hi dig,
if google does it, then my guess is it is probably done by web search behaviour.
in general, it's still an early stage science.
but i am lazy. i leave it to the quants to figure out. in spite of what the pundits will tell you, he doesn't provide a simple average. for instance, he doesn't take a rasmussen poll at face value. he adjusts for the house effect (ie bias). i am pretty sure he also prioritises polling organisations. in his shoes, i would simply ignore data from folks who were consistently far off the mean.
by the way, sam wang at princeton has also adopted 332 this morning.
so now we have some quants at 303 and some at 332.
| 