Saturday, October 11, 2008
A Failed Prediction?
As I write this, Barack Obama appears to be heading towards a clear victory over John McCain in the upcoming US presidential election. However, I predicted a McCain victory months ago, long before Obama even won the Democratic nomination.
I did so having no idea gas prices would spike as much as they have, and no idea that there would be a massive financial meltdown. And rightly or wrongly, American voters tend to view the Democrats as stronger than Republicans on domestic economic issues: these crises have significantly benefited Obama.
It seems now that for McCain to win, some new issue will have to induce greater fear than these economic crises, and McCain will have to be seen by voters as more likely to erase that fear. That seems to be how it works: which ever candidate voters believe is most likely to remove their most pressing fear, wins. In practical terms, I think this means that the only thing that could get McCain elected now is a brand new Osama bin-Laden video, or God forbid, a terrorist attack.
In any case, if my prediction fails, as it now seems certain to, it will be another testament to the difficulty of establishing an impressive track record of long-term political prediction (by impressive, I mean notably above what random guessing would yield). There is an inverse correlation between the possibility of accurate prediction and the amount of (uncontrollable) variables. And certainly, not just political science but all the social sciences - if indeed it is even accurate to call them "sciences" - labour under the difficulty of trying to understand phenomena which occur as the result of an infinity of uncontrollable, unknowable stimuli. And this is not even to factor in the possibility that humans have something like genuine free will. If we do - if volition in some way is not constrained by physical laws - then the possibility of ever being able to accurately predict social or political phenomena - history maybe I should say - seems even more daunting, if not impossible.
To attempt predictions where the variables are few and the relevant laws may be inferred, as Thales showed with his eclipse, makes sense; but to attempt predictions where the variables are, practically speaking, infinite, and the relevant "laws", if there even are any, are not only unknown but probably in principle unknowable, as in the case of an election many months in the future whose candidates have not even been selected - appears so impossible in principle that only a complete idiot would try :P.