I’ve been continuing to try to incorporate information about the candidates into my forecast. However, this is proving to be pretty difficult at this stage since the polling data that we have about direct matchups (e.g., HRC vs. Trump) has questionable predictive value. That much I already knew.
But it gets even worse, I think. In a forecast earlier today, I used meta-polling data from RealClearPolitics and concluded that, given the likelihood of the various matchups, the Democrats had only a very slightly better chance at the White House than the Republicans. However, Pollster’s meta-polling data tells a rather different tale than that of RealClearPolitics. Behold,
- Clinton beats Cruz, 49% to 44%
- Clinton beats Rubio, 47% to 46%
- Clinton beats Trump, 48 to 44%
- Bernie beats Cruz, 50% to 39%
- Bernie beats Rubio, 47% to 41%
- Bernie beats Trump, 51% to 41%
That covers (by my risk averse estimate) 97% of the likely matchups, and it paints a much rosier picture for the Democrats than the RealClearPolitics poll of polls. Sure, the margins are not that big in some of these matchups (e.g., a mere 1% for Clinton vs. Rubio), but I’d rather be leading with a small margin than trailing by a small margin.
But what’s really problematic (as far as I’m concerned) is that a forecast based on the RealClearPolitics and a forecast based on Pollster look so different, and I don’t really have any good reason to view one or the other of these sources as being (much) more reliable. So it’s not just what I already knew: that the present data has limited predictive value. It’s that even this data is quite ambiguous. What I take away from all of this is that I probably started to take the inside view too early. Long term data concerning the way in which Americans vote for representatives of the two major parties is probably the best bet right now.