one man’s reasoned argument is another man’s dogma. The long-running dispute between Yoram and myself must surely demonstrate that the Habermasian discursive ideal is nothing more than a pipe dream. Our original exchanges were patient and courteous but we have ended up just trading insults. The irony is that we are in complete agreement on 90% of sortition-related issues — we both agree that the prime political potential of sortition is descriptive representation (we differ here from Dowlen and Stone) and that politics is essentially an agonistic/majoritarian rather than an epistemic/consensual procedure (we differ here from Teleb, Landemore, Burnheim and Bouricius). I would go so far as to say that the only thing that separates us is our view as to the level of analysis at which representativity applies — to my mind it is clearly a statistical aggregate property, whereas Yoram appears to believe that sortition would privilege average individuals (the 99%) at the expense of the elite. This has important entailments for what a sortition-based assembly can or cannot do without damaging its original mandate. To my mind this indicates the need for a mixed constitution, rather than pure sortition; I would like to think that I’ve defended my position by careful, reasoned argument (it is the subject of my ongoing PhD thesis), but Yoram has ruled that this is nothing more than “electoralist dogma”. I’m glad that he is not my external examiner, otherwise I might as well give up and go fishing! It’s nothing less than tragic that this small disagreement has led to such a lot of animosity and that this has been transferred to new arrivals on this blog — particularly when they are so well qualified to help us in our understanding of these matters. Andre, I completely agree, that’s why I’m in favour of selecting legislative juries by lot, rather than election. I also agree we should “look before we leap”. Like 56. André Sauzeau says: *** Bernie finds “totally unsurprising that a business or cultural elite would appear to drive the legislative process and that groups with coherent and cohesive interests would appear to have a greater impact on legislation than categories of individuals with more diffuse interests” and he says “not very surprising or interesting at all”. Right. But, if we must acknowledge that in probably all the political systems there will be a difference in influence, we must see that the contemporary polyarchic system is exceptional by combining a high level of influence inequality and the myth of democracy. And we must consider that “democracy-through-sortition” would reduce the discrepancy between political reality and political myth, which would be a good thing by itself. *** Bernie says: “Your notion of a mini-populus may or may not come up with better options. I do not see any a priori reason to assume that it would be an improvement over either the status quo or the current reform.” The a priori reason is the following: “generaly , if not always, a decision taken by a body which is independent and has time for information and deliberation is more enlightened and related to the (material and moral) interests of its ordinary members than the decisions following other processes”. *** Bernie says: “As to what might be reasonable to a significant majority of ordinary citizens, there certainly were simple changes that would likely dramatically reduce costs but which would be fiercely opposed by powerful interest groups, most notably personal injury lawyers and small insurance companies”. In “democracy-through-sortition”, the citizen juries would be more able to resist “powerful interest groups” than congressmen who need to be reelected, therefore who need money for their campaigns, who are afraid of vicious media attacks by lobbies, and of its effects on an uninformed “public opinion”. *** Bernie says: “I am not sure as to the nature and source of your pre-determined set.” The better option can be one of the options which are pushed by a strong lobby, “the pre-determined set”. But a citizen jury, a mini-populus, could consider other options Like 57. André Sauzeau says: Keith Sutherland says “politics is essentially an agonistic/majoritarian rather than an epistemic/consensual procedure”. I object to this opposition as a democrat. The dêmos will always include groups with strong specific moral or material interests, and therefore intense feelings about some subjects; the democratic model implies that about most issues there are a majority of “ordinary citizens” able to hear the conflicting “orators”, to deliberate and to decide. For instance about health system I think that in France at least a majority of citizens can be considered as “ordinary citizens”, without strong specific interests. And even with a “basic” difference as between men and women, I think we can consider a majority of “ordinary citizens”, who are not in a mind of “war between the sexes” and would not behave in a “automatic” way about a “gendered issue”. Keith jokes “the long-running dispute between Yoram and myself must surely demonstrates that the Habermasian discursive ideal is nothing more than a pipe dream.” If the blog has the objective to get a consensus between Keith and Yoram, maybe the prospect is poor. But we must consider that the objective is to enlighten the community of the participants, including me for instance who does not feel as a warrior in one side or the other. Like 58. keithsutherland says: Andre That’s a fair point, and one that Terry has made before. It matters little whether partisan advocates like Yoram and myself agree (although it’s a serious problem when we just dismiss each others’ views as “dogma”), so long as our exchanges are judged by a representative sample of ordinary citizens who have no axe to grind. But this is a long way removed from the Habermasian search for consensus by either appeal to a transcendental standard of rationality or the warm and fluffy approaches favoured by Tom and Ahmed. It also indicates the need to ensure parity of partisan advocates — if it were left to purely endogenous factors then a randomly-selected forum might find itself with either a Yoram or a Keith hectoring them and might well end up supporting radically different policies. This is why democratic norms require a) exogenous, balanced advocacy and b) separation of the advocacy and judgment functions, even if this is at the cost of the epistemic deliberative ideal (and the privileging of knowledge elites). PS I think the long-running dispute between Yoram and myself, while possessing comedic aspects, is tragic if it means that well-informed people like Bernie and HH are driven from the blog. Personally I couldn’t care less what Yoram thinks of me, but others might have thinner skins or greater regard for deliberative decorum. Internecine warfare in a nest of ferrets is not very attractive to newcomers. Like 59. Yoram Gat says: > The irony is that we are in complete agreement on 90% of sortition-related issues Aside from the fact that this is not true, there would be nothing ironic about that if it were true. Our agreement or disagreement on sortition-related issues has nothing to do with our strained relationship. The real irony is that on this matter as on most other matters you have no clue what’s going on because you live in your own world where you don’t bother to pay attention to what other people say and the facts do not intrude. Adding to that your self-importance, obnoxiousness and bigotry, it is a wonder anybody is willing to converse with you. Like 60. keithsutherland says: >Adding to that your self-importance, obnoxiousness and bigotry, it is a wonder anybody is willing to converse with you. Gulp, I’m happy to leave it to a jury of my peers to judge the merit’s of Yoram’s latest piece of vitriol (and his rejection of my my sincere attempt at bridge-building). It would be good to hear from some of those who have hitherto taken a less vocal part — perhaps this is a case for Solon’s law — whereby those who sit on their hands during a stasis are stripped of citizenship. When good men do nothing . . . Like 61. HH says: @keith I really do not know what Yoram’s problem is. I had never seen this blog until 3 days ago, and so far have experienced everyone on here (except Yoram) as well-reasoned, sincere people. It is ironic I think that pretty much everyone agrees with Gilens opinion, but still gets accused of being a bigot and self-important etc. just for saying that Gilens model may not be all that great. Anyway, I may see and read a couple of the other pieces of the blog to get to know a little more about “sorting” @Yoram Regarding my my expectations towards the level of statistical analysis. Essentially, Gilens dataset would be great for a final report/longer homework question for an applied stats class (beginning PhD or advanced Master’s students). Having been a TA for such classes, my expectations towards Gilens article would have been to get something as good or better I would have expected to get from an average Stats Student. Nothing more, nothing less. I do not expect every researcher to have this level of knowledge – but in that case I am sure Gilen and Page could find a Stats PhD student happy to do the analysis part for them in exchange for authorship credit …. Like 62. André Sauzeau says: *** Coming back to the subject of this post, measuring the influence of the opinions of the citizens of different social classes is interesting, but we must consider that a big part of elite and lobby influence could be in the framing of opinion itself. And that dêmokratia is not the rule of opinion, anyway, as I insisted. But, if we speak of differential influence of opinions, an historical event is worth remembering. *** The Treaty establishing a Constitution for Europe (TCE), commonly referred to as the European Constitution, was rejected by the French people in the May 2005 referendum; with a not very wide margin, 55%, but a significant one, especially because all the main elites were for the YES and the media coverage clearly biased for the YES. What occurred? The substance of the rejected treaty was incorporated in a new treaty voted by the French Parliament, avoiding any referendum. Let’s speak of the consent of the people as basis of our political system ! But the more interesting fact is that the European governing circles, facing what was a deep distrust in the way Europe was ruled, did not display any hint of panic, and did not think about any change in their ways, even about feigning some change. Basically, they did not mind. Strange, because France is a big State in the heart of Europe, the no-consent of the French people would seem very upsetting for them. They did not mind, because they saw that the main elites of France were for YES (the money elite, the cultural elite, the political elite …) and therefore the NO of more ordinary citizens (up to 80 % among workers!) was not a problem. *** I suggest that the influence of ordinary citizens may be somewhat significant when there is a fight between different elites, for instance the money elite and the cultural elite (I am sorry if I simplify). The ordinary citizens can be used in these fights by good “framing”. When the main elites agree partially, the influence of the ordinary citizens decreases. If the main elites agree, the real influence of ordinary citizens becomes minimal, getting at most cosmetic changes in policies. Like 63. keithsutherland says: Andre, >The Treaty establishing a Constitution for Europe (TCE), commonly referred to as the European Constitution . . . An interesting point. The UK elites are, in general, also in favour of greater European integration, but UKIP (the one major party opposed to EU membership) is predicted to win the May 22 election for the EU parliament. The Conservative Party is committed to an in/out referendum during the next parliament, even though David Cameron and most of the party big-wigs are committed to EU membership. UKIP are focusing on the control of immigration (a policy very close to that of Marine Le Pen) and this is forcing all the major parties to move sharply to the right on this issue (even though EU membership means there is little they can do about it). Perhaps there’s something unusual about the UK parliamentary system that makes for a more populist approach — the general view amongst political scientists being that the three largest parties are in all out competition to win the backing of the median voter. I’m not denying that elites have considerable influence, but in the UK public policy is largely driven by voters, not elites (and the powerful elites are more cultural than economic), so Gilens and Page’s paper has little relevance. Like 64. HH says: Maybe I can add a little to that from the German perspective. Here, the big current policy initiative is a pension reform that promises substantial additional payments for mothers and certain blue collar workers. From a EU perspective, this is an unusual proposal as this time as especially Germany has been telling all other countries to save money by making cuts – also for pensions. The median voter is in favor of this proposal – the elites are dead set against it. And yet, despite all this the government is going ahead with it. Of course this is a single policy initiative, but it is a really big one. Also here, powerful elites do not always win – even if it is about a lot of money and they are trying quite hard. Like 65. Yoram Gat says: > Having been a TA for such classes, my expectations […] I see you have retreated again to the area of bravado and fluff. That is unfortunate. > what Yoram’s problem is I have many problems, I don’t doubt. That said, it seems that one of your problems is that despite being very liberal with your criticism of Gilens and Page, and despite setting the bar for the “job of a reviewer/critique” at a very low level, you apparently have difficulty when your own comments are met with disapproval. Like 66. HH says: Hi, I find the idea of political decisionmaking using randomness intriguing. But to be honest, being on this blog isn’t really fun given how one is treated by some people. It is sad to say, but if the Kleroterians have friends like Yoram Gat, I think you don’t need enemies anymore. Goodbye! Like 67. André Sauzeau says: *** Keith Sutherland tells us that “the UK public policy is largely driven by voters, not elites”, and the one instance he gives is about foreign immigration … and he adds that “EU membership means there is little they can do about it”. How powerful the voters! Funny! If European policy rules, is it driven by European voters? *** Keith Sutherland says “I’m not denying that elites have considerable influence, (and the powerful elites are more cultural than economic), so Gilens and Page’s paper has little relevance.” I think that on this point we can get to a consensus. *** I would like to know if a consensus is possible likewise on the following idea “modern dêmokratia does not mean that the opinion of the lower class voters has equal influence as the opinion of high class voters, because dêmokratia is not the rule of opinion, but the rule of a will enlightened by deliberation and expressed through a citizen jury, a mini-populus”. Like 68. keithsutherland says: Andre >Funny! If European policy rules, is it driven by European voters? Delegated governments always have executive prerogative rights, and these will always be abused and will corrupt the legislative function. This is particularly problematic in fused parliamentary regimes (an allotted legislature would simply remove ministers who overstep the mark by a censure motion). UK electors agreed to join the Common Market but this has been extended into a vision of ever-increasing European integration by the political class. But they can’t get away with it in the long run — hence the current malaise over immigration. >I would like to know if a consensus is possible likewise on the following idea . . . Yes on the second part: dêmokratia is not the rule of opinion, but the rule of a will enlightened by deliberation and expressed through a citizen jury, a mini-populus”. but the first part is anachronistic. It’s true that speech acts in a demokratia will belong to the better educated, but in modern social democracies there is no necessary connection between education and socio-economic class. In Harrington’s Commonwealth of Oceana, the Senate (an aristocracy of merit) has speech rights, but voting rights are restricted to the Prerogative Tribe (selected by sortition). This creates a self-regulating feedback loop, as Senators (Advocates in my scheme) would be obliged to defend the policies that meet with popular approval. As well as matching Harrington’s 17th century proposal, this also fits well with 5th century Athenian practice. Like 69. Terry Bouricius says: While I agree the paper doesn’t offer much evidence of any specific causality in terms of influence over government policy, or proving the existence of a wealth-based oligarchy, it DOES provide evidence of another important ABSENCE of causality. When the influence of the preferences of the middle 50% income decile is isolated from the influence of the top 10% and of special interests, they show pretty convincingly that there is essentially no independent influence by the middle. Since this is the fundamental claim of the high school civics view of electoral democracy (what the authors call “Majoritarian Democracy”), this is a very important point. While this comes as no surprise to anyone who has looked at how the system works, to my knowledge, this is the first quantitative study (rather than theoretical argument) to prove the falsity of this myth of our democracy. Like 70. keithsutherland says: Terry, The evidence certainly does indicate that, in the minority of cases where preferences differ, those of the top 10% predominate. But, in the absence of a hypothesised mechanism, why is this a “very important point”? As HH suggests, we need to know the salience of the issues involved. If voters consider these to be important issues the politicians supporting the preferences of the 10% would be committing electoral suicide. Given that the primary purpose of campaign contributions is to enable the recipient to buy votes, why risk alienating the votes of 2/3 of voters in order to receive funds from (a few of) the 1/3 in order to help compensate for the loss? The rational approach would simply be to pander to the preferences of the median voter all of the time. This would suggest that other causal mechanisms are involved, and we are none the wiser until we know what sort of policy areas lead to the divergence and how important these are to the median voter. It’s perfectly possible, for example, that politicians view the preferences of the 10% to be epistemically correct, and to consider this a sufficient reason for running the risk of electoral suicide. (This is just a hypothesis, we need a more fine-grained analysis of the dataset to put it to the test.) Like 71. Terry Bouricius says: Keith, The reason I say this is important is because it doesn’t matter WHY government action is not responsive to the preferences of the median respondent (median income, which this study shows is closely correlated with the median voter in terms of preferences as well)… The fundamental argument for electoral democracy is that politicians have to be responsive to median voters…but in fact the study shows they aren’t. This is revealed through looking at policies where there is a divergence of opinion between median and high income voters, but the statistical analysis indicates the apparent responsiveness of politicians to the public is merely an illusion, following responsiveness to high income voters on those policies where high and median income voters AGREE. We don’t need an hypothesis as to WHY government is non-responsive to median voters to understand that the basis of electoral democracy is a false mythology. Developing and testing a hypothesis is an obvious next challenge, but that does not make THIS study trivial. Now as to WHY politicians can ignore median voter preferences …There are a vast number of possible reasons … I would argue that in America (where the study was done) the policy process is so convoluted and biased towards the status quo that hardly anybody can say WHO is to blame for disapproved policy outcomes. VERY few voters can say how their Representative voted on any particular bill. And thus it is impossible to use elections as an accountability tool (even if the voter was offered a competing candidate with a different policy agenda, which often isn’t the case). Like 72. keithsutherland says: Terry, >The fundamental argument for electoral democracy is that politicians have to be responsive to median voters. There are two possible senses of the phrase “have to”: 1) The rational choice sense, whereby if politicians ignore the preferences of the median voter then they will be out of a job. This study shows that politicians follow these preferences 2/3 of the time and would suggest that on the minority of occasions that they don’t this is because the median voter doesn’t view the issues as salient (if they did then the non-pandering politician would be out of a job). If it is the case that voters cannot tell how their representative voted then accountability is in fact a myth, but in the UK this isn’t the case as representatives vote along party lines. I’m not an expert on US politics, but my understanding was that similar processes applied — Republicans have a majority in the House, so voters can make a reasonable guess as to the voting record of their own rep on most issues. 2) the normative sense “should”. I don’t think “have to” is the argument behind most normative theories of democracy, which require that politicians govern in the interests of the median voter / common good. It may well be that in some, or even many, cases well-informed politicians might believe that they know better than the median voter what is in her interests and this might explain the deviation in the 1/3 of cases. Without an analysis of the 1/3 cases and competing hypotheses of the mechanisms involved this study is of little value. >[politicians are responsive] to high income voters on those policies where high and median income voters AGREE. Not so. Given that the immediate currency of electoral democracy is votes rather than cash (votes are not on sale at Amazon, Ebay or Walmart), Occam’s Razor would suggest that 2/3 of the time politicians are responsive to the preferences of median voters. That is the received wisdom in political science and this study does nothing to contravene it (although you wouldn’t think that from the media firestorm). >I would argue that in America (where the study was done) the policy process is so convoluted and biased towards the status quo That is uncontroversially true, as it was the stated intentions of the majority (Federalist) faction in the ratification debate. Some might well argue that this built-in conservatism is the reason that the US political system is comparatively stable, compared to (say) France which is currently on its fifth republic (and counting). Like 73. Terry Bouricius says: Keith wrote: “Occam’s Razor would suggest that 2/3 of the time politicians are responsive to the preferences of median voters. That is the received wisdom in political science and this study does nothing to contravene it (although you wouldn’t think that from the media firestorm).” It depends on what you mean by “responsive.” I was using the word as having a causality rather than a mere correlation. The study showed that when high income and median income voters disagree, the median income voters virtually never get their way. The authors say that this indicates the median voters are having no independent effect on policy, but merely APPEAR to have an effect (when they happen to agree with high income voters.) It would seem unlikely that by chance every issue that the polling firms asked about, where the median income and the high income disagreed, happened to be low salience. That seems like a stretch. It is far more probable that the median income voters indeed have no INDEPENDENT influence on any issue regardless of salience. Like 74. keithsutherland says: Terry, What is your hypothesised causal mechanism, seeing that you can’t buy votes at Walmart, all you can do is use the money on advertising and persuasion in order to offset the electoral damage that results from ignoring the median elector. Are you suggesting that elected representatives are literally being bribed — that the money is enriching them personally rather than just their campaigns? That’s a serious charge, and I’m not sure if there is the evidence to back that up if what you are suggesting is true, ie 100% of the cases (remember that median voters have, according to Gilens, Page and Bouricius, no influence on policy outcomes.) I think the police might have found some evidence for this level of corrupt practice. Why is the salience hypothesis such a stretch? This is obvious corollary of the hypothesis that legislators will pander to the median voter whenever it is to their electoral advantage — this would predict that it will happen for all high-salience issues and not happen when legislators anticipate that such pandering will result in medium- or long-term damage (to the state and/or their own chances of reelection). Bear in mind that Galens and Page are seeking to overturn a huge volume of research findings, one of which found the enormous (.91) correlation between opinion and policy “remarkable” and “awesome” (Erikson, Wright an McIver, 1993, p. 80). Or “there exists about a one-to-one translation of preferences into policy” (Erikson et al, 2002, p.316) “Politicians are keen to pick up the faintest signals in their political environment. Like antelopes in an open field, they cock their ears and focus their full attention on the slightest sign of danger” (ibid. pp. 319-320). The view that politicians are chasing votes, not campaign contributions, is the prevailing view in political science (as opposed to political theory and communications studies), and the Galens/Page paper does not overturn it, notwithstanding all the hype resulting from misleadingly captioned figures and over-stated conclusions. References Erikson, R. S., Mackuen, M. B., & Stimson, J. A. (2002). The Macro Polity. Cambridge: Cambridge University Press. Erikson, R. S., Wright, G. C., & McIver, J. P. (1993). Statehouse Democracy: Public opinion and policy in the American states. New York: Cambridge University Press. Like 75. keithsutherland says: Terry, One simple question: If it is the case that elected officials ignore median public opinion then why do they devote so much time and money to conducting their own straw polls? If they are only interested in the views of the rich and powerful, all they need to do is go to their dining club and ask their friends. Even the presidents who denied that they followed public opinion employed their own pollsters. Truman (and to a lesser extent Eisenhower) was the only president from the polling era who genuinely took no interest in public opinion (this was consistent with his preference for one-armed economists). Like 76. HH says: This discussion around the Gilens paper reminds me a lot about the situation in a different setting – the Gender Pay Gap (GPG). In case of the GPG, it is also the case that if one does the straightforward analysis, one finds a rather big pay difference between men and women of 25%. In the Gilens case as well as GPG, this simple analysis certainly contains a substantial truth and completely hits the “Zeitgeist” of what people expect anyway. Both situations also have in common that there are no accepted models that straightforwardly would explain this. For Gilens, Keith has given all the arguments as to why a causal relationship is hard to infer. For GPG, the pay difference is simply too big to be explained by outright discrimination of women. However, if one takes the time to delve deeper into the data – for the GPG one finds out that the situation is much more complicated with a certain part of outright discrimination mixed with personal choices as well as real and/or perceived public pressures in certain life situations leading to the observed result of women being paid less. What I take from this is – the situation that Gilens describes is likely to have a kernel of truths – but as usual the real situation tends to be a lot messier than one might hope. One thing about Gilens that also surprised me: Despite trying to explain political decisions, Gilens apparently doesn’t need to include the majority parties in congress, the senate or the party membership of the president for his analysis. His model also seems to suggest that he does not think at all that there may be interactions between ruling party and how much they listen to the rich. That certainly goes against public opinon – as usually one party is perceived to be much more closely aligned with the rich than the other. Like 77. Terry Bouricius says: Keith asked: “If it is the case that elected officials ignore median public opinion then why do they devote so much time and money to conducting their own straw polls?” There are three common types of polls paid for by politicians: 1. Horse race candidate standing polls (“Am I likely to win? Is my campaign strategy working?”). 2. Pus Polls (intended to sway voters rather than elicit their opinion. “Would you be less likely to vote for candidate X if you knew he favored murdering babies?” 3. Framing Questions…testing whether calling them “inheritance taxes,” “estate taxes,” or “death taxes,” was most effective at gaining popular acceptance of a pre-selected policy position. It is extremely rare that any American politician asks genuine questions about policy preferences of voters in any poll they pay for. Like 78. keithsutherland says: >It is extremely rare that any American politician asks genuine questions about policy preferences of voters in any poll they pay for. That’s certainly the prejudice of communications theorists working in the tradition of Harold Lasswell. Not so with mainstream political scientists, who are more interested in evidence than psychoanalytic theory. For a refutation of Lasswell’s top-down argument I recommend John Geer’s From Tea Leaves to Opinion Polls (Columbia University Press, 1996). Philip Gould’s Unfinished Revolution is a good exploration of how New Labour used techniques derived from Clinton’s people on how to ascertain the preferences of the median voter and then to formulate policies around them. I’ve spent the last 3 or 4 months immersed in the empirical political science polling literature and could give you a long reading list if you want to evaluate the evidence for yourself. Like 79. bernie1815 says: HH: Did you see this, http://www.nytimes.com/2014/04/24/upshot/the-pay-gap-is-because-of-gender-not-jobs.html?_r=0 ? The money quote for me is ““The gender gap in pay would be considerably reduced and might vanish altogether if firms did not have an incentive to disproportionately reward individuals who labored long hours and worked particular hours,” she wrote in a paper published this month in The American Economic Review.” The mind boggles. Under specified models are the Achilles heel of much social science. Like 80. HH says: @bernie, no I didn’t. Thanks for the link. The results certainly make sense – as a lot of studies point to part-time work and taking time off are important factors as well. So I think that fits relatively well that people not willing/able to work long hours in certain jobs would be paid less – and also of course due to domestic roles women disproportionately being part of that role. Of course, in the end it all depends on whether business are right to act this way – reward extremely long hours disproportionately. And also in this case – a difference that first looks liek being due to gender – isn’t due to gender at all. All men not willing to work such long hours get punished as well. Regarding under-specification, she may have left out a couple of variables in her analysis, but it doesn’t look terribly under-specified to me, be honest. Which variables are you missing (time taken off for child-rearing comes to mind …)? Like 81. bernie1815 says: HH: I was thinking far more operationally in terms of the nature of the employers’ businesses and the reasons for long hours and off-hours. Moreover, I do not see it as “punishment”. Nor do I necessarily believe that an actual disproportionality exists – since you would have to measure actual work outputs as well as inputs to make that determination and some of these would be difficult to measure (e.g., a 24 hour pharmacy is not likely to be busy at 3 AM, but the value to the franchise of being open when needed may be more valuable than the revenue actually produced). In addition, off-hour premiums are largely a market mechanism and capture real labor supply factors. Bottom-line though and despite the head-line, surely it is harder to make a gender argument given her findings? Like 82. HH says: I agree. Today anything that is correlated to gender is framed as a gender issue – no matter the causality. It just makes for better headlines. Like 83. keithsutherland says: >Today anything that is correlated to gender is framed as a gender issue – no matter the causality. It just makes for better headlines. Yes, and the parallel you draw with the media reception of the Gilens paper is a good one. Without some hypothesis on the causal connection, correlations can be entirely misleading. I’m with Mark Twain/Disraeli on that. Like 84. keithsutherland says: Terry, Would appreciate an answer to my earlier question: “Are you suggesting that elected representatives are literally being bribed — that the money is enriching them personally rather than just their campaigns?” Like 85. keithsutherland says: Just been reading a very fair-minded review of the literature on the responsiveness of political elites to public opinion surveys (Manza and Lomax Cook, 2002). The authors argue the case for a via media, or a “contingency” approach between the true believers and the deniers: “Where measured public opinion expresses a coherent mood or view on a particular policy question (or bundle of policy or political questions) in a way that is recognizable by political elites, it is more likely than not that the movement of policy will tend to be in the direction of public opinion” (p.657) But this “prima facia” support for the “strong influence” thesis is constrained by a number of powerful factors (listed on pp. 652-653): 1. Policy making occurs through a variety of mechanisms, some of which (adminsitrative action, judicial decisions etc) are not directly subject to electoral accountability. 2. Some policy areas are viewed by the public as more salient than others, for example “lower salience of foreign policy issues often give politicians more room for maneuver than domestic policy domains”. Where attitudes are bimodal (e.g. abortion, gun control), “responsive outcomes are much more problematic than on issues where public opinion is unimodal” (p.653). 3. Fiscal constraints: “issues where the responsiveness requires significant budgetary expenditures may reduce the impact of public opinion. Budgetary politics can lock in spending commitments that cannot easily change even if public opinion shifts.” (ibid.) 4. Some policy domains attract more powerful interest groups than others, which are subsequently more open to policy innovation. (ibid.) 5. “Path-dependent policy-making processes restrict policy responsiveness in that policy outcomes at one point in time constrain what is possible — or even likely to be considered — at later points of time” (ibid.) In sum, in hyper-responsive democracies like the US, elected politicians will be motivated to track public opinion, but responsiveness is constrained by the above contingencies. The authors include an interesting reductio of the argument of the top-down thesis, especially its “critical theory” variant (surveys are part of the process whereby elites manufacture public opinion). The theory acknowledges “the extensive efforts undertaken by elites to shape public opinion” (p.648), but then raises the following rhetorical question: “Why bother trying to influence public opinion if it has little impact on policy outcomes? [as Galens and Page conclude]. Evidence that elites on both sides of an issues such as social security will often invoke particular readings of public opinion to support desired policy positions, or that business interests will seek to shape or influence public opinion in addition to direct lobbying activities provides, ironically, evidence of the extent to which political and economic elites take seriously the political impact of public opinion” (ibid.) Reference Manza, Jeff and Lomax Cook, Fay, A democratic polity? Three views of policy responsiveness to public opinion in the United States, American Politics Research, 30: 630-667. Like 86. Terry Bouricius says: Keith asked: “Are you suggesting that elected representatives are literally being bribed — that the money is enriching them personally rather than just their campaigns?” Quid pro quo bribery as cash in paper bags is probably rare, but through campaign contributions is fairly common though hard to prove. It doesn’t require that an elected official ALSO divert the campaign contribution to illegal personal use, for it to be a bribe. More often than quid pro quo bribery is the purchase of access and general ingratiation (which the U.S. Supreme Court has ruled to be protected by the Constitution.) Like 87. keithsutherland says: Thanks for the clarification Terry. But, granted that the bribe is primarily intended for political (rather than personal) use, it remains the case that the immediate currency of democratic politics is votes, so it makes no sense for a politician seeking (re)election to risk alienating voters by following the diverging policy preferences of those offering money designed for campaigns to attract the support of the self-same voters. It requires a more nuanced explanation (as Manza and Lomax Cook offer) in order to make sense of such behaviour. Like 88. Terry Bouricius says: Keith, It is exactly the claims of the median voter influence on policy discussed in the paper by Manza, Cooks and Fay, that Gilens put to a quantitative test. The Manza paper you discuss cites work by people like Shapiro and Erikson, (as does the Gilens paper) which find a strong correlation between public opinion polls and policy outcomes. The thing that Gilens does is to control for income, and finds that the correlation only persists for the high income decile. There is virtually zero correlation with the public opinion of middle-income voters when those opinions differ from the rich. Like 89. Terry Bouricius says: Keith wrote: “it makes no sense for a politician seeking (re)election to risk alienating voters by following the diverging policy preferences of those offering money designed for campaigns to attract the support of the self-same voters.” Think about it. If I am a politician who wants re-election, and I believe (for good reason) that voters are FAR more influenced by expensive campaign ads than by my voting record, I have a stronger incentive to do what I need to build a large campaign war chest, than to worry about opinions of ill-informed voters who don’t know how I vote any way. Like 90. keithsutherland says: Terry, What the Manza paper does offer is suggestions as to why median-voter and high-income preferences might, on occasion, differ and why legislators might choose to follow the latter. Gilens and Page offer no causal hypothesis at all, just innuendo (the votes of legislators are bought by the rich ‘n powerful). Both HH and Bernie have suggested that it might be more fruitful if Gilens’ dataset was interrogated from a different angle, examining the salience of different kinds of issues, and it is this sort of nuanced analysis that Manza and his colleagues offer. Recommended reading. Like 91. Terry Bouricius says: Keith, Yes it would be NICE to know WHY median income voters virtually never get their way when they disagree on policy with high-income people…but the point I am trying to make is that even if we have no idea WHY, we still now know that median income voters (who track closely with median ideology voters) do not independently influence public policy. Whether politicians are instead responding to the wealthy or Tarot Card readers would be nice to know… but the point is that we now have powerful evidence that the electoral theory (politicians being responsive to median voters) is a false myth. Like 92. keithsutherland says: >we now have powerful evidence that the electoral theory (politicians being responsive to median voters) is a false myth. Not so. In the 2/3 of cases when preferences agree policy alignment is to the median voter. Gilens’s implication (politicians are in fact responding in these cases to the preferences of rich voters, the preferences of median voters being epiphenomenal) cannot be established without a causal hypothesis (see my Janet and John argument, above). Sorry to keep restating this but the default causal hypothesis is that votes are the immediate currency of electoral politics so politicians will chase votes, unless there are pressing reasons not to (Manza et al suggest 5 possible reasons). This being the case you (and/or Gilens) need to establish a plausible competing hypothesis to substantiate your claim. Correlation is not enough. Frankly I’m surprised that this paper made it past the peer review stage as the methodology is so obviously flawed. Like 93. HH says: @Terry Please go to Gilens website and download the dataset before judging what it actually means. Here some data: – For 50% of cases, the favorability rating for the median and 90th percentile voter are within 5 percentage points of each other – For 80% of cases, the favorability rating for the median and 90th percentile voter are within 10 percentage points of each other – For 90% of cases, the favorability rating for the median and 90th percentile voter are within 12 percentage points of each other Meaning – for the vast majority of cases, the median and 90th percenile voter have essentially teh same opinion (or both are ambivalent). Like 94. Terry Bouricius says: HH, Yes, exactly…thus it is that unique set of cases which pollsters decided to survey about where the preferences of high and median income respondents DIVERGE that we have to look at if we want to know if one subset or another has INDEPENDENT influence on government action. If there are a million possible policy questions the fact that all voters agree on 900,000 doesn’t matter at all IN TERMS OF INFLUENCE, if in all 100,000 in which they disagree, only one subset gets its way. The question isn’t whether rich and poor agree or disagree on policy generally (they may all favor government efficiency, prosperity, and good health for all). the question is who is calling the shots when they disagree. Like 95. HH says: I could not agree more. The problem with his dataset is – there are few questions in it were subjects meaningfully disagree at all (i.e. > 10 or 15%), and even fewer where they actually end up on opposite sides. We would need a dataset like that with big disagreements between voters. The problem is – Gilens dataset is not it! Like 96. bernie1815 says: HH: I am having trouble figuring what the “median” or 90% voters actually are in this analysis. These statistical aggregates seem to have the same anthropomorphic characteristics as individual actors. Is the median voter a Democrat (or a Republican)? A taxpayer? How about the 90% voter? Like 97. HH says: @bernie If you read Gilens paper – he briefly refers to how he got the data: The survey data he has for each question is broken down to some extent by income brackets (but not the same across the surveys). So, using the results from the different income brachets, he fit a quadratic regression I think and then infers the favorability ratio at the 90th percentile income bracket. Same for the 50th percentile. In my earlier post – when I wrote 90th percentile voter – I was implying the favorability rating for the issue at that income percentile. And yes – all good questions. Party affiliation is not taken into account. In fact, governing party is not taken into account. Other variables correlated with income (e.g. age usually) – not taken into account. Education? Nope! Like 98. HH says: one more brief thought. Even if one believes the results of the statistical analysis of Gilens, there is a much simpler explanation than assuming politicians listen only to the rich. Given that most politicians are actually among the top 10% of earners or better, maybe they are not listening to anyone and simply follow their own conscience? In that case, favorability of the top 10% of earners would be a red hering with no causality, simply being highly correlated with personal opinions of congressmen and senators. Like 99. bernie1815 says: HH: I have read the paper a couple of times. One thought struck as a result of different viewpoints on this analysis: Keith talks about the mechanism, I talk about the unit of analysis and you raised the issue of the missing demographic variables – we may be all talking about the same underlying type of issue. Your final notion that it is the actual decision makers/politicians that may explain the general pattern of results seems to cover these concerns. Like 100. HH says: @bernie: I merely raised that as an additional explanation (don’t really believe it explains everrything, but may have an influnce). But I think we can all agree on one thing: The paper doesn’t proof these far-reaching clonclusions it draws: – The dat itself is limited and has issues – The unit of analysis is unclear – The statistical analysis leaves much to be desired – The is no good model of causality that would explain the conclusions etc. Like 101. keithsutherland says: Terry, >the question is who is calling the shots when they disagree. Sorry to be a bore, but I’ll keep asking the question until you answer it — how can you know whether it is Janet or John who are “calling the shots” (a rather vague mechanism from the perspective of political science) unless you have a testable causal hypothesis? (correlation tells you zilch). Manza et al. suggest five mechanisms for the (occasional) disconnect between median-voter preferences and policy outcomes, Bernie suggests another, but Gilens and yourself have no explicit hypothesis, just “money calls the shots”. I’ve pointed out repeatedly that votes, not money, are the immediate currency of electoral politics, so you need a detailed causal model as to how money can come to trump votes, but we are yet to be enlightened as to what it is. If you look to your history books you’ll find that the introduction of the secret ballot was in order to prevent the (literal) purchase of votes, so we need to know exactly how this corruption comes about. Critical theorists would put it down to the hegemony of rich media moguls who induce false consciousness in the masses through propaganda, but Galens’ dataset refutes this hypothesis (Rupert Murdoch is wasting his money as the masses stubbornly refuse to be corrupted by the rich ‘n powerful). You’ve denied that the top 10% (a very large number of citizens, most of who don’t make political donations) are bribing politicians, so what does that leave? P.S. I’m extremely relieved that HH and Bernie have courageously resisted the earlier attempts to bully them off the rhetorical life-raft so that I am no longer a solitary voice crying alone in the wilderness (I really must stop mixing my metaphors). Keep up the good work guys! Like 102. André Sauzeau says: *** Keith Sutherland says “Some might well argue that this built-in conservatism is the reason that the US political system is comparatively stable, compared to (say) France which is currently on its fifth republic (and counting).” *** Well, if I remember, the American Republic did not change number but solved (badly) some problems by one of the bloodiest civil wars in modern Western history. *** Speaking of “contemporary times” – last quarter of century – I think there is no so much real differences between the Western polyarchies. There is some kind of “convergence” between the systems. Like 103. André Sauzeau says: Keith Sutherland and Terry Bouricius seem to agree that there is a “buit-in conservatism” in US polity. It could lead to think about status-quo in the society. But, as far as I know, there were strong changes in the society in the last quarter of century, in my opinion some good ones and some bad ones – as a very strong surge in socio-economic inequality (which some critics describe as a peaceful counter-revolution by the moneyed classes). It is important to remember that few effective actions by elected powers do not mean status-quo, because it leaves open field to unelected social powers, and because any dynamic modern society will undergo changes – only, in a dêmokratia the dêmos would have at least some power to control them (I don’t think any political power can magically control everything, I believe only there is some possibility of more effective rational control than in our polyarchic systems). Like 104. André Sauzeau says: *** Keith Sutherland quotes Manza and al, 2002: “Why bother trying to influence public opinion if it has little impact on policy outcomes? [as Galens and Page conclude]. Evidence that elites on both sides of an issue (…) will often invoke particular readings of public opinion (…), or that business interests will seek to shape or influence public opinion in addition to direct lobbying activities provides, ironically, evidence of the extent to which political and economic elites take seriously the political impact of public opinion”. *** Very smart question, indeed. Here some possible answers. * Even if the specific impact opinion of the median voter is small, it is better not to overlook it. * If a big part of public opinion is strongly against you, it is always a potential source of practical difficulties. For instance top decisions could be badly enforced by administrative networks, juries or even some judicial lower courts. * If a big part of public opinion is strongly against you, it will create a bad temptation among the political class: some politicians, especially some alienated from their fellows for some reason, could be tempted to take the issue. * In an elite, the policy choices will not be always spontaneously homogeneous. Minorities can disagree, from sub-groups identities, from generational interests and sensitivities, from personal reasons. Strong propaganda is useful not only to convince the median citizen, but to unify ideologically the elite; without ideological unity, the strength of an elite is much lesser in a polyarchy. * A wide difference between the rulers and the common citizens is always dangerous, ultimately for the political system itself. And furthermore it creates a legitimacy problem given the “democratic myth”. *** The “rhetorical question” of Manza and Lomax Cook actually enunciates a real-life paradox. An example can be seen in the famous case of the “European Constitution”. As I said, the clear NO of the French citizens did not have any effect. The French and European elites did not take it into account, without problem. At the same time the propaganda against “the enemies of Europe unity” is always strong and strident. Is it an irrational waste of time and energy? I am not sure. The above mentioned reasons can “solve” the paradox. Especially in this case some fractions of the elites appeared to have voted NO, as the mass of workers, much more than in the previous Maastricht referendum. The political homogeneity of the elites on the subject was at stake Like 105. keithsutherland says: >The “rhetorical question” of Manza and Lomax Cook actually enunciates a real-life paradox. An example can be seen in the famous case of the “European Constitution”. As I said, the clear NO of the French citizens did not have any effect. Perhaps the UK is just an outlier and US and France are typical (in which case it’s more down to contingent factors than Yoram’s Iron Law of Electoralism). Let’s wait and see the results from the 22 May election, which the UK Independence Party is predicted to win. It’s hard to imagine that such a win will just be ignored by the three main parties. If they do, then this will provide a major fillip for UKIP’s performance in the 2015 general election. In the final analysis the demos has kratos, as confirmed by Manza and Lomax Cook. Like 106. André Sauzeau says: If the opinion of the median citizen in opposition to all the elites is able to win, it will be an exceptional case in polyarchy. Wait and see. But, I must repeat, the public opinion is not the same as the will of the dêmos through deep deliberation of citizen juries. Like 107. Yoram Gat says: > Why bother trying to influence public opinion if it has little impact on policy outcomes? That’s actually a good question. But before trying to answer it, the premise needs examining. What is the evidence that elite groups spend a lot of resources to try to sway public opinion regarding policy decisions? How much money is spent on influencing the public compared to lobbying decision makers directly? Like 108. Terry Bouricius says: Keith wrote: ” how can you know whether it is Janet or John who are “calling the shots” (a rather vague mechanism from the perspective of political science) unless you have a testable causal hypothesis?” It would be nice to know whether Jane or John was calling the shots (or Jacob, or Jill)…but it is still very useful to simply know that one of them is NOT calling the shots. The HYPOTHESIS that Gilens tested was the belief that median voters guide government policy (median income being a close proxy according to polling data) . In other words, the theory being tested is that “Janet” is calling the shots. Their study showed that Janet has no independent influence on government policy. They did not prove that John (the wealthy) are calling the shots, (though much of the press coverage suggests they did)…but they tested the civics class hypothesis and found that it was false. Like 109. HH says: @Terry Sorry, but they found no such thing. First, you cannot prove by testing that a coefficient in a regression is =0. You can only show by testing that it is not = 0. And the fact that it is *not* significantly different from 0 does not imply that it is actually = 0. In general, a problem with linear models always arises when 2 predictors are highly correlated (as is the case here). Estimates in these cases tend to be unstable. General advice is to use uncorrelated predictors where possible. This even would have been possible in this case. Instead of includeing the favorability of the 50th and 90th income percentile, they could have used – favorability of the 50th percentile – a predictor that measures the “slope” between the favorabilities at the 50th and 90th income percentile Advantage of this approach: These 2 predictors are likely to be uncorrelated (unless richer voters tend to dislike what median voters prefer and vice versa, which would be unusual and very interesting in itself). The simple outcome of such an analysis would be: a) the 50th income percentile is highly predictive of the overall outcome b) proposals are more likely to pass if they have a positive “slope”, i.e. richer voters like them A lot less inflammatory model that explains everything just as well and would have this rather simple causal explanation: – First, politicans listen to the median voter, because they are important for getting votes – Second, of course as they have more exposure to richer voters (more access) as well are richer themselves with similar background to richer voters, they can be more easily shifted in those direction as well. Like 110. keithsutherland says: Andre, Yoram, Terry, >But, I must repeat, the public opinion is not the same as the will of the dêmos through deep deliberation of citizen juries. I think we all agree on that as a normative goal, the only dispute is an empirical one (who currently rules?) >What is the evidence that elite groups spend a lot of resources to try to sway public opinion regarding policy decisions? How much money is spent on influencing the public compared to lobbying decision makers directly? Good question (I don’t know the answer) but the issue in hand is political donations (campaign contributions to enable politicians to ignore public opinion and still get reelected). I suppose in the US this mostly goes on TV advertising campaigns and I would imagine the sums are substantial as the cost of re-election has shot up over the last decade or so. Expenditure on lobbying firms would probably be a lot less. >it is still very useful to simply know that one of them is NOT calling the shots. That is only true for the minority (1/3) of cases where preferences diverge. Notwithstanding the media hype the study does not demonstrate whether elites or median voters are determining policy in the majority of cases. >but they tested the civics class hypothesis and found that it was false. Not so for the majority of cases (as demonstrated above), and the data is pretty weak for the minority of cases, as HH has indicated. Like 111. HH says: Hi, I thought it would be useful to visualize the data. I used the dataset ds1 from http://www.russellsage.org/research/data/economic-inequality-and-political-representation In the graph, I plotted the 50th income percentile favorabilities against those of the 90th income percentile. Each subgraph is for a specific policy area (so also giving an impression as to how the number of policy change proposals is distributed by policy area) with proposals that have resulted in policy changes in blue. Hope that clarifies things a bit. The notion of keith that people disagree in 1/3 of cases is a misunderstanding of the pearson correlation coefficient – it is actually a lot less than that were people disagree. http://postimg.org/image/e6ptk7j3n/ Like 112. HH says: Here another version where one can see the diagonal better (sorry for the slightly weird commercials on that site – i just looked for a free image host) http://postimg.org/image/obnyqpzlr/ Like 113. Yoram Gat says: HH, About everything in your answer to Terry is inaccurate or misleading. > you cannot prove by testing that a coefficient in a regression is =0. You can only show by testing that it is not = 0. And the fact that it is *not* significantly different from 0 does not imply that it is actually = 0. Yes, there is no way to show equality to a single value (either 0 or any other value), but the estimate does indicate the that value of the best fit must be close to zero. A small enough coefficient is the same as 0 for all practical purposes. (By the way, I actually get statistically significant negative values for the 50th percentile coefficient.) > In general, a problem with linear models always arises when 2 predictors are highly correlated (as is the case here). Estimates in these cases tend to be unstable. The instability is expressed in the statistical model via the estimate variance, and therefore through z-scores and p-values. The p-values of the estimates of the Gilens-Page model are highly significant, meaning that colinearity is not a problem in their model. > A lot less inflammatory model that explains everything just as well and would have this rather simple causal explanation: No – your model implies that for a fixed 90th percentile opinion, the median opinion would still have noticeable impact on the chance that a piece of proposed policy gets implemented. The data in fact indicates that this is not the case. The bottom line is that according to the evidence, as analyzed by Gillens and Page (and you have yet to present a different analysis that produces different results), Terry has things exactly right and Keith, Bernie and yourself are tying yourselves in knots trying to deny those findings without actually providing any valid argument. Like 114. HH says: @Yoram A picture says more than a 1000 words. If the effect is so clear and meaningful as you seem convinced that it is – you should be able to look at the plots I posted and point to where in those plots you think you detect that the 50th percentile opinion doesn’t matter. Like 115. Yoram Gat says: > the issue in hand is political donations No – those donations are supposed to help people get elected, not to change opinion on particular policy. The issue is expenditure aimed to change the median opinion on particular pieces of policy. >> it is still very useful to simply know that one of them is NOT calling the shots. > That is only true for the minority (1/3) of cases where preferences diverge. To state the obvious – no. According to your faulty logic, if I read someone’s blog (but never comment on the blog or interact with the blogger in any way) and I agree with them 90% of the time and disagree 10% of the time, then 90% of the time I “call the shots” about what’s written in the blog. Like 116. keithsutherland says: HH >The notion of keith that people disagree in 1/3 of cases is a misunderstanding of the pearson correlation coefficient – it is actually a lot less than that were people disagree. I’m happy to be corrected on this, although Yoram disagrees (and I’m not remotely qualified to adjudicate). I was merely reporting Galens’ conclusions. If your analysis of the data in the 1/3 of cases is correct, then it’s doubly puzzling that the article was accepted for publication, with or without the hyperbolic claims and misleading figures. I don’t know anything about Perspectives on Politics — is it likely that the journal would have accepted the article without peer review by a competent statistician? Or is it the journal’s policy to publish “perspectives” (opinion, ideology etc) as opposed to peer-reviewed political science papers? (needless to say the media can’t be expected to understand the difference). It all sounds a little too postmodern for my taste. Like 117. Yoram Gat says: HH, I actually think it is quite obvious from the pictures that the blue dots tend to be concentrated above the 45 degree line in almost all the plots. Looking at those pictures, I am guessing that adding the policy area to the model (as you suggest) would only increase the magnitude of the difference between the influence of 90th percentile opinion and the median opinion. Like 118. HH says: Hi, here another plot, with the data in a different format x-axis: Favorability in the entire population (across all income groups) y-axis: Favorability differential of 90th minus 50th percentile (could also use 90th vs 10th, but for consistency use 50th) http://postimg.org/image/s25f8f9g3/ Also, not how uncorrelated those predictors are … Like 119. Yoram Gat says: As for whether, given the 90th percentile opinion, there is any remaining influence to the median opinion – that is difficult to tell from those plots one way or another. Like 120. HH says: You are right as to insofar it is *one* way to interpret the data. Another way, just as logical, is – Overall favorability across the whole population has a very large impact – module that – the rich retain a minor influence the whole point I ever wanted to make is that both interpretations are valid – and therefore Gilens did not prove their conclusion. Like 121. keithsutherland says: Yoram, No – those donations are supposed to help people get elected, not to change opinion on particular policy. The reason that they need help to get elected is because (according to Gilens) they are failing to pander to the median voter in a minority of cases. Occams’ razor would suggest the simplest, and cheapest, strategy would be to adopt the median voter strategy, so a strong explanation (not provided by Gilens) is required for politicians to deviate from rational behaviour. >According to your faulty logic, if I read someone’s blog (but never comment on the blog or interact with the blogger in any way) and I agree with them 90% of the time and disagree 10% of the time, then 90% of the time I “call the shots” about what’s written in the blog. The problem with this analogy is that, in the case of the blog, you did not vote in the editor. In the case of elected politicians you are very much interacting, as she is only there on account of your vote. This is the whole rationale behind the economic theory of democracy. Blog sites operate on a zero cost basis, so market theories do not apply. Like 122. HH says: @keith In the end, how many people “disagree” on an issue depends on what you actually mean by it. – If one group has an 80% faveorability, the other 60% do they disagree? – If one group has 60%, the other 49%? – 55% vs 45%? For me, real disagreement at best starts in teh last case – and the number of cases in that category (as you can see on the plots) is really small. And there are no strong disagreements at all (i.e. points in the upper left or lower right hand corner of the original set of plots). Like 123. HH says: @Yoram Also – again referring to the plots (look at the last one Overall vs. difference), there are some interesting patterns here – The rich did not get their way on terrorism, or soc. welfare (where a lot of policies they strongly favored nontheless got defeated – The rich very much got their way on foreign policy – but then again, the average american doesn’t care about that all that much I think …. – Surprising to me – the picture for taxation is quite ambialent Which brings me back to a point I made a lot earlier – acting as if each policy proposal should have the same weight makes no sense. Some areas are more important than others. In this dataset, foreign policy has just too high a weight compared to what people really see as important. And just look at the budget case – that one is funniest of all. If look at in terms of actual favorabilities – the politicians pretty much did the opposite of what the voters wanted :-) just a funny side note Like 124. HH says: @Yoram I kept playing around with the ds1 data. I regressed (as you can see blow) on policy area, 50th income percentile and 90th income percentile. I did not apply the logit transformation Gilens did, but used the straight favorabilities. Outcome: pred50_sw has a significantly negative coefficient pred90_sw has a signicantly positive coefficient So, if I argue along the lines as to what Gilens do, does this now “prove” that politicians not only do what the rich (90th percentile) do, but also *actively* do the opposite of what the 50th income percentile wants (as a causal model)? In other words – does this model prove that politicians actively *spite* the median voter? > model_more summary(model_more) Call: glm(formula = CHANGE ~ XL_AREA + pred50_sw + pred90_sw, data = ds) Deviance Residuals: Min 1Q Median 3Q Max -0.8342 -0.3285 -0.1535 0.4194 1.1195 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -0.05800 0.09686 -0.599 0.549410 XL_AREAcamp finance -0.23054 0.10572 -2.181 0.029335 * XL_AREAcivil rts 0.02361 0.11140 0.212 0.832179 XL_AREAdefense 0.33443 0.10606 3.153 0.001641 ** XL_AREAecon & labor 0.08131 0.09996 0.813 0.416099 XL_AREAeducation 0.01997 0.10607 0.188 0.850719 XL_AREAenvironment 0.03065 0.10657 0.288 0.773672 XL_AREAforeign pol 0.32768 0.09441 3.471 0.000531 *** XL_AREAgovt reform -0.03800 0.11432 -0.332 0.739626 XL_AREAguns -0.15294 0.10065 -1.519 0.128826 XL_AREAhealth -0.12724 0.09712 -1.310 0.190287 XL_AREAimmigration -0.16791 0.12843 -1.307 0.191223 XL_AREAmisc others 0.05462 0.09974 0.548 0.584016 XL_AREArace -0.18244 0.13311 -1.371 0.170665 XL_AREAreligious -0.05770 0.10125 -0.570 0.568879 XL_AREAsoc welfare -0.12106 0.09828 -1.232 0.218176 XL_AREAtaxation 0.12914 0.09439 1.368 0.171439 XL_AREAterrorism 0.07664 0.10320 0.743 0.457832 XL_AREAwelf reform 0.18123 0.10594 1.711 0.087317 . pred50_sw -0.67673 0.13793 -4.906 1.01e-06 *** pred90_sw 1.28933 0.14445 8.926 < 2e-16 *** Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 (Dispersion parameter for gaussian family taken to be 0.1890751) Null deviance: 418.41 on 1835 degrees of freedom Residual deviance: 343.17 on 1815 degrees of freedom (27 observations deleted due to missingness) AIC: 2175.2 Like 125. HH says: was a linear model – same result for logistic Call: glm(formula = CHANGE ~ XL_AREA + pred50_sw + pred90_sw, family = binomial(), data = ds) Deviance Residuals: Min 1Q Median 3Q Max -1.9525 -0.8449 -0.5479 0.9934 2.6312 Coefficients: Estimate Std. Error z value Pr(>|z|) (Intercept) -2.85709 0.49477 -5.775 7.71e-09 *** XL_AREAcamp finance -1.48299 0.59353 -2.499 0.012469 * XL_AREAcivil rts 0.08664 0.57491 0.151 0.880213 XL_AREAdefense 1.53750 0.52793 2.912 0.003587 ** XL_AREAecon & labor 0.41169 0.49903 0.825 0.409376 XL_AREAeducation 0.10352 0.53472 0.194 0.846489 XL_AREAenvironment 0.11015 0.54424 0.202 0.839613 XL_AREAforeign pol 1.55055 0.47093 3.293 0.000993 *** XL_AREAgovt reform -0.25612 0.58044 -0.441 0.659029 XL_AREAguns -0.93626 0.52207 -1.793 0.072918 . XL_AREAhealth -0.74020 0.49671 -1.490 0.136168 XL_AREAimmigration -1.27706 0.87195 -1.465 0.143029 XL_AREAmisc others 0.25543 0.49685 0.514 0.607179 XL_AREArace -1.29534 0.87991 -1.472 0.140987 XL_AREAreligious -0.32268 0.51661 -0.625 0.532222 XL_AREAsoc welfare -0.78522 0.51335 -1.530 0.126121 XL_AREAtaxation 0.61593 0.46902 1.313 0.189101 XL_AREAterrorism 0.30998 0.51530 0.602 0.547470 XL_AREAwelf reform 0.78929 0.52496 1.504 0.132704 pred50_sw -3.75088 0.76410 -4.909 9.16e-07 *** pred90_sw 7.05084 0.82508 8.546 < 2e-16 *** Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 (Dispersion parameter for binomial family taken to be 1) Null deviance: 2380.4 on 1835 degrees of freedom Residual deviance: 2019.5 on 1815 degrees of freedom (27 observations deleted due to missingness) AIC: 2061.5 Number of Fisher Scoring iterations: 4 Like 126. Yoram Gat says: HH, A possible interpretation of the negative coefficient for the median opinion is that it is a proxy for the opinion of, say, the 99th percentile. The direction of change of opinion from the 50th to the 90th percentile persist as one moves from the 90th percentile to the 99th percentile. Schematically: opinion99 = opinion90 + \alpha * (opinion90 – opinion50). This is what happens when regressing pred90_sw on pred50_sw and pred70_sw: Call: lm(formula = apred90_sw ~ a$pred70_sw + a$pred50_sw)

Residuals:
Min 1Q Median 3Q Max
-0.08877 -0.01406 0.00052 0.01429 0.09039

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.02438 0.00152 16.04 <2e-16 ***
a$pred70_sw 2.25629 0.01742 129.54 <2e-16 *** a$pred50_sw -1.30566 0.01717 -76.03 <2e-16 ***

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.02242 on 1833 degrees of freedom
(27 observations deleted due to missingness)
Multiple R-squared: 0.9876, Adjusted R-squared: 0.9876
F-statistic: 7.319e+04 on 2 and 1833 DF, p-value: < 2.2e-16

How does this sound?

Like

127. HH says:

@Yoram,
Two main points
1) I completeley agree with you. In fact, If you use your equation (make it a little more general, you get something like:

opinion(y) = opinion(50) + alpha * (y-50) * (opinion90-opinion50)

Here – of course – opinion90 – opinion50 is just any estimates as to how much people of different income groups differ in their opinion. So, lets call alpha * (opinion90 – opinion50) “income slope” and write

opinion(y) = opinion(50) + (y-50) incomeslope

then a regression

logit(policy change) = intercept + beta_1 * opinion(50) + beta_2 * opinion(90) =
intercept + beta1 * opinion(50) + beta_2 * (opinion(50) + (90-50) * incomeslope) =
intercept + (beta_1 + beta_2) * opinion(50) + beta_2 incomeslope

In other words – the gilens regression is equivalent – as I have suggested before – to regressing on median voter favorability + a measure for disagreement between income groups

So, now – lets use that regression

logit(policy change) = gamma_1 * opinion(50) + gamma2 * incomeslope

for some gamma1 and gamma2, then this is

= gamma1 * (opinion(50) + gamma2/gamma1 * incomeslope) and so by earlier assumption of linearity

= gamma1 * opinion(gamma2/gamma1 + 50)

The point being – if you assume such a linear relationship between the opinon of the yth percentile (and it is a fairly reasonable simplyifying assumption in my mind) then you can find a percentile y_opt (with certain bounds on gamma2 and gamma1) such that when regressing on

opinion(y_opt) and opinion(50), then coefficient of opinion50 will be exactly 0.

However that just illustrates the entire facileness of the conclusions Gilens drew from their regression and why I have insisted that correlation of predictors is a problem here.

You can of course push this argument further. Add an “educationslope” to the “incomeslope”. If that is significant – woudl your conclusion then be that politicians not only listen exclusively to the rich, but among the rich only to those with a certain education. Add a “religionslope”. You see where this is going. Using the same kind of logic, one can always find ever more specialized voter percentiles along several dimesions that capture all the effect of the regression. And again – correlation does not imply causality. Absence of correlation does not imply independence (other than for normal random variables).

Like

128. HH says:

2)
Of course you can shift predictors now around. But why? I included the same predictors as Gilens. You were perfectly happy to interpret the coefficients in his earlier work – why not now? Earlier – a coeffient of 0 for opinion(50) meant that have no influence – why does a negative one now not imply that politicians actively work against them? After all, you are interpreting a positive coefficient on opinion90 still to move that politicians work in their favor, no?

And what if there is no opinion(99) high enough to compensate for the negative coefficient of opinion(50)? After all, opinion(99) should be bounded somewhere.

Like

129. keithsutherland says:

HH,

>If one group has an 80% faveorability, the other 60% do they disagree?
>If one group has 60%, the other 49%?
>55% vs 45%?

Exactly, Gilens is attempting to draw a binary conclusion from analogue data. This is generally the strategy employed by the media, and has no place in a (peer-reviewed?) political science journal. The authors cannot even blame the press coverage, it’s the way they interpret (and over-hype) their own data that is at fault.

>The rich very much got their way on foreign policy – but then again, the average american doesn’t care about that all that much I think ….

This is similar to a point made in the Manza/Cook literature review (although the latter expressed it as “the politicians got their own way on foreign affairs). Most Americans know (or care) very little about what happens overseas, unless it impacts on homeland security.

Like

130. HH says:

@keith
yes! And on “terrorism” they mostly got their way, where I pick terrorism as a synonym for homeland security.

Like

131. Yoram Gat says:

HH,

I am not quite sure what your point is. Yes – under the strong assumptions that you are making, you would have a y-percentile that would predict the outcome as well as all percentiles combined. What of it?

First, under those strong assumptions, you would have a colinearity problem, which would be expressed in non-significant p-values – which in the analysis of the real data shows is not the case.

But putting this issue aside, the real question is what y – the center-of-influence percentile – is. In a democracy you expect y to be somewhere near 50. In our case we are getting y that is higher than 90, which is evidence that the system is oligarchical.

Again – I think your arguments against the Gilens and Page analysis would carry weight if you offered a plausible alternative model that fits the data. I don’t think you have done so up until now.

Like

132. keithsutherland says:

Regarding the issue of whether or not this paper has been misreported, this is what Martin Gilens had to say in a recent (22 April) TPM interview:

“I’d say that contrary to what decades of political science research might lead you to believe, ordinary citizens have virtually no influence over what their government does in the United States. And economic elites and interest groups, especially those representing business, have a substantial degree of influence. Government policy-making over the last few decades reflects the preferences of those groups — of economic elites and of organized interests.”

He fails to preface the word “influence” by “independent”. The discussions that we have had on this forum would suggest that his data does not support his own interpretation. I’m currently reading Ben Page’s 1992 book The Rational Public and would be astonished if he agrees to his co-author’s rubbishing of “decades of political science research”. I believe that Page contributed little more to the current paper than the initial 4-fold typology of correlation hypotheses and that he was surprised by its conclusions.

http://talkingpointsmemo.com/dc/princeton-scholar-demise-of-democracy-america-tpm-interview

Like

133. HH says:

@Yoram
Actually – my assumption is only slightly stonger than Gilens. The calculate their 90th percentile based on a quadratic function, I assumed linearity for easier exposition. The underlying issues are similar. And wether using a quadratic approximation in their survey data is reasonable when (and I am guessing), they probably only have 5 data points to estimate 3 parameters, that is of course another discussion.

And the question I asked earlier remains. Why would you want to change predictors? Why not interpret the negative predictor as it is .

And when you bring up that pred90 is just a proxy for a different percentile. So why not a proxy for a poor percentile of the population with a negative coefficient. If you can find that (with 0 coefficient for median) would that mean that given the scorn and disregard politicians have for the poor, the median voter has no influence?

Like

134. Yoram Gat says:

> The calculate their 90th percentile based on a quadratic function

They actually say that Gilens “employed a quadratic logistic regression”. I am not sure what this means, but I doubt that this means fitting a quadratic function. Also, the difference between linear and quadratic may not be “slight” – I am not sure how your analysis about having a y-percentile center-of-influence would carry over even if a quadratic approximation was used.

> Why would you want to change predictors? Why not interpret the negative predictor as it is

There is no “as it is”. All causal interpretations rely on modeling assumptions that are external to the data. It makes more sense to me that the causality is as I describe it than to assume that the elected are actively trying to displease the median voter. Your preference may be different, but as far as I am aware the data in this set does not allow discriminating between the two models.

Like

135. HH says:

@ Yoram
> Your preference may be different, but as far as I am aware the data in this set does not allow discriminating between the two models.

So – if this dataset does not allow to discriminate between the two models – and the two models do have quite different implications – why are you arguing that the implications of one model (i.e. Gilens) is proven – if as you say, there is a second available model?

Like

136. Yoram Gat says:

To be clear: the two different models are (a) policy is set according to the preferences of the 99th percentile, and (b) policy is set so as to please the 90th percentile and displease the 50th percentile.

As I wrote, I find the first model (a) more plausible. Do you really prefer model (b)? Model (b) puts the electoral system in a light that is even more oligarchical and twisted than model (a) does.

As for the general question (and to repeat basic epistemological facts) – any data set will be consistent with an infinite set of models. If we waited until only one model would be consistent with the data we would never learn anything. In any learning, there is always a matter of a-priori preference toward some explanations.

BTW, I never said that the Gilens or Gilens-Page model is “proven”. In general, this is not how I see the relationship between empirical findings and theory. I do think that this data set provides interesting empirical evidence in support of the oligarchical nature of the electoral system.

Like

137. Terry Bouricius says:

HH and Yoram,

Since this paper has gotten a lot of international attention, I suspect there are other web sites where statisticians and political scientists are debating the validity of Gilens’ conclusions. Can either of you locate such a site to compare your views with?

Like

138. keithsutherland says:

Terry,

I’ve been scouring the internet for any other serious debate on the merits of this paper and have drawn a complete blank. All I can find is lazy journalism and trite comments from those who like to have their prejudices confirmed.

Like

139. HH says:

@Yoram

actually – I prefer my earlier model
(c) The politicians listen to the opinon of surveys, the media (represented here by the 50th percentile voter). Then, they talk to lobbyists and other influencial people if they disagree with the 50th percentile opinon (my incomeslope) and let themselves be moved to some degree by them.

Yoram, if really all you think is that this paper supports Gilens opinion, but also that there are many other models out there that explain the data equally well, then we are in agreement.

For Gilens, I would just prefer if they stated that their opinon is only one of serveral (with quite different outlooks) that are all consistent with the data and that using their data you cannot discriminate between them.

@Terry
I had hoped that the new site 538 puts up something along those lines as they are supposed to be more “statistical”, but no. Regarding this kind of debate for statistics, I am not aware of other sites, sorry.

@Yoram
As for the way how empirical research “proves” theories, I think the case of tobacco smoke being cancerous would be interesting. The reason it took forever to “prove” this, is not that there was not plenty of data that lung cancer is correlated with smoking tobacco. The reason it took so long is that lots of other studies had to be run to disprove the various other theories the tobacco industry put up. So yes, disproving other theories takes a long time and can be tedious. But if you want to learn, there is really no other way. From that point of view, empirical data actually never really “supports” anything. At best, it is not in disagreement with your theory – but hopefully in disagreement with someone elses so that you can discredit theirs. After years of study – whoever cannot be discredit has the theory that is accepted.That is also why hypothesis testing is all about “rejecting” an assumption – never about proving one.
In hard sciences, e.g. physics, it is exactly the same. A theory is never proven (this is why – to this day, they are still running experiments testing relativity – I think more specialized areas but still. And me dropping my phone is nothing but an experiment to see if gravity still works ;-))

Like

140. keithsutherland says:

>(c) The politicians listen to the opinon of surveys, the media (represented here by the 50th percentile voter). Then, they talk to lobbyists and other influencial people if they disagree with the 50th percentile opinon (my incomeslope) and let themselves be moved to some degree by them.

That strikes me as plausible — the default position is to defer to public opinion (in order, primarily, to secure reelection), unless a mixture of epistemic and financial (campaign-contribution) factors overrule. This fits well with both the empirical literature and a commonsense view of how elected officials might well be motivated.

Like

141. bernie1815 says:

HH:
I no longer have access to the tools to analyze the data set – unless there is a way to download the data to Excel. Can you create a simple crosstab that shows the numbers of issues where a clear majority (say 60/40) of non-elite and elite voters had opposing views on the issue – simply so I can get a sense of the size of the issue we are dealing with?

Like

142. Yoram Gat says:

Terry, I am not aware of a scientific discussion of Gilens and Page. I would certainly be interested if anyone is aware of such a discussion. By the way, the findings that we are discussing – connecting opinion at different income levels with policy, rather than the positions of interest and lobbying groups – are old, dating back to 2005.

Like

143. HH says:

@bernie
Sure, I can dump the whole dataset into excel – then you can mess around with it (e.g. calculate difference and sort). 60/40 is also not a good split – there are almost no policies where people disagree like that.

How should I get the dataset to you?

Like

144. Yoram Gat says:

> actually – I prefer my earlier model (c)

The Gilens data set provides evidence against the model as you describe it. Of course, I would not say that it “disproves” it, but it does create a problem for this model.

> The reason it took so long is that lots of other studies had to be run to disprove the various other theories the tobacco industry put up.

I disagree. The reason it took so long is that there were a lot of political power behind the opposing position. You seem to be over-estimating the importance of scientific evidence in setting public policy.

> That is also why hypothesis testing is all about “rejecting” an assumption – never about proving one.

This “science is about falsification” is the view that is often associated with Popper, I believe (I haven’t read him in the original). There may be something to that, but I think that depending on what is meant it is less true or less important than is usually asserted.

Like

145. HH says:

@Yoram
> logit(policy change) = intercept + beta_1 * opinion(50) + beta_2 * opinion(90) =
intercept + beta1 * opinion(50) + beta_2 * (opinion(50) + (90-50) * incomeslope) =
intercept + (beta_1 + beta_2) * opinion(50) + beta_2 incomeslope

I posted above. According to this, fitting my model c) (considering that beta_1 is close to 0 for Giles; in my own regression, as the pred90 coefficient is larger in absolute terms than pred50, that suffices as well) and beta_2 is significant, that under my model c), opinion50 is positive and significant and so is incomeslope. I don’t see where Gilens data poses a problem for this model.

About the tobacco data. I think you are right about setting policy. But going after big tobacco had a lot to do with big lawsuits of individual people and the subsequent penalties. And there you need better evidence than a simple correlation if you are up against an army of lawyers paid by big tobacco.

Like

146. Yoram Gat says:

Creating linear combinations of the regressors is neither here nor there.

We could boil it down the issue as follows: given that a piece of policy is supported by elite opinion at a given level, does your model predict that higher median-voter support for the policy would correlate with higher chance that the policy is adopted? If your answer is “yes”, and your description above implies that it is, then the data contradicts your model.

> And there you need better evidence than a simple correlation if you are up against an army of lawyers paid by big tobacco.

Again, I disagree. I think that what the courts accept as “credible evidence” is much more a function of power and convention than of scientific rigor. I am not saying that science plays no role, but it is far from being a determinant factor.

Like

147. HH says:

@Yoram
The only reason I am calculating my incomeslope like this is b/c all I have is the dataset Gilens provides. With the original survey questions, something more sophisticated would be possible.

So, assume a piece of policy is supported by a given elite opinon. If median voter acceptance increases and incomeslope stays fixed, then the likelihood increases (of courses, this implies increasing elite opinon). If incomeslope decreases at the same time (which would be the case if elite opinon stayed fixed), then it depends on the strength of the two effects.

In the end – witthout nailing me down on the importance of the incomeslope predictor, there is no way to test these models.

But even if I state something (and there is no reason to; out discussion is not if there is a way to show that Gilens is right – only about whether he has done so already) – the issue would still be that the underlying dataset has big problems as well.

Aside:
opinion(50) is of course only a proxy. Personally, I like overall acceptance much better. In Gilens dataset, correlation of pred50_sw and PREDALL_SW is
> cor(ds$pred50_sw, ds$PREDALL_SW, use=”pairwise.complete.obs”)
[1] 0.9933848

So, I think i could replace that :-)

Furthermore:
Regarding the court room. I desperately hope that science is *a* determining factor, but I agree that it is not the *only* determining factor.

Like

148. Yoram Gat says:

> So, assume a piece of policy is supported by a given elite opinon. If median voter acceptance increases and incomeslope stays fixed, then the likelihood increases (of courses, this implies increasing elite opinon). If incomeslope decreases at the same time (which would be the case if elite opinon stayed fixed), then it depends on the strength of the two effects.

But as you defined incomeslope, i.e., as opinion90 – opinion50, then given opinion90, incomeslope determines opinion50. E.g., given opinion90, if opinion50 increases then incomeslope must decrease.

Like

149. HH says:

> If incomeslope decreases at the same time (which would be the case if elite opinon stayed fixed)

as I said ….

Like

150. Terry Bouricius says:

Gilens and Page were interviewed on the American satire news program “The Daily Show With John Stewart.” the first five minutes of the interview was aired, and the remaining ten minutes is posted on the Comedy Central network web site here:
http://thedailyshow.cc.com/extended-interviews/urccsf/martin-gilens-and-benjamin-page-extended-interview

Like

151. Yoram Gat says:

HH,

We seem to have some miscommunication here so let me try to formalize things:

Define f(x,y) = P(policy changed | median support = x, elite support = y) = P(policy changed | median support = x, incomeslope = y – x).

Now, according to your model, holding y = y0 fixed, is f(x,y0) an increasing function of x?

Like

152. HH says:

@Yoram
Well – you downloaded the same data and have it in R as well.

As far as I can see, using the predictors I have in Gilens data, it is actually a decreasing function :-)

But with different coefficients, it could be constant or increasing, depending on the strength of the incomeslope effect.

Besides, as I have tried to point out before, this is all *not* about the question if I like my model, or if it could be tested or if it is likely or what it actually says. To be honest, I don’t really have a model. I am not a political scientist, I am just a statistician. The only question ever was: Have Gilens and Page proven their claim to an extent so that they can make their strong conclusions? Have they proven it enough to go onto “the daily show” with it?

Anyway, I think we have said all there is to say and are starting to talk in circles. For myself, I will close this thread for now.

Like

153. keithsutherland says:

Terry, HH, Yoram

Here are the Comedy Central URLs for the UK; I think it’s the whole interview:

http://www.comedycentral.co.uk/shows/featured/the-daily-show/videos/exclusive-martin-gilens-amp-benjamin-page-extended-interview-pt-1-the-daily-show-1039365/

http://www.comedycentral.co.uk/shows/featured/the-daily-show/videos/exclusive-martin-gilens-amp-benjamin-page-extended-interview-pt-2-the-daily-show-1039366/

http://www.comedycentral.co.uk/shows/featured/the-daily-show/videos/exclusive-martin-gilens-amp-benjamin-page-extended-interview-pt-3-the-daily-show-1039367/

>Have Gilens and Page proven their claim to an extent so that they can make their strong conclusions?

No

>Have they proven it enough to go onto “the daily show” with it?

Yes (self-evidently), but they have yet to undergo trial by statisticians (other than on this blog).

>We could boil it down the issue as follows: given that a piece of policy is supported by elite opinion at a given level, does [HH’s] model predict that higher median-voter support for the policy would correlate with higher chance that the policy is adopted?

Yes it does, because if politicians regularly ignore median voter opinion on issues that are deemed salient by the self-same persons they will be removed from office. It is only possible to ignore majority opinion in a minority of cases and/or on issues that are not seen as important by median voters. Gilens’ data does nothing to contradict this claim — the data that confirms it is the electoral cycle, whereby unpopular governments are removed (thereby correcting the rupture between majority opinion and policy outcomes).

Like

154. Yoram Gat says:

HH,

> But with different coefficients, it could be constant or increasing, depending on the strength of the incomeslope effect.

No – as you should know, this is not a matter of “coefficients”. This is a property of the distribution itself, not of its representation. And the Gilens data set indicates that, indeed, f(x, y0) is not increasing in x as your model would have it.

> To be honest, I don’t really have a model.

Quite obviously, you do have a model. It is the model you described above as model (c). It is the model that motivated you, as well as others on this thread, to cast doubts about the Gilens and Page paper since their data presents a problem for this model. This model may also be described as the standard elitist model of government – the people rule, but with some tweaking by the responsible elites and the occasional undue influence by irresponsible lobbyists or pressure groups.

> Have Gilens and Page proven their claim to an extent so that they can make their strong conclusions?

Again, the term “proven” is wrong. They have provided interesting evidence in support of their conclusions and against the standard “people rule in elections based systems” claim. That’s all that can be reasonably asked. Again, you raising the burden of “proof” to unreasonable levels merely reflects your bias.

Like

155. keithsutherland says:

Yoram,

HH is right to reference the dominant Popperian model of science; there is nothing in this paper that refutes the standard model of democratic governance that you accurately describe. But why do you choose to label a model of government in which the demos has kratos “elitist”?

Like

156. keithsutherland says:

Early in the campaign, UKIP’s manifesto was for a flat tax and charging people who visited their doctor; when they discovered they actually stood a chance of challenging the three mainstream parties they quietly dropped these two policies. This strikes me as good prima facie evidence that mass political parties will always mirror the preferences of the median voter over those of a rich minority, even when this means going against their own better judgment. What other explanation fits the facts?

Like

157. keithsutherland says:

For evidence of UKIP’s turn to the left, see: http://www.newstatesman.com/politics/2014/04/how-ukip-turning-left-economy

All political parties, even those who claim to be breaking the mould/anti-establishment etc., will jettison the ideological baggage of their original founding elites and move towards the median voter position. This is an inevitable consequence of electoral arithmetic, so the middle ground is going to become exceedingly crowded. The problem is that voters are denied genuine choice as this depends on articulate political elites putting their competing offerings up for sale. The only reason that sortition will make any difference is that elites will campaign on individual issues (to be voted up/down by the allotted parliament) and will have no need to make the grubby compromises that are required by the policy aggregation process.

Andre, is there any evidence yet that Marine Le Pen’s Front National is following a similar trajectory to UKIP?

Like

158. Naomi says:

Keith: >”All political parties, even those who claim to be breaking the mould/anti-establishment etc., will jettison the ideological baggage of their original founding elites and move towards the median voter position. This is an inevitable consequence of electoral arithmetic, so the middle ground is going to become exceedingly crowded.”

If and only if you have a majoritarian (or majority-favoring) electoral system. Countries with highly proportional systems generally have very well-represented and influential parties that are perfectly content to eke out an existence far from the middle ground. This is one of the reasons why I don’t believe elections are as elitist as many people here seem to think. They are elitist, sure, but if you can gain real power (in the form of both policy concessions and even cabinet positions) off a whopping 5% of the public’s support, they aren’t all that bad from an elitism standpoint.

>”The only reason that sortition will make any difference is that elites will campaign on individual issues (to be voted up/down by the allotted parliament) and will have no need to make the grubby compromises that are required by the policy aggregation process.”

The lack of post-electoral horse-trading is one of the things that bothers me most about a sortition-first approach. You can’t as easily steam-roll over a minority when they hold a fair share of the country’s political power. Trade takes place when it is beneficial to all those involved. The way in which they were elected/selected determines what is beneficial to those involved. It logically follows that policy trading between the representatives of different factions in a society generally results in more people being more satisfied overall result than if you were to simply go with the median preference. Sure, those closest to the median preference will end up less happy, but their wishes are not – should not be – the only ones that matter in a free country.

Like

159. Naomi says:

Politicians are rational actors; they will behave in a fashion consistent with success in whatever system they find themselves in. When it comes to Westminster elections that means being the lesser of the (viable) evils in the eyes of the average person in as many districts as possible. All parties that believe they have a real shot at winning seats will make a mad dash to the center. Any other strategy would be downright silly.

Like

160. Yoram Gat says:

> All parties that believe they have a real shot at winning seats will make a mad dash to the center. Any other strategy would be downright silly.

If by “mad dash to the center” you mean “say whatever maximizes the number of votes they get”, then naturally. In terms of policy, post or pre-election, this has negligible impact, as experience shows and Gilens and Page’s data corroborates.

Like

161. keithsutherland says:

Naomi,

I’m unclear as to whether you view elections as elitist [why? given the rational actor/median voter theorem] or why you believe that a PR system that gives some power to 5% of voters would be less elitist. In Britain the only recent election that resulted in a coalition gave significant power to a small metropolitan elite (aka the Liberal Democrat party) and allowed them to impose a number of policies that pleased their activists but alienated everyone else (Clegg has just been voted the least popular leader in modern British history). And why should the leader of a tiny minority get to be the King maker — the choice of who to anoint is generally self-serving, as seems to have been the case in the UK (Clegg very nearly chose to keep Labour in power but Cameron came up with a better deal).

I agree that compromises need not necessarily be grubby, but the 5% scenario you outline rarely happens. Only minorities that enter the winning coalition have any power and this will exclude most of them. In the event of a grand coalition (of left and right) arguably no-one wins, and any kind of resolute activity is ruled out.

The biggest problem with coalitions is policy aggregation. Around 50% of those who voted for UKIP say they will vote the same way in the next general election. The only common denominator amongst UKIP voters is opposition to the EU, but if they gain 20-30 seats at the next general election and end up holding the balance of power, neither Conservatives nor Labour will give them the keys to the Foreign Office; it’s far more likely they will be given a small ministry in which to implement some of their more obscure policies, which next to nobody voted for.

Yoram,

>If by “mad dash to the center” you mean “say whatever maximizes the number of votes they get”, then naturally. In terms of policy, post or pre-election, this has negligible impact, as experience shows and Gilens and Page’s data corroborates.

Rather than just repeating yourself it would be better to engage with the arguments that Gilens’ data demonstrates no such thing. As for “experience” how do you explain the fact that UKIP quietly dropped the flat-tax and charging to see GPs policies as soon as they had a whiff of the sweet smell of success. You can’t just ignore evidence because it supports the median voter theorem.

Like

162. keithsutherland says:

“I have a heuristic: the very fact that a piece of research appears in the papers indicates that it is nonsense.” David Spiegelhalter, Professor for the Public Understanding of Risk, Cambridge University.

Like

163. Naomi says:

Keith: >”I’m unclear as to whether you view elections as elitist [why? given the rational actor/median voter theorem] or why you believe that a PR system that gives some power to 5% of voters would be less elitist.”

Political parties are oligarchical in nature. They have an entrenched elite within them. This isn’t meant as criticism of parties. Elitism doesn’t generally bother me the way it seems to bother Yoram. What bothers me is that majoritarian electoral systems make realignments difficult, entrenching a particular set of elites. At least as long as they behave themselves. What they (and the government as a whole) can get away with without long-term consequence ends up being much greater than if there was stiffer interparty competition and a lower threshold for entry for upstarts.

>”And why should the leader of a tiny minority get to be the King maker — the choice of who to anoint is generally self-serving, as seems to have been the case in the UK (Clegg very nearly chose to keep Labour in power but Cameron came up with a better deal).”

It was my understanding that a Labour-LibDem coalition wasn’t really viable because it would have depended upon the support of several very small regional parties.

In any case, the king maker issue is a very serious problem. Absolutely. We have a need to maximize representativity, choice, and competition while at the same time minimizing the number of effective actors. We also need to maximize the stability and the democratic mandate of the government and its policy positions.

Here’s how I would construct an elected government (or an elected wing of a government), if anyone cares. I’d have a simple unicameral assembly. I would have voters rank parties the same way they rank individual candidates in IRV/STV. The party that wins the smallest share of the vote gets eliminated, and all their votes are transferred. Then the next lowest party is eliminated and so on until only the top seven remain. Or the top five. Pick your favorite number. Seats would then be distributed proportionally to the remaining parties. After the seat totals are assigned, I would have the process of eliminating parties and redistributing votes begin again. I would have it continue until there is a majoritarian winner. Had the election been a pure majoritarian contest, this party would, of course, have won outright. I would give this party’s caucus in the legislature the authority to appoint and dismiss the chief executive with there being no dependence on full parliamentary confidence. Move away form pure parliamentarianism, in other words. I wouldn’t give this chief executive a veto or any other powers beyond what a typical prime minister holds. A coalition would still be needed. It’s just that there would be a clear winning party. The winning party and it’s proposed compromise platform would have a mandate (albeit a weak one, as it would not likely be the first choice of most voters), and the voter-preferred coalition would likely be clear based on the way the votes were redistributed from the smaller parties.

> “I agree that compromises need not necessarily be grubby, but the 5% scenario you outline rarely happens.”

It’s a bit of an extreme example, but Israel, New Zealand, Finland, Sweden, Belgium, Romania, Italy and a few other countries have coalition governments with such minor partners right now. It’s far from rare. If Germany’s threshold were 0.2% lower Merkel’s CDU/CSU would be in coalition with the Free Democrats (who received 4.8% of the vote) right now instead of the SPD.

Yoram,
I would point out, again, that if you change the rules of a game you change the way the game is played. You can not generalize elections based on the American example. Institutional configuration is everything. It would be like studying blackjack and concluding that card games are played against a dealer.

Like

164. Yoram Gat says:

Naomi,

> I would point out, again, that if you change the rules of a game you change the way the game is played. You can not generalize elections based on the American example. Institutional configuration is everything. It would be like studying blackjack and concluding that card games are played against a dealer.

What electoral configurations do you think are producing results that are significantly better than the U.S. system? What is the evidence for this?

Like

165. keithsutherland says:

Naomi,

>Political parties are oligarchical in nature. They have an entrenched elite within them.

Why is it that something that is taken to be of value in any other profession (most of us would prefer to be operated on by the best surgeons) does not apply to politics? Surely the issue is not the personnel, it’s the degree to which parties follow elite preferences/interests rather than those of the public. The median voter theorem applies specifically to a two-party system; the more parties there are then the further the political elites will depart from average-voter preferences.** At some point preferences will have to be aggregated, but if this occurs at the coalition-building stage, then the greater chance that the outcome will be random, and/or reflecting the interests of the elite personnel involved (deals done behind closed doors).

Selecting legislative juries randomly, by contrast, will lead to entirely deterministic outcomes. Assuming exogenous agenda-setting (as in 4th century Athens) and balanced information/advocacy, every random sample will return the same decision and the decision will reflect the informed preferences of the median voter (I retain the use of the term “voter” as exogenous agenda-setting assumes some form of voting/direct democracy/votation). The protection of minority interests is best secured by extra-democratic safeguards, as opposed to leaving it the chance results of secret negotiations between politicians. In sortition randomness plays a benign and deterministic role; not so with the politics of coalition building, which is more like tossing a coin. Sortition also has the value of allowing each issue to be decided on its merits, as opposed to obliging (say) fiscal conservatives and economic libertarians to swallow socially-conservative policies (and vice-versa). Just watch UKIP and the Front National over the next year to see the difficulty they have putting together policy packages that reflect the disparate preferences of their supporters — this is when the compromises become really grubby.

** an update on the theorem (originally devised in the 1950s) is that when both parties converge to the extent that voters can no longer exercise a genuine choice, then a new party will emerge and the convergence process will start again. The political elite can only get away with imposing its own preferences at the no-choice stage, but this doesn’t last very long — in the end the median voter rules. Recent events in European politics bear this out, whereas the US is at a different stage in the democratic cycle.

Like

166. Naomi says:

Keith: >”At some point preferences will have to be aggregated, but if this occurs at the coalition-building stage, then the greater chance that the outcome will be random, and/or reflecting the interests of the elite personnel involved (deals done behind closed doors).”

If a party acts in a fashion inconsistent with the wishes of its voters it will lose votes. Period. All the parties that enter a coalition are gravely concerned about the coalition preferences of their voters. The SPD actually sent out postal ballots to all its members asking them if they wanted to go into coalition with Merkel’s CDU/CSU after the last German election. Opinion polls are a more common strategy. On top of all that coalitions in most countries are predictable if you know the politics and the political history of the country. As long as there are a modest number of parties you can usually look at the result and predict the resulting coalition.

Sure, voters don’t end up directly endorsing the compromise before it goes into effect. Then again, voters don’t directly endorse anything in an assembly chosen by sortition. That’s clearly not a deal-breaker here.

Let’s say we have a market. And let’s say we have only two companies competing in the market. Will they both seek to meet the needs of the average consumer? Sure. Will there be certain areas they avoid competing on? Sure. If they end up being too similar will another company be come up and challenge them? Sure. If demand is great enough that they are able to break into the system.

Having more competitors is healthy. The way in which a debate is framed matters. Most people don’t go to the polls knowing exactly what they want years in advance. They look at their options and pick whichever option suits them. Having more options presented and defended in the public sphere is a good thing. Competition is healthy in a market and it’s healthy in a democracy.

> “The protection of minority interests is best secured by extra-democratic safeguards, as opposed to leaving it the chance results of secret negotiations between politicians. In sortition randomness plays a benign and deterministic role; not so with the politics of coalition building, which is more like tossing a coin.”

There is nothing random about coalition building. I can recommend a few titles about coalition dynamics if anyone cares. For what it’s worth, coalitions in countries with plurality elections are usually awkward, brittle things. Minor party seat totals are generally too fragile for minor parties to be willing to force an early election, vote splitting is usually a serious problem and so a majority coalition does not necessarily have the support of a majority of voters. Not that majority governments necessarily have the support of a majority of the people. Modi’s party won an outright majority off 31.0% of the vote in the last Indian election, for example. I just want to caution against generalizing based on the UK’s recent coalition experience. Politics is also very different when coalitions are expected. Parties often spend decades establishing positive relationships with other parties that they may find themselves in coalition with at some point in the future. This is why you see consensus, rather than adversarial, politics in these countries. See pretty much everything ever written by Lijphart for a far more detailed analysis.

You say that minority interests are best served by non-democratic safeguards. Like the Supreme Court? The Supreme Court has done a fine job, but it has a tendency to follow trends in the popular will. What happens if the popular will is soundly against a minority? How about political needs that go beyond simply avoiding government oppression? What if addressing needs of a minority are very unpopular with the median voter? A minority voter is not somehow less deserving of having a place at the table than a median voter. They are both citizens. A minority’s concerns are not less worthy of being heard.

If you want the assembly to serve as a proxy for conflict in society, then every significant faction in society needs a seat at the table. Every major faction needs to have their leaders at the table.

If we look at plurality presidential elections in Latin America we see something very strange. Very often you see very powerful presidents being elected off around a third of the vote. The current President of Honduras was recently elected with 37% of the vote, for example. This is not at all uncommon in Latin America. Many countries see similar results election after election. Why don’t the parties rally together and propose compromise candidates? According to Shugart and Carey, there are too many divisions within these societies. It’s literally not possible to propose compromise candidates, as there’s no such thing as a true median voter. You have a Condorcet’s paradox type problem between factions. Everyone just has to go to the polls and hope for the best.

Yoram,
In addition to the conclusions that can be drawn from the arguments above, the US lacks clear responsibility in its institutional structure. Most of your previous comments about democratic accountability are absolutely true in the American context. If that were the only model of representative democracy I would absolutely agree with you. However, the problem with accountability in the US government is fairly simple: no one is “in charge.” Who is “in charge” right now? Is it Obama? Not really. There’s relatively little he can do without congressional approval. Is it Boehner? No. If one party wins the presidency and a majority in both houses is that party in charge? Not really. US parties are too poorly disciplined to be responsible for the actions of their members. Even if they were highly disciplined, then is an inherent separation between an entrenched president and his/her party. Who do you punish if the government acts against your wishes? The politicians can all blame each other and say they did the best they could given their colleagues. And you know what? They can all be right about that. Even if you were to elect the most brilliant, most honest, most capable people in the world, the behavior of the government as a whole could still be terrible simply because each individual member is only responsible for their own actions, no one is responsible for the aggregate effects of the actions of all the members.

There is no accountability in a committee. The US government is a committee of committees. It’s not surprising that it is relatively unresponsive. How do you vote to get what you want?

In order to achieve good government in an electoral framework there must be someone who is objectively in charge. There must be consensus on this. In a parliamentary system, for example, the party or parties that form the government are in charge. Period. If an action is taken then it is taken because they decided to act. If no action is taken when one was needed it is because they decided against acting. Whether or not you support the government is irrelevant: there is no ambiguity about who is in charge.

——————-
Sorry for the length of the post. Touched a few nerves is all. Coalition dynamics and proportional representation happen to be passions of mine. Yes. I do need a life. We chemists take pride on being the craziest scientists. Something to do with all the fumes. Sorry for any typos, or confrontational comment bits or anything like that. I may have had a bit too much Jack tonight. Jack beats fumes any day.

Like

167. Yoram Gat says:

Naomi,

> In a parliamentary system, for example, the party or parties that form the government are in charge. Period.

This is an unrealistic view of how things work. In reality the coalition parties blame each other, blame the opposition, blame the media, blame the economy, blame Putin or blame the terrorists, or point out (quite justly) that the alternative is not any better. The public has no real ability to know whether any of this is true, and even if it did it would not be able to elect a better government since all the credible candidates are of a kind.

Again – can you offer any examples of electoral systems that you think are well functioning? What is your evidence that they are doing any better than the US or the UK?

Like

168. keithsutherland says:

Naomi,

>If a party acts in a fashion inconsistent with the wishes of its voters it will lose votes. Period.

That’s certainly what has happened to the UK LibDems — they have upset their activists due to the need to compromise in government, have upset left-leaning voters by supporting a Conservative government and have upset Conservative voters by tying their hands. They may well be wiped out at the next general election.

>If you want the assembly to serve as a proxy for conflict in society, then every significant faction in society needs a seat at the table.

Agreed, but this is not the case for small parties that fail to be included in the coalition. If the population they represent is so small and unpopular that the supreme court cannot protect them they have no chance to become part of a governing coalition. They are just left crying in the wilderness.

>Sure, voters don’t end up directly endorsing the [PR] compromise before it goes into effect. Then again, voters don’t directly endorse anything in an assembly chosen by sortition. That’s clearly not a deal-breaker here.

Agreed. That’s why, unlike most commentators on this blog, I advocate either PR-based elections or DD initiative for policy proposals. The job of the randomly-selected assembly is only to exercise their informed judgment on these democratically-mandated proposals. As the judgment is exercised on a case-by-case basis, it’s not subject to the problems of policy aggregation. Allotted jurors are accountable to no-one, hence the need for (Bayesian) accountability for the factions making the proposals and (occupational) accountability for the government ministers (who can be removed by confidence motion).

Accountability requires mixed government — electoral systems (plurality or PR) will not suffice. I agree with you that the US system is an unaccountable committee of committees, but it is not mixed government, it’s just an artificial division of powers. Individual secretaries may be brilliant or dreadful at their jobs but will still be out of a job after the election of a new commander in chief. Parliamentary systems are only accountable over a very short term (and ministers generally only last a year or so) and political parties are very good at changing their leader/name etc. in order to jettison the baggage of past mistakes. Most elected ministers would not even make the shortlist for a job interview in any other organisation.

Like

169. Terry Bouricius says:

Naomi,

Your last post sounds very much like what I would have written not long ago. I worked for over a decade as a senior policy analyst for FairVote: The Center for Voting and Democracy pushing single transferable vote proportional representation (STV PR). The slogan we used for PR was that while the majority should have the right of decision, everybody had a right to representation at the table. In fact it was the process of consideration of STV by the randomly selected Citizens’ Assembly in British Columbia (to which I submitted testimony) that convinced me that sortition might well be a superior tool for running a democracy.

But I want to challenge one of your statements that on the surface seems so obvious it is above challenge… You wrote: “Competition is healthy in a market and it’s healthy in a democracy.” Yes, in an electoral system some competition is always better than no competition. But competition for votes in a mass election (where rational ignorance reigns) can distort society’s agenda in terrible ways. Parties seek out hot-button issues to inflame core supporters or appeal to ill-informed median voters, while begging for vast amounts of money from sources with ulterior motives to win competitive elections. Reducing “democracy” to a horse race is not healthy.

A randomly allotted genuinely representative sample of society will, of course, have conflicting values and interests, but these are distinct from the imperatives created by electoral competition. An allotted group has an interest in seeking and occasionally finding win-win compromises that would never see the light of day in a winner-take-all, nor in a PR party-based competitive election system.

Thus even ignoring for the moment the elite manipulation issue (though I consider it VERY important), sortition without the competitive incentives of competitive elections may produce better, and even more popular policies than a competitive system. (To be clear, this is not WITHOUT competition, as advocates for and against various policies will compete for their preferred policies within an allotted body, but without the society-wide hatred-inducing competition of mass elections).

Like

170. Naomi says:

Terry,
I don’t disagree. I just don’t believe highly competitive elections need to be so dirty. Look at classic Lijphartian consensus democracy. You can’t exactly spend months slandering a rival party and its policies if you might need to enter into coalition with that party at some point in the future. Your political fortunes and your ability to deliver at least some of your campaign promises could easily hinge on your ability to form a coalition with a party that absolutely thrashed you in the polls. You end up having to vote for some of the things they campaigned on… and that you campaigned against as a part of an overall package that you can sell to your voters as having been worth their vote. This is the norm in a lot of countries. Politics ends up being rather cordial. At least in comparison to adversarial politics like we see in the US and the UK. Cordial enough for the system to be dubbed “consensus democracy” in the literature. Sure, parties still exploit hot-button issues and voter ignorance, but we’re on a whole other level here.

Then there’s the Swiss example. Elections in Switzerland are downright boring. They make Scandinavian elections look almost adversarial by comparison. There are many competitors vying for seats, but every policy of consequence will go before a direct vote anyway so why bother getting worked up? There’s been a grand coalition continuously for almost 60 years. Few people care. Swiss referendums have all the standard problems with mass politics. Sure. That’s a great reason to largely replace them with a sortition-based alternative.

Will a minority with an established leadership feel represented if their chosen leaders are not allowed a seat at the table?

I’ve read your sortition-based legislature proposal but I’m still unclear on how you would construct the executive. I apologize if I just missed that bit somewhere.

Like

171. keithsutherland says:

Terry:> Sortition without the competitive incentives of competitive elections may produce better, and even more popular policies than a competitive system. (To be clear, this is not WITHOUT competition, as advocates for and against various policies will compete for their preferred policies within an allotted body, but without the society-wide hatred-inducing competition of mass elections).

That’s clearly true, but without some form of society-wide competition (election/DD/votation) how will the competitive advocates be selected? If there is no democratic mechanism for generating policy options and (discounting tautological arguments on the interests of the microcosm automatically reflecting the interests of the demos), how can you assume that the resultant policies will be popular (I leave aside epistemic considerations)? Whilst you argue that cleavages are artificially manufactured by election, Naomi and I would take the view that competitive politics is a way of civilizing existing political differences that would otherwise be settled by war. A good analogy here is the criminal justice system that civilized retributive passions and revenge feuds. Subvert the ritualised battle of democratic politics and the likely result is bloodshed (vide Ukraine).

Naomi:> Then there’s the Swiss example. There are many competitors vying for seats, but every policy of consequence will go before a direct vote anyway so why bother getting worked up? . . . Swiss referendums have all the standard problems with mass politics. Sure. That’s a great reason to largely replace them with a sortition-based alternative.

That’s the sort of balance between election and sortition that I have been advocating. Both are necessary — elections/DD to generate policy options and sortition to subject them to informed democratic judgment. “Votation”, during which the public make a selection from a list of DD policy initiatives, is also taken from Swiss democracy.

Like

172. Terry Bouricius says:

Naomi,
As to the executive branch…you are correct about the fact that my paper didn’t deal with that at all. David Schecter and I have written a paper that is currently in peer review on a proposal for the executive, but I can summarize it concisely. the Executive should not have policy-making function (though may propose policy to an allotted legislative branch, which might be organized in various ways…such as Keith’s, Yoram’s or my scheme). The Executives would be essentially hired by an allotted search committee (sort of like a city manager). Periodically each appointed executive would also face a retention jury, which could remove an executive, but leaving it to a separate appointment jury to find a replacement. A key point is that the executives should be judged on how well they carry out the democratic policy of the legislature. With current election schemes there is usually a conflation of the two VERY distinct issues of the candidates POLICY desires, and COMPETENCE… often leaving voters with an impossible task (Do I vote for the better administrator whose policy I hate, or the one with better policy who is corrupt and incompetent?)

Like

173. Naomi says:

Terry,
I know it’s off-topic, but decentralized cabinets make me very uncomfortable. Who’s in charge in a crisis? What happens when some of the ministers favor one (perfectly reasonable) course of action and others favor a different (also reasonable) course of action while the rest favor keeping a low profile in the hopes that the more vocal ministers end up taking most of the blame when things go south? What happens when there’s interpersonal squabbling between ministers? What if the balance of the members in their respective review boards just don’t understand what is involved in the ministers’ roles well enough to be able to judge who acted more appropriately, if there objectively is such a thing? This is why CEOs are typically responsible for appointing other important officers. The board appoints the CEO, and the CEO appoints (and can dismiss) everyone else, including the company’s president. Usually.

Can a sortition-based legislature really move fast enough, and provide enough detail to replace the sort of executive autonomy we see in modern governments? I would be very surprised if modern legislatures could manage it. At least ones with poor party discipline would probably struggle.

I feel like I’m coming across as overly critical and I really don’t mean to. I do like your overall proposal.

Like

174. […] choose between the theoretical and empirical literature and not to seek to reconcile the two. The recent thread on this blog discussing Gilens and Page’s claim to have disproved the median voter t… is a good indication of the sharp divide between the two literatures. The honourable exception to […]

Like

175. […] has been a lot of debate on this blog recently as to whether or not political outcomes are unduly influenced by the rich; in this post I want to consider the influence of other nondemocratic agents – activists, civil […]

Like

176. […] to retain membership of this percentile for ten consecutive years (p. 6). The oft-cited research of Gilens and Page (G&P) focuses on the top 10%, and H&R’s findings indicate that over half of the U.S. […]

Like

177. I wonder if anyone has spent much time considering the tiny R square value of the model. at 0.07, it suggests that only 7% of the variation in policy outcomes is explained by *all* of the stated preferences combined. Now if I were of the same mindset at the popularisers of this paper, I’d say the remaining 93% is presumably proof that policymakers do as they damn well please.

Like

178. keithsutherland says:

Emmanuel,

For the benefit of the statistically illiterate (such as myself) could you expand on this please?

Like

179. Robert says:

I am concerned with, among other issues, that multicollinearity is masking the significance of the preferences of average citizens variable in the full model. The correlation between preferences of average citizen snd preferences of economic elites is quite high (.78) and both are significantly associated with the dependent variable in bivariate analysis.

Like

180. Yoram Gat says:

Robert,

The point of the analysis is to understand what happens when the elite and the average citizen disagree. The data show that average citizen opinion carries essentially no weight.

It is true that in general collinearity means that one regressor can be approximated by a combination of other regressors and in this way the uncertainty associated with the coefficients of particular regressors is increased. However, in this particular case, the data does show a statistically significant result.

Like

181. HH says:

Hi Robert,

I remember having this same discussion with Yoram a while back (see above). Don’t even argue with him – it is completely pointless. The analysis says what he wants to hear and therefore any criticism of it or even just scepticism or fear of over-interpretation will fall on completely deaf ears. The small R-square, the issue of multicollinearity, the simplicity of the model, the problems of the data collection all don’t matter.

Like

182. keithsutherland says:

Like

183. HH says:

Hi keith, nice to hear from you as well. Oh well – that is not unexpected that we are still at the same place. It is one reason that I haven’t been here for a while ….

Like

184. keithsutherland says:

I don’t think I would be here either were not this forum required reading for my day job (a PhD on sortition). Having said that, there have been some extremely valuable contributions by Naomi, Andre and others who are prepared to engage in genuine discursive exchange. Fortunately for me Yoram ignores everything I say on principle, so I don’t even to follow your recommendation to Robert.

Like

185. bernie1815 says:

Hi Keith, HH and Robert:
I agree with Robert. As my first comment, oh so long ago, indicated – complex, multifaceted variables should not be entered into a multiple regression because of exactly Robert states. Interpreting the results from models specified thusly remains for most of us a statistical and conceptual challenge because of this issue and what has been left out. Perhaps in the last year the authors have refined and sharpened their analysis. Or perhaps they are continuing to saw a piece of wood with a hammer!

Like

186. Yoram Gat says:

Wow! Welcome back, everybody.

Yes – Robert, by all means, please read the thread and see if you find anything useful. (But if, like Sutherland, HH and bernie1815, you arrived with your mind already made up and damn the data, just be upfront about it and don’t pretend, like they do, to be having a reasonable discussion.)

Like

187. HH says:

Yes, please read the thread. But I reject the idea that I arrived with my mind already made up. During the discussion last year I went so far as to actually get the data and analyzed it. The main problem – the underlying data is really not all that impressive. They acknowledge that their income percentiles have high correlation. But it is actually worse, because most of the time median voter and high income voter are on the same side of 50%. If the rich agree with something 80% and the median voter 60% and the politicians do it – did they now do what the rich want? Gilens concludes yes. Hardly any decisions have median and rich voters on opposite sides of a subject. Anyway – it is all in the thread up there. Have fun!

Like