What's with the Rasmussen Reports?

Why is the Rasmussen Reports always different from all of the other polls, most times even Fox News? What do they know that no other polls know?

From what I can tell, Rasmussen uses more automated polling methods than competitors, which some will say is less accurate than those using live polling, but others say give less biased results. The results seem pretty good to me though, Rasmussen polls are usually pretty good so I think traditional polling methods are just behind the times. In 2004 Slate pretty much crapped all over Rasmussen before and during the elections then afterward admitted that they were among the most accurate. You’ll see those kinds of reactions often to Rasmussen polls over their history.

Rasmussen is rated a C+ by fivethirtyeight.com for being correct 79% of the time. It does has a bias toward Republicans, but as that chart indicates the vast majority of firms have biases toward one side or the other.

In the short term, remember that polls do have a tendency to swing toward the party having the convention. The polls over the next two weeks are likely to have large name recognition sways. Try not to pay attention anything any poll comes out with until mid-August, and then use that only as a baseline for future trends.

Or wait until the polls come out on Nov 9th … those are usually the best and most accurate.

Hmm… you’d think as a time traveler and con artist, he’d be able to do better than 79%… :wink:

Polling hasn’t really figured out that the number of landline phones has declined precipitously since 2012… except among Republicans.

Rasmussen has a significant “house effect.”

In 2010 they were simply awful.

And in 2012

From the 2010 election the explanation was:

They give different and often outlier results because they use (cheap) non-standard techniques. Mind you not all that is non-standard is poor. Internet polling is nonstandard but some of them have been performing well. But Rasmussen has had fairly few good years (2004 and 2006) and quite a few poor ones.

Will their non-standard cheap techniques more accurately reflect the make-up of the actual voting population than more standard approaches this cycle?

It’s not impossible. The miracle of aggregation though keeps an outlier in its place.

Polling is more an art that a science.

The problem is to figure out how well your polling population matches the real population. If, say, your poll has 5% Blacks, and the population in that state is 10% Black, you have to make adjustments to compensate. There are also different way polls define likely voters, which becomes important as election day approaches.

Rassmussen is not doing anything differently from other polls, but their adjustments tend to skew Republican. (Note: I’m not saying their deliberately skewing the data; it’s just that their formulas just work out that way.)

That much is clearly not true. See cites above.

The differences are much deeper than how they correct the demographics.