What do you make of the significant discrepancy between small media election polls and big media election polls?

It is just oo hard to tell. It could well be a difference in the quality of the pollsters the smaller outlets can afford.

It could also be caution on the big media site’s parts because of the problem in calling some states last election (which they took heat for) and the Obama factor that turned up in the primaries this go round, which they are all now compensating for. See this MSNBC article.

Excerpt: “NEW YORK – Barack Obama’s tendency through the Democratic primaries to perform better in exit polls than he actually does at the ballot box has some media organizations nervous heading into Election Night.

Television networks want to avoid having their performance become an issue for the third straight presidential election. Their political experts hope that experience gained during the primaries will help things run smoothly Nov. 4.”

Colbert (or Stewart) had a guest pollster on the show. The pollster actually specializes in analyzing baseball statistics, but decided to look at 2008 election polls. According to him:
a All polls are biased (either intentionally or by accident)
b Fox TV polls are least biased of the large television polls. (Not talking commentators)

The pollster jvgr is referring to Nate Silver, who runs FiveThirtyEight which is a remarkable website that analyzes to polls to make meta projections. He has some sort of complicated system that ranks polls according to how recently they were taken and how often polls from that source have been accurate. It is one of my favorite sites on the web recently—I check it daily.

Incidentally, I don’t know who’s correct, but FiveThirtyEight contests that phenomenon that Marina’s MSNBC article mentions, saying:

“As we have described here before, polling numbers from the primaries suggested no presence of a Bradley Effect. On the contrary, it was Barack Obama—not Hillary Clinton—who somewhat outperformed his polls on Election Day.”

Most of the time, it would seem that larger polls with more of a sample group would necessarily be more accurate. The trouble with smaller polls is that sometimes they’re from a somewhat biased group, too, which can slant the results more, especially without a large enough sample to balance that out.