Why so much Stability?

Who do you find more Convincing, Riker or Mackie?

Riker claims that cycles and manipulation are pervasive in politics, and that the only reason they’re not more obvious is the paucity of data on voter preferences (since many electoral systems only record first preferences, and can’t distinguish whether what’s expressed is sincere or sophisticated). This possibility calls into question the meaningfulness of democratic politics, and allows scope for politicians skilled in ‘heresthetic’ (the art of political manipulation) to secure their favoured outcomes by altering the agenda or issue dimensions.

Despite the difficulties of observing such instability and manipulation, Riker claims to have identified certain historical examples – such as the 17th Amendment, Wilmot Proviso and Lincoln election – that are best explained by postulating voter cycles and heresthetic manipulation.

Gerry Mackie’s historical analysis of these cases calls Riker’s findings into question[1]. For example, Riker assumes in the 1860 election that Lincoln won, Lincoln voters ranked John Bell above Stephen Douglas, and generates a cycle based on this assumption. Mackie points out that this is implausible. While this shows there may not have been a Bell, Douglas, Lincoln cycle (as Riker would have liked), it doesn’t harm the essential point of Riker’s story, however – which is that Lincoln was not a Condorcet winner, but won through a combination of the institutional system and heresthetics[2].

As McLean points out, “Riker’s case studies are falsifiable, and some of them have been falsified. But they are not verifiable”[3]. If Riker’s stories weren’t falsifiable, then they would have little meaning – at least, they would be untestable hypotheses, and as Popper observes, a prediction that can’t be falsified isn’t useful in social (or other) science. In trying to give any historical example, then, Riker is always vulnerable to the claim that he has got the analysis wrong, and that the case doesn’t reveal what he wants. On the other hand, no matter how many particular cases he has got wrong, one can’t prove that Rikerian cycles never take place.

Mackie’s more powerful criticisms go beyond simple quibbling with Riker’s examples, and strike to the heart of the Rochester methodology – arguing that Riker’s analysis fails or is self-defeating on its own terms. For example, Mackie points out that Riker places unrealistically high standards – for those of us outside of epistemology classes – on what we need to know someone’s preferences. Riker says we can’t know someone’s real preferences just from how they vote.

It’s true that we can distinguish someone’s action (what they were trying to do, in their self-description) from their observed movements. As Mackie says, “When someone buys a Cadillac, what choice has been revealed: a means of transportation, a status symbol, a dating ploy, a nostalgic memory of the buyer’s father, a tax dodge, a mistake?”[4].

Judging someone’s inner state from their outward movements is always a matter of interpretation, and always leaves the possibility of mistakes. In everyday life, however, we do think we can make reasonable assumptions about what may have motivated another’s movements. For example, if I see you reach across the table and pour a glass of water, it is natural to assume you wanted a drink. There may, of course, be other explanations, but we are surely ordinarily justified in taking the most obvious.

If we apply such a strict standard of knowledge, however, it doesn’t just block the possibility of inferring preference from how one votes, it seems to block the very possibility of recognising a vote from one’s movements: “raising one’s hand might be a sincere vote, might be a strategic vote, might be a mistake, might be a yawn and a stretch, might be a sign of a follower of St. John the Baptist, might be a joke, might be an involuntary reflex”[5].

Riker’s point is that we can’t know, just from someone casting a certain order of votes (say, in an STV election) what their real preferences are – we can’t know the person they put second really is their second preference, or if they are put there strategically. The ad hominem reply is that on such a strict requirement for knowledge, we can’t even know – simply from someone walking into a booth and writing numbers next to different names – that they are intentionally voting at all, rather than merely doing something to alleviate boredom. Our electoral officials will take it as voting, because that is the natural explanation of what they do. Similarly, the natural explanation of why someone puts 1 next to one name and 5 next to another is that they prefer the first.

Moreover, the idea that we could recognise someone as voting strategically is parasitic on the idea that we at least understand (even if we can’t always recognise) sincere voting. As Mackie puts it, “We could not discover that choices may strategically misrepresent preferences unless we have information from beyond the single instance… we know that choices sometimes misrepresent preferences only because we know that choices sometimes do represent preferences”[6].

If we thought what someone wrote on their ballot paper was simply arbitrary, we’d have no reason to pay their marks any attention. We interpret their pen strokes as expressing preferences because that’s the usual explanation. We assume they actually rank the first choice higher than the fifth, because again that’s the normal case. It is only against this background assumption that we can understand strategic misrepresentation. If we took the marks as meaningless, then swapping two votes wouldn’t have any effect, because they would be meaningless. Strategic voting only works because the sophisticated voter knows his/her marks will be interpreted as expressing sincere preferences. Compare a feint (in football, or sword-fighting, say) – faking a move to your right (before actually going left) only works if others believe from the move that you’re actually going to your right.

All these points by Mackie are well made, I believe. To say we can never infer someone’s preferences, and that all votes must be taken as meaningless, is surely too strong. From the fact that we can be wrong about any particular case, it doesn’t necessarily follow that we can be wrong about every case[7]. I might believe P and ¬P. Either of those beliefs could be false, but they can’t both be.

So, it seems Riker tries to draw conclusions that go beyond his evidence. We can’t simply discard all our beliefs about votes, preferences and outcomes, simply because some of them are called into question. On the other hand, he still seems to have enough resources to draw more modest conclusions, sufficient for his purposes. The universal possibility of error doesn’t entail the possibility of universal error, but it does mean, while we might be right to trust our intuitions as a whole, we should be wary of relying on any one in particular – we must always acknowledge that the outcome of any particular vote could be meaningless.

A similar point can be made about Riker’s obsession with cycles. Perhaps he simply liked their elegance, or the fact they could undermine the very possibility of a Condorcet winner[8], but Riker was always looking, not just for instability or manipulation, but cycles. Mackie may be right that actual cycles are far less prevalent than Riker’s sometimes-fanciful examples would have us believe. These findings, however, don’t challenge Riker’s fundamental conclusions as McLean observes, the cycles are inessential elements of the stories, “There are heresthetic moves that do not entail cycles: some of them are in Riker’s own stories… All that a Rikerian needs to show is that multidimensional issue space offers the potential to construct a new winning majority”[9].

Thus far, it seems Mackie’s criticisms should have some bite, in limiting the scope of Riker’s conclusions, and forcing him to be more modest in his attacks on democracy. Nonetheless, we haven’t seen anything that forces us to give up the Rikerian project, or deny that it’s lessons can be of great practical import or be applied to new (past or future) cases[10]. We still have perhaps Mackie’s most powerful ad hominem point to come, however.

Riker’s methodological approach seems to presuppose the falsity of the very claims he is trying to support!
“Riker’s further analysis of the Powell amendment makes an implicit auxiliary assumption that elected representatives represent the interests of their districts. He goes on to confidently identify five “natural political groups”… and the preferences of each group over three alternatives (Riker 1986, 118-122). He seems to have forgotten, among other things, that on his account there is no such thing as a district interest that could be discovered by electing a representative. It is a delicious irony that his analysis is forced to assume that Congressional districts have identifiable interests… [I]n his attempts to show the empirical relevance of Arrow’s logical possibility result, Riker is forced to assume the empirical irrelevance of Arrow’s result”[11]

As a rhetorical point, this is brilliant – that someone is forced (by their self-defeating argument) to deny the very thing they’re trying to establish is more fatal than simply observing that they beg the question by supposing its truth, since at least the latter allows that the proposition could still be true (though the argument for it fails), whereas the former suggests an incoherence in what is to be proved. It’s therefore no surprise that this point comes at the end of Mackie’s chapter 2, on ‘The doctrine of democratic irrationalism’, as the key point in his refutation of Riker’s general argumentative strategy.

The question is, however, whether it’s as effective as it first seems. If Riker stuck to his position that it was impossible ever to know a group’s preferences, then it would indeed establish that Riker’s analytic narratives were incoherent, for he is forced to give up that commitment and postulate the very thing he denies to explain cycles. If he gave up the search for cycles, however, it’s not clear he would have to engage in such problematic postulations. Moreover, we’ve already established that Riker’s conclusion are too strong – he should give up the claim that democratic outcomes are always arbitrary, irrational and meaningless, in favour of the weaker claim that we can never know they aren’t. If this is so, then it seems his stories aren’t so implausible – we can suppose certain groups of voters who have stated preference patterns, and see how the final result is still chaotic and manipulable. If we deny that the voting groups can even be given such ordered preferences, then in a sense we make the final conclusion even stronger. If the parts are arbitrary and meaningless, so a fortiori is the whole derived – albeit itself in an arbitrary and meaningless manner – from them.

There is much truth in Mackie’s criticisms, and it would be right to be more cautious in stating Rikerian conclusions than Riker himself was, but Mackie has not refuted Riker’s analysis. What Mackie has shown is that democracy is not impossible – that society can choose, and stable collective orderings may be possible. Riker should never have denied this, for the Arrow result doesn’t say it is impossible to reach a collective decision (it can be done, if e.g., voter preferences are relatively homogeneous without violating Condition U). Rather Arrow showed it is impossible to guarantee an ordering – that is, and ordering is possible but not necessary. What we learn from this is not that democratic outcomes can’t occur, but that they needn’t always occur – no matter how we design our institutions.

The search for historical cases demonstrating cycling and instability may be inconclusive. Indeed, given limits of what we know, it may even be necessarily so – it may be that we can’t possibly find a definite real-world case of such. This only demonstrates that stable outcomes are possible, not that cycles and instability can’t occur. So even if Mackie is right about the historical instances Riker cites, Riker’s conclusions are still relevant and important – particularly when it comes to designing future institutions (the most normatively important question).

[1] His results are summarised by I. McLean (2002) ‘Review Article: William H. Riker and the Invention of Heresthetic(s)’ British Journal of Political Science 32 table 1 (p.549) and G. Mackie (2003) Democracy Defended table 1.6 (pp.18-9). Since I am not qualified to comment on these historical examples, the following discussion draws more on abstract methodology, rather than interpretation of these data.
[2] I. McLean (2002) p.553.
[3] I. McLean (2002) p.555.
[4] G. Mackie (2003) p.38.
[5] G. Mackie (2003) p.38.
[6] G. Mackie (2003) p.39.
[7] It doesn’t follow from the fact that anyone could be an eldest sibling that everyone can be.
[8] The identification of which Riker took as an essential normative requirement of an acceptable voting system.
[9] I. McLean (2002) p.555, c.f. p.553.
[10] See I. McLean (2001) Rational Choice and British Politics, who argues various possible explanations of stability “do not invalidate Riker’s central claim. Once in a while there comes a politician who… can see opportunities where others do not, in opening up or closing down political dimensions” p.231.
[11] G. Mackie (2003) p.43.