@georgeob1,
georgeob1 wrote:
I Googled "Bradley Effect to see what this bit of jargon means. Apparently it refers to excuses pollsters offer for the failures of their polls & statistical analyses to accurately predict outcomes. How scientific of them to blame the subjects of their sampling & analysis for the failures of their, at best, approximate forecasting methods.
Believe it or not, Georgeob1, pollsters are extremely aware of the approximaty (sp?) of their results. That's why they attach all those conditionalities about margins of error (and that's just the most basic one), and that's why they always go on about how their polls do
not, in fact, constitute "forecasting methods" and are
not, in fact, intended to "predict" any outcome, but merely measures of the current opinion at the moment they were taken. (How the media runs with them is a different story.)
This near-obsession, in fact, with the "known unknowns and unknown unknowns", so to speak, involved in polling has led pollsters, statisticians etc to analyse their data and analyse any divergent patterns. The Wilder/Bradley effect, as a thesis, is one of those analyses. It's based on extensive comparisons between all the ways in which polling data diverge from actual results; they can diverge in countless ways, but when there's a sudden, collective spike of divergency in one direction, across all pollsters, they will look again. At some point, it turned out that the same, non-random, collective divergence took place time after time when a black candidate ran - different from all the regular ways in which results diverged from polls. Not just statistical noise.
Anyway, you're not interested in any of this. Which is perfectly your right. But if you prefer not to learn about the analyses underpinning a thesis, could you please just be honest and say, hey, I dont know bout all this and frankly, I dont care -- so I cant say anything about this? Rather than first, choose to be ignorant about the subject, and then sweepingly claiming that anyway, it's all BS anyway? I mean, that's like me quickly Googling some economic theory, reading a para or two and saying - ah - I'm not interested in this stuff and I'm not going to bother about learning about it - but it's obviously all bunkum anyway.
Re the Bradley/Wilder effect, in any case, for those who
are interested -- it should be kept in mind that there is no conclusive evidence at hand that this effect is
still at work today. The theory dates, as the name indicates, from races that took place fifteen, twenty, thirty years ago; data from recent elections are much more ambiguous.