Quote:
Originally Posted by swinginglory
Hi Nich,
Don't blame the messenger. First I didn't say anything about predicting election margin. There seems to be a perfect 12 for 12 correlation (for some reason FDR's1944 election isn't in the data although he had a 70% rating in 1944) with an incumbent's favorables and his re-election prospects. I'm not a statistician but I don't see how 12 for 12 isn't significant.
12 for 12 does suggest a high correlation between approval ratings and re-election chances. I don't think anyone is disputing that. What we are disputing is that elections are a "thumbs up or thumbs down affair".
For instance, suppose we were doing a similar analysis but for the correlation between speed and fatalities for car crashes. However, we could only get data on 12 crashes. In 9 of those crashes, the cars were travelling above 100km/hr at the time of the crash and all occupants died. In the other 3, the cars were travelling below 100km/hr and none of the occupants died. Could we conclude that speed is correlated with car crashes? Most likely. But could we conlude that car crashes above 100km/hr are guaranteed to result in death but those below 100km/hr are guaranteed not to? Not at all. Even a weaker conclusion like "crashes are much more likely to be fatal when a car goes beyond 100km/hr" isn't supported.
The margin is important because when you're got so few data points of close elections, it provides additional information. If elections really are a "thumbs up or thumbs down affair", then approval ratings should be perfectly correlated with election margins. They aren't. If it's possible for the margins of victory of incumbents with approvals in the mid-50s to vary by double digits, then it should also be possible for candidates with approvals in the mid-40s.
Quote:
Cliffs: NO incumbent below 49% has ever won and no incumbent above 50% has ever lost. Could an incumbent win with a 48-47% approval, yeah probably, but the farther he drops below 50% the more difficult his task becomes.
And , of course, it goes without saying that Nate is a big lib, so he puts the best face on horrid data for his team. Totally standard, but he has the intellectual chops to present the data, which doesn't lie.
And I don't think anyone's disagreeing with the bolded part either. Is it hard to win with 42% approval? Yes, very. But it's possible. And one of the ways in which it may be possible is with an opponent who is even more unpopular. Do you agree? If you do, are you repudiating the simplistic "thumbs up or thumbs down affair" mantra?