Talk Elections

The record of constituency polling in all countries that have single-member districts verges on the outright appalling. This is as true of the U.S as everywhere else.

Not saying that we should ignore all the polls of House districts, but it's something worth keeping in mind.

Sssh....don't give it away....

In 2006 and 2008, House polls taken within 15 days of the election were within 5 points of the final outcome 81% of the time (http://www.huffingtonpost.com/thomas-m-holbrook/are-polls-junk_b_750318.html).  That's lower than gubernatorial (88%) and Senate (90%) polling - but still significantly more correct than not.

Also, looking at the Pollster.com database this information was taken from, polls used to create this info for 2008 House races are 45.1% partisan/internal polls, while 2008 Gubernatorial race polls are only 28.8% partisan/internal.

In addition, while I was only counting through to determine the above, I did notice that the House elections had a much higher occurrence of random uni pollsters and such compared to the Gubernatorial elections.

I didn't look through the Senate election polls but I'd imagine they'd be very similar in composition to the list of Gubernatorial polls.

I think the reduced accuracy of House polls found in your link, cynic, can be mostly explained by pollster quality.

I know that single-member district polls throughout the world always have some accuracy problems, which is to be expected when polling a district of, say, around 100,000 people. I think this error would be much more reduced in US House elections, however, compared to elsewhere, with the average Congressional district population being greater than the four least populous states- and I don't think polls in AK, ND, VT, or WY are any less accurate than NY or CA or what-have-you.

Most other countries also have far fewer constituency polls... and let's face it, only by doing it repeatedly can pollsters get any good at it.