The point is not that we all know politics equally well.
Thr point is that, in politics, a credentialed expert doesn't really have any advantage in predictive power over a reasonably well-informed layman.
Statistician and market researcher Jim Manzi made this point roughly ten years ago. It's not a knock against the social sciences either. The point is that you can't generalize in the social sciences the way you can with harder sciences because the causal factors are so complex.
For prediction? No. For understanding (some of) the dynamics of what is going on? Hopefully. That's what much of our discipline is actually aimed at. (Source: am political scientist)
> For prediction? No. For understanding (some of) the dynamics of what is going on? Hopefully.
For some reason, it seems to me that social scientists undersell themselves (why is that? it seems like they almost go along with anti-intellectualism rather than have to deal with the conflict). Understanding the dynamics leads to better predictions and thus better outcomes; it's hard to overlook that mechanism.
Because the systems under study are not isolated external systems that follow well-order rules like a laboratory science. Prediction of human systems is like prediction of weather systems one week ahead: there are ways to model and improve methods for such a task, but our predictive success will continue to be just a little bit better than playing slots.
It would be folly and vain to misunderstand the difficulty of the problem at hand in an effort to signal that one is confident and optimistic. Social science from a century+ ago suffered a lot from folks with that frame of mind, sometimes ending in utter disaster.
A major difficulty is that understanding one part of a complex system does not necessarily lead to testable predictions about the system as a whole. Take elections, for example: we have really good models for voter behaviour under different conditions, so can make good estimates for how, say, a sitting President's votes will vary as a function of employment, or the differential effect of various changes on results in states with different demographics. But what is the electoral consequence (say) of Bill Clinton being impeached? There's no real way to know in advance: individual political scientists will have intuition, but it's too novel a situation with too little existing data for solid model building or prediction. And it turns out that politics is rammed full of new situations, unexpected events, or changes in the environment. How will voting patterns in the South change as a result of GOP realignment after civil rights? What will happen to the British government's popularity when the Queen dies?
A case in point: a couple of years ago I made an HN comment about parliamentary manoeuvring over Brexit and the likely outcomes [0]. I am skilled and knowledgeable about British parliamentary culture and history, many of the subtleties of the Standing Orders (which were suddenly relevant), and had been closely following the proceedings and thinking hard about the implications. I missed the actual outcome completely, by factoring in the possibility of no deal, or of different kinds of compromise, or of the rebels winning a second referendum. But I didn't consider that Johnson could knife Theresa May, supplant her as PM and then successfully sell a clearly defective alternative (and 'harder') deal as a triumph. Even though I knew lots about Conservative politics, and the strengths of populist rhetoric in changing public perceptions. I assumed that if May was deposed her successor would be in the same bind. But Johnson changed the terms of debate unexpectedly successfully and escaped it. My mostly-rationalist colleagues basically missed calling it too.
FWIW I'm big on intellect, but I'm also big on epistemic modesty. There are lots of things we can show are basically true or false (the 'filter bubble' hypothesis for online political information has been basically debunked, for example) but too many unknowns for our mere opinions to be provably correct.
The same is equally or more true of basically everyone else, of course, which is why I find the false certainties of the newspaper opinion pages so depressing.
> Even though I knew lots about ... the strengths of populist rhetoric in changing public perceptions
It seems to me - with no expertise or data beyond reading a lot of public affairs, history, and political science - that from the start almost everyone has had (or put) their heads in the sand about it, that such a response is the norm for insurgent populists (perhaps not unlike the response of the incumbent status quo to disruptive challengers in every field, including IT (think of all the 'disruption', or just Blackberry's response to Apple/Google smartphones)), and that those twin patterns were the biggest threat. And I think that has played out.
AND IT CONTINUES TO PLAY OUT: The competitors to the populists ignore the true political power, the populist reactionary messaging which can sell anything to the public (so far, they can fool enough of the people enough of the time), and try to focus on where they feel safe - legislation. I think they are afraid of the populist rhetoric, and I think people see that they are afraid and ineffectual. I predicted in November and predict now that it will be the downfall of Biden and the Democrats. Nobody votes for ineffectual and afraid - they have nothing to offer - they adjust to where the power lies and the future holds.
Is that just a bunch of BS? (I really did say those things and do think them, but that doesn't rule out a bunch of BS.) And if not, why the heck am I the only one to see the obvious? Is everyone just trying to ignore the disruptive challenger?
I definitely think there's an impedance mismatch between centrist politicians, who believe in Median Voter Theorem[0], a modest policy offer, and a very specific kind of competence; and various kinds of populist disruptive challengers (this latter could be from the left, e.g. Jeremy Corbyn; the right, e.g. Silvio Berlusconi; or the centre, e.g. Beppe Grillo).
The former approach has demonstrably worked for those politicians for decades. Bill Clinton and Blair, for example, were crushingly successful electorally and marginalised their parties' radical wings. So their followers and successors (and lots of political scientists) hold the empirical success of this kind of politics to be proven and obvious.
But the latter has also worked electorally and politically: Berlusconi, M5S, Trump, whatever. And when will they work again in the future? We don't really know. Ross Perot didn't get far in the 90s - how did Trump win? There are lots of theories, but given how few elections you get per year, not enough data. Which environments support which kinds of politics? We have an idea, but not a functioning predictive model.
So people tend to argue for what they personally believe in - which is obviously informed by professional knowledge but shouldn't be confused with science. Many of my colleagues were astonished and shocked by the rise of Corbyn, for example, while I'm more sympathetic to disruptive politics (and critical of the technocratic status quo) and the possibility of something populist breaking open the existing system was obvious to me. That the breakthrough candidate was Corbyn (a long standing politician with no previous evidence of either high competence or desire for leadership) astonished me too.
[0]: MVT is an example of the difficulties of formal modeling in political science. It's provably true within its assumptions (a one-dimensional political spectrum and voters who support the closest candidate) but those assumptions are only weakly true for politics in reality (and the system becomes unstable if the first in particular is false).
Anthony Downs just died. His Median Voter Theorem organises a lot of understanding about political competition. Short version: elections get won in the centre.
That is what has always confused me about "prediction markets".
The reason why non-experts perform as well as experts is because there is a huge level of randomness in actual outcomes, and very little information that can help predict beyond randomness.
And prediction really isn't the goal in practice, but robustness. Even if you were a "superforecaster", you are still going to be wrong most of the time. So surely it is more important to be robust to outcomes and stop trying to predict something that can't be predicted.
(To compare this with large skill gaps to expertise...sports betting, these events are random, some are more random than others but the gap with expert knowledge exists because there is information that exists that will help you predict outcomes. That gap doesn't exist in politics because there is no real information that can help you...it is just totally different, it makes no sense to compare them imo).
Thr point is that, in politics, a credentialed expert doesn't really have any advantage in predictive power over a reasonably well-informed layman.
Statistician and market researcher Jim Manzi made this point roughly ten years ago. It's not a knock against the social sciences either. The point is that you can't generalize in the social sciences the way you can with harder sciences because the causal factors are so complex.